Evolutionary hijacking of fatty acid biosynthesis drives insecticidal acylsugar diversity in tomato

Abstract

The evolution of plant defensive specialized metabolites often involves repurposing existing primary metabolic pathways for novel roles. However, the relative contribution of changes in protein function versus gene regulation and localization in driving such divergency remains less well understood. Here, we investigate the evolutionary origin of long-branched-chain acylsugars—a class of insecticidal metabolites found in wild tomatoes but absent from their domesticated counterpart. We show that this chemical divergency is enabled by a single enzyme, an acyl-CoA synthetase (SpBACS1), that was repurposed to bridge two distinct primary metabolic pathways. We demonstrate that SpBACS1 activates products of amino acid catabolism and primes them as non-canonical starters for elongation by the plastidial fatty acid synthase machinery for acylsugar assembly. Strikingly, the loss of this trait during domestication was not due to impaired enzyme function. Instead, we describe two distinct regulatory mechanisms. First, the cultivated ortholog, SlBACS1, lacks expression in acylsugar-producing trichomes due to promoter sequence divergence, despite retaining enzyme catalytic activity. Concurrently, their paralogs, BACS2, were neolocalized to the mitochondria, functionally isolating them from the chloroplast-based primary fatty acid metabolic elongation pathway. These findings demonstrate how divergence in cis-regulatory elements and subcellular targeting, in the absence of protein function modification, were potent drivers of metabolic evolution, providing a strategy for re-engineering valuable chemical diversity into cultivated crops.

Long-branched-chain acylsugars are insecticidal metabolites found in wild tomatoes but not their domesticated counterpart. Here the authors show changes in expression and localization of a key enzyme is responsible for this difference.

Introduction

The co-evolutionary arms race between plants and insects has fueled the evolution of a vast and complex chemical repertoire, from the bitter alkaloids in coffee to the fiery capsaicin in chili peppers. A central question in evolutionary biology is how this chemical novelty arises. Rather than inventing entire biochemical pathways de novo, evolution often relies on the principle of “evolutionary tinkering”—the recruitment and repurposing of conserved enzymes and pathways from primary metabolism^1,2. This metabolic innovation is achieved through genetic changes that alter the function, expression pattern, or subcellular localization of existing enzymes. As a result, ubiquitous primary metabolites are transformed into potent defensive compounds, providing plants with an adaptable toolkit to navigate relentless herbivore pressure^3,4.

Glandular trichomes, specialized secretory structures on the plant surface, are hotspots of this evolutionary innovation, functioning as biochemical factories⁵. The striking diversity of specialized metabolites produced in the glandular trichomes of cultivated tomato Solanum lycopersicum and its wild relative Solanum pennellii makes this system, including the introgression lines (ILs) and the backcross inbred lines (BILs) derived from them^6,7, exceptional tools for dissecting the genetic and biochemical basis of metabolic divergency. Acylsugars are a prominent Solanaceae (Nightshade family) trichome^8,9 and root^10,11 defensive compounds —a class of metabolites consisting of a sugar core esterified with multiple short- to medium-length acyl chains. These sticky, insecticidal compounds deter a wide range of arthropod pests^12–16 and can also protect against fungal pathogens and desiccation^17,18.

The structural diversity of acylsugars results from variation in the number, length, and branching patterns of their acyl chains, and this chemical variation is a major contributor to the efficacy of acylsugar-mediated plant defense¹⁵. The final assembly of these molecules is a two-step process: fatty acyl groups are first activated to their high-energy acyl-coenzyme A (acyl-CoA) forms, which then serve as substrates for a suite of Acylsugar Acyltransferases (ASATs) that sequentially attach the chains to the sugar core^8,18–20. The types of acyl-CoA precursors are therefore a primary determinant of acylsugar structure. In both cultivated and wild tomatoes, acylsugars contain a mixture of short, branched chains (e.g., C4, C5) and medium-length chains (e.g., C10, C12), the latter of which can be either straight or branched^21,22. This diversity raises fundamental questions about the biosynthetic origins of these acyl chain building blocks.

The biosynthesis of the short, branched acyl chains is well understood. These chains, such as isovaleryl (iC5) and isobutyryl (iC4) moieties, are produced through the catabolism of branched-chain amino acids (BCAAs)^23,24. This conserved pathway begins with a deamination step catalyzed by a branched-chain amino acid aminotransferase (BCAT), followed by oxidative decarboxylation via the branched-chain α-keto acid dehydrogenase complex located in the mitochondria, which directly yields the corresponding acyl-CoAs^23,25,26. While this pathway supplies the C4 and C5 precursors, it does not explain the medium- and long-branched chain fatty acids constituents of the acylsugars across the tomato clade.

The biosynthetic pathways responsible for these longer acyl chains are more complicated and have not been fully characterized. Medium-chain fatty acids (MCFAs; C8–C12) are valuable precursors for industrial chemicals but are typically rare in plant vegetative tissues, as their accumulation can disrupt membrane integrity. They are, however, abundant in the seed storage oils of certain species, such as California bay (Umbellularia californica) and coconut (Cocos nucifera), where MCFA production is achieved by specialized acyl-ACP thioesterases that catalyze the premature termination of the plastidial fatty acid synthase (FAS) cycle^27–29. In tomato trichomes, by contrast, evidence for a similar thioesterase-driven mechanism for medium-chain biosynthesis is lacking. While silencing genes for the FAS components 3-ketoacyl-ACP synthase (KAS) and 3-ketoacyl-ACP reductase in the wild tomato S. pennellii confirms the involvement of the core fatty acid synthesis machinery^26,30,31, a recent study supports a model where these medium chains are generated via the β-oxidation-mediated shortening of abundant C16/C18 fatty acids³².

Beyond straight-chain medium acyl groups, wild tomatoes—particularly S. pennellii—produce acylsugars with distinctive long-branched-chain fatty acyl moieties²². These structures are unusual in plants but have analogs in bacterial membranes, where they increase membrane fluidity without the easily oxidizable double bonds found in unsaturated fatty acids³³. The long-branched chains were previously hypothesized to be derived from mitochondrial BCAA catabolism products^23,26,34, but until now, no genes or proteins specifically involved in their biosynthesis were identified. The terminally-branched medium-length fatty acyl chains iC10 and iC11 within acylsugars present a particularly interesting case—these structures suggest a specialized elongation pathway that preserves terminal branching during carbon chain extension. Understanding the biosynthetic origin of these compounds represents a significant knowledge gap in deciphering acylsugar diversity and its evolutionary significance.

In this study, we identified an S. pennellii branched acyl-CoA synthetase (SpBACS1) that activates the short branched-chain fatty acid iC5 to form iC5-CoA, and provided strong evidence that these products serve as a primer for fatty acid elongation machinery to produce the medium-length branched chains iC11 and iC13. These elongated branched chains are subsequently incorporated into acylsugars, generating the defensive compounds characteristic of the wild tomato S. pennellii. We discovered that the unique capacity of SpBACS1 to generate branched-chain acylsugars stems from its localization in the chloroplast. This specific location enables it to directly provide iC5-CoA as a primer to the plastidial fatty acid elongation pathway. The cultivated tomato ortholog has undergone promoter sequence divergence that alters its expression pattern, resulting in the absence of these unique trichome acylsugars. In contrast, the paralogous gene SpBACS2 in S. pennellii was unable to recapitulate the IL2-5 phenotype, with differences in subcellular targeting potentially contributing to this outcome. Our findings reveal how evolutionary changes in enzyme expression and localization without changes in catalytic function can drive metabolic divergency, by repurposing primary metabolism for the biosynthesis of specialized defensive compounds.

Results

The tomato introgression line IL2-5 produces acylsugars containing long-branched-chain moieties

Trichome metabolite screening was performed on tomato ILs derived from crossing S. lycopersicum M82 with S. pennellii LA0716. This analysis revealed that IL2-5 generates a distinct acylsugar profile compared to the M82 parent. While total acylsugar levels in IL2-5 remained similar to M82, with the most abundant acylsugars largely unchanged (Figs. 1A and S1A), four additional acylsugars showed significant increases in IL2-5 (Figs. 1A and S1B–E). Mass spectrometry fragmentation analysis identified these compounds as S3:21(5,5,11), S4:23(2,5,5,11), S3:23(5,5,13), and S4:25(2,5,5,13), each containing an odd-numbered long acyl chain with 11 or 13 carbons, as supported by LC-MS/MS fragmentation patterns (Fig. S1F). In this acylsugar nomenclature, S3:21(5,5,11) represents a sucrose with 3 acyl chains totaling 21 carbons, with chain lengths of 5, 5, and 11 carbons.

Fig. 1. The tomato introgression line IL2-5 produces long-branched-chain-containing acylsugars.

A Comparative LC/MS analysis of trichome acylsugars from cultivated tomato M82 and IL2-5. Left panel: Representative electrospray ionization negative (ES−) mode extracted ion chromatogram (EIC) of the main acylsugars from M82 and IL2-5. Right panel: Magnified EIC showing four acylsugars—S3:21(5,5,11), S4:23(2,5,5,11), S3:23(5,5,13), and S4:25(2,5,5,13)—present in IL2-5 but not detectable in M82. Chromatograms are normalized to the largest peak in each comparison. Selected ion masses of the corresponding acylsugars are shown in the parentheses following the EIC, with maximum ion counts indicated in the upper left of each panel. B Analysis of acylsugar acyl chain structures. Left panel: Workflow for tomato leaf acylsugar characterization, including extraction, concentration, ethyl esterification, and GC-MS analysis. Middle and right panels: GC-MS electron ionization (EI) total ion chromatograms (TIC) of ethyl esters derived from C11 and C13 acyl chains, respectively. The chromatograms of the terminal-branched fatty acid (iC11 and iC13) and straight-chain fatty acid (nC11 and nC13) ethyl ester standards are shown in the third and fourth panels. The bottom chromatograms show the ethyl esters derived from IL2-5 acylsugars spiked with iC11 and iC13 ethyl ester standards, respectively. Light green bands highlight peaks corresponding to ethyl esters of C11 and C13 acyl chains from IL2-5 acylsugars. Part of the left panel in this figure (B) was created in BioRender. Fan, P. (2026) https://BioRender.com/x8dzwga.

To determine the acyl chain indentity in the original C11 and C13 acylsugars, we analyzed their structures through ethyl transesterification of acyl chains from IL2-5 and parental M82 leaf acylsugars, followed by GC-MS analysis (Fig. 1B). The resulting C11 and C13 ethyl esters from IL2-5 co-eluted with standards for the ethyl esters of the terminal branched fatty acids—9-methyl decanoic (iC11) acid and 11-methyl lauric (iC13) acid—rather than with straight-chain nC11 and nC13 standards. To further exclude matrix-related retention-time shifts, authentic iC11 and iC13 standards were spiked directly into the IL2-5 acylsugar-derived samples. In each case, the spiked samples showed a single co-eluting peak at the same retention time as the corresponding endogenous IL2-5-derived peak and authentic standard (Fig. 1B). These results support the conclusion that the C11 and C13 acyl chains in IL2-5 acylsugars are terminally branched iC11 and iC13 chains, rather than straight-chain nC11 and nC13 structures.

As supplementary evidence, we tested whether exogenous application of straight-chain fatty acids (nC11 or nC13) could generate acylsugars with chromatographic behavior distinct from the endogenous C11- and C13-containing species. After application of nC11 or nC13 fatty acids to IL2-5 leaves, we detected new peaks with m/z values matching those of the endogenous S3:21(5,5,11)/S3:23(5,5,13) and S4:23(2,5,5,11)/S4:25(2,5,5,13) acylsugars, respectively, but with slightly later chromatographic elution times (Fig. S2). Although these feeding experiments were limited by low signal intensity, the distinct chromatographic behavior of the nC11- and nC13-derived products is consistent with the GC-MS and spike-in results, further supporting the branched-chain identities of the endogenous IL2-5 acylsugars. The biosynthetic origin of these long-branched acyl chains and the genetic basis underlying this chemical variation between cultivated and wild tomatoes were subjected to further investigation.

SpBACS1 mediates long-branched-chain acylsugar production in IL2-5

As a first step toward elucidating the biosynthetic origin of these long-branched acyl chains and their underlying genetic basis, we conducted fine mapping in IL2-5 to identify the responsible genes. As shown in Fig. 2A, fine mapping initially localized the region to the overlapping segment between IL2-4 and IL2-5, which was subsequently refined to a 960-kb interval using BIL6661 and BIL6351. The specific breakpoint positions of these two BILs—both of which produced the characteristic long-branched-chain acylsugars—enabled the narrowing down of the candidate region within the IL2-4/IL2-5 overlap, which was further refined by self-designed INDEL (insertion-deletion) genetic markers (Fig. 2A and Supplementary Data 6).

Fig. 2. SpBACS1 is responsible for the IL2-5 long-branched-chain acylsugar phenotype.

A Genetic mapping of loci controlling long-branched-chain acylsugar production in IL2-5. The responsible chromosomal region was initially mapped using IL2-4, IL2-5, and selected backcross inbred lines (BILs) that have an acylsugar phenotype similar to IL2-5. The region was further narrowed using custom INDEL markers. Candidate genes were then filtered using S. pennellii LA0716 wild tomato trichome transcriptomic datasets, revealing two acyl-CoA synthetases with significantly higher expression in trichomes compared to shaved stems, designated Branched-chain-acylsugar-related Acyl-CoA Synthetase1 and 2 (BACS1 and BACS2). The red-blue color gradient represents relative expression levels measured as Fragments Per Kilobase of transcript per Million mapped reads (FPKM). B Representative LC/MS extracted ion chromatograms (EICs) of the four long-branched-chain acylsugars from M82, IL2-5, and M82 transgenic plants expressing either SpBACS1 (two independent lines, #13, #24) or SpBACS2 (two independent lines, #3, #4) driven by their native promoters. Chromatograms are normalized to the largest peak across all samples. Selected ion masses of the target acylsugars are shown in parentheses following the EIC, with maximum ion counts indicated in the upper right of the figure. C Quantitative comparison of the four specific acylsugars across M82, IL2-5, and transgenic independent lines expressing SpBACS1 (#13, #24) or SpBACS2 (#3, #4). Values represent peak area ratios of acylsugar to internal standard (IS), normalized by leaf dry weight (DW). Data presented as means ± SEM, n represents biologically independent replicates. Statistical significance determined by unpaired two-tailed t-test with Welch’s correction.

Because previously characterized acylsugar biosynthetic enzymes are expressed in cultivated tomato and S. pennellii trichome tip cells, we screened our candidate region for trichome-enriched genes using available tomato trichome transcriptomic datasets³². These published datasets compared expression among isolated trichomes and shaved stems (stems with trichomes mechanically removed)³². This analysis revealed two candidate acyl-CoA synthetases with significantly higher Fragments Per Kilobase of transcript per Million mapped reads (FPKM) in trichomes compared to stems with trichomes removed (“Shaved stem” in Fig. 2A). We named these candidates Branched-chain-acylsugar-related Acyl-CoA Synthetase1 and 2 (BACS1 and BACS2) (Fig. 2A and Supplementary Data 1).

To determine the inheritance pattern of the branched-chain acylsugar trait, we analyzed F₁ plants derived from crosses between BIL6661 and M82. These F₁ plants produced branched-chain acylsugars similar to BIL6661, indicating dominant gene action (Fig. S3). We took advantage of this observation to functionally validate these candidates by transferring the wild tomato alleles (SpBACS1 or SpBACS2) into cultivated tomato M82 under the control of their native promoters. Transgenic M82 plants expressing proSpBACS1::SpBACS1 (two independent lines #13 and #24) recapitulated the IL2-5 phenotype, producing the four iC11 or iC13-containing acylsugars (S3:21, S4:23, S3:23, and S4:25) in amounts comparable to that seen in IL2-5, as supported by LC-MS/MS fragmentation patterns (Figs. 2B, C and S4). In contrast, transgenic plants harboring proSpBACS2::SpBACS2 (two independent lines #3 and #4) did not increase these branched-chain acylsugars, which remained at trace or near-undetectable levels (Fig. 2B, C).

These results demonstrate that SpBACS1 is sufficient to promote the accumulation of iC11- and iC13-containing long-branched-chain acylsugars in the M82 background. This finding raised the question of the underlying molecular mechanism—specifically, how the SpBACS1 enzyme contributes to the biosynthesis of these unique acyl chains.

SpBACS1 generates iC5-CoA as a primer for fatty acid chain elongation

Given the role of SpBACS1 in generating long branched-chain iC11 and iC13 in IL2-5, we initially hypothesized that SpBACS1 uses iC11/iC13 acid and coenzyme A (CoA) as substrates to produce iC11-CoA/iC13-CoA, which would be subsequently incorporated into acylsugars by ASATs. To test this hypothesis, we measured SpBACS1 enzyme activity using iC11 or iC13 as substrates. Surprisingly, SpBACS1 showed no detectable activity toward these long branched-chain fatty acids (Fig. 3A).

Fig. 3. Biochemical analysis and in vivo isotope labeling indicate that SpBACS1 generates iC5-CoA for fatty acid chain elongation.

A Substrate specificity of SpBACS1 across a broad spectrum of fatty acids varying in chain length and branching pattern with Coenzyme A (CoA) as co-substrate. Enzyme activities shown as means ± SEM (n = 3 independent experiments). B Enzyme kinetic analysis of SpBACS1 with iC5 and iC4 fatty acid substrates. Data presented as means ± SD, n represents independent replicates. C Abundance ratios of one- and three-¹³C-labeled acylsugars versus their non-labeled counterparts ([M + 1]⁻/M⁻ and ([M + 3]⁻/M⁻)) in IL2-5 plants fed with isotope-labeled [1-¹³C]-iC5 or non-labeled iC5 (CK). Data presented as means ± SEM, n represents biologically independent replicates. Statistical significance determined by unpaired two-tailed t-test with Welch’s correction. D LC-MS/MS fragmentation analysis of acylsugars S3:21(5,5,11) (upper panel) and S3:23(5,5,13) (lower panel), showing MS spectra of one-carbon labeled ([M + 1]⁻) versus non-labeled (M⁻) iC11 or iC13 ions. Spectra were obtained from IL2-5 plants fed with [1-¹³C]-iC5 or non-labeled iC5. Hypothesized positions of isotopic incorporation in iC11 or iC13 are highlighted by black dashed boxes. E Abundance ratios of two- and four-¹³C-labeled acylsugars compared with their non-labeled counterparts ([M + 2]⁻/M⁻ and [M + 4]⁻/M⁻) in IL2-5 plants fed with isotope-labeled 1,2-¹³C₂-sodium acetate or non-labeled sodium acetate (CK). Data presented as means ± SEM (n = 6 biologically independent samples). Statistical significance determined by unpaired two-tailed t-test with Welch’s correction. F LC-MS/MS fragmentation analysis of acylsugars S3:21(5,5,11) and S3:23(5,5,13), showing MS spectra of two-carbon labeled ([M + 2]⁻) versus non-labeled (M⁻) iC11 or iC13 ions. Spectra were obtained from plants fed with 1,2-¹³C₂-sodium acetate or non-labeled sodium acetate. Possible positions of isotopic incorporation in iC11 or iC13 are highlighted by black dashed boxes. G LC-MS/MS fragmentation analysis of acylsugars S3:21(5,5,11) from dual-isotope feeding experiments. Spectra show unlabeled (M⁻) and one-, two-, and three- carbon labeled ([M + 1]⁻, [M + 2]⁻ and [M + 3]⁻) iC11 fragment ions. Spectra were obtained from IL2-5 plants co-fed with [1-¹³C]-iC5 and [1,2-¹³C₂]-sodium acetate or with non-labeled iC5 and sodium acetate. H Schematic comparison between proposed SpBACS1-mediated long-branched-chain fatty acid synthesis for acylsugar production and classic de novo fatty acid biosynthesis.

To elucidate the substrate specificity of SpBACS1, we screened a broad spectrum of fatty acids varying in chain length and branching pattern. SpBACS1 exhibited minimal activity toward long-chain fatty acids with more than seven carbons, but displayed particularly strong activity toward the terminally branched isovaleric acid (iC5 acid) and isobutyric acid (iC4 acid) (Fig. 3A). These compounds are the free-acid counterparts to the CoA thioesters generated by the BCAA catabolic pathway previously documented in tomato trichomes^26,35. While SpBACS1 also showed in vitro activity toward short straight-chain fatty acids such as butyric acid (nC4) and valeric acid (nC5), these compounds, unlike their branched-chain counterparts, have not been identified as natural constituents of tomato leaves. Michaelis–Menten enzyme kinetic analysis revealed that SpBACS1 has lower K_m for iC5 (K_m = 36.2 μM, k_cat = 12.5 min⁻¹) compared to iC4 (K_m = 672.7 μM, k_cat = 86.4 min⁻¹) (Fig. 3B). These results indicate that SpBACS1 preferentially activates short branched-chain fatty acids to their corresponding CoA thioesters but shows no detectable activity toward long-chain fatty acids, including iC11 and iC13. Therefore, we hypothesized that the iC11 and iC13 chains are not activated directly, but are instead synthesized via elongation of an iC5-CoA primer generated by SpBACS1.

To test whether iC5 is incorporated into the iC11 and iC13 moieties of IL2-5 acylsugars in vivo, we fed IL2-5 leaves with [1-¹³C]-iC5, using unlabeled iC5 as a control, followed by acylsugar characterization. LC-MS/MS analysis revealed significantly higher ratios of ¹³C-labeled S3:21(5,5,11)ᴹ⁺¹, or S3:23(5,5,13)ᴹ⁺¹ versus non-labeled compounds ([M + 1]⁻/M⁻) in treated plants compared to the unlabeled control (Fig. 3C). Because these acylsugars contain two directly incorporated iC5 chains and one elongated iC11 or iC13 chain predicted to derive from iC5, all three acyl chains can theoretically incorporate the ¹³C label. Consistent with this expectation, the [M + 3]⁻/[M]⁻ ratio of S3:21(5,5,11) and S3:23(5,5,13) were also significantly increased, and the relative abundance of [M + 2]⁻ and [M + 3]⁻ isotopologues increased after [1-¹³C]-iC5 feeding (Fig. S5A). Negative-mode MS/MS fragmentation analysis of S3:21(5,5,11)^M+1 and S3:23(5,5,13)^M+1 revealed fragment ions at m/z at 186.16 and 214.19, corresponding to one-¹³C-labeled iC11 or iC13 [M + 1]⁻ ions, respectively (Fig. 3D). The intensity of these [M + 1]⁻ ions in [1-¹³C]-iC5-treated plants was significantly higher than the unlabeled control plants, providing evidence that carbon from iC5 is incorporated into the iC11 and iC13 chains of IL2-5 acylsugars.

We next compared labeling patterns among acylsugars with different acyl-chain compositions. Our model predicts that S3:21(5,5,11) and S3:23(5,5,13), which contain two iC5 chains and one iC5-elongated iC11 or iC13 chain, should incorporate more [1-¹³C]-iC5 than S3:22(5,5,12), which contains two iC5 chains and one nC12 chain not predicted to derive from iC5. As expected, we observed significantly higher [M + 1]⁻/M⁻ abundance ratios for the molecular ion of S3:21(5,5,11) and S3:23(5,5,13) compared to that of S3:22(5,5,12) after [1-¹³C]-iC5 feeding (Fig. 3C). MS/MS analysis further showed that, in S3:21(5,5,11)^M+2 and S3:23(5,5,13)^M+2, both an iC5 fragment and the elongated iC11 or iC13 fragment carried a single ¹³C label (Fig. S5B). By contrast, analysis of S3:22(5,5,12)^M+1 detected a labeled iC5 fragment at m/z 102.06, but no labeled nC12 [M + 1]⁻ ion (Fig. S5B). These results support the conclusion that iC11 and iC13 chains, unlike the nC12 chain, are biosynthetically derived from an iC5 precursor in IL2-5.

As a genetic-background control, we also fed [1-¹³C]-iC5 to M82 leaves. In M82, neither the unlabeled M⁻ signals nor the [M + 1]⁻ signals of iC11/iC13-containing acylsugars were detected after [1-¹³C]-iC5 feeding (Fig. S5G). However, the iC5 fragment of S3:22(5,5,12) was labeled, whereas its nC12 fragment was not (Fig. S5F). These results indicate that exogenous [1-¹³C]-iC5 was taken up by M82 and incorporated into endogenous iC5 acyl chains, but was not converted into detectable iC11 or iC13 acyl chains in the absence of the SpBACS1-dependent pathway.

To determine whether iC5-derived precursors undergo two-carbon elongation to form iC11 and iC13 chains, we performed stable isotope labeling with [1,2-¹³C₂]-sodium acetate. Because iC5 contains six or eight fewer carbons than iC11 or iC13, respectively, conversion of iC5 to these long-branched-chain products would require repeated two-carbon elongation of the fatty acid biosynthetic machinery. After applying ¹³C₂-sodium acetate to IL2-5 plants, we observed significantly increased ratios [M + 2]⁻/M⁻ and [M + 4]⁻/M⁻ for both S3:21(5,5,11) and S3:23(5,5,13) compared to the control (Figs. 3E and S5C). In contrast, the [M + 2]⁻/M⁻ ratio of S3:15(5,5,5), an acylsugar composed exclusively of C5 chains, was not increased (Fig. 3E). Negative-mode MS/MS analysis of S3:21(5,5,11)^M+2 and S3:23(5,5,13)^M+2 using Q-TOF further detected fragment ions at m/z 187.16 and 215.19, corresponding to two-¹³C-labeled iC11 or iC13 [M + 2]⁻ ions, respectively (Fig. 3F). These data indicate that acetate-derived two-carbon units are incorporated into the long-branched-chain acyl moieties.

Although Q-TOF MS/MS was sufficient to detect [M + 2]⁻ labeled iC11 and iC13 fragments, it was not sensitive enough to reliably detect the lower-abundance [M + 4]⁻ labeled long-chain fragments. To improve detection of higher-order labeled fragments, we extended the [1,2-¹³C₂]-sodium acetate feeding period to 4 days and analyzed the samples using high-sensitivity TIMS-TOF MS. This analysis detected iC11 fragment ions at m/z 187.16 and 189.17 from S4:23(2,5,5,11), corresponding to [M + 2]⁻ and [M + 4]⁻ labeled iC11 fragments, respectively (Fig. S5D). The [M + 4]⁻ iC11 fragment provides evidence that multiple acetate-derived two-carbon units can be incorporated into the same long-branched-chain acyl group. Fully labeled [M + 6]⁻ or [M + 8]⁻ fragments, which would require three or four consecutive labeled elongation cycles, were not detected, likely because their abundance remained below the detection limit even with TIMS-TOF analysis. Nevertheless, the detection of both [M + 2]⁻ and [M + 4]⁻ labeled iC11 fragments supports repeated incorporation of two-carbon units during elongation of the branched-chain precursor.

To further determine whether two-carbon units were incorporated during the elongation of iC5 precursor and rule out potential parallel biosynthetic pathways, we performed a dual-isotope feeding experiment in which IL2-5 leaves were simultaneously supplied with [1-¹³C]-iC5 and [1,2-¹³C₂]-sodium acetate. LC-MS analysis revealed increase relative isotopologue distributions of [M + 3]⁻ (Fig. S5E) and a diagnostic [M + 3]⁻ iC11 fragment ion in negative-mode MS/MS fragmentation analysis of S3:21(5,5,11)^M+3, demonstrating simultaneous incorporation of carbon from both labeled iC5 and labeled acetate into the same elongated branched acyl chain (Fig. 3G). Together with the single-labeling experiments, these data support a model in which iC11 and iC13 acyl chains are generated through repeated two-carbon elongation of an iC5-derived precursor.

Collectively, our results reveal a non-canonical fatty acid biosynthetic pathway in tomato trichomes, in which iC5-CoA generated by SpBACS1 serves as an elongation primer to produce long-branched-chain fatty acids, which are ultimately incorporated into acylsugars of specialized metabolism (Figs. 3H and S6).

Promoter-mediated expression differences, rather than enzyme activity, account for interspecific allelic variation between SpBACS1 and SlBACS1

The elucidation of the SpBACS1-mediated pathway raised a critical evolutionary question: why does the cultivated tomato accumulate almost undetectable level of these long-branched-chain acylsugars despite the presence of the orthologous SlBACS1 gene? The most straightforward hypothesis was that SlBACS1 had lost the enzymatic ability to activate short-chain fatty acids. However, this theory was not supported by enzyme assays: SlBACS1 showed a similar catalytic profile to SpBACS1, preferring short-chain over long-chain fatty acids (Fig. 4A). Although SlBACS1 had higher K_m values for iC5 and iC4 than SpBACS1 (170.4 and 756.5 μM, respectively), indicating lower apparent affinity for these branched-chain substrates, its catalytic efficiency was comparable to that of SpBACS1 (Fig. 4B). For example, the k_cat/K_m values for iC5 were 0.61 μM⁻¹ min⁻¹ for SlBACS1 and 0.35 μM⁻¹ min⁻¹ for SpBACS1(Fig. 4B). These results indicate that the biochemical activities are similar between SlBACS1 and SpBACS1, suggesting that a characteristic other than enzymatic differences account for the observed allelic variation.

Fig. 4. Promoter divergence, rather than enzyme activity, underlies the allelic variation of SpBACS1 and SlBACS1.

A Substrate specificity of SlBACS1 across a broad spectrum of fatty acids varying in chain length and branching pattern with CoA as co-substrate. Data presented as means ± SEM (n = 4 independent experiments). B Enzyme kinetic analysis of SlBACS1 with iC5 and iC4 fatty acid substrates. Data presented as means ± SD, n represents independent replicates. C Heatmap showing expression levels of two orthologous gene pairs, BACS1 and BACS2, in shaved stems, total trichomes, and type I/IV trichomes. Expression values represent averaged log₂(FPKM + 1) that are row-scaled. Because S. pennellii contains predominantly type I/IV trichomes¹⁶, its total trichome dataset was used to estimate type I/IV trichome expression. D Confocal fluorescence microscopy showing GFP expression driven by the S. pennellii BACS1 promoter (proSpBACS1::GFP in M82) localized to tip cells of type I/IV trichomes, while the S. lycopersicum BACS1 promoter (proSlBACS1::GFP in M82) does not confer detectable expression in these cells. Both promoters drove the GFP expression in the type VI trichomes. White arrowheads indicate tip cells of type I/IV trichomes. Representative confocal images are shown from three independent biological experiments that produced similar results. E Representative LC-MS extracted ion chromatograms (EICs) of four long-branched-chain acylsugars from M82, IL2-5, and three independent M82 transgenic lines expressing SlBACS1 under the SpBACS1 promoter (proSpBACS1::SlBACS1 in M82 #9, #15 and #18). Chromatograms are normalized to the most abundant peak to facilitate comparison. Selected ion masses are shown in parentheses following the EIC, with maximum ion counts indicated in the upper right of the figure. F Quantification of the four long-branched-chain acylsugars in M82, IL2-5, and independent M82 transgenic lines (proSpBACS1::SlBACS1 in M82 #9, #15, #18). Peak area ratios (acylsugar/internal standard) normalized by leaf dry weight (DW) are shown. Data presented as means ± SEM (n = 5 biologically independent samples). Statistical significance determined by unpaired two-tailed t-test with Welch’s correction.

We next asked whether SlBACS1 gene expression patterns mirror those of SpBACS1, which is enriched in wild tomato trichomes. Indeed, SlBACS1 showed enriched expression in total trichomes compared to shaved stems³², similar to SpBACS1 (Fig. 4C and Supplementary Data 2). However, analysis of M82 type I/IV trichome transcriptomic datasets³⁶—the specific trichome types responsible for acylsugar production—revealed that cultivated SlBACS1 has significantly lower expression compared to SpBACS1 in wild tomato type I/IV trichomes (Fig. 4C and Supplementary Data 2). Interestingly, SpBACS2, a paralogous gene that did not recapitulate the acylsugar phenotype when transgenically expressed in M82, showed similarly high expression in type I/IV trichomes of both species (Fig. 4C and Supplementary Data 2), highlighting the specific reduction of SlBACS1 expression in these acylsugar-producing structures. These expression differences suggested that SlBACS1 has undergone altered tissue-specific regulation compared to its wild tomato ortholog.

To directly visualize the expression patterns of SlBACS1 and SpBACS1, we generated transgenic M82 plants expressing GFP under the control of either SpBACS1 (proSpBACS1::GFP) or SlBACS1 (proSlBACS1::GFP) promoters (Fig. 4D). In proSpBACS1::GFP plants, fluorescence was detected in tip cells of both type I/IV and type VI trichomes (Fig. 4D). In contrast, proSlBACS1::GFP plants showed GFP expression only in type VI trichomes, which primarily produce terpenoids³⁷, but not in acylsugar-synthesizing type I/IV trichomes (Fig. 4D). This expression pattern explains SlBACS1’s enrichment in total trichome samples despite its absence from type I/IV trichomes.

The differential expression pattern of SlBACS1 and SpBACS1 in an isogenic cultivated tomato M82 background strongly indicates that sequence divergence in the promoter regions of these two BACS1 orthologs (Fig. S7) underlies their differential tissue-specific expression and the consequent differences in acylsugar profiles. To test this hypothesis, we attempted to recapitulate this phenotype in M82 plants by expressing SlBACS1 driven by the SpBACS1 promoter (proSpBACS1::SlBACS1). We isolated the 1.8 kb region upstream of the predicted SpBACS1 translational start codon from S. pennellii and used it to drive SlBACS1 expression in cultivated tomato. Indeed, the four iC11/iC13-containing acylsugars—S3:21(5,5,11), S3:23(5,5,13), S4:23(2,5,5,11), and S4:25(2,5,5,13), present in M82 only at trace levels under our analytical conditions, all consistently accumulated significantly higher iC11/iC13-containing acylsugars in these transgenic plants (three independent lines #9, #15, and #18), as supported by LC-MS/MS fragmentation patterns and similar retention time with those in IL2-5 (Figs. 4E, F and S8).

Comparative analysis of the promoter regions revealed multiple putative transcription factor binding motifs that differ between SpBACS1 and SlBACS1, including MYB and ERF/AP2 binding sites (Fig. S7 and Supplementary Data 3). These sequence differences likely account for the differential expression patterns observed between the two orthologs. These results collectively demonstrate that promoter sequence divergence, rather than enzymatic activity differences, underlies the allelic variation and reduced acylsugar diversity in cultivated tomato—a diversity that can be restored by reintroducing the wild tomato promoter sequence.

Divergent subcellular localization underlies functional differentiation between SpBACS1 and SpBACS2 paralogs

SpBACS1 and SpBACS2 are tandem duplicates on chromosome 2 of wild tomato S. pennellii; both are preferentially expressed in trichomes. Although paralogous genes with similar expression patterns typically function redundantly, our genetic complementation studies revealed that only SpBACS1 contributes to long-branched-chain acylsugar production in IL2-5 (Fig. 2B, C). To understand this functional differentiation, we investigated their biochemical properties. Enzyme activity characterization showed that SpBACS2 has slightly lower activity toward short-chain fatty acids compared to SpBACS1, with a K_m of 973.3 μM and k_cat of 44.2 min⁻¹ for iC4, as well as a K_m of 290.2 μM and k_cat of 27.8 min⁻¹ for iC5 (Fig. 5A, B). Despite these modest differences, the overall biochemical profiles of SpBACS1 and SpBACS2 were similar, suggesting additional factors contribute to functional divergence.

Fig. 5. Functional divergence of SpBACS1 and SpBACS2 paralogs in catalysis and subcellular localization in S. pennellii.

A Substrate specificity of SpBACS2 across a broad spectrum of fatty acids varying in chain length and branching pattern with CoA as co-substrate. Data presented as means ± SEM (n = 4 independent experiments). B Enzyme kinetic analysis of SpBACS2 with iC4 and iC5 fatty acid substrates. Data presented as means ± SEM (n = 6 independent experiments). C Confocal microscopy showing subcellular localization of GFP-tagged BACS proteins expressed in Nicotiana benthamiana leaf epidermal cells. SpBACS1-GFP and SlBACS1-GFP co-localize with chloroplast autofluorescence (ChAF), while SpBACS2-GFP co-localizes with the mitochondrial marker Mt-signal-RFP. White arrowheads indicate mitochondria identified by Mt-RFP fluorescence. Scale bars = 25 μm. Representative confocal images are shown from three independent biological experiments that produced similar results. D Validation of reduced gene expression of SpBACS1 and SpBACS2 following virus-induced gene silencing (VIGS) in S. pennellii by RT-qPCR. Plants infiltrated with empty pTRV2 vector served as controls. Data presented as means ± SEM (n = 10 biologically independent samples). Statistical significance determined by unpaired two-tailed t-test. E Quantification of selected acylsugars (iC10-containing and non-iC10-containing) in VIGS lines with silenced SpBACS1 or SpBACS2. Peak area ratios (acylsugar/internal standard) normalized by leaf dry weight (DW) are shown. Data presented as means ± SEM (n = 10 biologically independent samples). Statistical significance determined by unpaired two-tailed t-test. F Abundance ratios of one-¹³C-labeled acylsugars versus their non-labeled counterparts ([M + 1]⁻/M⁻) in S. pennellii plants fed with isotope-labeled [1-¹³C]-iC4 or non-labeled iC4 (CK). Data presented as means ± SEM (n = 4 biologically independent samples). Statistical significance determined by unpaired two-tailed t-test. G LC-MS/MS fragmentation analysis of S. pennellii acylsugars S3:18(4,4,10) (left panel) and S3:19(4,5,10) (right panel), showing MS spectra of one-carbon labeled ([M + 1]⁻) versus non-labeled (M⁻) iC10 ions. Spectra were from S. pennellii plants fed with [1-¹³C]-iC4 or non-labeled iC4. Hypothesized positions of isotopic incorporation in iC10 are highlighted by black dashed boxes.

To ask whether subcellular localization differences contribute to the functional divergence between SpBACS1 and its paralog, SpBACS2, we employed transient expression of GFP-BACS1 fusion proteins in Nicotiana benthamiana leaves. Both SpBACS1 and SlBACS1 fusions are localized to chloroplasts, the organelle where fatty acid elongation occurs (Fig. 5C). In contrast, SpBACS2 was targeted to mitochondria (Fig. 5C). This dramatic difference in subcellular localization is consistent with the requirement for the access to the plastidial fatty acid elongation machinery and may limit the ability of SpBACS2 to contribute to long-branched-chain acylsugar biosynthesis. In addition, SpBACS2 showed a higher K_m and lower catalytic efficiency toward iC5 than SpBACS1 and SlBACS1 (Figs. 3B, 4B, and 5B), suggesting that both subcellular localization and catalytic properties may contribute to the functional divergence between these paralogs.

Given the distinct subcellular localization of SpBACS1 and SpBACS2, we next examined whether these two paralogs contribute differently to long-branched-chain acylsugar accumulation in their native wild tomato background (S. pennellii LA0716). Notably, S. pennellii normally accumulates acylsugars containing an even-numbered branched-chain iC10 moiety²², differing from IL2-5. Indeed, using virus-induced gene silencing (VIGS), we found that silencing SpBACS1 in S. pennellii significantly reduced the levels of these iC10-containing acylsugars, including acylsucroses S3:18 (4,4,10) and S3:19 (4,5,10), as well as acylglucoses G3:18 (4,4,10) and G3:19 (4,5,10) (Figs. 5D, E and S9A). In contrast, silencing SpBACS2 had no effect on the acylsugar profile (Fig. 5E), consistent with our transgenic study results demonstrating that introducing SpBACS2 could not restore long-branched-chain acylsugars in cultivated tomato (Fig. 2B, C). Together with the localization data, these results support a model in which chloroplast localization is a key feature enabling SpBACS1 to access the plastidial fatty acid elongation machinery and promote long-branched-chain acylsugar biosynthesis. Conversely, mitochondrial targeting likely separates SpBACS2 from this plastidial pathway, explaining why it does not detectably contribute to long-branched-chain acylsugar accumulation in either native S. pennellii or transgenic M82, although SpBACS2 may have other biological functions outside the contexts examined here.

To investigate the ecological relevance of SpBACS1-mediated long-branched-chain acylsugar production, we performed insect bioassays using SpBACS1- and SpBACS2-silenced S. pennellii plants, as well as proSpBACS1::SpBACS1 transgenic M82 plants. In western flower thrips (Frankliniella occidentalis) preference assays, the thrips showed a consistent but non-significant trend toward increased feeding damage on SpBACS1-silenced S. pennellii leaves compared with empty-vector control leaves (p = 0.0908), whereas silencing SpBACS2 had no detectable effect (Fig. S10A). In the cultivated tomato background, the introduction of proSpBACS1::SpBACS1 into M82 did not significantly alter thrips feeding preference (Fig. S10B). We also tested cotton bollworm (Helicoverpa armigera) larval performance using no-choice feeding assays. Neither silencing SpBACS1 or SpBACS2 in S. pennellii nor expressing proSpBACS1::SpBACS1 in M82 caused a significant change in larval weight gain (Fig. S10C, D). These results indicate that altering SpBACS1-mediated long-branched-chain acylsugars alone has limited impact on herbivore resistance under the conditions tested, consistent with a model in which tomato trichome-based defense depends on the collective effects of multiple specialized metabolites.

The question remained as to how iC10, rather than iC11/iC13 acylsugars, are produced in S. pennellii LA0716 through the SpBACS1-mediated pathway. It was previously demonstrated that an enzymatically functional and feedback insensitive isopropylmalate synthase (IPMS3) C-terminal deletion mutant is involved S. lycopersicum of iC5 acylsugar chain biosynthesis. In contrast, an enzymatically inactive larger deletion mutant IPMS results in the accumulation of iC4 acylsugar chains in S. pennellii³⁵. These results are consistent with the hypothesis that wild tomato S. pennellii and cultivated tomato trichomes have predominant iC4 and iC5 pools, respectively. We hypothesized that in S. pennellii, SpBACS1 converts the iC4 to iC4-CoA in chloroplasts, which then undergoes three rounds of two-carbon elongation to produce iC10 for incorporation into acylsugars. To test this, we used stable isotope tracking by exogenously applying one-¹³C-labeled iC4 to S. pennellii plants. Negative-mode MS/MS fragmentation analysis of labeled S3:18 (4,4,10)ᴹ⁺¹ and S3:19 (4,5,10)ᴹ⁺¹ revealed a mass spectrum with m/z of 172.14, corresponding to one-¹³C-labeled iC10 [M + 1]⁻ ion (Fig. 5G). Support for the hypothesis that iC4 serves as the precursor for iC10 synthesis in S. pennellii was obtained because the abundance ratios in the molecular ion for S3:18 (4,4,10)^M+1 and S3:19 (4,5,10)^M+1 versus non-labeled forms ([M + 1]⁻/M⁻) were significantly higher in the labeled plants compared to the control (Figs. 5F and S9B). These results validate our hypothesis for the function of SpBACS1 in trichomes of its native S. pennellii and suggest that the presence of SpBACS1 in different genetic backgrounds can yield varied phenotypes due to substrate availability differences.

To investigate the evolutionary divergency of ACS genes that gave rise to the long branched-chain acylsugar-producing SpBACS1 in S. pennellii, we performed phylogenetic analysis of all annotated ACS genes from S. lycopersicum and S. pennellii (Fig. S11 and Supplementary Data 4 and 5). The trichome-enriched ACS genes, including several orthologous pairs, formed a distinct phylogenetic clade (Figs. 6 and S11 and Supplementary Data 2 and  4). Within this clade, SpAACS1 and SlAACS1—which participate in acylsugar long straight-chain biosynthesis but show no activity toward iC4 and iC5³²—are positioned evolutionarily further from the other three gene pairs. While most of these genes are enriched in type I/IV trichomes, only SpBACS1 and SlBACS1 are targeted to chloroplasts (Figs. 5C and 6 and Supplementary Data 2). Among this trichome-enriched ACS clade, only SpBACS1 satisfied all conditions—type I/IV trichome expression, iC4/iC5 activity, and chloroplast targeting (Figs. 3A, 4D, 6, and  7 and Supplementary Data 2)—ultimately enabling its participation in long branched-chain acylsugar biosynthesis in S. pennellii.

Fig. 6. Phylogenetic analysis of acyl-CoA synthetases (ACS) in tomato species reveals factors enabling long-branched-chain acylsugar biosynthesis.

Four orthologous pairs of ACS genes from S. lycopersicum and S. pennellii form a distinct clade within the broader ACS phylogenetic tree. Among the members of this trichome-enriched ACS clade, only SpBACS1 fulfills all three necessary conditions for long-branched-chain acylsugar biosynthesis in S. pennellii. These are high expressions in type I/IV trichomes, catalytic activity toward iC4/iC5 substrates, and chloroplast localization. The presence (“+”) or absence (“−”) of these key features is indicated for each gene. Trichome enrichment is defined as log₂ fold change (log₂FC) > 2 comparing trichome vs. shaved stem expression. High type I/IV trichome expression is defined as FPKM > 50. Subcellular localization of BACS1/2 proteins was experimentally validated in this study (Fig. 5C), while 03gACS1 localization was computationally predicted using TargetP 2.0. Enzyme activity and localization data for AACS1 were previously reported³².

Fig. 7. Proposed model for BACS gene divergence determining long-branched-chain acylsugar production in wild versus cultivated tomato.

The figure illustrates the proposed molecular mechanism underlying the presence of long-branched-chain acylsugars in S. pennellii (wild tomato) and their absence in S. lycopersicum (cultivated tomato). The left panel depicts a type I/IV trichome tip cell from S. pennellii or IL2-5, where chloroplast-localized SpBACS1 activates branched-chain amino acid (BCAA) breakdown products (iC4/iC5) by converting them to corresponding CoA thioesters. These activated intermediates enter the fatty acid elongation pathway to generate iC10, iC11, and iC13 acyl chains, which are ultimately incorporated into specialized acylsugars. Although SpBACS2 is also expressed in trichome tip cells, its mitochondrial localization prevents participation in the chloroplast-localized fatty acid elongation process. In cultivated tomato, promoter divergence led to the absence of SlBACS1 expression in type I/IV trichome tip cells, thereby preventing the production of long-branched-chain acylsugars despite the enzyme’s catalytic competence. Chloroplast and mitochondrial sizes are not drawn to scale relative to the cell but are enlarged for clarity of the illustrated biochemical processes. BCAT branched-chain amino acid transferase, BCKDH branched-chain-α-ketoacid dehydrogenase.

Discussion

This study reveals how two fundamental primary metabolic pathways—branched-chain amino acid catabolism and plastidial fatty acid synthesis (Fig. 3H)—can be bridged by a single repurposed enzyme to create a new group of defensive compounds. Unlike glucosinolate biosynthesis, which co-opts methionine elongation enzymes for precursor diversification³⁸, or benzylisoquinoline alkaloid pathways that hijack tyrosine catabolism machinery for alkaloid assembly^39,40, the SpBACS1-mediated pathway represents a different evolutionary strategy. Here, a single acyl-CoA synthetase functions as a metabolic bridge, activating short branched-chain products from amino acid catabolism and channeling them as non-canonical primers into the core fatty acid elongation machinery. This creates a non-canonical biosynthetic route where the conserved plastidial FAS—normally restricted to straight-chain fatty acid production—is co-opted to generate terminally branched MCFAs that would otherwise be toxic to normal cellular function⁴¹. Compartmentalization within glandular trichomes is crucial, as these specialized cells function as metabolic factories that safely accumulate compounds detrimental to vegetative tissues, enabling plants to explore otherwise inaccessible chemical space for defense. This pathway represents a textbook case of François Jacob’s concept of evolution and tinkering¹, where evolution creates novel chemical scaffolds not by inventing new enzymatic chemistry, but by connecting existing metabolic modules through a single bridging enzyme with specific tissular and subcellular targeting.

Our findings imply an underexplored system of organellar metabolite trafficking within plant trichome cells. Tomato acylsugar assembly, catalyzed by acylsugar acyltransferase (ASAT) enzymes, likely occurs in the cytosol of trichome tip cells, as these ASATs lack subcellular targeting signals. This cytosolic localization suggests a substantial pool of activated short branched-chain acyl-CoAs (e.g., iC4-CoA and iC5-CoA) must be available in this compartment. However, our discovery that a chloroplast-localized BACS1 is essential for producing the elongated chains suggests that these cytosolic acyl-CoAs are not directly available to the plastidial FAS machinery (Fig. 7). This aligns with the common barrier property of intracellular membranes to acyl-CoA transport, where fatty acid movement typically involves deacylation to free fatty acids before membrane passage and subsequent reactivation to acyl-CoAs in the destination compartment⁴². Our model posits that free iC4/iC5 fatty acids, likely derived from the hydrolysis of CoA thioesters produced during BCAA catabolism in the mitochondria, are transported into the chloroplast. Our results indicate that, once inside, SpBACS1 re-activates them into their CoA esters, delivering them directly to the fatty acid elongation machinery (Fig. 7). This compartmentalized activation of branched-chain acyl substrates by SpBACS1 represents a control point for directing carbon flux specifically into elongated acylsugar precursors in trichome specialized metabolism.

Although our data identify SpBACS1 as a key entry point for channeling short branched-chain fatty-acid-derived carbon into plastidial fatty acid metabolism, additional components are likely required to define the final chain lengths of these unusual acyl groups. One plausible mechanism is premature termination of elongation by specialized acyl-ACP thioesterases. Such enzymes are known to influence chain-length output in plastidial fatty acid biosynthesis by releasing acyl groups from ACP^43,44, and a thioesterase with activity toward branched medium-chain acyl-ACP intermediates could help generate discrete iC10, iC11, or iC13 products for subsequent incorporation into acylsugars. Another unresolved step is primer entry into the FAS cycle. In canonical de novo fatty acid synthesis, KAS III catalyzes the initiating condensation between an acyl-CoA primer and malonyl-ACP^43,44. Thus, the use of iC4-CoA or iC5-CoA as non-canonical primers may require a KAS III or KAS III-like activity capable of accepting branched short-chain acyl-CoA substrates. Therefore, SpBACS1 likely provides the metabolic bridge into fatty acid elongation, while additional, possibly co-evolved activities at the primer-entry and chain-termination steps help define the final chain-length distribution.

The comparative analysis of wild tomato S. pennellii and the cultivated tomato S. lycopersicum provides insights into the genetic mechanisms underlying the loss of chemical diversity during domestication. Wild relatives of modern crops are increasingly recognized as a vital reservoirs of robust defensive traits that have been inadvertently selected against during breeding for yield and quality^45,46. Our work pinpoints two distinct evolutionary mechanisms that can lead to the change of such traits. The first, more prevalent mechanism involves functional divergence through protein coding sequence changes, exemplified by the paralogs SpBACS1 and SpBACS2. These genes likely arose from a tandem duplication event, and their functional differentiation was achieved not by altering their catalytic properties, but by changing their N-terminal subcellular targeting signals (Fig. 5). This led to neolocalization, with SpBACS1 being directed to the chloroplast and SpBACS2 to the mitochondria, effectively isolating SpBACS2 from the plastidial FAS pathway and dedicating SpBACS1 to its role in acylsugar diversification (Fig. 5C). This illustrates how gene duplication followed by changes in subcellular trafficking can be a potent driver of metabolic evolution.

More strikingly, this study reveals a less frequently documented mechanism for attenuation of a metabolic trait: divergence in cis-regulatory elements leading to altered tissue-specific expression. Although the cultivated tomato ortholog SlBACS1 maintains enzymatic activity comparable to SpBACS1, cultivated tomato M82 accumulates little, if any, of these long-branched-chain acylsugars under our analytical conditions (Fig. 4A, B). The key distinction lies in its promoter sequence, which has diverged to eliminate expression in acylsugar-producing type I/IV trichomes while retaining expression in terpenoid-producing type VI trichomes^47,48 (Figs. 4C, D and S7). This finding exemplifies how changes in gene regulation, without altered protein function, can drive phenotypic evolution. The experimental recapitulation of the wild tomato metabolic trait in cultivated tomato by expressing the SlBACS1 coding sequence under control of the S. pennellii promoter underscores the substantial impact of cis-regulatory evolution (Fig. 4E, F). These results have direct practical implications for crop improvement, suggesting that targeted promoter modification or replacement could serve as an effective strategy to reintroduce valuable traits from wild relatives while minimizing pleiotropic effects.

Furthermore, this study highlights that the output of a biosynthetic pathway is also determined by the availability of precursors within a given genetic background. In the S. lycopersicum or IL2-5 background, which has relatively high levels of iC5-containing acylsugars³⁵, SpBACS1 facilitates the synthesis of iC11 and iC13 chains (Fig. 2B, C). In contrast, in its native S. pennellii context, where iC4 is the more abundant precursor³⁵, SpBACS1 primarily primes the synthesis of iC10 chains (Fig. 5). This plasticity demonstrates that the introduction of a single gene into a new metabolic network can yield distinct chemical phenotypes, a critical consideration for metabolic engineering studies.

This model also clarifies why C5-containing acylsugars in S. pennellii do not necessarily imply efficient iC5-dependent elongation. Previous structural analyses assigned the C5 moieties in major S. pennellii acylsugars such as S3:19(4,5,10) and G3:19(4,5,10) as anteiso-C5 (aiC5), rather iC5²². Because SpBACS1 showed strong activity toward iC5 but only weak activity toward aiC5 (Fig. 3A), these C5 acyl chains are unlikely to represent an efficient substrate pool for iC11/iC13 biosynthesis. Thus, the predominant production of iC10-containing acylsugars in S. pennellii likely reflects the combined effects of iC4 availability, weak utilization of aiC5 by SpBACS1, and the substrate flexibility of the enzyme. While low-abundance iC11/iC13 products in S. pennellii cannot be completely excluded, they are not major detectable products under our analytical conditions.

The evolutionary analysis of the trichome-enriched ACS clade further provides a framework for interpreting our findings. The capacity of SpBACS1 to produce long-branched-chain acylsugars appears to be associated with the combination of three key properties: trichome-enriched expression, catalytic activity towards short branched-chain acids, and chloroplast localization. The co-occurrence of these properties in SpBACS1 provides a mechanistic explanation for its ability to support long-branched-chain acylsugar production, a trait that is prominent in wild tomato S. pennellii but reduced in cultivated tomato (Fig. 6).

Our findings investigate the native biological role of SpBACS1-mediated long-branched-chain acylsugars in plant defense. In M82, iC11- and iC13-containing acylsugars had no effect on insect resistance (Fig. S10B, D). Silencing SpBACS1 in S. pennellii, where it produces iC10-containing acylsugars, led to a modest trend toward increased thrips feeding preference, although the variation is not significant (Fig. S10A). These results indicate that acylsugar-mediated defense is likely determined by the combined action of multiple acylsugar species rather than by any single acyl chain class alone. This chemical redundancy may contribute to the robustness of plant defense. Additionally, acylsugar compositional diversity has been proposed to provide adaptive advantages in natural environments⁴⁹. Such effects may be more relevant under ecologically complex conditions with multiple herbivores and environmental variables, which are not fully captured in simplified laboratory bioassays.

In conclusion, this study illuminates fundamental principles of metabolic evolution while providing insights on exploiting the genetic variations of wild crop relatives. The divergence of the SpBACS1-mediated pathway showed that simply altering promoter sequences or subcellular targeting can dramatically reshape specialized metabolism (Figs. 4, 5, and 7). As biotechnologists and synthetic biologists seek to enhance crop resilience, understanding the basis of specialized metabolic traits in wild relatives provides both inspiration and practical tools for metabolic engineering.

Methods

Plant materials and growth conditions

Seeds of cultivated tomato (S. lycopersicum cv. M82), wild tomato (S. pennellii LA0716), tomato ILs, and backcrossed inbred lines (BILs) were originally obtained from the C.M. Rick Tomato Genetic Resource Center (https://tgrc.ucdavis.edu/) and subsequently self-propagated for this study. F₁ hybrids were generated through hand-pollination using M82 as the pollen donor and BIL6661 as the female parent. These F₁ plants were used to characterize acylsugar traits and determine their dominance relationships. Seeds were germinated on moistened filter paper at room temperature before being transplanted to plastic pots containing nutrient soil. All plants were maintained in a controlled growth chamber at 60% relative humidity with a 14/10 h light/dark photoperiod (light intensity: 500 μmol m⁻² s⁻¹) and corresponding 25/20 °C day/night temperature schedule.

Acylsugar extraction and LC-MS analysis

Acylsugars from leaves of 4-week-old tomato were extracted following the protocol available at Protocols.io https://doi.org/10.17504/protocols.io.xj2fkqe. Briefly, one fully developed leaflet was collected from 4-week-old tomato plants and immersed in 1 mL of acetonitrile/isopropanol/water (3:3:2, v/v/v) containing 0.1% formic acid with 1 μM telmisartan as internal standard. Samples were gently agitated at room temperature for 2 min, and extracts were transferred to glass vials for LC-MS analysis.

Acylsugar extracts were analyzed using an Agilent 1290 Infinity II UHPLC system coupled to an Agilent 6545B Q-TOF mass spectrometer (Agilent Technologies). Mass spectrometer conditions were as follows: gas temperature, 275 °C; gas flow rate, 8 L/min; nebulizer pressure, 35 psig; sheath gas temperature, 300 °C; sheath gas flow rate, 11 L/min; capillary voltage, 3500 V; nozzle voltage, 0 V; fragmentor voltage, 150 V; and skimmer voltage, 65 V. Chromatographic separation was performed on a reverse-phase C18 column at 40 °C with 5 μL injection volume. Binary gradient elution using water (A) and acetonitrile (B) was applied at 0.3 mL/min flow rate: 0–3 min, 5–60% B; 3–15 min, 60–100% B; 15–18 min, 100% B; 18–18.1 min, 100–5% B; 18.1–21 min, 5% B. Mass spectrometric data were acquired in Auto MS/MS mode (m/z 50–1700) or Target MS/MS mode for isotopic labeling experiments. Data were analyzed using Agilent MassHunter Workstation software (version 10.0).

For MS/MS data of S4:23(2,5,5,11)^M+4, acylsugar were analyzed using an Agilent 1290 Infinity II UHPLC system coupled to Bruker timsTOF Pro 2 mass spectrometer. Chromatographic separation was performed on the same condition as other acylsugars. MS data were acquired using electrospray ionization in negative ion mode over 100-1350 m/z. The TIMS setting include: 1/k₀ from 0.55 to 1.9 V·S/cm³, ramp time 100 ms, accumulation time 100 ms. Other MS settings include: End plate offset 500 V, capillary 4500 V, nebulizer 2 bar, dry gas 8 L/min, dry temp 230 °C, sheath gas temperature 400 °C, sheath gas flow 4 L/min. Target MS data was acquired in prm-PASEF mode. Prm-PASEF settings include: m/z 769.4048, isolation width (Da) 2.0, RT (min) 10.2 min, RT range 0.2 min, 1/k₀ from 1.1 to 1.4 V·S/cm³. Collision energy was 40 eV. Raw data were processed using DataAnalysis (Bruker) for peak detection and integration.

Acylsugar compositions were assigned based on high-resolution LC-MS, negative-mode MS/MS fragmentation, and comparison with previously reported tomato acylsugar structures. The acylsugar nomenclature follows the standard format in which the letter indicates the sugar core, the first number indicates the number of acyl chains, the second number indicates the total number of acyl-chain carbons, and the numbers in parentheses indicate individual acyl-chain lengths. For newly detected C11- and C13-containing acylsugars in IL2-5, LC-MS/MS was used to determine the acyl-chain composition, while the branching patterns of the C11 and C13 chains were further confirmed by GC-MS analysis of released acyl chains using authentic straight-chain and terminally branched fatty acid standards.

GC-MS analysis of acyl-chain composition of acylsugar

The analysis of acylsugar acyl-chain composition was performed based on a modified protocol from a previous report³⁵. In brief, six tomato plants at 4 to 5 weeks old were selected, and their leaves were submerged for 3 min in an extraction solution consisting of isopropanol:acetonitrile (1:1, v/v) containing 0.01% formic acid. The extractions were evaporated to dryness under a gentle nitrogen stream. To derivatize the acyl-chains as ethyl esters, 400 μL of 20% (v/v) sodium ethoxide solution (Macklin, cat. no. S817522) was added to the dried sample. The reaction was carried out at room temperature for 30 min with vortexing every 5 min. After derivatization, 400 μL of n-hexane was added and vortexed for 30 s, followed by addition of 500 μL saturated sodium chloride solution. The mixture was thoroughly mixed and centrifuged at 13,000 × g for 5 min. The upper n-hexane layer was carefully transferred to a glass vial with glass insert for subsequent GC-MS analysis. Standards nC11:0 (Aladdin), iC11:0 (Jiangsu Aikon), nC13:0 (Aladdin), and iC13:0 (Aladdin) were used as controls for acyl-chain structure verification. To further exclude matrix-related retention-time shifts, iC11:0 and iC13:0 standards were spiked into the IL2-5 acylsugar-derived samples.

The fatty acid ethyl esters were analyzed using an Agilent 7890B gas chromatography system coupled with a 5977B mass selective detector (GC/MSD, Agilent Technologies). Samples were separated using a DB-5MS capillary column (40 m × 0.25 mm × 0.25 μm; Agilent Technologies). The injection volume was 2 μL with injector temperature set to 250 °C. Helium was used as carrier gas at a constant flow rate of 1.19 mL/min. The temperature program was: 36 °C for 4 min; 15 °C/min to 150 °C (no hold); 20 °C/min to 310 °C; 310 °C for 2 min. The mass spectrometer operated in electron ionization mode at 70 eV with a 6 min solvent delay. Ion source temperature was 230 °C, quadrupole temperature 150 °C, and interface temperature 250 °C. Data were acquired in full scan mode over m/z 30–500.

Plasmid constructs and generation of transgenic plants

The sequences of SpBACS1 (Sopen02g027670; XM_015208428), SpBACS2 (Sopen02g027680; XM_015208253), and SlBACS1 (Solyc02g082870; XM_004232503) were retrieved from the Sol Genomics Network database.

For the tissue-specific expression vectors, approximately 1.8 kb promoter regions upstream of the SlBACS1 and SpBACS1 genes were amplified using the primers SlBACS1-pro-F/R and SpBACS1-pro-F/R, respectively. The amplified promoter fragments were cloned into the pENTR/D-TOPO vector, followed by recombination into the vector pKGWFS7 via GATEWAY cloning.

To ectopically express SpBACS1 and SpBACS2 genes in the cultivated tomato (S. lycopersicum cv. M82) background, the genomic regions starting from ~1.8 kb upstream of the start codon and encompassing the entire gene sequences of SpBACS1 and SpBACS2 were amplified using the primers SpBACS1-pro-gene-F/R and SpBACS2-pro-gene-F/R, respectively. The fragments were cloned into pENTR/D-TOPO and then transferred into the expression vector pK7WG via GATEWAY cloning.

Additionally, to express SlBACS1 under the control of the SpBACS1 promoter in the M82 background, the full-length SlBACS1 gene was amplified using primers SlBACS1-gene-F/R. The ~1.8 kb promoter region upstream of SpBACS1, cloned with ProSp-Sl-F/R, was fused to the 5’ end of the SlBACS1 gene sequence. The resulting construct was cloned into pENTR/D-TOPO and subsequently introduced into the pK7WG vector via GATEWAY cloning.

All recombinant constructs were introduced into Agrobacterium tumefaciens strain GV3101 and transformed into cultivated tomato M82 using an Agrobacterium-mediated transformation protocol as previously described in refs. ^50,51. All transgenic plants were genotyped using self-designed primers. All primers described above are listed in Supplementary Data 6.

Protein expression and BACS enzyme activity assay

To express His-tagged SpBACS1/2, SlBACS1/2, Sp03gACS1, and Sl03gACS1 proteins in a prokaryotic system, specific gene primers were used to amplify the full-length open reading frames of each gene (Supplementary Data 6). The PCR products were cloned into the BamHI- and XhoI-linearized pET28b expression vector using One Step Infusion Mix (TOROIVD). The resulting constructs were transformed into Escherichia coli Rosetta 2 (DE3). When the culture reached an OD₆₀₀ of 0.4, protein expression was induced by adding 0.05 mM isopropyl β-D-1-thiogalactopyranoside, followed by incubation at 16 °C with shaking at 120 rpm for 12 h. His-tagged proteins were purified using a Ni-NTA affinity column (Smart-Lifesciences) following the manufacturer’s instructions.

BACS enzyme activity assays were conducted as previously described³². Fatty acid substrates (5 mM) were pre-dissolved in solution containing 5% Triton ×-100. The reaction premix contained: 0.1 M Tris–HCl (pH 7.5), 2 mM DTT, 5 mM ATP, 10 mM MgCl₂, 0.5 mM CoA, 0.8 mM NADH, 250 μM fatty acid substrate, 1 mM phosphoenolpyruvate monopotassium salt (Roche, cat. no. 10108294001), 20U myokinase (Sigma-Aldrich, cat. no. M3003), 10U pyruvate kinase (Roche, cat. no. 10128155001), and 10U lactate dehydrogenase (Roche, cat. no. 10127230001). Reactions were carried out in 96-well UV-transparent microplates. Each well contained 95 μL of premix and 5 μL of purified protein (approximately 1–2 μg) to initiate the reaction. NADH consumption was monitored at 340 nm at 30 °C using a SpectraMax iD3 multi-mode microplate reader (Molecular Devices), with absorbance recorded every 1 min for 30 min. The molar extinction coefficient for NADH was 6.22 cm²·μmol⁻¹. The oxidation of 2 mol of NADH corresponds to the formation of 1 mol of acyl-CoA. To measure the enzyme kinetics, the concentrations of iC5:0 or iC4:0 fatty acid substrates were adjusted according to enzyme affinities, ranging from 0–500 μM or 0–2000 μM. NADH concentrations were either 1 mM or 2 mM, depending on the concentrations of fatty acid substrates. The fatty acid substrates used in the assays included: sodium propionate (C3:0), sodium butyrate (C4:0), isobutyric acid (iC4:0), sodium hexanoate (C6:0), sodium octanoate (C8:0), sodium nonanoate (C9:0), sodium decanoate (C10:0), undecanoic acid (C11:0), sodium dodecanoate (C12:0), tridecanoic acid (C13:0), 11-methyllauric acid (iC13:0), sodium myristate (C14:0), sodium palmitate (C16:0), and sodium stearate (C18:0) (from Aladdin); 2-methylbutyric acid (aiC5:0) and isovaleric acid (iC5:0) (from Sigma-Aldrich); sodium valerate (C5:0), sodium pivalate (2,2-DiMe-C3:0), and sodium heptanoate (C7:0) (from Macklin); 9-methyldecanoic acid (iC11:0) (from Jiangsu Aikon); pentadecanoic acid (C15:0) (from Shanghai Yuanye); 4-methylpentanoic acid (iC6:0) (from Xiya Reagent); and 5-Methylhexanoic acid (iC7:0) (from ThermoFisher Scientific).

Stable isotope labeling analysis

To determine the precursor substrates for acyl-chain biosynthesis, stable isotope-labeled compounds were used to label IL2-5 and S. pennellii leaf acylsugars. The following isotope-labeled compounds were applied: sodium acetate-¹³C₂ (Sigma-Aldrich, cat. no. 282014), [1-¹³C]-isovaleric acid (iC5) (ISOREAG, cat. no. IR-34890), and [1-¹³C]-sodium 2-methylpropionate (iC4) (ISOREAG, cat. no. IR-22567).

Compound leaves from 4 to 5-week-old plants were harvested by making a diagonal cut across the stem, preserving as much petiole length as possible. Small leaflets near the petiole were removed, leaving only the terminal leaflet and one pair of lateral leaflets. All cutting procedures were performed with leaves submerged in purified water using sharp blades to avoid mechanical damage or air bubble entry into the vascular system. The excised compound leaves were placed in tubes containing 1 mM stable isotope-labeled substrate and maintained under the same environmental conditions as plant growth. Feeding experiments with isotope-labeled [1-¹³C]-iC5, [1-¹³C]-iC4 or sodium acetate-¹³C₂ were conducted for 3 days. For TIMS-TOF analysis aimed at detecting higher-order acetate-derived labeling, the sodium acetate-¹³C₂ feeding period was extended to 4 days. The feeding solution was replenished as needed to ensure the cut surfaces remained fully submerged. Isovaleric acid solutions were adjusted to pH 7 using sodium hydroxide solution. Acylsugars from isotope-labeled leaves were extracted and analyzed using the same methods and instrumentation as described above.

Isotope labeling was analyzed using Agilent MassHunter Workstation. The molecular mass M⁻ of unlabeled acylsugars was used to extract ion chromatograms for non-labeled compounds. Subsequently, the theoretical masses of isotope-labeled acylsugars—[M + 1]⁻ for compounds labeled with [1-¹³C]-isovaleric acid or [1-¹³C]-sodium 2-methylpropionate, or [M + 2]⁻ for sodium acetate-¹³C₂—were used to extract chromatograms of isotope-labeled acylsugars. The ratio of peak areas between labeled and unlabeled compounds was calculated. Changes in these ratios between isotope-treated and control plants indicated successful incorporation of labeled substrates into acylsugars. For isotopologue distribution analysis, the ion intensities of [M]⁻, [M + 1]⁻, [M + 2]⁻, [M + 3]⁻, and [M + 4]⁻ for acylsugar were extracted from LC-MS data. For each biological replicate, the summed intensity of these isotopologues was defined as 100%, and the relative proportion of each isotopologue was calculated accordingly. Final values represent the mean isotopologue proportions across biological replicates. To further investigate the ¹³C incorporation within the acyl-chains, the labeled acylsugars were analyzed using targeted MS/MS in Collision-Induced Dissociation mode on the same instrument. Collision energies of 35 V were applied. Fragment ions corresponding to the labeled [M + 1]⁻ or [M + 2]⁻ acyl-chains and their unlabeled counterparts M⁻ were extracted and analyzed.

Confocal microscopy

Images were acquired using a Nikon A1 laser scanning confocal microscope equipped with NIS-Elements Advanced Research software. GFP fluorescence in transgenic tomato glandular trichomes and Nicotiana benthamiana mesophyll cells was detected with 488 nm excitation and 505–525 nm emission. RFP signals were captured using 561 nm excitation and 580–630 nm emission, while chlorophyll autofluorescence was detected at 561 nm excitation and 652–721 nm emission.

Protein subcellular localization

To investigate BACS subcellular localization, full-length coding sequences (excluding stop codons) of SlBACS1, SpBACS1, and SpBACS2 were amplified from cultivated and wild tomato cDNA using gene-specific primers (Supplementary Data 6). The amplified fragments were cloned into the pro35S::GFP binary vector using One Step Infusion Mix (TOROIVD), generating pro35S::SlBACS1-GFP, pro35S::SpBACS1-GFP, and pro35S::SpBACS2-GFP constructs. Constructs were transformed into Agrobacterium tumefaciens strain GV3101. For transient expression, bacterial cultures were centrifuged and resuspended in infiltration buffer (10 mM MES, 10 mM MgCl₂, 200 μM acetosyringone, pH 5.6) to OD₆₀₀ = 0.1, then infiltrated into 4-week-old N. benthamiana leaves. Confocal imaging was performed 3 days post-infiltration. To confirm mitochondrial localization, leaves were co-infiltrated with GV3101 harboring a mitochondrial-RFP marker⁵². Subcellular localization predictions for Sl03gACS1, Sp03gACS1, and SlBACS2 were performed using TargetP 2.0⁵³ with the plant organism setting.

VIGS and qRT-PCR

VIGS was performed using a Tobacco rattle virus (TRV)-based system consisting of vectors pTRV1 and pTRV2-LIC^54,55. Target fragments of SpBACS1 and SpBACS2 were designed using the Solanaceae Genomics Network VIGS tool (http://vigs.solgenomics.net/) and amplified with the primers SpBACS1-VIGS-F/R and SpBACS2-VIGS-F/R. The PCR products were cloned into the pTRV2-LIC vector using One Step Infusion Mix (TOROIVD). The constructs were transformed into A. tumefaciens strain GV3101. Agrobacterium strains harboring pTRV1 co-transformed with pTRV2 (empty vector) or the corresponding pTRV2 constructs were cultured overnight at 28 °C in medium containing kanamycin (50 μg·mL⁻¹) and rifampicin (25 μg·mL⁻¹). Cells were harvested by centrifugation (6000 × g, 8 min, 4 °C) and then resuspended in infiltration buffer (10 mM MES, 10 mM MgCl₂, 200 μM acetosyringone, pH 5.6) to OD₆₀₀ = 1.0. After incubation at 28 °C for 3 h, equal volumes of pTRV1 and pTRV2 suspensions were mixed and infiltrated into the cotyledons of ~10-day-old wild tomato seedlings following the protocol described previously⁵⁶. Silencing of phytoene desaturase (PDS) served as a positive control. Sampling of target gene-silenced plants was performed after the characteristic leaf-bleaching phenotype was observed in PDS-silenced plants. The collected leaves were divided into two halves: one half was used for acylsugar extraction and analysis, and the other for gene expression analysis by qRT-PCR.

Total RNA was extracted from silenced leaves using the Easy Plant RNA Rapid Extraction Kit (Easy-Do) and digested with DNase I. The first-strand cDNA was synthesized by the TOROBlue All-in-One qRT Mix with dsDNase (TOROIVD). Subsequently, the transcript levels of SpBACS1 and SpBACS2 were analyzed by qRT-PCR using primers SpBACS1-qRT-F/R and SpBACS2-qRT-F/R with EF1α serving as the internal reference gene. qPCR reactions were performed using TOROGreen qPCR Master Mix (TOROIVD) and detected on a LightCycler 480 II system (Roche). Gene expression levels were calculated using the relative quantification method (2^−ΔΔCt). All primers described above are listed in Supplementary Data 6.

Insect feeding assays

To characterize the ecological function of SpBACS1-mediated long-branched-chain acylsugars, plants were tested for resistance against the western flower thrips (Frankliniella occidentalis) using feeding-preference assays, and against the cotton bollworm (Helicoverpa armigera) using larval performance assays. Both insect rearing and feeding assays were conducted under controlled conditions of 50% relative humidity, 26 °C, 50,000 lux light intensity, and a 16 h light/8 h dark photoperiod. Approximately 10 mL of 0.75% agar solution was poured into tilted Petri dishes. After solidification, the agar surface was covered with plastic wrap to prevent water evaporation. Leaves for insect were inserted into small slits made in the plastic wrap within the same Petri dish. For western flower thrips feeding preference assays, one leaf from the experimental line and one leaf from the corresponding control plant with similar size and developmental stage were placed in the same Petri dish. Twenty 1st/2nd instar western flower thrips were released equidistant between the two leaves and allowed to feed freely for 3 days. Leaves were subsequently collected and subjected to 3,3′-diaminobenzidine staining, in which feeding-damaged regions appeared as brown-colored areas. Damaged leaf area was quantified with ImageJ 1.54 g software to evaluate thrips feeding. For cotton bollworm assays, individual second instar larvae were placed in separate Petri dishes and fed exclusively with leaves from a single plant. Larval weights were recorded before and after the feeding period. Weight gain was used to evaluate insect performance on different plant materials. Sex of the larvae was not determined or considered because the feeding assays were designed to assess overall insect feeding responses rather than sex-dependent effects. This study involved only invertebrate insect species (F occidentalis and H. armigera), for which ethical approval was not required.

Exogenous application of fatty acids

To assess the effects of fatty acids on acylsugar profiles, four fatty acids were applied exogenously: straight-chain C11:0 and C13:0, and terminally branched iC11:0 and iC13:0. Each fatty acid was dissolved in 0.05% (v/v) Tween 20 to a final concentration of 0.1 mM. Four-week-old IL2-5 plants were spray-treated with approximately 15 mL of solution per plant until all leaf surfaces were thoroughly wet. Applications were performed every 12 h for 3 days. Following treatment, leaves were harvested for acylsugar extraction and analysis.

Phylogenetic tree construction

Homologs of SpBACS1 and SpBACS2 were identified from S. lycopersicum and S. pennellii using BLAST⁵⁷. Nucleotide sequences were aligned using MUSCLE⁵⁸ with default parameters. A phylogenetic tree was constructed using RAxML/8.0.6⁵⁹ with the maximum likelihood method and 1000 bootstrap replicates (parameters: -f a -x 12345 -p 12345 -# 1000 -m PROTGAMMAAUTO –auto-prot = bic). The tree and the corresponding heatmap were visualized with midpoint rooting using ITOL⁶⁰.

Promoter analysis

Promoter sequences of SlBACS1 and SpBACS1 were aligned using MUSCLE⁵⁸ in Geneious Prime 2023.1 with default settings. Transcription factor binding sites were predicted using PlantRegMap (https://plantregmap.gao-lab.org/)^61,62 with species-specific settings for S. lycopersicum and S. pennellii. This platform uses TF binding motifs from PlantTFDB and the default FIMO scanning algorithm for binding site prediction. Motif occurrence p-values were calculated from log-likelihood ratio scores, and q-values were generated following false discovery rate correction. A p-value threshold of <1e-5 was applied for predictions. For figure clarity, predicted sites showing minor differences between promoters were omitted.

Statistical analysis

Statistical analyses were performed using GraphPad Prism v10.0. Between-group comparisons were conducted using unpaired Student’s t-tests with Welch’s correction. Kinetic parameters were derived by fitting initial velocities against substrate concentrations to the Michaelis–Menten equation using nonlinear regression in the same software. Best-fit values with 95% confidence intervals (CI) are reported. All heatmaps were generated using TBtools software⁶³, except for the phylogenetic analysis heatmap.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Supplementary information

Supplementary Information ^{(2MB, pdf)}

41467_2026_76062_MOESM2_ESM.pdf ^{(291.8KB, pdf)}

Description of Additional Supplementary Files

Supplementary Data 1 ^{(19KB, xlsx)}

Supplementary Data 2 ^{(11.1KB, xlsx)}

Supplementary Data 3 ^{(12.6KB, xlsx)}

Supplementary Data 4 ^{(46.5KB, xlsx)}

Supplementary Data 5 ^{(17KB, xlsx)}

Supplementary Data 6 ^{(13.5KB, xlsx)}

Reporting Summary ^{(1.6MB, pdf)}

Transparent Peer Review file ^{(676.8KB, pdf)}

Source data

Source Data ^{(3.2MB, xlsx)}

Author contributions

P.F. and J.W. designed the experiments and analyzed the data; J.W., J.J., X.F., J.L., L.H., T.L. and W.Zhang performed most of the experiments; J.J., X.F. and J.L. assisted with data analysis; J.W., X.Xu, and S.F. performed acylsugar analysis; J.W., J.P. and S.F. performed GC/MS analysis; W.Zhou provided guidance for insect bioassays; Z.H., J.Z., X.Xia, J.Y., and R.L. provided guidance for the experiment design; J.W. and P.F. wrote the first draft of the manuscript; P.F. and R.L. revised the manuscript with input from all the authors.

Peer review

Peer review information

Nature Communications thanks Zhenhua Liu and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

This study was funded by National Key Research and Development Project of China (Grant No. 2024YFD2001003 to P.F. and Grant No. 2023YFC2604500 to W. Zhou), the National Natural Science Foundation of China (Grant No. 32472718 to P.F. and Grant No. 325B2079 to W. Zhang), the U.S. National Science Foundation (Grant No. 1546617 to R.L.L.), Natural Science Foundation of Zhejiang province (Grant No. LZ22C150005 to P.F.), Starry Night Science Fund of Zhejiang University Shanghai Institute for Advanced Study (SN-ZJU-SIAS-0011 to P.F.), Organized Scientific Research in Horticulture Discipline of Zhejiang University (Grant No. B231220.0005-26 to P.F.), and Innovative Development of Horticulture Discipline of Zhejiang University (B231220.0005-25 to P.F.).

Data availability

The raw mass spectrometry data supporting this study, including the LC-MS, LC-MS/MS, GC-MS analyses, have been deposited in a public repository Figshare (https://figshare.com/s/d8ece53f660c98464934). Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-76062-w.