Decoding 4-vinylanisole biosynthesis and pivotal enzymes in locusts

Aggregation pheromone, 4-vinylanisole (4VA), is specifically released by gregarious migratory locusts, and is crucial in forming locust swarms that cause destructive plagues1. Control of locust plagues relies heavily on the extensive application of chemical pesticides, which has led to severe environmental and health issues2. As pheromones are primary mediators of insect communication and behaviour3, exploring their biosynthesis can provide important cues to develop innovative behavioural regulators, potentially reducing the reliance on chemical pesticides. Here we resolve the biosynthesis of 4VA and behavioural responses of locusts when enzymes in the 4VA biosynthetic pathway are manipulated. The process initiates with phenylalanine derived from food plants and proceeds through three precursors: cinnamic acid,p-hydroxycinnamic acid and 4-vinylphenol (4VP). Notably, the conversion from 4VP to 4VA through methylation is unique to gregarious locusts. This step is catalysed by two crucial methyltransferases, 4VPMT1 and 4VPMT2. Guided by the X-ray co-crystal structure of 4VPMT2 bound with 4VP andS-adenosyl-l-methionine, we developed 4-nitrophenol as a substrate surrogate. We identified several chemicals that can block 4VA production by inhibiting the enzymatic activities of 4VPMT proteins, thereby suppressing locust aggregative behaviour. The findings uncover the chemical logic behind 4VA biosynthesis and pinpoint two crucial enzymes as novel targets for locust swarm management.

Destructive and frequent locust plagues threaten the safety of agriculture, the economy and the environment globally. To control the outbreaks of locust plagues and other pests, several hundred thousand tonnes of chemical pesticides are applied to the agricultural and grassland environment annually4. The widespread use of chemical insecticides has raised severe concerns about food security, biodiversity loss, environmental pollution and human health2. Therefore, there is an imperative to develop behaviour regulators to supplement and reduce reliance on chemical insecticides.

Pheromones act as primary mediators of insect chemical communication and elicit most behaviours3. Conspecific aggregation is profoundly important for insect survival and reproduction5,6. The use of pheromones has become a widespread strategy in the prevention and control of pest insects7,8. In addition to chemical approaches to pheromone synthesis, the production of insect pheromones by organisms such asEscherichia coli9, yeast10and plants11has been considered a cost-effective and environmentally friendly strategy. Thus, identifying the pathway and critical enzymes involved in pheromone biosynthesis in insects lays the groundwork for synthetic biology approaches and opens up innovative methods for disruption of pheromone biosynthesis. However, the development of behavioural regulators that target these biosynthetic enzymes has been limited owing to the unknown biosynthetic pathways of many insect pheromones.

4VA has recently been identified as the aggregation pheromone of the migratory locust (Locusta migratoria) and has crucial roles in recruiting individuals and forming swarms of locusts1. 4VA also has an essential function in promoting the synchronic sexual maturation of female locusts12and the social interactions of conspecifics13. OR35 is a specific odorant receptor for detecting 4VA, and knockout ofOr35results in the loss of attraction behaviour1. Deficiency of 4VA perception throughOr35knockdown prevents the formation and maintenance of gregarious behaviour13. Thus, 4VA is a primary factor of locust aggregation, and blocking its production is a potential strategy for direct prevention of locust swarms and migration. Therefore, a thorough understanding of the biosynthetic pathway of 4VA in locusts is crucial for identifying pivotal targets and designing potential inhibitors to disrupt its production. However, how 4VA is biosynthesized in locusts remains largely unknown.

The aggregation pheromone 4VA is only released by gregarious locusts. To investigate the source substrate of 4VA biosynthesis, we first compared the amount of released 4VA between hungry gregarious locusts and gregarious locusts fed with wheat seedlings. The release of 4VA in gregarious locusts decreased significantly after starvation for 2 h. The concentration of 4VA approached zero after starvation for 8 h. By contrast, fed locusts released a significant amount of 4VA (Fig.1a). These results suggest that locusts synthesize 4VA using building blocks derived from food plants.

a, Release of 4VA in gregarious locusts after starvation treatment.n= 6 biological replicates.b, The main compound classes in locust host plants, wheat seedlings, include lignin, amino acids, cellulose and hemicellulose.c, 4VA release in locusts after feeding with artificial diet (AD) containing lignin.n= 11 biological replicates.d,e, The release of 4VA-d4 in gregarious locusts after injection of Tyr-d4 (d) or Phe-d5 (e) for 12 h, 24 h and 48 h.n= 6 biological replicates. kcps, thousand counts per second.f,g, Release of 4VA-d4 in gregarious locusts after feeding with artificial diet containing Tyr-d4 (f) or Phe-d5 (g) for 12 h, 24 h and 48 h.n= 6 biological replicates.h,i, Release of 4VA-d4 in gregarious locusts after feeding on plants containing Tyr-d4 (h) or Phe-d5 (i) for 12 h, 24 h and 48 h.n= 6 biological replicates.d–i, Right, representative chromatograms. Different letters indicate statistically significant differences between groups using one-way ANOVA (Tukey’s multiple comparisons test,P< 0.05) (a,c);Pvalues in chemical assays were determined by a two-tailed unpairedt-test (d–i). Data are mean ± s.e.m.

The main compound classes in the wheat seedlings used as host plants include lignin, amino acids, cellulose and hemicellulose (Fig.1b). Owing to the presence of a phenyl moiety in 4VA, lignin, tyrosine and phenylalanine were considered potential biosynthetic precursors of 4VA. To determine whether lignin participates in the biosynthesis of 4VA, we compared the emission levels of 4VA in different groups of locusts fed with artificial diet containing varying amounts of added lignin. 4VA emissions did not significantly increase with increasing lignin content compared with the control group (Fig.1c). Thus, lignin is unlikely to be a major contributor to 4VA biosynthesis in locusts.

To investigate whether tyrosine and phenylalanine are involved in the biosynthesis of 4VA, deuterated tyrosine (Tyr-d4) or deuterated phenylalanine (Phe-d5) were directly injected or fed to the locusts. Injection of Tyr-d4 or Phe-d5 into the locusts (Fig.1d,e) or feeding on an artificial diet containing Tyr-d4 or Phe-d5 (Fig.1f,g) did not result in the production of deuterated 4VA (4VA-d4). Locusts that were fed on host plants sprayed with deuterated tyrosine did not produce 4VA-d4 (Fig.1h). However, feeding on host plants sprayed with Phe-d5 significantly induced the release of 4VA-d4 compared with the control group (Fig.1i). Thus 4VA is biosynthesized from plant-derived phenylalanine.

To explore the biosynthetic intermediates from phenylalanine to 4VA, we summarized phenylalanine metabolism annotated in the Kyoto Encyclopedia of Genes and Genomes (KEGG) and predicted two chemically reasonable biosynthetic routes from phenylalanine to 4VA. The first route is phenylalanine→phenylacetaldehyde→phenethyl alcohol→styrene→4VP→4VA. The second route is phenylalanine→cinnamic acid→p-hydroxycinnamic acid (p-HCA)→4VP→4VA (Fig.2aand Extended Data Fig.1a). Phenylacetaldehyde and phenethyl alcohol have been identified in locusts previously14, and cinnamic acid andp-HCA can also be detected in locusts using the standard compounds as references (Extended Data Fig.1b,c).

a, Two proposed biosynthetic routes of 4VA in locusts.b–d, Release of 4VA-d5 in the gregarious locusts after injection of deuterated phenylacetaldehyde (PAH-d6;b) or feeding with artificial diet containing PAH-d6 (c) or plants sprayed with PAH-d6 (d).n= 5 biological replicates.e–g, Release of 4VA-d5 in gregarious locusts after injection with CA-d6 (e) or feeding with artificial diet containing CA-d6 (f) or plants sprayed with CA-d6 (g).n= 5 biological replicates.h, Release of 4VA-d4 in gregarious locusts afterp-HCA-d4 injection.n= 6 biological replicates.i, Release of 4VA-d4 in gregarious locusts after injection of 4VP-d4.n= 5 biological replicates.j, Release of 4VA-d4 in gregarious locusts after feeding with plants sprayed withp-HCA-d4.n= 5 biological replicates.k, The release of 4VA-d4 in gregarious locusts after feeding with plants sprayed with 4VP-d4.n= 5 biological replicates.b–k, Right, representative chromatograms.l, Phe-d5, CA-d5,p-HCA-d4 and 4VP-d4 content of plants after spraying with Phe-d5.n= 6 biological replicates.Pvalues were determined by a two-tailed unpairedt-test (b–l). Data are mean ± s.e.m.

To determine the actual biosynthetic pathway of 4VA, we applied stable isotope feeding studies in locusts and measured the production of deuterated 4VA. Injection and feeding of deuterated phenylacetaldehyde, either with an artificial diet or plants, did not result in production of 4VA-d5 (Fig.2b–d). Injection of hexadeuterated cinnamic acid (CA-d6) also did not induce the production of 4VA-d5 (Fig.2e). However, feeding with CA-d6 in an artificial diet or plants induced the production of 4VA-d5 (Fig.2f,g). Thus, 4VA is likely to be synthesized through the second proposed biosynthetic route. To further verify this proposed biosynthetic route, we injected deuteratedp-HCA (p-HCA-d4) and 4VP (4VP-d4) directly into locusts—both compounds markedly induced the production of 4VA-d4, indicating thatp-HCA and 4VP are indeed biosynthetic intermediates in 4VA biosynthesis (Fig.2h,i). Furthermore, feeding of locusts withp-HCA-d4 and 4VP-d4 in plants significantly induced the production of 4VA-d4 (Fig.2j,k). Together, these results show that locusts can independently convert cinnamic acid,p-HCA and 4VP to 4VA.

Because the transformation from phenylalanine top-HCA via cinnamic acid is conserved in lignin biosynthesis in plants15, we searched for these four compounds in wheat seedlings sprayed with Phe-d5 to determine whether plants can provide these intermediates for locusts. We detected Phe-d5, pentadeuterated cinnamic acid (CA-d5) andp-HCA-d4—but not 4VP-d4 (Fig.2l). Thus, locusts can also directly obtain phenylalanine, cinnamic acid andp-HCA, but not 4VP or 4VA, from host plants.

Because solitary locusts cannot produce 4VA, we further determined which of these four precursors results in the differential production of 4VA between gregarious and solitary locusts. However, phenylalanine, cinnamic acid,p-HCA and 4VP can all be detected in different tissues of gregarious locusts and solitary locusts (Fig.3a), suggesting that the absence of 4VA in solitary locusts is not due to the lack of these four precursors. To further verify this assumption, we applied the deuterated compounds in solitary locusts using the same method as in gregarious locusts. Feeding solitary locusts with plants containing Phe-d5 and CA-d6 did not induce the production of 4VA-d4 or 4VA-d5 (Fig.3b,c). Moreover, the injection ofp-HCA-d4 and 4VP-d4 did not induce the production of 4VA-d4 in solitary locusts (Fig.3d,e). Thus, the chemical transformations from phenylalanine to 4VP are not rate-determining steps for 4VA biosynthesis in locusts. It instead seems to be the step from 4VP to 4VA that contributes to the distinct emissions of 4VA between gregarious and solitary locusts.

a, Phenylalanine, cinnamic acid (CA),p-HCA and 4VP content in different tissues of gregarious and solitary locusts.n= 5 biological replicates. G, gut; H, haemolymph; L, legs.b,c, Feeding with Phe-d5 (b) or CA-d6 (c) did not result in deuterated 4VA in solitary locusts.n= 5 biological replicates.d,e, Injection ofp-HCA-d4 (d) or 4VP-d4 (e) did not result in production of 4VA-d4 in solitary locusts.n= 5 biological replicates.b–e, Right, representative chromatograms.f, Transcriptome analysis identified nine genes with differential expression patterns in hind legs of gregarious (G1–G3, leg shown on the left) versus solitary (S1–S3, leg shown on the right) locusts.n= 3 independent biological samples of gregarious and solitary hind legs for RNA sequencing.g, qPCR analysis shows six genes with higher expression in gregarious locusts than in solitary locusts.n= 8 biological replicates.h,i, Emission of 4VA after RNAi knockdown ofLOCMI16699(ds16699;h) orLOCMI02868(ds02868;i). Locusts injected with double-stranded RNA (dsRNA) targeting GFP expression (dsGFP) were used as controls for all genes.n= 5 biological replicates.j,k, 4VPMT1 (j) and 4VPMT2 (k) exhibit enzyme activity that converts 4VP to 4VA. Mcps, million counts per second.Pvalues were determined by a two-tailed unpairedt-test (a–e,g–i). Data are mean ± s.e.m.

Chemical logic suggests that the transformation from 4VP to 4VA may involve a methyltransferase-mediated methylation reaction. Given that the highest emission of 4VA is in the hind legs1, we performed transcriptome analysis of methyltransferase genes that were differentially expressed in hind legs of gregarious versus solitary locusts. Nine genes annotated as methyltransferases exhibited significantly higher expression levels in gregarious locusts compared with solitary locusts (Fig.3fand Supplementary Table1). Six of them showed significantly higher expression in gregarious locusts, as verified by quantitative PCR (qPCR) analysis (Fig.3g). We next performed RNA-mediated interference (RNAi) experiments with the identified genes. Knockdown ofLOCMI16699andLOCMI02868significantly decreased the production of 4VA in gregarious locusts (Fig.3h,iand Extended Data Fig.2a). However, the production of 4VA did not change after the knockdown ofLOCMI17143,LOCMI17606,LOCMI16705andLOCMI03758(Extended Data Fig.2b–e). These results suggest thatLOCMI16699andLOCMI02868mainly control 4VA production in locusts.

To further verify the function of LOCMI02868 and LOCMI16699, we heterologously expressed these two proteins using a pET28a-derived vector inEscherichia coliBL21 (DE3) strain and tested their enzymatic activities in vitro. We found that both LOCMI16699 and LOCMI02868 catalysed the methylation of 4VP to yield 4VA in the presence ofS-adenosyl-l-methionine (SAM), and we therefore refer to them as 4VP methyltransferase 1 (4VPMT1) and 4VP methyltransferase 2 (4VPMT2), respectively (Fig.3j,k). The kinetic characterization of 4VPMT1 revealed that its Michaelis constant (Km) and catalytic constant (kcat) with respect to 4VP were 22.42 μM and 4.68 × 10−4s−1, respectively (Extended Data Fig.2f). In addition, 4VPMT1 showed decreased activity when the concentration of 4VP was gradually increased, with an inhibition constant (Ki) of 94.07 μM (Extended Data Fig.2f).Kmandkcatof 4VPMT2 with respect to 4VP were 306.8 μM and 1.45 × 10−6s−1, respectively (Extended Data Fig.2g). Compared with 4VPMT2 (kcat/Km= 0.00473 ± 0.00097 M−1s−1), 4VPMT1 (kcat/Km= 20.9 ± 9.98 M−1s−1) showed much higher catalytic efficiency, indicating that 4VPMT1 is the major contributor to the methylation of 4VP in locusts. Thus, 4VPMT1 and 4VPMT2 catalyse the production of 4VA in locusts in vivo and in vitro.

To reveal how these two methyltransferases control 4VA production in locusts, we first examined their tissue-specific expression patterns and responses to population density. Both transcripts were widely expressed in different tissues, with4VPMT1being most highly expressed in the hind legs and4VPMT2being most highly expressed in the gut (Fig.4a). In transverse sections of hind legs, 4VPMT1 and 4VPMT2 co-localized in the cells approaching the epidermis (Fig.4b). The mRNA levels of4VPMT1and4VPMT2increased significantly when crowding solitary locusts, and decreased after isolation of gregarious locusts (Fig.4c,d). Protein levels of 4VPMT1 and 4VPMT2 aligned with the mRNA levels during the crowding of solitary locusts and the isolation of gregarious locusts (Fig.4e,fand Supplementary Fig.1). Thus, 4VPMT1 and 4VPMT2 regulate the production of 4VA in locusts by responding to changes in population density.

a, Tissue-specific expression of4VPMT1and4VPMT2. Br, brain; HI, head integument; Te, tergum; F+M leg, fore- and mid-leg; H leg, hind leg; AI, abdominal integument; Mu, muscle. FB, fat body; AD, abdomen; An, antenna; Th, thorax; TG, thoracic ganglia; G, gut; Ov, ovary; Te, testis.n= 8 biological replicates.b, Cellular localization of 4VPMT1 and 4VPMT2 in locust hind leg.c,d, Relative mRNA levels of4VPMT1and4VPMT2during crowding of solitary locusts (c) and isolation of gregarious locusts (d).n= 8 biological replicates.e,f, Relative protein levels of 4VPMT1 and 4VPMT2 during crowding of solitary locusts (e) and isolation of gregarious locusts (f).n= 6 biological replicates.g, 4VA emissions in gregarious locusts after double RNAi knockdown of4VPMT1and4VPMT2.n= 5 biological replicates.h, Behavioural states of gregarious locusts after double RNAi knockdown of4VPMT1and4VPMT2(ds4VPMT1+2).n= 25 locusts (dsGFPand ds4VPMT1+2).Pgreg, probabilistic metric of gregariousness. Arrows indicate medianPgregvalues. Different letters indicate statistically significant differences between groups using one-way ANOVA (Tukey’s multiple comparisons test,P< 0.05) (a,c–f);Pvalues were determined by a two-tailed unpairedt-test (g); medianPgregwere compared using Mann–WhitneyU-test (two-sided) (h). Data are mean ± s.e.m.

Furthermore, after RNAi knockdown of gene expression of both4VPMTgenes, 4VA production was significantly decreased after 24 h, and approached zero at 48 h and 72 h (Fig.4g). The knockdown of both4VPMTgenes significantly reduced the production of four compounds, but these four compounds did not elicit any preference for locusts (Extended Data Figs.3and4). We next evaluated the behavioural states of locusts after the knockdown of4VPMTgenes, which are directly related to the aggregation state, in a well-established behavioural assay paradigm13,16. Gregarious locusts exhibited significant behavioural changes in solitary behavioural features, indicating that they lost gregarious behavioural features (medianPgreg= 0.77 for dsGFP; medianPgreg= 0.18 for ds4VPMT1+2) (Fig.4h). These results indicate that these two4VPMTgenes are essential for maintaining gregarious behaviours of locusts by controlling 4VA biosynthesis.

To gain more insights into its catalytic function, we obtained a co-crystal structure of 4VPMT2 bound with 4VP and SAM at a resolution of 3.40 Å (Protein Data Bank (PDB) ID8ZSA; Extended Data Table1). The overall structure of 4VPMT2 is very similar to that of its homologue, juvenile hormone acid methyltransferase (JHAMT)17(PDB ID7EC0), which contains similar binding sites for the cofactor SAM and its substrates (Extended Data Fig.5a–d). 4VP was located in a hydrophobic pocket of 4VPMT2 consisting of I19, Y25, H137, L138, I164, M167, Y173, F192, F238, V242 and A244, with the phenol group of 4VP pointing towards the methyl group of SAM with a distance of 4.14 Å (Fig.5a). Among these residues, H137 forms an edge-to-face stacked π–π interaction with 4VP (approximately 3.40 Å), and disrupting this interaction by mutating H137 to A resulted in a markedly reduced enzymatic activity (Fig.5b). Several bulky residues, including F238, V242 and F192, were also found to be critical for the enzymatic activity (Fig.5b), indicating that the hydrophobic interactions between them and 4VP are also crucial for the binding of the substrate. By contrast, the 4VPMT2-L138A variant exhibited improved activity compared with the wild type. Notably, despite the long distance (4.53 Å) between Y25 and 4VP, the 4VPMT2-Y25F variant barely showed any enzymatic activity, indicating the crucial role of the phenol group in Y25.

a, Structure of the 4VPMT2–4VP–SAM ternary complex. Amino acid side chains and SAM are shown as brown sticks and 4VP is shown in cyan. The distance between the phenol group of 4VP and Y25 or the methyl group of SAM is indicated by red dashed lines. The black dashed line indicates the π–π interaction between 4VP and H137.b, Enzyme activities of wild-type (WT) 4VPMT2 and indicated variants.c, Predicted structure of the 4VPMT1–4VP–SAM ternary complex. Amino acid side chains and SAM are shown as brown sticks and 4VP is shown in cyan. The distance between the phenol group of 4VP and Y61 or the methyl group of SAM is indicated by red dashed lines. The π–π interactions are indicated by the black dashed line.d, Enzyme activities of wild-type 4VPMT1 and indicated variants.e, Top, structures of 4VP and thepara-substituted styrenes2–5. Bottom, enzyme activity of 4VPMT1 in the presence of compounds1–5.f, Top, structures of thepara-substituted phenols6–22. Bottom, enzyme activity of 4VPMT1 in the presence of compounds6–22.g,h, IC50plots from in vitro enzymatic assay of 4VPMT1 versus 4NP (21) (g) and 4-trifluoromethylphenol (22) (h).i, In vitro enzyme activity assays of 4VPMT1 with varying 4NP.j, Catalytic rate of 4VPMT1 for the conversion of 4VP and 4NP.k, 4NA production after adding 4NP to 4VPMT1.l, 4VA production after addition of 4NA to 4VPMT1. Ctrl, control.m, Predicted structure of the 4VPMT1–4NP–SAM ternary complex. Amino acid side chains and SAM are shown as brown sticks and 4NP is shown in cyan.Pvalues were determined by a two-tailed unpairedt-test (b,d–f,j–l).n= 3 biological replicates (b,d–l). Data are mean ± s.e.m.

To gain mechanistic insight into the difference in activity between 4VPMT1 and 4VPMT2, we attempted to solve the crystal structure of 4VPMT1 but failed despite many efforts. Instead, we modelled the structure of 4VPMT1 using AlphaFold, and docked its substrate 4VP in the active site. After 500 ns molecular dynamics simulations to relax the protein–ligand binding complex, we identified an energetically favourable binding conformation, in which the oxygen atom of the phenol is close enough to attack the methyl group of cofactor SAM (3.88 Å) (Fig.5c). Owing to the high sequence identity (Extended Data Fig.5e), the binding mode of 4VP in 4VPMT1 is very similar to that in 4VPMT2 and the key protein–substrate interactions observed in 4VPMT2 are also formed in 4VPMT1 (Fig.5c). For example, similar to Y25 in 4VPMT2, the corresponding Y61 was also demonstrated to be necessary for the activity (Fig.5d), possibly owing to its ability to form a hydrogen bonding interaction with 4VP (Fig.5c). However, an additional edge-to-face stacked π–π interaction was observed in 4VPMT1 between 4VP and W174, corresponding to L138 in 4VPMT2. Breaking this π–π interaction reduced the enzymatic activity (Fig.5d), indicating the importance of W174 in maintaining the activity. To further determine the role of this site in the activity of 4VPMTs, we replaced L138 with the corresponding residue in 4VPMT1 (W). The enzymatic activity of this 4VPMT2-L138W variant was significantly higher, by more than 300-fold (Fig.5b), indicating that the higher catalytic efficiency of 4VPMT1 can be attributed to this additional π–π interaction between W174 and 4VP in 4VPMT1. Together, 4VPMT1 shares a similar catalytic mechanism to 4VPMT2, but the W138L mutation of 4VPMT2 decreases its binding affinity for 4VP owing to the loss of a vital π–π interaction, and makes it a less effective methyltransferase compared with 4VPMT1. These results reveal the important role of this tryptophan residue in the activity enhancement of 4VPMT1.

To confirm the pivotal role of 4VPMTs in 4VA biosynthesis and develop an effective inhibitor of 4VPMTs, we systematically evaluated the activities of 4VP analogues as a substrate competitor to inhibit the enzymatic methylation of 4VP. First, styrene derivatives with otherpara-substitutions (such as amino, nitro and iodo groups) rather than the phenol hydroxyl group were tested by the enzymatic assays. We found that even in the presence of 10 equivalents of the styrene derivatives (2–5), the enzymatic activities of 4VPMT1 were decreased by less than 50% (Fig.5e). These results consistently indicate that the phenol hydroxyl group on 4VP has a vital role in substrate binding and recognition for 4VPMTs.

We then focused onpara-substituted phenols to develop 4VPMT inhibitors. Seventeenpara-substituted phenols with different functional groups (6–22) were screened in the enzymatic assay. Among these compounds, 4-nitrophenol (4NP) and 4-trifluoromethylphenol were the most effective inhibitors, exhibiting 339-fold and 348-fold decreases in the enzymatic activity of 4VPMT1, respectively, in the enzymatic assay (Fig.5f). Further biochemical assays showed that the half-maximal inhibitory concentration (IC50) values of 4NP for 4VPMT1 and 4VPMT2 were 190.9 and 177.7 nM, respectively. The IC50values of 4-trifluoromethylphenol for 4VPMT1 and 4VPMT2 were 407.3 and 550.1 nM, respectively (Fig.5g,hand Extended Data Fig.6a,b).

The kinetic characterization of 4VPMT1 was used to evaluate the activity of 4NP. TheKmandkcat/Kmvalues of 4VPMT1 with respect to 4NP were 15.13 μM and 1.02 M−1s−1, respectively (Fig.5i). TheKmandkcat/Kmvalues of 4VPMT2 with respect to 4NP were 278.8 μM and 0.11 × 10−3M−1s−1, respectively (Extended Data Fig.6c). The conversion rates of 4NP by 4VPMT1 and 4VPMT2 were much lower than that of natural substrate 4VP (Fig.5jand Extended Data Fig.6d). Although 4NP can be methylated to 4-nitroanisole (4NA) by both 4VPMT1 and 4VPMT2 (Fig.5kand Extended Data Fig.6e), 4NA can successively inhibit the production of 4VA (Fig.5land Extended Data Fig.6f). To probe the mechanism of inhibition by 4NP, we conducted computational calculations to identify the binding mode of 4NP in 4VPMT1 and 4VPMT2. Owing to a hydrogen bonding interaction between W174 and 4NP (Fig.5m), 4NP may have a more robust interaction with 4VPMT1 and thus exhibit a lowerKmvalue than 4VP. For 4VPMT2, we found that Y25 and H137 formed hydrogen bonding interactions with 4NP, which were not observed in the original substrate–enzyme interaction study with 4VP (Extended Data Fig.6g). Overall, 4NP shows higher binding affinity but lower catalytic efficiency with the 4VPMTs, making it an effective competitive inhibitor to block 4VA production.

To investigate the effects of 4NP on 4VPMTs in vivo, we injected 4NP into locusts and quantified 4VA production. Injection of 4NP at doses of 0.1 nmol to 100 nmol significantly inhibited 4VA production (Fig.6a). Next, we injected locusts with 1 nmol 4NP and detected 4VA emission at different times. 4NP injection (1 nmol) significantly inhibited 4VA production within 2 h and 4 h, but the inhibitory effects were not sustained beyond 8 h and 12 h (Extended Data Fig.7a). Following feeding on wheat seedlings sprayed with 4NP, 4VA production in gregarious locusts significantly decreased compared with the control group (Fig.6b).

a, Effects of varying 4NP on 4VA production in locusts.n= 6 biological replicates.b, 4VA production in gregarious locusts after feeding on plants sprayed with 4NP.n= 5 biological replicates.c, Preference of locusts for volatiles emitted by control gregarious locusts compared with those fed with 4NP.n= 28 locusts.Pvalues were determined by Wilcoxon signed-rank test (two-sided).d, Behavioural state of gregarious locusts after feeding on plants sprayed with 4NP.n= 29 locusts (ctrl) andn= 30 locusts (4NP). Arrows indicate medianPgregvalues.e, 4VA production of crowded solitary locusts after feeding on plants sprayed with 4NP.n= 5 biological replicates.f, Behavioural state of crowded solitary locusts after feeding on plants sprayed with 4NP.n= 25 locusts (control and 4NP). Arrows indicate medianPgregvalues.Pvalues were determined by a two-tailed unpairedt-test (a–c,e). Comparisons of medianPgregwere analysed using the Mann–WhitneyUtest (two-sided) (d,f). Data are mean ± s.e.m.

We then examined the behavioural preference of locusts exposed to volatile compounds emitted by gregarious locusts fed wheat seedlings sprayed with or without 4NP. Locusts showed a significant preference for the volatile emissions from the control group, compared with the 4NP-treated group (Fig.6c). In the phase behaviour assay, gregarious locusts significantly changed their behaviour to solitary features (medianPgreg= 0.83 for controls; medianPgreg= 0.46 for 4NP) (Fig.6d). We also evaluated the effects of 4NP on the formation of gregarious behaviour in solitary locusts. 4VA production decreased significantly in solitary locusts when crowded compared with the control group (Fig.6e). The crowded solitary locusts fed with 4NP maintained solitary behavioural features at 72 h, whereas control locusts showed gregarious-like behaviour at the same time point (medianPgreg= 0.75 for control; medianPgreg= 0.16 for 4NP) (Fig.6f).

To evaluate other probable effects of 4NP on locusts, we first tested whether 4NP elicited behavioural responses. We found that the locusts did not show any attraction or repulsion responses to 4NP (Extended Data Fig.7b). We then quantified volatile organic compounds (VOCs) produced after feeding on wheat seedlings sprayed with 4NP, and found that two compounds,m-xylene and nonane, were significantly decreased in VOCs from 4NP-fed locusts compared with those from the control group (Extended Data Fig.3d,e). The dual-choice behavioural assay showed thatm-xylene and nonane did not elicit any behavioural responses (Extended Data Fig.4a,b). Next, we tested locust perception of 4VA, and found that the electrophysiological responses of basiconic sensilla did not exhibit any differences in 4NP-fed locusts compared with control locusts (Extended Data Fig.7c). Therefore, the exogenous application of 4NP can effectively block the aggregation of locusts by inhibiting the production of 4VA.

Given that using 4NP may be unsafe in field conditions18, we performed virtual screening to identify inhibitors of 4VPMTs with higher potency and biosafety profiles from the library of preclinical and clinical drugs in our lab. We identified seven compounds as candidate inhibitors for activity testing (Extended Data Fig.8a). Among the seven compounds, tolcapone, a drug used clinically to treat symptoms of Parkinson’s disease, exhibited an inhibitory effect (approximately 85% inhibition) on the enzymatic activity of 4VPMT1 (Extended Data Fig.8b,c). Moreover, injection and feeding of tolcapone significantly inhibited the production of 4VA in gregarious locusts and crowded solitary locusts (Extended Data Fig.8d–f). Thus, we identified tolcapone as a candidate small-molecule inhibitor of 4VPMTs.

Locusts exploit common plant metabolites and modulate the expression of 4VPMTs to enhance the efficiency of 4VA biosynthesis. Given that the transformations of phenylalanine to cinnamic acid and subsequently top-HCA are conserved in the lignin biosynthesis pathway of plants15, locusts can quickly obtain a substantial amount of the biosynthetic precursorp-HCA, which facilitates further conversion to 4VA. By using plant-derived intermediates, locusts can obtain 4VA in just two steps and rapidly switch 4VA emission on or off at the last step of 4VA biosynthesis by controlling the expression of 4VPMTs. The transformation from the -OH group of 4VP to the -OCH3group of 4VA decreases hydrophilicity and increases volatility, facilitating the release of 4VA. Consequently, these adaptations significantly reduce the energy and material costs, and markedly improve the synthetic efficiency of the aggregation pheromone 4VA.

Two 4VPMTs are crucial enzymes for 4VA biosynthesis and targets for inhibiting locust aggregation. 4NP inhibits 4VA biosynthesis by competitively binding to 4VPMTs. As a substrate analogue, the nucleophilicity of the phenol group in 4NP is markedly reduced owing to the existence of the electron-withdrawing nitro group. Thus, 4NP is a less reactive substrate and exhibits lower reactivity than 4VP for 4VPMTs. Moreover, 4NP shows higher binding affinity to 4VPMTs than 4VP, therefore competitively binding to the enzymes in preference to 4VP. As dictated by the protein structures, the specific interactions between 4NP and 4VPMTs ensure the selectivity of 4NP and minimize off-target effects. As an alternative to small-molecule inhibitors, RNAi insecticides targeting 4VPMTs could also be developed to control locust swarming behaviours. In summary, manipulation of 4VA biosynthesis is an effective and sustainable strategy for locust control.

Gregarious and solitary locusts (L. migratoria) used in the experiments were maintained at the Institute of Zoology, Chinese Academy of Sciences, Beijing, China. In brief, gregarious locusts were reared in cages (30 cm × 30 cm × 30 cm) with 800 to 1,000 first-instar insects per cage in a well-ventilated room. Solitary locusts were raised in another room, each in a separate ventilated cage (10 cm × 10 cm × 25 cm). Gregarious and solitary locusts were maintained for at least three generations before the experiments. All locusts were cultured under the following conditions: photoperiod light 14 h:dark 10 h, temperature 30 ± 2 °C, relative humidity 60 ± 5%, and a diet of fresh greenhouse-grown wheat seedlings and bran.

The volatiles of fifth-instar gregarious and solitary nymphs were collected by static solid phase microextraction (SPME) and detected by a Bruker GC system (456-GC) coupled with a triple-quadrupole mass spectrometer, as described1. In brief, a fibre (PDMS/DVB 65 μm, 57310-U) was introduced into a glass jar (10.5 cm high × 8.5 cm internal diameter) approximately 1 cm above a stainless steel lid (9 cm in diameter with holes of 2 mm diameter and 2 mm apart), which served as a barrier to confine a group of five fifth-instar locusts. The SPME volatiles collected from an empty glass jar for 30 min served as a control. The fibres with adsorbed odours were subjected to chemical analyses.

Fifth-instar gregarious nymphs were cultured in perspex cages (15 cm × 15 cm × 15 cm) for starvation treatment. The locusts were first fed with sufficient wheat seedlings, and the starvation time of the locusts was 1 h, 2 h, 4 h, 6 h and 8 h. After treatments, the volatiles of locusts were collected using SPME, and gas chromatography–mass spectrometry (GC–MS) analysis was used to quantify 4VA emissions. The experiment consisted of six biological repetitions with four locusts per repetition. A Bruker chemical analysis MS workstation (v.8.0) was used to analyse and process the data.

Phe-d5, Tyr-d4, PAH-d6, CA-d6,p-HCA-d4 and 4VP-d4 were dissolved with sterile water at the concentration of 10 μg μl−1; the deuterated compounds (2 μl) were injected into the abdomen cavity of early fifth-instar gregarious nymphs. The control group was injected with 2 μl sterile water. The volatiles of locusts were collected by SPME and quantified by GC–MS at 12 h, 24 h and 48 h after injection.

Locust artificial diet was made according to Supplementary Tables2and3. In brief, 400 ml ultra-pure water and all compounds without vitamin C and vitamin B1 were thoroughly mixed and then placed in a high-temperature sterilizer for 20 min to fully dissolve the agar powder. The dissolved solution was placed at room temperature to 30–40 °C; vitamin C and vitamin B1 were added and the solution was stirred thoroughly until uniform. One, three and nine grams of lignin powder (Sigma-Aldrich, 370959) was added to 50 ml of artificial diet before being cooled and coagulated. The early fifth-instar gregarious nymphs were selected and starved for 6–8 h. The locusts in the experimental group were fed an artificial diet containing lignin, while the control group locusts were fed an artificial diet without lignin. A fresh artificial diet was applied once per 12 h, and the locusts were fed for 24 h. SPME was used to collect volatiles of locusts, and GC–MS was used to quantify the release of 4VA.

Phe-d5, Tyr-d4, PAH-d6 and CA-d6 were dissolved in sterile water at 10 μg μl−1. Phe-d5, Tyr-d4, PAH-d6 and CA-d6 solutions were mixed with an artificial diet or sprayed uniformly on the stems and leaves of wheat seedlings. The fifth-instar gregarious nymphs were fed with artificial diet or wheat seedlings supplemented with deuterated compounds. The control group was fed with an artificial diet of wheat seedlings sprayed with sterile water. The volatiles of gregarious locusts were collected after 12 h, 24 h and 48 h of feeding, respectively, and GC–MS was used to detect 4VA release by the locusts.

A Bruker GC system (456-GC) coupled with a triple-quadrupole mass spectrometer (Scion TQ MS/MS, Bruker Daltonics) equipped with a DB-1MS column (30 m × 0.25 mm i.d. × 0.25 μm film thickness, Agilent Technologies) was used to quantify the volatile compounds in the SPME samples as described in the previous studies1,14. A Bruker chemical analysis MS workstation (MS Data Review, Data Process, v.8.0) was used to analyse and process the data. Standard compounds (purities ≥ 95%, Sigma-Aldrich) with different dosages (0.1, 1, 10, and 100 ng μl−1) were used to develop the standard curves to quantify the volatiles. The same thermal program and MRM method were used.

The volatile profiles of locusts after RNAi knockdown of 4VPMTs and feeding of 4NP were quantified using an Agilent GC system (5973N) coupled with a mass spectrometry system equipped with a HP-5MS column (60 m × 0.25 mm i.d. × 0.25 μm film thickness, Agilent Technologies) to quantify the volatile compounds in the SPME samples. The initial temperature of the oven was maintained at 40 °C for 1 min, and the ramp was at 20 °C min−1to 300 °C (hold 2 min). The injector temperature was maintained at 250 °C with a constant flow rate of 1.0 ml min−1of helium. The GC–MS electron impact source was operated in full scan mode with the MS source temperature at 240 °C and MS Quad at 150 °C. Volatile compounds were identified by comparing their retention times with the synthetic standards in the same column. The referenced mass spectra were from the NIST11 library (Scientific Instrument Services).

Phenylalanine, cinnamic acid,p-HCA and 4VP in guts, haemolymph and legs of gregarious and solitary locusts were derivatized and quantified by GC–MS/MS. Approximately 100 mg of fresh tissue samples were weighed and transferred into a centrifuge tube containing 500 μl of extraction reagent (ethanol: acetonitrile, 9:1) and homogenized thoroughly in a grinder (JXFSTRP-32L, Shanghai Jingxin Industrial Development, 60 Hz, 2 min). After stewing for 15 min at room temperature, the samples were centrifuged at 12,000 rpm for 10 min at 4 °C. The supernatants were transferred to new centrifuge tubes for vacuum drying. Methoxyaminutese hydrochloride (20 mg ml−1, dissolved in pyridine, 50 μl each sample) was mixed thoroughly in the dried samples, and the samples were kept at 37 °C for 90 min.N-Methyl-N-(trimethylsilyl) trifluoroacetamide (70 μl each sample) was subsequently added to the samples, and the samples were placed at 37 °C for 90 min. The samples were centrifuged at 12,000 rpm for 10 min at 4 °C. The supernatants were stored at −20 °C for detection.

Total RNA was extracted from hind legs (3 samples with 5–6 individuals per sample) using Trizol reagent following the manufacturer’s instructions. The quantity and purity of the total RNA were determined in an Agilent 2100 Bioanalyzer (Agilent) to verify RNA integrity. After the QC procedure, the RNA with poly-A in eukaryotic total RNA was enriched by TIANSeq mRNA Capture Kit (TIANGEN). Then, using the captured RNA as the starting sample, the TIANSeq Fast RNA Library Kit (Illumina) was used to construct the transcriptome sequencing libraries. In brief, the transcriptome sequencing library was built through RNA random fragmentation, cDNA strand 1 and strand 2 synthesis, end repair, A-tailing, ligation of sequencing adapters, size selection and library PCR enrichment. Library concentration was first quantified using a Qubit 2.0 fluorometer (Life Technologies) and then diluted to 1 ng μl−1before checking insert size on an Agilent 2100 and quantifying to greater accuracy by qPCR (library activity > 2 nM). The clustering of the index-coded samples was performed on a cBot Cluster Generation System using TruSeq PE Cluster Kit v3-cBot-HS (Illumina) according to the manufacturer’s instructions. After cluster generation, the library preparations were sequenced on an Illumina sequencing platform and 150 bp paired-end reads were generated. The raw reads of 6 samples are available for download from the NCBI Sequence Read Archive (SRA) server (accession number:PRJNA1046724).

Eight samples of locust hind legs (2–3 individuals per sample) were collected and homogenized in Trizol reagent (Life Technology) and the total RNA was extracted following the manufacturer’s instructions. DNase was applied to eliminate DNA contamination in the RNA samples. We reverse transcribed 2 µg of total RNA in every sample using MMLV reverse transcriptase (Promega) by the manufacturers’ instructions to analyse the expression levels of mRNAs. PCR amplification was conducted with the Roche Light Cycler 480 using a miRcute miRNA qPCR Detection Kit (Tiangen) and a Real Master-Mix (SYBR Green) kit (Tiangen), respectively. RP49 was used as an endogenous control for mRNAs. The amplification procedure followed the manufacturers’ protocols, and the melting curve was detected to confirm the amplification specificity of the target genes. All PCR amplifications are sequenced to verify the specificity of primers. The primer sequences for the qPCR assay are provided in Supplementary Table4.

After designing the fragments ofLOCMI16699,LOCMI02868,LOCMI17143,LOCMI17606,LOCMI16705andLOCMI03758sequences for RNAi, we blasted theLOCMI16699fragments against theLOCMI02868sequences of the migratory locust to detect sequence homologies. We selected the non-homologous fragment with other genes in the genome database to avoid non-specificity during RNAi knockdown. dsRNAs of GFP,LOCMI16699,LOCMI02868,LOCMI17143,LOCMI17606,LOCMI16705andLOCMI03758were prepared using the T7 RiboMAX Express RNAi System (Promega). Fifth-instar gregarious locusts were selected for injection. Ten micrograms of double-stranded RNAs ofLOCMI16699,LOCMI02868,LOCMI17143,LOCMI17606,LOCMI16705,LOCMI03758and double-stranded RNAs of GFP (as a control), respectively, were injected into the second abdominal segment of gregarious locusts. The effect of RNAi on relative mRNA and 4VA production levels was investigated by qPCR and GC–MS/MS after injection for 72 h. The primers for RNAi are provided in Supplementary Table5.

Hind leg samples from six groups of crowded solitary locusts and isolated gregarious locusts were collected and homogenized in TRIzol reagent (Life Technology), and protein for western blot analysis was extracted following manufacturers’ instructions, respectively. Custom-made affinity-purified polyclonal antibodies against 4VPMT1 (mouse) (ABclonal), 4VPMT2 (rabbit) (ABclonal) and GAPDH19were used for protein analyses. The specificity of the two antibodies were verified by western blot after the knockdown of4VPMT1and4VPMT2, respectively (Supplementary Fig.1a,b). The protein samples (10 μg μl−1) were separated by gel electrophoresis and then transferred onto polyvinylidene difluoride membranes (Millipore). Non-specific binding sites on the membranes were blocked with 5% skim milk. The blots were separately incubated with the primary antibody (mouse anti-4VPMT1, 1:5,000; rabbit anti-4VPMT2, 1:1,000; rabbit anti-GAPDH, 1:5,000) in 5% skim milk overnight at 4 °C. After incubation; the membranes were washed by PBST (PBS with 0.1% Tween), then incubated with anti-rabbit IgG secondary antibody (1:5,000) (EASYBIO Technology) for 1 h at room temperature, and then washed three times. An ECL kit subsequently detected the immunological blot (Thermo Fisher, 34096). The intensities of the western blot signals were quantified using ImageJ (v.1.51k).

The dissected hind legs were embedded in freezing medium Tissue-Tek O.C.T. Compound (Sakura Finetek) and rapidly frozen at –70 °C. Sections (10 μm) were prepared at –20 °C (Leica CM1950), thaw mounted on SuperFrost Plus slides (Menzel-Gläser) and air dried for 15 min. The immunofluorescence was performed according to the protocol described in a previous study20with slight modifications. After fixed in 4% formaldehyde at room temperature for 1 h, the sections were washed with 0.1 M PBS (pH 7.2) twice for 15 min each, and then incubated in 0.1 M PBS containing 5% normal goat serum (NGS, Boster) for 1 h at room temperature. The primary anti-4VPMT1 and 4VPMT2 antibody (custom made, ABclonal) was diluted at 1:1000 and 1:500 in 0.1 M PBS containing 2% NGS, respectively. Incubation with primary antibodies lasted for 24 h. The tissues were washed with 0.1 M PBS three times for 15 min each and subsequently incubated with the secondary antibodies, goat anti-rabbit antibody Alexa fluor 488 (1:500, Life Technology, A11034) and goat anti-mouse antibody Alexa fluor 546 (1:500, Life Technology, A11030) for 1 h at room temperature. After washing three times, the tissues were mounted in anti-fade fluorescence mounting medium. The negative serum was used as the negative control. The nucleus of locust hind leg was labelled by Hoechst33342 (Life Technology). Confocal images were obtained by a Zeiss LSM 710 confocal microscope (Zeiss) equipped with ZEN 2012 software.

The DNA fragments encoding 4VPMT1 and 4VPMT2 were amplified by PCR using the primers pET28a-4VPMT1-F/R and pET28a-4VPMT2-F/R (Supplementary Table6), respectively, and cloned into a pET28a-derived vector. The constructs with an N-terminal His tag were transformed into theE. coliBL21 (DE3) strain, and the bacteria were cultured in LB medium at 37 °C. When the OD600 of the culture reached 0.5, the expression of 4VPMT1 and 4VPMT2 was induced by the addition of 0.4 mM isopropyl-β-d-thiogalactopyranoside at 16 °C for 14 h. The cells were collected by centrifugation at 4,500gfor 15 min and then resuspended in a binding buffer (30 mM imidazole, 30 mM Tris-HCl, pH 8.0, 50 mM NaCl). The supernatants were transferred into new centrifuge tubes after ultrasonication and centrifugation at 15,000gfor 15 min. Ni2 + -NTA agarose resin was first equilibrated with 30 mM imidazole buffer. After adding supernatant, the target protein was washed with 50 mM imidazole buffer and then eluted with 250 mM imidazole buffer. The purity of the eluted protein was estimated by SDS–PAGE, and the protein sample was stored at −80 °C. The mutant proteins were expressed, purified, and stored in the same manner as the wild-type protein.

The enzymatic activity of 4VPMTs was measured by quantifying the amount of product in the reactions using SPME headspace, as described above, with some modifications. To determine the kinetic parameters of 4VPMT1 for 4VP, 399 μl of the reaction mixture was pipetted into a sample vial (RY-10100, 5 ml for headspace adsorption) containing buffer (20 mM Na2HPO4-NaH2PO4, 50 mM NaCl, pH 7.5), 20 μg 4VPMT1 protein and 4 μl SAM (100 mM). The reaction was started by adding 1 μl substrate 4VP (0.04–100 mM) at various concentrations, and the vial was then sealed with a cap (RY-10100, Sample bottle with clamp cap). The handle rod was held in a horizontal position with the sample vial, then the extraction head was slowly pushed into the sample vial at a distance of 3–4 cm from the bottle mouth. The fibre was pushed out to expose the headspace volatiles generated by the reaction, and the adsorption time was 10 min at 30 °C. To determine the kinetic parameters of 4VPMT2 for 4VP, 399 μl of the reaction mixture was pipetted into a sample vial (RY-10100, 5 ml for headspace adsorption) containing buffer (20 mM Na2HPO4-NaH2PO4, 50 mM NaCl, pH 7.5), 180 μg 4VPMT2 protein and 4 μl SAM (100 mM); the reaction was started by adding 1 μl 4VP (4–320 mM) at various concentrations, and the vial was then sealed with a cap (RY-10100, Sample bottle with clamp cap). The products were collected using SPME for 20 min at 30 °C. The amounts of product in the reactions were calculated according to the standard curve of 4VA. Data fitting was performed using GraphPad Prism 8, andKm,kcat,kcat/Kmandkivalues represent the mean ± s.e.m. of three independent replicates.

The inhibition assay of 4VPMT1 against 4VP using 4VP analogues was carried out by pipetting 399 μl of the reaction mixture into a sample vial (RY-10100, 5 ml for headspace adsorption) containing buffer (20 mM Na2HPO4-NaH2PO4, 50 mM NaCl, pH 7.5), 20 μg 4VPMT1 protein, 4 μl SAM (100 mM) and 1 μl analogues (4 mM); the reaction was started by adding 1 μl 4VP (0.4 mM), and the vial was then sealed with a cap (RY-10100, Sample bottle with clamp cap). The products were collected using SPME for 10 min at 30 °C and analysed by GC–MS. The amount of product in the reactions with or without 4VP analogues was calculated according to the standard curve of 4VA. Data were analysed by a two-tailed unpairedt-test using GraphPad Prism 8 software.

The IC50values of 4NP and 4-trifluoromethylphenol to the enzymatic activity of 4VPMT1 were determined exactly as previously described. Specifically, 1 μl of 4NP or 4-trifluoromethylphenol at various concentrations (0.00004–20 mM) was added to the reaction buffer containing 20 mM Na2HPO4-NaH2PO4(pH 7.5), 50 mM NaCl, 20 μg 4VPMT1 protein and 4 μl SAM (100 mM). The enzymatic reaction was initiated by adding 1 μl of 0.4 mM 4VP, and the adsorption time was 10 min at 30 °C. To determine the IC50values of 4NP and 4-trifluoromethylphenol to 4VPMT2, 1 μl of 4NP or 4-trifluoromethylphenol at various concentrations (0.00004–20 mM) was added to the reaction buffer containing 20 mM Na2HPO4-NaH2PO4(pH 7.5), 50 mM NaCl, 180 μg 4VPMT2 protein and 4 μl SAM (100 mM). The enzymatic reaction was initiated by adding 1 μl of 30 mM 4VP, and the adsorption time was 20 min at 30 °C. The IC50values were calculated by plotting the relative activity against various concentrations of testing compounds with a dose-response-inhibition function in GraphPad Prism 8 software.

To determine the kinetic parameters of 4VPMT1 for 4NP, 399 μl of the reaction mixture was pipetted into a sample vial (RY-10100, 5 ml for headspace adsorption) containing buffer (20 mM Na2HPO4-NaH2PO4, 50 mM NaCl, pH 7.5), 200 μg 4VPMT1 protein and 4 μl SAM (100 mM). The reaction was started by adding 1 μl substrate 4NP (0.04–40 mM) at various concentrations, and the vial was then sealed with a cap (RY-10100, Sample bottle with clamp cap). The products were collected for 30 min at 30 °C. After adsorption, the fibre extraction head was retracted and then the protective needle was withdrawn from the sample vial to prepare for GC–MS analysis, including 3 biological replicates. Different concentrations of 4NA standard sample (1–100 ng μl−1) were configured and analysed using GC–MS. The amounts of product in the reactions were calculated according to this standard curve of 4NA. Data fitting was performed using GraphPad Prism 8, andKm,kcat, andkcat/Kmvalues represent the mean ± s.e.m. of three independent replicates. To determine the kinetic parameters of 4VPMT2 for 4NP, 399 μl of the reaction mixture was pipetted into a sample vial (RY-10100, 5 ml for headspace adsorption) containing buffer (20 mM Na2HPO4-NaH2PO4, 50 mM NaCl, pH 7.5), 1,800 μg 4VPMT2 protein and 4 μl SAM (100 mM), and the reaction was started by adding 1 μl 4NP (0.4–400 mM) at various concentrations, and the vial was then sealed with a cap (RY-10100, Sample bottle with clamp cap). The products were collected using SPME for 40 min at 30 °C. The amounts of product in the reactions were calculated according to the standard curve of 4NA. Data fitting was performed using GraphPad Prism 8, andKm,kcat,kcat/Kmandkivalues represent the mean ± s.e.m. of three independent replicates.

The DNA fragment encoding 4VPMT2 was amplified by PCR from the pET28a-4VPMT2 plasmid using the primers pQLinkH-4VPMT2-F/R (Supplementary Table6) and cloned into a pQlinkH vector. The construct with an N-terminal His tag was transformed into theE. coliMC1061 strain. The overexpression and purification of 4VPMT2 in the MC1061 strain followed the same protocol as the BL21 (DE3) strain described above. To prepare the protein sample for crystallization, N-His-4VPMT2 was further purified by the size exclusion chromatography using a Superdex increase 200 10/300 GL column (GE Healthcare) with buffer (30 mM Tris, 200 mM NaCl, pH 8.0). The peak fraction was collected and concentrated to 10 mg ml−1. Crystals were grown using a sitting drop vapour diffusion method at 18 °C. Pre-incubated 4VPMT2 with SAM was crystallized in 0.1 M MMT buffer (malic acid, MES and Tris buffers), pH 7.0, and 25% PEG1500. The crystals were soaked for about 24 h in the respective reservoir solution with a final concentration of 4VP of 2.5 mM. Crystals were cryoprotected in a reservoir solution supplied with 20% ethylene glycol and flash-frozen immediately in liquid nitrogen. X-ray diffraction data were collected at the Shanghai Synchrotron Radiation Facility (beamline BL02U1) with an X-ray wavelength of 1.0 Å at a temperature of 100 K. The data were processed using HKL2000 (HKL Research). The structure was solved by a molecular replacement method using Phaser, with the predicted structure from AlphaFold v2.0 (refs.21,22) as the search model and refined using Coot and Phenix.

The full-length amino acid sequence of 4VPMT1 was used for structure prediction by AlphaFold v2.0. The monomeric structure of 4VPMT1 was predicted. Parameters -m is model 1, model 2, model 3, model 4, model 5, -t is 2022-08-21, and -g False. The models with the highest confidence level ‘ranked_0.pdb’ were selected as the predicted structure for 4VPMT1. The predicted structures share a backbone similar to the resolved structures of homologous JHAMT (PDB ID:7EBS,7EBXand7V2S) and 4VPMT2 (8ZSA), indicating high predictive accuracy. The per-residue confidence score (pLDDT) of amino acids 1 to 54 of 4VPMT1 is below 70, and thus this region was removed in further structural analysis.

As 4VPMT1 shares analogous structural backbones to 4VPMT2, the positions of SAM and 4VP in 4VPMT1 were constructed according to their locations in 4VPMT2 (8ZSA). Similarly, the substrate competitor 4NP was docked into the pocket of 4VPMTs according to the position of the substrate 4VP. Molecular docking was accomplished by Schrödinger 2021-2. Preparation of the ligand was carried out via the default parameters of ligPrep. Protein preparation was carried out by Protein Preparation Wizard, with the force field as OPLS_2005 in restrained minimization23. According to the reaction mechanism, the cofactor SAM is involved in the reaction process, so the docking pocket for the protein was chosen as a 20 Å × 20 Å × 20 Å size with SAM as the centre of the grid box. There is no constraint on small molecules.

The complexes of 4VPMTs with their ligands obtained from the docking calculations were used to perform molecular dynamics simulations. Molecular dynamics simulations were performed in desmond (Schrödinger). The solvent was predefined using the TIP3P parameters. The solution simulated was a 0.15 M NaCl solution. NaCl was used to neutralize the charge of the system. The simulations were carried out at 300 K. The simulations were performed using Desmond’s standard NPT relaxation protocol. No restrictions are added to the ligands during this process. A total of three sets of 500 ns computational simulations were carried out. The results of the simulations were subsequently analysed for protein–ligand interaction, from which information on PL-RMSD and PL-Contact was obtained. The distance between the phenolic hydroxyl group of the substrate and the carbon atom on the sulfur of the SAM on the protein was also measured, and a reasonable conformation was selected to present the results given the reaction principle.

4VPMT1 variants (S277A, I200A, M203A, V278A and Y61F) and 4VPMT2 variants (L138A, I164A, Y25F and H137A, F192A) were engineered by site-directed mutagenesis (QuikChange, Stratagene) using the primers in Supplementary Table6. 4VPMT1 variants (F228A, H173A and W174A) and 4VPMT2 variants (L138W, Y173A, F238A and V242A) were engineered by SynbioB company. Expression and purification of all these variants were done according to the same protocol described for the wild type.

The activity assay of 4VPMT1 variants was carried out by pipetting 399 μl of the reaction mixture into a sample vial (RY-10100, 5 ml for headspace adsorption) containing buffer (20 mM Na2HPO4-NaH2PO4, 50 mM NaCl, pH 7.5), 20 μg 4VPMT1 variants and wild-type protein, and 4 μl SAM (100 mM); the reaction was started by adding 1 μl 4VP (0.4 mM), and the vial was then sealed with a cap (RY-10100, Sample bottle with clamp cap). The products were collected using SPME for 10 min at 30 °C and analysed by GC–MS. To determine the activity assay of 4VPMT2 variants, 399 μl of the reaction mixture was pipetted into a sample vial (RY-10100, 5 ml for headspace adsorption) containing buffer (20 mM Na2HPO4-NaH2PO4, 50 mM NaCl, pH 7.5), 180 μg 4VPMT2 variants and wild-type protein, and 4 μl SAM (100 mM); the reaction was started by adding 1 μl 4VP (30 mM), and the vial was then sealed with a cap (RY-10100, Sample bottle with clamp cap). The products were collected using SPME for 20 min at 30 °C. The amounts of product in the reactions were calculated according to the standard curve of 4VA.

Two microlitres of 4NP aqueous solution with different concentrations (0.005 mM, 0.05 mM, 0.5 mM, 5 mM and 50 mM) were injected into the second abdominal segment of gregarious locusts. The effects of 4NP on 4VA production in locusts were investigated by GC–MS at 2 h after injection. Furthermore, 2 μl 4NP (0.5 mM) was injected into gregarious locusts, and 4VA was detected by GC–MS at 2 h, 4 h, 8 h, and 12 h. All control groups were injected with 2 μl H2O.

4NP was dissolved in sterile water at 2 mM concentrations. One millilitre 4NP solution was sprayed uniformly on the stems and leaves of wheat seedlings. The fifth-instar nymphs were fed with wheat seedlings supplemented with 4NP. The control group was fed with wheat seedlings sprayed with sterile water. The volatiles of migratory locusts were collected and 4VA were quantified by GC–MS/MS after 24 h, 48 h and 72 h of feeding, respectively. The behavioural states of fifth-instar nymphs after 72 h of feeding were tested in a well-established behavioural assay arena.

The dual-choice behaviour assay was performed in a vertical airflow olfactometer as described1,24. To test the responses of locusts to the identified compounds, the diluted chemical was coated on filter paper (3 cm × 3 cm; Whatman No. 1), and mineral oil was used as a control agent in another funnel. To test the responses of locusts to the volatiles of 4NP-fed locusts, 30 4NP-fed gregarious locusts and 30 H2O-fed gregarious locusts were used as the odour sources of two sides of the chamber, respectively. A video camera (Panasonic), combined with VCR software (v.2, Noldus Information Technology), recorded locust behavioural activities within 10 min. The video was analysed using EthoVision XT software (v.11.5, Noldus Information Technology) to measure the total time spent on each side.

The phase behaviour assay was performed as previously described13,16. The experiment was performed in a rectangular perspex cage (40 cm length × 30 cm width × 10 cm height), where the top was clear and the walls were opaque. One of the separated chambers (7.5 cm length × 30 cm width × 10 cm height) contained 30 migratory locusts as the stimulation group, and the other was empty. The middle of the cage was the detection area. The locusts were released into the detection arena through a tunnel connected to the steel pipe at the centre of the detection area. The EthoVision video tracking system (Noldus Information Technology) automatically records individual behaviour. Each migratory locust was monitored for 6 min and tested only once. Five different behavioural parameters were extracted from the video: TDM, TDMV, total duration of stimulation area, total duration of relative area, and AI (AI = total duration of stimulation area-total duration of relative area). A binary logistic regression model described in the previous study was used to assess the phase status of the migratory locust. The regression model is as follows:Pgreg= eη/(1 + eη), whereη= 2.11 + 0.005 AI + 0.012 TDM + 0.015 TDMV. Among them,Pgregindicates the probability of a locust in the gregarious phase (Pgreg= 1 means the locust is fully gregarious, whereasPgreg= 0 means individuals display solitary behaviour).

To investigate the effects of 4NP on locust perception for 4VA, we conducted single-sensillum recordings of fifth-instar nymphs. Chemical substances such as single-sensillum recording stimulants included mineral oil as the blank, which was used to dilute 4VA by 1/10 (v/v). A piece of filter paper (Whatman) was placed in a 15 cm Pasteur glass tube and 10 μl of volatile solution was added to the filter paper. The responses of basiconic sensilla in 4NP-fed and control locusts were recorded and analysed as previously described1.

We conducted virtual screening on a library of 3,807 active small molecules using the GNINA software25to identify potential target compounds. During the screening process, we first ranked all the small molecules based on their scores from low to high. To enhance the diversity of the screening results, we analysed the structural similarity of the top-scoring small molecules and excluded compounds sharing more than 0.55 similarity with previously selected compounds based on the ECFP fingerprint of the molecules calculated by RDkit. This ensured the selected compounds exhibited chemical diversity. Ultimately, 80 candidate molecules were chosen for the next round of screening.

In the second round of screening, we used CREST26to determine the optimal conformation of each molecule and to compare these optimal conformations with their bound conformations. This step aimed to evaluate the conformational stability of the candidate molecules’ binding model and, in particular, exclude compounds with excessive torsional strain that might render them unstable. After rigorous screening, seven small molecules emerged as the most promising candidates and were selected for enzymatic activity assays following the same procedure described above.

4VA levels in starvation treatment experiments were analysed using one-way ANOVA with Tukey’s multiple comparisons tests. 4VA levels in lignin feeding experiments were analysed using a two-tailed unpairedt-test. The differences in 4VA-d4, Phe-d5, CA-d6,p-HCA-d4 and 4VP-d4 between treatment and control groups were compared using a two-tailed unpairedt-test. The differential expressions of genes between gregarious and solitary locusts and 4VA levels in gene RNAi knockdown experiments were compared using a two-tailed unpairedt-test. The expression of 4VPMT1 and 4VPMT2 during the crowding and isolation process were analysed using one-way ANOVA with Tukey’s multiple comparisons tests. The enzymatic activities of 4VPMT1 and 4VPMT2 between mutants and wild type were compared using a two-tailed unpairedt-test. The inhibition effects of 4VP analogues on 4VPMT1 enzymatic activity were analysed using a two-tailed unpairedt-test. The emissions of 4VA in gregarious locusts injected with 4-nitrophenol and H2O were compared using a two-tailed unpairedt-test. The behavioural data of locusts in dual-choice assay were analysed by Wilcoxon signed-rank test (two-sided). The phase states of locusts (Pgreg) were analysed by the Mann-WhitneyU-test (two-sided). Differences were considered significant atP< 0.05. Randomization and blinding details are provided in theReporting Summary. Each experiment was repeated three times independently with similar results. All statistics were analysed using SPSS 18.0 (SPSS Inc., Chicago, IL, USA). All the statistical results throughout this study are listed in Supplementary Table7.

Animal experiments were approved by the Animal Care and Use Committee of the Institute of Zoology, Chinese Academy of Sciences, or Peking University.

Further information on research design is available in theNature Portfolio Reporting Summarylinked to this article.

All data are available in the manuscript or the supplementary materials. RNA-seq data have been deposited in the NCBI SRA server (accession number:PRJNA1046724). The structural factors and coordinates of 4VPMT2 are deposited in the Protein Data Bank (PDB) under ID8ZSA. The reported structures of homologous JHAMTs can be found in the Protein Data Bank under ID7EC0,7EBS,7EBXand7V2S.Source dataare provided with this paper.

Guo, X. et al. 4-Vinylanisole is an aggregation pheromone in locusts.Nature584, 584–588 (2020).

ArticleADSCASPubMedGoogle Scholar

Tang, F. H. M., Lenzen, M., McBratney, A. & Maggi, F. Risk of pesticide pollution at the global scale.Nat. Geosci.14, 206–20 (2021).

Shorey, H. H. Behavioral responses to insect pheromones.Annu. Rev. Entomol.18, 349–380 (1973).

Schulz, R., Bub, S., Petschick, L. L., Stehle, S. & Wolfram, J. Applied pesticide toxicity shifts toward plants and invertebrates, even in GM crops.Science372, 81–84 (2021).

ArticleADSCASPubMedGoogle Scholar

Wertheim, B., van Baalen, E. J. A., Dicke, M. & Vet, L. E. M. Pheromone-mediated aggregation in nonsocial arthropods: An evolutionary ecological perspective.Annu. Rev. Entomol.50, 321–346 (2005).

Yew, J. Y. & Chung, H. Insect pheromones: an overview of function, form, and discovery.Prog. Lipid Res.59, 88–105 (2015).

Ryall, K. L., Silk, P., Thurston, G. S., Scarr, T. A. & de Groot, P. Elucidating pheromone and host volatile components attractive to the spruce beetle, (Coleoptera: Curculionidae), in eastern Canada.Can. Entomol.145, 406–415 (2013).

Witzgall, P., Kirsch, P. & Cork, A. Sex pheromones and their impact on pest management.J. Chem. Ecol.36, 80–100 (2010).

Hu, C. Y., Yang, J., Guo, W., Pan, H. & Guo, D. Y. De novo biosynthesis of 4-vinylanisole in engineeredEscherichia coli.J. Agr. Food Chem.72, 4334–4338 (2024).

Ding, B. J., Wang, H. L., Al-Saleh, M. A., Löfstedt, C. & Antony, B. Bioproduction of (Z,E)−9,12-tetradecadienyl acetate (ZETA), the major pheromone component of Plodia, Ephestia, and Spodoptera species in yeast.Pest. Manag. Sci.78, 1048–1059 (2022).

Ding, B. J. et al. A plant factory for moth pheromone production.Nat. Commun.5, 3353 (2014).

Chen, D. et al. Aggregation pheromone 4-vinylanisole promotes the synchrony of sexual maturation in female locusts.eLife11, e74581 (2022).