Abstract Background The extent to which herbivorous insects depend on gut bacteria and the molecular mechanisms by which these microbes help overcome host plant chemical defenses remain controversial. This study explored how the gut symbiont Acinetobacter sp. AS23 of a Camellia weevil (Curculio chinensis) (CW) enhances the weevil’s tolerance to toxic tea saponins in host plants. Methods We first conducted toxicity assays in which third-instar CW larvae were exposed to fermentation filtrates containing tea saponin degradation products from the weevil’s gut bacterium, the AS23 strain. A combination of metabolomic and transcriptomic analyses was used to investigate the degradation pathway and key genes used by the AS23 strain in tea saponin metabolism. We then re-inoculated axenic larvae with bacterial mutants generated through CRISPR-Cas9 and verified gene functions in tea saponin degradation. Results Toxicity assays demonstrated that the AS23 strain exhibited time-dependent tea saponin degradation capabilities. The benzoate degradation pathway emerged as a core metabolic pathway enriched during tea saponin degradation, with the involvement of four key enzyme genes confirmed through qPCR and functional studies. Knockout strains exhibited a significantly reduced detoxification capacity and increased larval mortality when reintroduced into CWs’ gut. Conclusion Our findings elucidated the key role of the AS23 strain in mediating CW larvae tolerance to tea saponins through the benzoate degradation pathway. This study highlights the potential of leveraging microbial saponin degradation pathways for developing environmentally friendly pest control strategies. Supplementary Information The online version contains supplementary material available at 10.1186/s40168-025-02131-9. Keywords: Acinetobacter sp. AS23, Curculio chinensis, Toxic secondary metabolites, CRISPR-Cas9, Benzoate degradation pathway, Tea saponin Introduction In addition to olfactory avoidance [[29]1], plant chemical defenses primarily exert their toxicity in the digestive tract after being ingested by herbivorous insects [[30]2]. Consequently, insect gut tissue and gut microbiota are directly exposed to a variety of plant secondary metabolites [[31]3]. Studies have shown that gut microbiota, as a multi-species community with rapid reproduction and iteration, assist or even enable insects to cope with various stresses [[32]4, [33]5]. The gut microbiota is also a crucial ally of herbivorous insects for countering plant defenses, and can facilitate the host insects’ rapid adaptation to plant chemical resistance [[34]6, [35]7]. Hammer and Bowers hypothesized that the adaptive variations exhibited by herbivorous insects when feeding on toxic plants may be partially attributed to differences in their gut microbiota, leading to the proposal of the “gut microbial facilitation hypothesis” [[36]8]. Extensive research demonstrated that different gut microbes specifically helped insects tolerate and degrade plant toxins [[37]9–[38]11]. These beneficial bacteria have been shown to help pests resist various compounds, including alkaloids, cyanogenic glycosides, cardenolides, amino acids, glucosinolates, oxalate, and saponins [[39]4, [40]12, [41]13]. However, the gut microbiota comprises a community of dozens or even hundreds of species. It remains controversial whether the response of an insect to a specific plant defense compound derives from the collaboration of multiple bacterial strains or from certain core bacterial species. Therefore, more case studies and mechanistic investigations are needed. Camellia oleifera, commonly known as oil tea, is an important economic plant [[42]14]. However, it is susceptible to damage from the Camellia weevil (Curculio chinensis) (hereafter referred to as CW), which can cause significant economic losses [[43]12, [44]15]. Our previous research showed that tea saponins are crucial secondary metabolites in the defense of C. oleifera against insect pests [[45]12]. Triterpene saponins are unique to Camellia species and are abundantly distributed in the kernels; their concentration often determines the insect resistance level of a plant [[46]16]. Moreover, the saponins can be isolated and purified for development as botanical pesticides [[47]16]. The insecticidal activity of saponins primarily involves disrupting ecdysteroid synthesis or acting as protease inhibitors that exert direct toxicity on insect cells [[48]17]. Studies have demonstrated that tea saponins exhibit strong repellent and lethal effects against lepidopteran pests by acting as stomach poisons that damage the intestinal tissue of the insect upon ingestion [[49]16, [50]17]. The CW, the only boring insect pest of oil tea seeds, has co-evolved with its host plant to develop specialized feeding structures and adaptive traits specific to tea seeds [[51]18]. Our research has revealed a high tea saponin content in Camellia fruits [[52]18]. The gut microbiota of the CWs was dominated by the Acinetobacter sp. AS23 strain (hereafter referred to as AS23), which accounted for 43.83% of the microbial community, and this abundance was positively correlated with the saponin content [[53]27]. A genome analysis of this strain identified 47 genes associated with triterpene degradation [[54]18]. However, the key question remains to be addressed: What are the crucial genes and pathways involved in tea saponin degradation by the AS23 strain? In this study, third-instar CW larvae were fed with artificial diets containing tea saponin degradation products of the AS23 strain at different times to determine their toxicity. Then, a metabolomic analysis of the tea saponin was performed to delineate the degradation products produced by the AS23 strain, combined with a liquid chromatograph-mass spectrometer (LC–MS) analysis to elucidate the transcriptional regulation and product compositions during the degradation process. We also performed a temporal transcriptomic analysis throughout the degradation process and integrated the data with the metabolomic results to construct a gene expression regulatory network of the degradation of tea saponins by the AS23 strain, and identified core functional genes. Using CRISPR-Cas9 gene editing technology, we generated key gene knockout mutants and complementation strains to specifically analyze the tea saponin degradation capacity of the AS23 strain and the function of key detoxification genes. In addition, we reintroduced both wild-type and mutant strains into “germ-free” larvae to verify the role of the AS23 strain in mediating the insect’s tolerance to tea saponin toxicity. The findings will elucidate the mechanism of tea saponin degradation by Acinetobacter from both metabolic pathway and key gene perspectives, revealing the crucial role of gut microbiota in host insect adaptation to plant chemical defenses. Furthermore, the results provide valuable insights into bacterial saponin degradation pathways and mechanisms, and offer new perspectives for plant pest control. Results Metabolites from the co-cultivation of Acinetobacter sp. AS23 and tea saponins exhibit varying toxicity against Camellia weevil larvae Toxicity assays revealed significant differences (P < 0.05) in survival rates when third-instar CW larvae were fed artificial diets containing fermentation filtrates from tea saponin degradation at 24 h, 48 h, and 72 h in the absence of the AS23 strain (Fig. [55]1B). The 24 h tea saponin degradation filtrate showed the highest toxicity, and produced an 85% mortality rate of CW larvae (Fig. [56]1C). After 48 h of degradation, the toxicity of the fermentation filtrate significantly decreased, with larval survival rates maintained at 60% during the first 25 days, with no significant difference (P > 0.05) from those of the control group (Fig. [57]1D). Similarly, after 72 h of degradation, the larval survival rates were comparable to those at 48 h, about 65%, with no significant difference (P > 0.05) from those of the control group (Fig. [58]1E). Fig. 1. [59]Fig. 1 [60]Open in a new tab Toxicity assessment of tea saponin against axenic Camellia weevil third-instar larvae at different degradation times. A Experimental design for toxicity assessment. B–E Comparison of the 0 h, 24 h, 48 h, and 72h degradation products of tea saponin on the axenic CW third-instar larvae's survival curves. B Comparison of survival curves at 0 h, 24 h, 48 h, and 72 h tea saponin degradation groups. C Survival curves of control group versus 24 h tea saponin degradation group. D Survival curves of control group versus 48 h tea saponin degradation group. E Survival curves of control group versus 72 h tea saponin degradation group Temporal transcriptome analysis reveals the involvement of the benzoate metabolism pathway in tea saponin degradation Transcriptome sequencing was performed on samples from the AS23 strain at 0 h, 24 h, 48 h, and 72 h during tea saponin degradation. An analysis of gene expression correlations between samples within the groups showed that the 24-h and 48-h treatment groups had stronger correlations with each other than with the 72-h group, with correlation coefficients approaching 1. The control group (CK) exhibited notable differences from all of the other treatment groups (Fig. [61]2A). Overall, the number of differentially expressed genes (DEGs) between the treatment and control groups was significantly higher than between treatment groups (Fig. [62]2B). No significant differences were observed when comparing the three treatment groups (24 h, 48 h, and 72 h) with the control group in either the total number of DEGs or the number of up- and down-regulated genes (Fig. [63]2B). Fig. 2. [64]Fig. 2 [65]Open in a new tab The transcriptome analysis of tea saponin degradation by Acinetobacter sp. AS23. A Correlation analysis of gene expression patterns among samples. CK, H24, H48, and H72 represent the degradation time points of tea saponin by the AS23 strain at 0 h, 24 h, 48 h, and 72 h, respectively. The gradient change from red to blue indicates the degree of correlation between samples, with red indicating high correlation and blue indicating low correlation. Asterisks denote: ***P < 0.001. B Volcano plot of differentially expressed genes (DEGs) between control and treatment groups. The threshold for significant difference is set at P < 0.05. Red dots represent upregulated genes, blue dots represent downregulated genes, and grey dots represent genes without significant differences. C Ternary plot of top 10 abundant DEGs. Only genes with abundance in the top 10% were included. The size of each point represents the relative abundance of genes, with colored points indicating the 10 most abundant genes. The position of each point reflects its contribution to each group. D Ternary plot of DEG group distribution. Only genes with abundance in the top 10% were included. The size of each point represents the relative abundance of genes. Red points represent CK, green points represent 24 h + 48 h group, blue points represent 72 h group, and grey points represent ungrouped genes. The position of each point reflects its contribution to each group. E KEGG enrichment analysis of DEGs. From outer to inner circles: First circle: Enriched categories with gene count coordinates on the outside. Different colors represent different categories; Second circle: Number of background genes in each category and P-values. Longer bars indicate more genes, and redder colors indicate smaller P-values; Third circle: up- and downregulated gene ratio bars, with dark purple representing upregulated genes and light purple representing down-regulated genes, with specific values shown below; Fourth circle: RichFactor values for each category (number of foreground genes divided by number of background genes), with grid lines representing intervals of 0.1 The majority of DEGs were shared between the CK and treatment groups, with 24 unique DEGs in both the 24-h and 48-h treatment groups and 34 unique DEGs in the 72-h group (Fig. [66]2D). The 10 most abundant DEGs were all found among the genes shared between the treatment and CK groups (Fig. [67]2C). To further analyze the functions of DEGs in the AS23 strain following tea saponin treatment, we performed a Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis. The results showed that the DEGs were mainly distributed across four categories: metabolism, cellular processes, environmental information processing, and genetic information processing. Among the top 20 enriched pathways, metabolism-related pathways were most prevalent (13 pathways), while genetic information processing had the lowest representation (one pathway). The pathway with the highest number of DEGs was Ko02010 (ABC transporters) with 47 genes, while Ko00362 (benzoate degradation) showed the most significant enrichment (Fig. [68]2E). Temporal metabolomic analysis revealed extensive enrichment of benzoate compounds during tea saponin degradation To elucidate the tea saponin degradation capability of the AS23 strain, we conducted a metabolomic analysis using untargeted LC–MS. A principal coordinate analysis (PCoA) analysis revealed distinct separation between the control and AS23-treated groups. Among the three treatment groups, the 24-h and 48-h groups' samples clustered together but were separated from the 72-h group' samples, with minimal differences among the samples. Overall, the AS23 treatment resulted in three distinct clusters: the control group, 24-h and 48-h groups, and 72-h group, which was consistent with the transcriptome sequencing clustering results (Fig. [69]3A). Fig. 3. [70]Fig. 3 [71]Open in a new tab The metabolomic analysis of tea saponin degradation by Acinetobacter sp. AS23. A PCoA based on Bray–Curtis distances. CK, 24 h, 48 h, 72 h represents 0 h, 24 h, 48 h, 72 h of tea saponin degradation by the AS23 strain, respectively. The closer the distance between samples, the more similar their metabolic compositions. B Volcano plots of differentially expressed metabolites. Numbers 1–6 represent six comparison groups: 24 h vs CK, 48 h vs CK, 72 h vs CK, 24 h vs 48 h, 24 h vs 72 h, and 48 h vs 72 h. Each point represents a metabolite, with red points indicating adjusted P value < 0.0001 and black points indicating adjusted P value > 0.0001. The labeled metabolites represent the top five metabolites in each comparison group. C, D, E KEGG pathway enrichment analysis of differential metabolites at 24 h, 48 h, and 72 h compared to CK. The size of each point represents the number of metabolites, while the color indicates the significance level of enrichment (decreasing from red to blue). The 20 most significantly enriched pathways are shown on the left, with different colors representing different pathway categories The number of differentially expressed metabolites (DEMs) among the 24-h, 48-h, and 72-h groups was relatively small (Fig. [72]3B (1, 2, 4)). In contrast, the comparisons between each treatment group with the control group revealed significantly more DEMs (Fig. [73]3B (3, 5, 6)). A KEGG enrichment analysis of the DEMs between each treatment group with the control group identified the top 20 enriched metabolic pathways for each comparison. Notably, “degradation of aromatic compounds” was the most significantly enriched pathway across all three comparisons. Although tea saponins are triterpene glycosides that typically lack benzene rings in their main structure, extensive degradation of aromatic compounds may be associated with a specific step in the tea saponin degradation process. We also identified another crucial metabolic pathway: benzoate degradation. Benzoate is an intermediate metabolite of saponin degradation. Its precursor, benzoic acid, and subsequent product, phenol, are both toxic compounds that should not accumulate in organisms for extended periods. Importantly, the benzoate degradation pathway (Ko00362) was also shown to be enriched in the differential gene expression analysis from transcriptome sequencing (Fig. [74]3C). The benzoate metabolic pathway is the core degradation pathway of tea saponins A transcriptomic analysis of the benzoate degradation pathway revealed three distinct expression patterns among the DEGs: decreased expression (gene2442 - gene2444), increased expression during the initial treatment phase (gene1547 - gene2266), and increased expression at 72 h of treatment (gene1580 - gene1700). The majority of genes (gene1547 - gene2266) showed increased expression at 24 and 48 h during the initial phase, while 10 genes (gene1580 - gene1700) followed the other two patterns (Fig. [75]4A). Similarly, the metabolomic data showed three patterns in the abundance of metabolites related to benzoate degradation: (1) an increased abundance during initial treatment (including pyrocatechol and 3,4-dihydroxybenzoicacid) (neg_4493 - neg_4495), (2) an increased abundance at 72 h (including p-salicylic acid and gentisic acid) (neg_2913 -neg_2853), and (3) a consistently low abundance throughout the treatment period (including 4-hydroxybenzoic acid, 3-hydroxybenzoic acid, and phenol) (neg_5922 - neg6666) (Fig. [76]4B). Fig. 4. [77]Fig. 4 [78]Open in a new tab Heatmap of expression abundance and the correlation network analysis of selected DEGs. A Heatmap showing transcriptome fragments per kilobase of exon model per million mapped fragments (FPKM) values of selected DEGs. B Heatmap showing expression abundance of selected DEMs of selected DEGs. Each square represents a gene or a metabolite, with red indicating upregulated genes or metabolites and blue indicating downregulated genes or metabolites. The color intensity reflects the significance level. C Correlation network analysis between selected DEGs and DEMs. Red lines indicate positive correlations, while green lines indicate negative correlations. Node sizes are proportional to the number of connected DEGs (degree). Different colors of nodes represent different differential metabolites. [79]C00146 (neg_6666): Phenol, [80]C00156 (neg_2913): p-salicylic acid, [81]C00156 (neg_5922): 4-hydroxybenzoic acid, [82]C00230 (neg_4495): 3,4-Dihydroxybenzoate, [83]C00628 (neg_2853): gentisic acid, [84]C00587 (neg_6101): 3-hydroxybenzoic acid, [85]C00090 (neg_4493): pyrocatechol To elucidate the complex regulatory relationships between the enriched genes and compounds, we constructed a correlation network. Pyrocatechol correlated with the highest number of DEGs, while the remaining six differential metabolites also exhibited substantial gene associations (Fig. [86]4C). Among the DEGs, gene2441 was negatively correlated with six metabolites, while gene2532 was positively correlated with six metabolites, which represented the highest number of correlations. The DEGs gene2530, gene1547, gene1542, gene1546, and gene2529 each positively correlated with five metabolites, while gene2442 exhibited four negative correlations. These core genes and metabolites collectively constitute the most crucial components of the benzoate degradation pathway (Fig. [87]4C). Four key enzymes in the benzoate metabolic pathway are involved in benzoate hydrolysis Through integrated transcriptomic and metabolomic analyses combined with a correlation analysis between the DEGs and DEMs, we proposed a schematic diagram of the core benzoate metabolic pathway (Fig. [88]5). This pathway involves five crucial metabolites and four enzyme genes, and qPCR results confirmed increased expression of these genes following tea saponin treatment. To explore potential interactions between key metabolites and related enzymes in the benzoate degradation pathway, we performed molecular docking to predict and visualize the binding modes. Molecular docking identifies optimal binding positions between substrates and receptors based on shape and property complementarity principles. The binding energy values for the following pairs were all negative: K16243 and [89]C00180; K05549 and [90]C00180; K16243 and [91]C00146; K05783 and [92]C00230; and [93]K01055 and C0090. Fig. 5. [94]Fig. 5 [95]Open in a new tab Role of benzoate degradation pathway in tea saponin degradation. A–D Expression changes of K16243, [96]K01055, K05549, and K05783 across different treatment groups. The benzoate degradation pathway diagram consists of compound structures and enzyme protein structures, surrounded by molecular docking illustrations between enzymes and compounds. The outermost layer shows the expression abundance changes of four enzyme genes across different treatment groups. CK, H24, H48, H72 represent tea saponin degradation times of 0 h, 24 h, 48 h, and 72 h, respectively. Different lowercase letters indicate significant differences in expression level. E Gene knockout experiments in the AS23 strain. A two-plasmid system was used for gene editing, and the mutant strains were plated on tea saponin screening medium. CK: wild-type AS23 strain; K16243, K05549, K05783, K16243 + K05549, K16243 + K05783 represent corresponding enzyme gene knockout strains. F Tea saponin degradation experiments with wild-type and mutant strains. Red line represents K16243 + K05549 double knockout mutant; green line represents K16243 + K05783 double knockout mutant; CK represents wild-type AS23 strain. The scatter plot shows [97]C00180 compound concentration (y-axis) versus degradation time in minutes pyrocatechol (x-axis) As shown in Fig. [98]5A, in one degradation pathway, both phenol and catechol bound to K16243 (phenol/toluene 2-monooxygenase) through two hydrogen bond interactions. In the alternative pathway, cis-1,2-dihydroxycyclohexa-3,5-diene-1-carboxylate bound to K05549 (benzoate/toluate 1,2-dioxygenase) via three hydrogen bonds, while catechol bound to K05783 (dihydroxycyclohexadiene carboxylate dehydrogenase) through one hydrogen bond. Both pathways converged at catechol as a nodal point. Subsequently, 3-oxoadipate bound to [99]K01055 (3-oxoadipate enol-lactonase) through two hydrogen bonds. At this stage, benzoate completely lost its benzene ring structure and transformed into a carbon chain, marking the completion of the core benzoate metabolic process. To validate the functions of key genes identified through omics screening, we generated the AS23 strain gene knockouts using a dual-plasmid gene editing system. The results showed that single knockouts of K16243, K05549, and K05783 retained the ability to utilize tea saponins and grow on a tea saponin screening medium. However, the double knockout mutants K16243 + K05549 and K16243 + K05783 lost their ability to utilize tea saponins (Fig. [100]5E). To verify whether these double knockout mutants were still able to degrade benzoate ([101]C00180), we cultured them in liquid medium containing [102]C00180 for 24 h and then measured the [103]C00180 content (Fig. [104]5F). The results showed significantly higher [105]C00180 levels in both double knockout mutants than in the control strain, indicating a substantially reduced degradation capability (Fig. [106]5F). These findings confirmed that K16243 with K05549 and K16243 with K05783 are key genes in the two tea saponin degradation pathways, with functional redundancy. Camellia weevil fed with Acinetobacter sp. AS23 mutant strain filtrate lost the ability to tolerate tea saponin toxicity A statistical analysis revealed that the survival rates of CW larvae were approximately 10% when fed with fermentation filtrate containing [107]C00180 from K16243 + K05549 or K16243 + K05783 double knockout mutant strains (Fig. [108]6A–D), while those fed with an artificial diet maintained a survival rate of around 70% (Fig. [109]6A–D). Upon reintroduction of the wild-type AS23 strain, CWs regained their tolerance to benzoate toxicity (Fig. [110]6E–H). In contrast, CW larvae associated with the complemented strains of the K16243 + K05549 and K16243 + K05783 double knockout mutants had a mortality rate of 100% (Fig. [111]6E–H). Fig. 6. [112]Fig. 6 [113]Open in a new tab Effects of Acinetobacter sp. AS23 mutant strains and wild-type strain reintroduction on Camellia weevil larvae survival rates under the exposure of tea saponin or its degraded products. A Experimental workflow for toxicological assessment using fermentation filtrates from different strains. B–D Survival curves of third-instar CW larvae under different treatment groups. E Experimental workflow for verifying synergistic benzoic acid tolerance through recolonization with different strains. F–H Survival curves of third-instar CW larvae after recolonization with different strains and subsequent exposure to benzoic acid-containing diet Discussion Chemical pesticides have long been the predominant means of controlling phytophagous pests [[114]19]. However, with growing concerns over environmental pollution caused by chemical pesticides, there is an increasing demand for environmentally friendly biopesticides. Recently, there have been numerous reports on the development of botanical pesticides based on pyrethrins, alkaloids, ketones, and terpenoids, which are becoming increasingly important in pest control [[115]20]. Tea saponins are triterpene saponins [[116]21], which are commonly used as botanical pesticides against lepidopteran larvae such as diamondback moth (Plutella xylostella) [[117]22] and tea geometrid (Ebrechtella tricuspidata) [[118]16], and they have considerable development potential. Our previous studies revealed that some CWs completed their life cycle in tea fruits containing high levels of tea saponins, primarily due to their association with soil-derived AS23 strain. We also found that the AS23 strain genome is highly enriched in pathways related to aromatic compound degradation and contains 47 annotated genes associated with terpene degradation [[119]18]. In recent years, various soil bacteria have demonstrated efficient tea saponin degradation; for example, Stenotrophomonas bacteria isolated from Panax notoginseng rhizosphere soil showed extremely high total saponin and ginsenoside degradation rates [[120]23]. In the human gut microbiota, genera such as Bifidobacterium, Lactobacillus, Bacteroides, and Prevotella metabolize traditional Chinese medicine saponins [[121]24–[122]26]. These findings collectively highlight the remarkable ability of these bacteria to degrade saponin compounds, although research into their degradation mechanisms remains limited. To elucidate the molecular mechanism of tea saponin degradation by the AS23 strain in CWs, we applied integrated temporal transcriptomics and metabolomics to investigate the transcriptional responses of the AS23 strain coupled with its tea saponin degradation capability. The results revealed a significant increase in DEGs related to transcription in the AS23 strain treated with tea saponin, with the benzoate degradation pathway (Ko00362) showing the most significant enrichment (Fig. [123]2E). Moreover, a metabolomic analysis indicated that the most enriched pathway was aromatic compound degradation, followed by benzoate degradation. Notably, the benzoate degradation pathway was also found to be enriched in the transcriptional differential gene analysis (Fig. [124]3C). These findings differed from Li et al.’s steroid degradation pathway results, possibly because they exclusively focused on a transcriptome analysis without considering tea saponin metabolism, which may have led to biased conclusions [[125]27]. The enrichment of the benzoate degradation pathway explains the substantial enrichment of genes involved in aromatic compound degradation, and may potentially affect subsequent stages of aromatic compound metabolism during tea saponin degradation, including benzoate degradation. The critical step in tea saponin degradation is the breakdown of toxic intermediates. Benzoate, a first-generation product in tea saponin degradation, has been most extensively studied in two forms: methyl benzoate and benzyl benzoate [[126]28]. Methyl benzoate naturally occurs in snapdragon [[127]29] and petunia [[128]30], with insecticidal effects against various pests, including Halyomorpha halys, P. xylostella, Manduca sexta, and Drosophila suzukii[[129]31]. Benzyl benzoate, naturally present in Nematus prasinus [[130]32], shows toxic effects against various mites and Sogatella furcifera [[131]33, [132]34]. The insecticidal mechanism of benzoate is similar to that of tea saponins: it inhibits acetylcholinesterase (AChE) in the insect nervous system, causing insect mortality [[133]35], and readily penetrates the cell membrane lipid layers of insects and their gut symbionts[[134]28], causing membrane damage. Benzoate is evidently toxic to CW larvae and requires further degradation to alleviate the toxicity. Benzoate degradation typically occurs through multiple pathways, generally forming two key intermediates: catechol and protocatechuic acid [[135]36, [136]37]. In Acinetobacter sp. strain ADP1 [[137]38] and A. baumannii DU202 [[138]39], benzoate is usually converted to catechol by benzoate 1,2-dioxygenase, while in Rhodopseudomonas palustris, it forms protocatechuic acid via 3-hydroxybenzoate [[139]36]. In our study, benzoate was transformed into catechol by benzoate/toluate 1,2-dioxygenase subunit alpha (K05549) and dihydroxycyclohexadiene carboxylate dehydrogenase (K05783). Catechol, a second-generation product in tea saponin degradation, was reported to severely inhibit acetylcholinesterase, carboxylesterase, and glutathione S-transferase activity in Spodoptera litura larvae [[140]40]. Moreover, catechol strongly inhibited potential insect gut symbionts such as Bacillus subtilis [[141]41] and Pseudomonas putida [[142]42], similar to benzoate’s mode of action. Given these findings, catechol in insects must be further degraded. Microbial degradation of catechol typically follows two pathways: intradiol cleavage initiated by catechol 1,2-dioxygenase (catA) or extradiol cleavage initiated by catechol 2,3-dioxygenase. The key enzymes in both pathways are highly conserved, and perform catechol oxidation across different species [[143]43–[144]47]. In our study, catA initiated the first step of catechol cleavage, followed by muconate cycloisomerase (catB) and muconate cycloisomerase (catC), and finally, 3-oxoadipate enol-lactonase ([145]K01055) cleaved catechol into 3-oxoadipate, a third-generation product without a benzene ring and no reported toxicity to insects. At this point, we considered that the key steps in strain AS23-mediated tea saponin degradation were complete. Four enzyme genes have crucial roles in the benzoate degradation pathway. After elucidating the key steps in tea saponin degradation by the AS23 strain, we identified four essential enzyme genes—K16243, K05549, K05783 and [146]K01055 —based on the results of the toxic metabolic intermediate analysis during degradation, transcriptional abundance, and qPCR (Fig. [147]4A, [148]B). Gene editing experiments revealed that introducing single mutations into the genes of these enzymes had minimal effects, but double knockouts significantly reduced the ability of the strain to metabolize the toxic substances (Fig. [149]5 and [150]6). Additional experiments demonstrated that by reintroducing the wild-type AS23 strain (Fig. [151]5 and [152]6), CW larvae regained their tolerance to benzoate toxicity, possibly due to certain protective mechanisms or metabolic pathways provided by the wild-type strain (Fig. [153]5 and [154]6). In contrast, the CW larvae group in which double knockout mutant strains were reintroduced had a 100% mortality rate, further confirming the critical role of these specific genes in benzoate degradation and weevil tolerance (Fig. [155]5 and [156]6). In conclusion, we propose that CW larvae detoxify tea saponins through the benzoate degradation pathway mediated by their gut symbiont AS23 strain, with four key enzymes working in combination. Materials and methods Sample collection and pretreatment The AS23 strain used in this study was previously isolated from the gut of CWs [[157]18]. The strain was stored at – 80 °C and activated in LB medium (10 g/L tryptone, 5 g/L yeast extract, 10 g/L sodium chloride, Shanghai Sangon Biotech) and subsequently preserved at 4 °C for further use. To prepare temporal transcriptomic and metabolomic samples, the following experimental procedures were performed. The experimental treatment consisted of inoculating 2 mL of an AS23 strain bacterial suspension (OD[600] = 2) into 50 mL of liquid medium containing 5 g/L tea saponin (purity 98%, Wuhan Yuancheng Technology Development Co., Ltd.) as the sole carbon source. The medium composition was as follows: 5 g of (NH[4])[2]SO[4], 2.5 g of NaCl, 0.3 g of KH[2]PO[4], 0.05 g of FeSO[4], 0.5 g of MgSO[4], 5 g of tea saponin, and sterile water to 1 L (pH 7.2). The control group was inoculated with 2 mL of sterile water instead of the bacterial suspension. Both the treatment and control groups included five replicates [[158]18]. All of the samples were incubated in a shaking incubator at 200 rpm and 37 °C under dark conditions. Aliquots (5 mL) were collected from each sample every 24 h and stored at – 80 °C. Sampling was performed four times (0 h, 24 h, 48 h, and 72 h). Adult CWs were collected from the field and reared to sexual maturity with honey water (10%) supplemented with four antibiotics, tetracycline, ampicillin, rifampicin, and daunorubicin (Sangon Biotech, Shanghai Co., Ltd., China), each at a concentration of 50 mg/L. After mating and oviposition, the eggs were placed in feed containing the same four antibiotics and maintained until the third-instar stage. The rearing conditions were 25 ± 1 °C with 75–80% relative humidity, without light. To ensure the effectiveness of the antibiotic treatment, during the rearing process, three larvae were randomly selected every 5 days for gut DNA extraction. The target sequence was amplified using strain AS23-specific primers (ACI381 F: 5′-CACAATGACATTGCAAGCAATTG-3′ and ACI382R: 5′-CCAATTTTTCATACGAATCTGG-3′) [[159]48]. The larvae were used for experiments only after their axenic status was confirmed. Toxicity determination of tea saponin against Camellia weevil larvae following incubation with Acinetobacter sp. AS23 Briefly, 50 mL of liquid culture medium containing 5 g/L tea saponin as the sole carbon source was inoculated with 2 mL of an AS23 strain bacterial suspension (OD[600] = 2). The fermentation broth was collected and filtered after 0 h, 24 h, 48 h, and 72 h of degradation. Then, 10 mL of filtered fermentation broth was added to 500 g of artificial feed (0.2% vitamin C, 2% sucrose, 0.05% vitamin B, 0.1% cholesterol, 0.5% potassium sorbate, 1.6% AGAR powder, 0.5% Wechsler salt, 10% Camellia seed powder) for third-instar CW larvae. The feed was prepared using fermentation filtrate instead of water, while other ingredients remained unchanged. A suitable volume of feed was placed into a 9-cm dish. After the feed solidified, it was used as a food source for the axenic larvae. There were 10 axenic larvae that were placed in each culture dish, with 10 replicates. The growth of axenic larvae was observed daily, and the mortality rate was calculated. Temporal transcriptome analysis of Acinetobacter sp. AS23 following tea saponin exposure Fermentation broth samples of the AS23 strain incubated with tea saponin or the control collected at different time points (0 h, 24 h, 48 h, and 72 h) (n = 3) and stored at − 80 °C were processed for total RNA extraction using a DP430 bacterial RNA extraction kit (Tiangen Biotech Beijing, China). RNA sequencing was performed on an Illumina MiSeq platform. The raw sequencing data were filtered using Cutadapt to remove 3′ adapter sequences, and reads with average quality scores below Q20 were removed to obtain high-quality sequences. The filtered sequences were mapped to a reference genome using Bowtie2 ([160]http://bowtie-bio.sourceforge.net/index.shtml) after establishing reference genome indices. Gene expression levels were calculated based on the mapping results. Principal component analysis was performed to cluster samples. DEGs between treatment groups were identified using the DESeq2 R package (P < 0.05, log2 (fold change) > 1) [[161]49–[162]51]. Heat maps were generated using the pheatmap (ver. 1.0.12) R package to visualize gene expression patterns and differences between treatments[[163]51]. A functional enrichment analysis of DEGs was performed using the KEGG ([164]https://www.kegg.jp/) to understand the functional differences among the DEGs. In addition, transcriptome structural analysis was performed to obtain information on the operon structures, untranslated regions (UTRs), coding single nucleotide polymorphisms (cSNPs), and insertion and deletion (InDels). We predicted the transcription start sites and transcription termination sites of operons using Rockhopper software ([165]http://cs.wellesley.edu/~btjaden/Rockhopper). The 25-bp sequences of predicted genes upstream of the 5′ UTR were analyzed using a 6-bp sliding window with ELFH software ([166]http://ccb.jhu.edu/software/ELPH/index.shtml) to identify potential ribosome-binding sites (also known as Shine–Dalgarno sequences). TransTermHP ([167]http://transterm.cbcb.umd.edu/) was used to predict rho-independent transcription terminators in the 3′ UTR. Additionally, SNP and InDel variants were identified using the VarScan program. Unannotated expression regions in the database were annotated, and the results were visualized. Temporal metabolomic analysis of Acinetobacter sp. AS23 following tea saponin exposure Fermentation broth samples of the AS23 strain incubated with tea saponin or the control collected at different time points (0 h, 24 h, 48 h, and 72 h) (n = 5) and stored at – 80 °C were extracted using an extraction solution (methanol/acetonitrile = 1:1, internal standard concentration 2 mg/L). The samples were sonicated in an ice-water bath for 10 min, rested at − 20 °C for 1 h, and then centrifuged at 12,000 rpm and 4 °C for 15 min. Then, 500 μL of supernatant was transferred to EP tubes and dried in a vacuum concentrator. The dried metabolites were reconstituted with 160 μL of extraction solution (acetonitrile/water = 1:1), vortexed for 30 s, and sonicated in an ice-water bath for 10 min. Finally, 120 μL of supernatant was transferred to 2-mL sample vials, and 10 μL from each sample was pooled to create QC samples for analysis. The LC–MS system for metabolomics analysis consisted of a Waters Acquity I-Class PLUS UPLC coupled with a Waters Xevo G2-XS QTOF high-resolution mass spectrometer. Chromatographic separation was performed on a Waters Acquity UPLC HSS T3 column (1.8 μm, 2.1 × 100 mm). For positive ion mode, the mobile phase A was 0.1% formic acid in water, and mobile phase B was 0.1% formic acid in acetonitrile. The same mobile phases were used for negative ion mode. The injection volume was 1 μL [[168]52]. Raw data collected using MassLynx V4.2 was processed for peak extraction and alignment using Progenesis QI software. Metabolites were identified using the online METLIN database in Progenesis QI and the in-house database from Biomarker Technologies (Beijing, China). Theoretical fragment recognition was conducted with mass tolerances of 100 ppm for parent ions and 50 ppm for fragment ions. Metabolite data from both positive and negative ion modes were collected to enhance the metabolite coverage and accuracy. After database searching, the overall data quality was evaluated. PCoA was performed for sample clustering of the treatment groups. Partial least squares discriminant analysis (PLS-DA) was used to determine differences between treatments. A functional enrichment analysis of differential metabolites was conducted using KEGG analysis to understand functional differences. A correlation analysis between differential genes and metabolites was performed using R Studio (environment: R-3.1.1, packages: ggplot2, enrichplot, clusterProfiler) [[169]53], and the correlation network was visualized using Gephi (version 0.9.2) software. The R package Psych was used to analyze the nodes and edges of the network. The network nodes represented individual DEGs or metabolites, while edges indicated statistically significant correlations between nodes within the network. The networks parameters included average path length, clustering coefficient, and connectivity, along with co-concurrent (positive) and mutually exclusive (negative) correlation numbers, modularity, and network diameter. The role of individual nodes was determined by considering their degree and closeness centrality topological features. qPCR analysis of genes in the benzoate degradation pathway Ko00362 of Acinetobacter sp. AS23 The expression levels of K16243, K05549, K05783, and [170]K01055 in the differentially enriched benzoate degradation pathway Ko00362 were further validated by qPCR. The AS23 strain total RNA was extracted using a DP430 bacterial RNA extraction kit (Tiangen Biotech Beijing, China) and treated with RNase-free DNase I (Thermo Scientific, USA) to remove DNA contamination. The total RNA from each sample was then reverse transcribed using a FastKing cDNA first strand synthesis kit (Tiangen KR116, Beijing) and used as a template. qPCR was performed on a fluorescence quantitative PCR instrument (Bio-Rad CFX96) using AceQ® qPCR SYBR® Green master mix (Vazyme, China). The relative expression of each DEG was evaluated using the 2^−ΔΔCt method. Three biological replicates and three technical replicates were used for each gene and sample. Generation of mutant strains via CRISPR-Cas9 dual-plasmid system targeting the Ko00362 benzoate degradation pathway Following the method of Wang et al. (2019), a CRISPR-Cas9 two-plasmid system was constructed in the AS23 strain to edit the genes K16243, K05549, K05783, K16243 + K05549, and K16243 + K05783 in the benzoate degradation pathway Ko00362 for functional verification. First, two plasmids were constructed: the first plasmid, pCasAb-apr, contained the Cas9 protein gene fragment, RecAb recombination system gene fragment, aprR apramycin resistance gene fragment, sacB sucrose screening gene fragment, and a tac promoter fragment to regulate the expression of the RecAb recombination system and Cas9 gene. The second plasmid, pSGAb-km, expressed sgRNA and contained the kmR kanamycin resistance gene fragment, sacB sucrose screening gene fragment, targeting sequence, and Cas9 nuclease recruitment sequence [[171]54]. After plasmid construction, electrocompetent cells of the wild-type strain AS23 were prepared, which were transformed with the pCasAb-apr plasmid. Subsequently, the AS23 strain containing pCasAb-apr was rendered electrocompetent. The pre-prepared pSGAb-target gene-out plasmids containing Ko00362 gene knockout sequences (K16243, K05549, K05783, K16243 + K05549, and K16243 + K05783) and corresponding homologous repair templates were mixed and transformed into the electrocompetent AS23 strain cells harboring pCasAb-apr. The transformants that grew normally on double-antibiotic medium were verified by colony PCR and sequencing to confirm successful target gene deletions. For gene insertion, the insertion fragment and upstream and downstream homologous recombination templates were joined into a complete DNA fragment by overlap PCR. The purified product was mixed with the prepared pSGAb-target gene-in plasmid and transformed into electrocompetent AS23 strain cells containing pCasAb-apr. The transformants were verified by colony PCR and sequencing to confirm successful gene insertions. Finally, pCasAb and pSGAb plasmids were cured. Both plasmids contained the sacB gene, which expresses secretory levansucrase that catalyzes sucrose hydrolysis into glucose and fructose and polymerizes fructose into the toxic high-molecular-weight levan, which causes bacterial death. The edited bacterial culture was plated on LB agar containing 5% sucrose. The resulting colonies were further verified by streaking on LB agar containing apramycin (100 mg/L, Sangon, China) and kanamycin (50 mg/L, Sangon, China) and on antibiotic-free LB agar. Colonies growing only on sucrose-containing and antibiotic-free media were confirmed to be plasmid-free and were preserved or used for further experiments. The mutant strains with deletions in K16243, K05549, K05783, K16243 + K05549, and K16243 + K05783 were spot-plated on solid medium containing 5 g/L benzoate as the sole carbon source (composition 5 g of (NH[4])[2]SO[4], 2.5 g of NaCl, 0.3 g of KH[2]PO[4], 0.05 g of FeSO[4], 0.5 g of MgSO[4], 5 g of benzoic acid sodium benzoate (purity 99.5%, Macklin, China), sterile water to 1 L, pH 7.2; 15 g of agar) to verify their ability to utilize benzoate. Subsequently, benzoate-intolerant strains were preserved in glycerol at − 80 °C. Comparison of the benzoate degradation capacity of mutant strains and the wild-type strain The five AS23 mutant strains, preserved at − 80 °C, were activated in LB medium (containing 10 g/L tryptone, 5 g/L yeast extract, 10 g/L sodium chloride, Sangon, China) and stored at 4 °C for subsequent use. In the treatment group, 2 mL of an AS23 mutant strain culture (OD[600] = 2) was inoculated into 50 mL of liquid medium containing 5 g/L benzoate as the sole carbon source. In the control group, 2 mL of sterile water was added to 50 mL of the same medium. Five replicates were used for each group. Both the treatment and control groups were incubated in a shaking incubator at 200 rpm and 37 °C under dark conditions. After 24 h of degradation, 30 mL of fermentation broth was collected from each replicate and filtered to remove bacteria and impurities. The filtered samples were stored at − 80 °C for subsequent analysis. A high-performance liquid chromatography system was equipped with a C18 column, and the mobile phase consisted of a methanol–water mixture at a flow rate of 1 mL/min, with the column temperature maintained at 35 °C. The samples underwent liquid–liquid extraction and rotary evaporation, and then were dissolved in methanol to a final volume of 1 mL to ensure consistent sample concentrations. Using an autosampler, 50 μL of each sample was injected into the chromatography system. The components were separated by the column, and their UV absorption was detected at 254 nm using a UV detector. Chromatograms were recorded and analyzed using the chromatography workstation software. Benzoic acid peak areas were calculated and converted to concentrations using a pre-established standard curve for quantitative analysis. After the experiment, the column was washed to remove residual samples, and regular maintenance was performed on the chromatography system to ensure stable performance. Verification of the degradation product toxicity and in vivo detoxification capacity of the mutant strain To verify the toxicity of the benzoate degradation products of the mutant strains, we mixed the fermentation filtrate (obtained after a 24-h treatment of mutant strains) with artificial CW feed. Third-instar axenic CWs larvae were fed with this mixture, and their survival was monitored for 30 days. To determine the in vivo detoxification capacity of the mutant strains, we mixed [172]C00180 with artificial feed and administered it to third-instar axenic CW larvae. The survival of these weevils was monitored over a 30-day period. Statistical analysis All of the graphs (bar, line, bubble, and heat maps) were created using the OmicShare Tools web site and R Studio (environment: R-3.1.1, software package: ggplot2, enrichplot, clusterProfiler) [[173]51]. Survival curves were generated using GraphPad software, and statistical differences between treatment groups were analyzed using the log-rank test. The images in the figures were merged using Adobe Photoshop 2024. There were significant differences among all of the experimental treatments. Univariate analysis of variance (ANOVA) and Tukey tests were performed using IBM SPSS 22.0 software, with a significance level of P < 0.05. Supplementary Information [174]Supplementary Material 1^ (25.4KB, docx) Acknowledgements