The L. decemlineata beetles were routinely reared using a previously described method (Meng et al., 2019). Briefly, the beetles were maintained in an insectary at 28°C, under a 16:8 h (light/dark) photoperiod, and 50–60% relative humidity using potato foliage at the vegetative growth or young tuber stages to assure sufficient nutrition. At this feeding protocol, the larvae progressed the first-, second-, penultimate-, and final-instar stages with approximate periods of 2, 2, 2, and 4 days, respectively. On reaching full size, the final larval instars stopped feeding, dropped to the ground, burrowed to the soil, and entered the prepupal stage. The prepupae spent ~3 days to pupate. The pupae lasted about 5 days and the adults emerged.

All solvents used for UPLC-Q-TOF-MS analysis were of analytical or HPLC grade and purchased from Sigma-Aldrich Co., Ltd. (Shanghai, China).

The detailed procedure of sample preparation for UPLC-Q-TOF-MS was as follows: ten (5 males and 5 females) 2-day-old fourth-instar larvae, ten 4-day-old pupae, and ten 5-day-old adults were collected as a replicate. The collection continued for three consecutive generations to generate three biologically independent replicates. The specimens were individually weighed and then crushed in a pre-chilled mortar using a pellet pestle. Each sample was lyophilized, added with 1 ml of methanol/water (7:3, v/v), vortexed, and sonicated twice on ice for 30 min. Then the metabolite was incubated at −20°C for 1 h and centrifuged at 13,000 rpm at 4°C for 15 min. The supernatant was lyophilized and stored at −80°C.

Functional prediction was carried out based on the operational taxonomic units (OTU) data downloaded from the NCBI Sequence Read Archive (SRA) database (Accession Number: PRJNA613266) (Kang et al., 2021). The OTU data were obtained by high-throughput sequencing of 16S rRNA genes from three biologically independent collections from three consecutive generations. Each collection included ten 2-day-old fourth-instar larvae, ten 4-day-old pupae, and ten 5-day-old adults with the sex ratio of 1:1 (5 males and 5 females). The PICRUSt algorithm was used to infer the functions of the bacterial communities through the Kyoto Encyclopedia of Genes and Genomes (KEGG) database (Kanehisa et al., 2012).

The supernatant was reconstituted with 500 μl of methanol, vortexed, and centrifuged at 20,000 rpm at 4°C for 20 min. Then the supernatant was filtered through a 0.22 μm (nylon) syringe filter and analyzed by UPLC-Q-TOF-MS system (Waters). Each sample was analyzed six times (i.e., 2 μl aliquot of each sample was injected six times).

Since the retention times or even elution order in analytical system may vary during the UPLC-Q-TOF-MS analysis, it is necessary to monitor the system consistency. In this study, a 60 μl mixture of the control (30 μl) and treatment samples (30 μl) was used as a quality control (QC) sample for method validation. A QC sample ran four times prior to beginning the whole sample list. Moreover, a QC sample was run every six samples during the analytical run.

The analysis was performed on an HSS T3 column (2.1 × 100 mm, 1.8 μm; Waters Acquity), using an ACQUITY UPLC I-Class PLUS System (Waters) coupled with a Xevo® G2-XS QT of High-Definition Mass Spectrometer (Waters). Mobile phase A was water and mobile phase B was 0.1% formic acid acetonitrile solution. The gradient elution procedure was set as follows: 0–1 min, 0–2% B; 1–2 min, 2–25% B; 2–4 min, 25–60% B; 4–7.5 min, 60–90% B; 7.5–9.5 min, 90–99% B; 9.5–12.5 min, 99% B; 12.5–13 min, 99–2% B; and 13–16 min, 2% B. The flow rate was 0.4 ml/min, and the column temperatures were held constant at 45°C.

Each sample was detected by positive and negative ion modes using an electrospray ionization mass spectrometer (ESI-MS). The product ion scan was acquired using the first- and second-level mass spectrometry data acquisition method based on the Photodiode Array (PDA) detector.

The ESI conditions were as follows: nitrogen was used as cone gas and desolvation gas at a flow rate of 50 and 800 l/h, respectively. The source temperature was 120°C and desolvation gas temperature was 450°C. Quality scanning range was set m/z 50–1,200. In positive ion mode, capillary, cone, and extraction cone voltages were 3.0 kV, 40 V, and 5.0 V, respectively. In negative ion mode, capillary, cone, and extraction cone voltages were 2.0 kV, 40 V, and 5.0 V, respectively. MS data were acquired in full-scan mode from 100 to 1,000 Da.

The raw data detected by UPLC-Q-TOF-MS were loaded on the commercial metabolites database Progenesis QI (Waters Corporation, Milford, USA) for peak detection, alignment, and normalization, as well as the main information, such as the mass, retention time, and intensity of the peaks in each chromatogram. The metabolites were identified by comparing their retention times, m/z values, and MS fragmentation patterns with those of commercial standard compounds. Fragmentation patterns collected in online databases, such as MycompoundID (http://www.mycompoundid.org), MassBank (http://www.massbank.jp), ChemSpider database (www.chemspider.com), and METLIN (http://metlin.scripps.edu) were also considered, especially when no authentic standard compounds were available.

Before multidimensional statistical analysis, the data were processed: the missing values of the original data >50% were excluded. The processed data were then imported into SIMCA-P14.1 software (Umetrics, Umea, Sweden) for pattern recognition, and Pareto scaling was used to preprocess the data for principal component analysis (PCA) and orthogonal PLS-DA analysis (OPLS-DA). According to the Variable Importance for the Projection (VIP) obtained by the OPLS-DA model and Max Fold Change (MFC) from Progenesis QI, the influence intensity and explanatory ability of each metabolite on the classification and discrimination of each group of samples were evaluated, and the biologically significant differential metabolites were mined. The larger the VIP and MFC values, the greater the contribution of the metabolite in the differentiation of the sample, and the variable with VIP > 1 and MFC > 2 was generally considered to have a significant difference. In the experiment, based on the screening criteria of VIP > 1 and MFC > 2, the substances between the groups were initially screened. Next, the univariate statistical analysis was used to verify whether the selected metabolites had significant differences. Ions meeting VIP > 1, MFC > 2, and 0.05 < p < 0.1 were considered differential metabolites; VIP > 1, MFC > 2, and p < 0.05 were regarded as significantly different metabolites.

To highlight differential biomarker role, the resulting significant differential metabolites were analyzed in KEGG (http://www.kegg.jp) to resolve the topological trait of metabolic pathways.

The same method was used to expose the larvae to an antibiotic mixture (Löfmark et al., 2010; Xia et al., 2020). Briefly, a mixture containing three antibiotics (1 mg/ml ciprofloxacin, 1 mg/ml levofloxacin, and 2 mg/ml metronidazole) in 1% Tween 20 aqueous solution was used to immerse potato foliage. Tween 20 in sterile water (1%) was set as the control group. Ten newly ecdysed fourth-instar larvae were confined in a Petri dish (9 cm diameter and 1.5 cm height) containing five treated leaves. The larvae were allowed to feed on the treated leaves until they reached the prepupal stage. The foliage was replaced with freshly treated ones each day.

The total microbial DNAs were individually extracted from 4-day-old pupae having fed on the antibiotic mixture or control solution as larvae, using E.Z.N.A.® Tissue DNA Kit (Omega). The removal of bacteria was examined using a pair of universal primers, the forward primer 5′-TCCTACGGGAGGCAGCAGT-3′ and reverse primer 5′-GGACTACCAGGGTATCTATCCTGTT-3′, of 16S rDNA from the Domain Bacteria (Nadkarni et al., 2002; Silkie and Nelson, 2009). Quantitative DNA measurements were performed by real-time quantitative reverse transcription PCR (qRT-PCR) in technical triplicate. Relative expression level of 16S rDNA was calculated by the 2^−ΔΔCT method, using the geometric mean of two internal control genes (LdRP4 and LdARF1) (Shi et al., 2013) for normalization.

To further examine the removal of bacteria, the dissected guts from newly emerged adults in control and treatment groups were homogenized. The supernatants separated by centrifugation at 500 g were spread on a plate culture medium (Luria–Bertani) after diluting for 10 times. The dishes were inoculated at 30°C for 17 h and then the bacterial spots were detected.

The metabolites in control and bacteria-removed pupae were tested using UPLC-Q-TOF-MS system and the data were analyzed using the method described above.

Using SPSS for Windows (Chicago, IL, USA), one-way analysis of variance (ANOVA) with a Tukey–Kramer test, or Student's t-test was performed to determine significant difference between average values (±SD). Results were considered statistically significant when p < 0.05.

The OTUs of bacteria obtained by PCR amplification and sequencing of culture-independent 16S rRNA (Kang et al., 2021) were analyzed using KEGG enrichment analysis. At KEGG level 1, metabolism was the most active one, especially in the pupa group (Figure 1A). The highly expressed genes associated with the catabolism of aromatic compounds (at KEGG level 3) were observed in the pupa group, followed by those in the larvae, and the levels were lower in the adult collection. One-way ANOVA revealed that a statistical significant difference was present between pupa and adult groups (p < 0.05) (Figure 1B).

Correspondingly, the actively expressed genes (at KEGG level 3) involved in the degradation of benzoate, aminobenzoate, styrene, toluene, xylene, nitrotoluene, ethylbenzene, bisphenol, fluorobenzoate, naphthalene, dioxin, steroid, polycyclic aromatic hydrocarbon, caprolactam, atrazine, furfural, chloroalkane, chloroalkene, chlorocyclohexane, and chlorobenzene were significantly richer in the pupa group than those in the larva and adult collections (Figure 1C).

To obtain comprehensive information on the metabolome, small molecule metabolites in the larval, pupal, and adult samples were analyzed by UPLC-Q-TOF-MS. A preliminary analysis of all samples was performed by PCA. The larva and adult groups were biased compared with the pupa collection in both positive and negative ion maps, whereas the larva and adult groups were overlapped (Figures 2A,B).

By OPLS-DA model, 43 potential aromatic biomarkers were screened in positive ion mode. Out of them, a subset of 39 was greater in the pupal specimen compared with those in the larval sample, while another subset of 40 was more in the pupal specimen than those in the adult group (Table 1). A total of 11 potential aromatic biomarkers were obtained in negative ion mode. Among them, subsets of 9 and 8 were greater in the pupal sample than those in the larva and adult groups, respectively (Table 2).

Biological pathway analysis revealed that five active pathways were involved in the metabolism of aromatics in pupae compared with those in the larva/adult groups. These ways contained the catabolism of styrene, naphthalene, fluorobenzoate, aminobenzoate, and benzoate, respectively (Figures 2C,D).

Feeding a mixture of three antibiotics by the fourth-instar larvae almost completely removed culturable aerobic bacteria (Figure 3B vs. Figure 3A), and all other bacteria (Figure 3C) in the resultant pupae. In the PCA graphs, the aposymbiontic groups were biased compared with the control groups in both positive and negative ion maps (Figures 3D,E).

In both positive and negative ion modes, the differences in xenobiotic biomarkers between control and aposymbiontic pupal samples included a great number of aromatic compounds (Tables 3, 4).

Metabolic pathway analysis revealed that a total of seven active pathways were enriched in either control or aposymbiontic pupal samples (Figures 3F,G). These ways were involved in the degradation of styrene, naphthalene, benzoate, fluorobenzoate, aminobenzoate, xylene, and dioxin.

Identified metabolites and biological pathways were imported into the KEGG (http://www.kegg.jp/) to find interactions. Among networks for catabolism of aromatics, benzoates, 4-methoxybenzoate, vanillin, and benzamide were finally transferred into succinyl-CoA and acetyl-CoA, which were completely degraded to CO₂ through tricarboxylic acid cycle (Figures 4A,B,E). Fluorobenzoates, for instance, 3-fluorobenzoate and 4-fluorobenzoate, were finally converted to 2-fluoro-cis, cis-muconate, 3-fluoro-cis, cis-muconate, and 4-fluoromuconolactone (Figure 4F). Styrene was oxidized to phenylacetic acid (Figure 4D). In contrast, the degradation of (2-naphthyl)methanol (naphthalene) was only partially annotated by KEGG (Figure 4C).

Comparison of metabolite enrichment between control and aposymbiontic pupal samples revealed that 4-methoxybenzoate (Figure 4A), vanillin (Figure 4B), (2-naphthyl)methanol (Figure 4C), styrene oxide, and phenylacetaldehyde (Figure 4D) were higher in the control group. These findings indicated that the soil aromatic precursors could be actively converted into these compounds by pupa bacteria. Conversely, benzoate, styrene, and 3,4-dihydroxybenzoate were accumulated in the aposymbiontic pupal group (Figures 4A,B,E). These results suggested that pupa bacteria were associated with the catabolism of the three chemicals. In contrast, the contents of 3-fluorocatechol and 4-fluorocatechol in control group were similar to those in the aposymbiontic pupal group (Figure 4F). These data implied that pupae may metabolize fluorobenzoates.

Incorporation of plant residues and organic fertilizers into soil brings about some secondary chemicals, e.g., alkaloids, terpenoids, cardenolides, glucosinolates, and oxalates (Zhang et al., 2020), and their metabolites such as monoaromatics (e.g., BTEX and phenol) and polycyclic aromatics (e.g., naphthalene, dioxin, and steroids) (Chen et al., 2017; Steinmetz et al., 2019). These compounds often exert deleterious effects to soil-dwelling pupae when accumulated to high concentrations within the bodies. Insects are hypothesized to specify their microbial community compositions to degrade these substances to avoid intoxication, if suitable ecological conditions are satisfied (Zhang et al., 2020). However, the experimental evidence to support the hypothesis is very limited.

In the present study, we analyzed the bacterial OTU data (Kang et al., 2021) by KEGG and found that a subset of bacterial genes was abundantly expressed at the pupal stage. These genes were associated with the catabolism of aromatics (Figure 1). Consistently, metabolomic analysis displayed that dozens of monoaromatics, polycyclic aromatics, and steroids were richer in the pupal sample than those in the larval and adult specimens (Tables 1, 2).

According to the identified metabolites and biological pathways, a total of seven active ways were enriched. These active pathways are mainly involved in the degradation of various benzoates and their precursors, which include benzoate, 4-methoxybenzoate, 3-fluorobenzoate, 4-fluorobenzoate, vanillin (aminobenzoate), and benzamide (Figure 4). Most of these benzoates were finally transferred into succinyl-CoA and acetyl-CoA, which were completely degraded to CO₂ through tricarboxylic acid cycle. Similarly, Dechloromonas sp. strains RCB and JJ can completely break down aromatic compounds into CO₂, coupled with the reduction of nitrate (Coates et al., 2001). Dechloromonas belongs to proteobacteria. Among the 18 pupa-specific genera identified (Kang et al., 2021), Escherichia_Shigella, Acinetobacter, Pseudomonas, Lysobacter, and Stenotrophomonas are also proteobacteria. One or several genera listed above may be responsible for the catabolism of the benzoates in L. decemlineata pupae.

Specifically, we uncovered that pupa bacteria were involved in the aerobic degradation of styrene to phenylacetic acid in L. decemlineata (Figure 4). In agreement with our result, the way in other documented results involves epoxidation of the vinyl side chain, followed by isomerization of the epoxy styrene to form phenylacetaldehyde. This compound is subsequently oxidized to phenylacetic acid through the action of either an NAD⁺- or phenazine methosulfate-dependent dehydrogenase (O'Connor and Dobson, 1996; O'Leary et al., 2002). Conversion of styrene to phenylacetic acid has been documented in various bacterial strains; examples include Pseudomonas putida CA-3 (O'Connor et al., 1995), P. fuorescens ST (Marconi et al., 1996), Pseudomonas sp. VLB120 (Panke et al., 1999), Pseudomonas sp. Y2 (Utkin et al., 1991; Velasco et al., 1998), Xanthobacter sp. 124X (Hartmans et al., 1989), and Xanthobacter sp. S5 (Hartmans et al., 1990). Pseudomonas is among the 18 pupa-specific genera identified recently (Kang et al., 2021), and the genus may be responsible for the catabolism of styrene in L. decemlineata pupae.

At present, polycyclic aromatic hydrocarbons produced by all vertebrates as well as some invertebrates have been considered one of the most important environmental problems (Xiong et al., 2019; Chiang et al., 2020). In this study, we discovered that the polycyclic aromatic hydrocarbon (2-naphthyl)methanol (naphthalene) can be biodegraded by pupa-specific bacteria in L. decemlineata (Figure 4). It has been reported that naphthalene can be catabolized by P. fluorescens AH-40 (Mawad et al., 2020), P. putida BS3701 (Pozdnyakova-Filatova et al., 2020), and Stenotrophomonas sp. S1VKR-26 (Patel and Patel, 2020). Naphthalene is converted via salicylate and catechol to the intermediates of tricarboxylic acid cycle in P. putida PpG1 (Yen and Gunsalus, 1982), catalyzed by four key enzymes, namely, naphthalene 1,2-dioxygenase, salicylate hydroxylase, catechol 2,3-dioxygenase, and catechol 1,2-dioxygenase (Izmalkova et al., 2006). In this survey, we determined that 3- and 4-fluorocatechol, 4-hydroxymethylcatechol, and 4-methylsalicylate were higher in the pupae than those in the larvae and adults. It appears that the same naphthalene transformation way is present in the L. decemlineata pupae. Consistent with the metabolomic result, our microbiome analysis revealed that both Pseudomonas and Stenotrophomonas are among the 18 pupa-specific genera identified recently (Kang et al., 2021).

Both biogenic (natural) and anthropogenic steroids are frequently detected in soils and aquatic environments in China (Chiang et al., 2020). For example, oestrogens, androgens, progestogens, glucocorticoids, and mineralocorticoids are detected in the surface water of urban rivers in Beijing (Chang et al., 2009). Bacteria are responsible for mineralizing polycyclic aromatic hydrocarbons from the biosphere (Chiang et al., 2020). Among 18 pupa-specific bacterial genera (Kang et al., 2021), Nocardia (Coombe et al., 1966), Rhodococcus (Fernandez de las Heras et al., 2009; Li et al., 2018), and Stenotrophomonas (Juhasz et al., 2000; Tachibana et al., 2003; Guan et al., 2019; Xiong et al., 2019) have been documented to degrade steroids from other environments. Whether the three bacterial genera in L. decemlineata can break down steroids deserves further research.

Saponins are a class of secondary plant metabolites which includes triterpenoids, steroids, and steroidal alkaloids glycosylated with one or more sugar chains. They are produced by many plant species (Zhang et al., 2020). Saponins provoke molting defects in, and exert deleterious effects on insects (De Geyter et al., 2007; Podolak et al., 2010; Cai et al., 2016; Dolma et al., 2017). Therefore, tea saponin has been widely used as an insecticide in China (Cai et al., 2016). We herein demonstrated that the contents of saponin H and assamsaponin D were higher in the pupae compared with those in the larvae and adults in L. decemlineata (Figure 4).

Although a considerable amount of saponins is indicated in the soil, L. decemlineata pupae could still develop into adults, indicating that other factors may help L. decemlineata resist saponins. Some bacteria, for instance, Acinetobacter calcoaceticus and A. oleivorans, are known to detoxify saponins (Zhang et al., 2020). Moreover, the mixed cultures of Methanobrevibacter spp. and Methanosphaera stadtmanae in the crop of the avian foregut fermenter in Opisthocomus hoazin are able to reduce the hemolytic activity of Quillaja saponins by 80% within a few hours (García-Amado et al., 2007). Consistent with these results, Acinetobacter is a pupa-specific genus (Kang et al., 2021), and it may be responsible for the metabolism of saponins in the L. decemlineata pupae.

Potato plants contain many aromatics, where some exert noxious effects when accumulated to high concentrations within insect bodies (Gandia-Herrero and Garcia-Carmona, 2013; Kostyn et al., 2020). In this survey, we discovered that only 6 and 7 aromatics were accumulated in the larvae and adults, respectively, in contrast to 42 cumulated aromatics in the pupae (Tables 1, 2). The less accumulation implies the more active elimination of aromatics in the L. decemlineata larvae and adults.

Consistently, the microbiota are widely distributed in the larval or/and adult guts (Jing et al., 2020). These gut microbiota are involved in the breakdown of noxious compounds in numerous insect species in Coleoptera (Ceja-Navarro et al., 2015; Berasategui et al., 2017; Zhang et al., 2020), Lepidoptera (Vilanova et al., 2016; Zeng et al., 2020), Diptera (Griffin and Reed, 2020), Hymenoptera (Wu et al., 2020), and Isoptera (Van Dexter and Boopathy, 2019).

In this survey, the bacterial OTU data (Kang et al., 2021) revealed that the catabolism of aromatics was less active in the larvae and adults compared with that in L. decemlineata pupae (Figure 1). Therefore, the bacterial biodegradation of aromatics only partially contributes to the removal of excessive aromatics. Intestinal excretion should be another route to eliminate superfluous aromatics (Rozman, 1985). In fact, insecticides can be excreted by insects, directly or indirectly (modified forms) (Quesada et al., 2020). In L. decemlineata larvae and adults, superfluous aromatics in food may be excreted through the guts directly, or transferred to more hydrophilic forms. Only those potato aromatics absorbed by L. decemlineata larvae and adults need to be biodegraded by bacteria. Conversely, the alimentary canal is not well-developed in L. decemlineata pupae and cannot actively remove excessive aromatics. The pupae mainly depend on bacteria to catabolize the noxious substances.

In summary, we uncovered the stage-dependent alterations in bacterial degradation of aromatics in L. decemlineata. The candidate bacterial genera contributing to aromatic catabolism were Nocardia, Rhodococcus, Enterococcus, Acinetobacter, Pseudomonas, and Stenotrophomonas, among others. This study provides new insights into the adaptation of L. decemlineata to different environmental niches and offers a better understanding of the relationship between ONS and a shift of bacterial flora. Moreover, since removal of the symbiotic bacteria inhibited the breakdown of superfluous aromatics (this study) and results in a decrease in the emergence rate and adult weight (Kang et al., 2021), disruption of bacterial communities may be a potential strategy to control L. decemlineata.