Background

Morphological traits of larvae and adults are genetically decoupled in favor of adaptation to selective forces, which possibly explains the advantage of metamorphosis [1]. Symbiotic bacteria, such as Lactobacillus plantarum, positively affect many aspects of host physiology and are therefore passed on to host offspring [2, 3]. Holometabolous insects are at risk of losing symbiotic bacteria, leading to opportunistic infections, as the gut is renewed during complete metamorphosis [4, 5]. Hence, the passage of symbiotic bacteria through metamorphosis is key to ensuring vertical transmission to host offspring [5], representing a fundamental problem encountered by most holometabolous insects. To minimize the risk of opportunistic infections caused by symbiotic bacteria during metamorphosis, the strategies that holometabolous insects have evolved remain obscure.

Insect immune responses, including antimicrobial peptide (AMP) synthesis, are initiated rapidly following the recognition of microbes by pattern recognition receptors (PRRs), such as peptidoglycan recognition proteins (PGRPs) and C-type lectins (CTLs) [6, 7]. Insect immunity plays crucial roles in sha** gut microbiota and minimizing opportunistic infections [8,9,10]. In mosquitoes, depletion of PGRP-LD results in the upregulation of several AMPs in the midgut, thus killing a large number of gut microbes [8]. In Drosophila, the loss of AMPs and lysozymes (Lys) contributes to an increase in the abundance and changes in the composition of gut microbiota [9]. Downregulation of AMPs resulting from fungal infection leads to dysbiosis of gut microbiota and translocation of symbiotic bacteria from the gut to hemocoel [10]. Although PRRs, AMPs, and Lys are essential for maintaining gut homeostasis, few studies have demonstrated how immune pathways regulate gut microbiota at the effector level, especially during complete metamorphosis.

Host development, such as metamorphosis, can be modulated or influenced by gut symbiont [3, 11, 12]. The gut bacterium Acetobacter pomorum controls Drosophila developmental rate by modulating insulin or insulin-like growth factor signaling [11]. Gut bacterium L. plantarum promotes Drosophila systemic growth by influencing ecdysone and insulin signaling [3]. These indicate that symbiotic bacteria are beneficial to host development in the gut. However, Ma et al. [13] have shown that gut microbiota retards larval weight growth and serves as a burden in the beetle’s development. Gut microbiota has the potential to migrate into the hemocoel and attenuate host development, representing opportunistic pathogens [14, 15]. For example, Enterococcus faecalis changes its role from resident symbiont in the midgut to a pathogen in the hemocoel [14]. Injection of gut bacteria Enterocuccus mundtii into the hemocoel reduces larval body size and delays pupariation [15]. It has been shown that the immune activation is energetically costly, leading to a decrease in growth and nutrient stores [16]. The mechanisms by which opportunistic infections affect host development, especially insect metamorphosis, remains largely elusive, although gut microbial dysbiosis and penetration of gut microbes into the circulation sometimes occur [10, 17].

The steroid hormone ecdysone is not produced and stored prior to release, but is produced on demand from its precursor cholesterol esters stored in lipid droplets (LDs) [18]. As an endocrine regulator, ecdysone and its active form 20-hydroxyecdysone (20E) not only trigger metamorphosis [19], but positively regulate insect immunity [20]. Activation of PGRP-SC2 or AMP expression extends Drosophila lifespan [21, 22], suggesting that immunity affects the endocrine system and development. Nunes et al. [23] have revealed that ecdysone signaling impacts pupariation by modulating the immune system, emphasizing the importance of the association between the endocrine system and immunity.

The lepidopteran insect Helicoverpa armigera is one of the most destructive and polyphagous pests worldwide [24]. Here, we used H. armigera as a model and compared the profiles of bacteria and immunity-related genes in the midgut between the feeding and wandering stages. We found that 20E enhanced the expression of midgut CD209 antigen-like protein 2 (CD209, as a CTL, GenBank no. XM_021344387) and its downstream gene Lys1 (GenBank no. XM_049840569). RNAi-mediated depletion of CD209 or Lys1 abolished the inhibitory capacity toward midgut Lactobacillus during the wandering stage. Translocation of overgrowing Lactobacillus and its possibly associated peptidoglycan (PGN) from the midgut to hemocoel led to the decreased triacylglycerol (TAG) and cholesterol storage, thereby reducing the availability of energy and ecdysone for pupariation. This study revealed the importance of CTL − Lys in gut homeostasis, which was crucial for minimizing the risk of opportunistic infections and ensuring timely metamorphosis.

Results

A significant decrease in Lactobacillus load in the midgut transitioned from the feeding to wandering stage

We have previously defined the developmental stages of H. armigera [25]. Sixth-instar at 24 h post-ecdysis (PE) is an active feeding stage, and sixth-instar at 72 h PE represents a wandering stage (the initial stage of metamorphosis). To compare the composition and diversity of microbiota, 16S rRNA gene sequencing was performed using the midguts of sixth-instar larvae at 24 h PE (feeding stage) and 72 h PE (wandering stage). In total, 455 operational taxonomic units (OTUs) were obtained, which were annotated into 24 phyla, 49 classes, 113 orders, 183 families, 294 genera, and 382 species (Additional file 1: Table S1). The number of bacterial genera in the wandering stage was higher than that in the feeding stage, with 169 unique genera in the former, 23 unique genera in the latter, and 102 genera shared (Fig. 1A).

Fig. 1
figure 1

The abundance of Lactobacillus is significantly decreased in the midgut and significantly increased in hemolymph when larvae transition from the feeding to wandering stage. A Venn diagram of bacterial genus abundance in the larval midgut. Midguts were collected from sixth-instar larvae at 24 h PE (feeding) and 72 h PE (wandering). B Principal coordinates analysis (PCoA) of bacterial communities from the midgut based on genus level. C Relative abundance (%) of bacterial genera between the feeding and wandering stages. D, E Quantification of midgut Lactobacillus (D) and hemolymph Lactobacillus (E) in the feeding and wandering stages. Quantification was by qPCR analysis using primers of Lactobacillus-specific 16S rRNA gene. Statistical differences were analyzed using the Mann–Whitney U test or Student’s t test (*p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001)

Full size image

Principal coordinate analysis (PCoA, Bray–Curtis) revealed that the structure of midgut microbiota is different between the feeding and wandering stages (Fig. 1B). Of the most abundant 8 genera, the abundance of Lactobacillus decreased the most, from 86.34% in the feeding stage to 15.15% in the wandering stage (Fig. 1C). Quantitative PCR (qPCR) revealed that the load of midgut Lactobacillus largely decreased, whereas the load of hemolymph Lactobacillus dramatically increased when larvae transitioned from the feeding to wandering stage (Fig. 1D, E). These results are consistent with our previous findings that the abundance of Lactobacillaceae in hemolymph is markedly elevated after larvae enter the wandering stage [15]. Similarly, the total bacterial load largely decreased in midgut and dramatically increased in hemolymph when larvae entering from the feeding to wandering stage (Additional file 2: Fig. S1).

A significant increase of CD209 and Lys1 expression in the midgut entering from the feeding to wandering stage

To obtain differentially expressed genes (DEGs), we performed transcriptomic sequencing of midguts from sixth-instar larvae at 24 and 72 h PE. A total of 5429 DEGs were identified, including 2712 up- and 2717 downregulated genes (Additional file 1: Table S2). Of the 5429 DEGs, 87 immunity-related DEGs were screened, including 24 genes encoding PRRs, five encoding AMPs and Lys, 12 involved in signaling, 37 encoding serine proteases, and nine encoding serpins (Additional file 1: Table S3). The transcript abundance of most of the PRR, AMP, and Lys genes increased during the wandering stage compared with the feeding stage (Fig. 2A, B).

Fig. 2
figure 2

CD209 is dominantly expressed in the midgut of the wandering stage. A, B Heatmap showing the expression profiles of PRRs (A), AMPs, and Lys (B) in the midgut between the feeding (MgF) and wandering (MgW) stages. CD209 and Lys1 are highlighted in red. C Predicted domain architecture of CD209 protein. Characterization of a predicted signal peptide (SP, 1–19 aa) and a carbohydrate recognition domain (CRD, 23–162 aa) from http://smart.embl.de/ website. D Expression and purification of recombinant CD209 (rCD209), as well as rCD209 recognized by the anti-CD209 antiserum. Lanes 1 and 2, crude protein extracts from bacteria (carrying the CD209-pET32a plasmid) before (lane 1) and after (lane 2) IPTG induction. Lane 3, purified rCD209 protein. Lane 4, rCD209 recognized by the anti-CD209 antiserum. E Spatial and temporal expression profiles of CD209 analyzed by RT-qPCR. Total RNAs were extracted from sixth-instar larvae at 24 h PE (feeding) and 72 h PE (wandering). F Tissue-specific expression of CD209 detected using western blotting. Antiserum against CD209 or β-actin was used for detection. Protein samples were collected from the wandering stage. Pl, plasma. Hc, hemocytes. Fb, fat bodies. Mg, midgut. In, integument

Full size image

Among PRR, AMP, and Lys genes, CD209 and Lys1 exhibited their respective maximal increase (Additional file 1: Table S3). CD209 contains a signal peptide and a carbohydrate recognition domain (CRD) (Fig. 2C), representing a single-CRD CTL. To analyze its function, we expressed and purified recombinant CD209 (rCD209, Fig. 2D; Additional file 3). The rCD209 has an expected molecular weight of ~ 39 kDa, including ~ 18 kDa N-terminal recombinant thioredoxin (rTrx, His-tagged). The antiserum against rCD209 could recognize the recombinant protein specifically (Fig. 2D; Additional file 3). RT-qPCR analysis indicated that CD209 was dominantly expressed in the midgut of the wandering stage (Fig. 2E); this result was validated by western blotting (Fig. 2F; Additional file 3). Considering that the anti-CD209 antiserum could recognize rTrx, the lower band detected in the fat body, midgut, and integument might be endogenous Trx protein (Fig. 2F; Additional file 3).

CD209 and Lys1 expression in the midgut during the wandering stage is elevated by 20E

In insect hemolymph, 20E titers are significantly higher at the wandering stage than that at feeding stage of the last instar larvae [26]. Because the transcript abundance of CD209 and Lys1 increased during the wandering stage (Fig. 2A, B), we hypothesized that endogenous 20E might be involved. To test the hypothesis, we injected 20E into the hemocoel of sixth-instar larvae at 48 h PE, when endogenous 20E titers are low [26]. RT-qPCR analysis revealed that at 24 h post-20E injection, CD209 and Lys1 expression was significantly elevated in the larval midgut (Additional file 2: Fig. S2A, B). Western blot analysis indicated that CD209 expression in the midgut increased significantly at 24 h post-20E injection (Additional file 2: Fig. S2C, D; Additional file 3). Thus, endogenous 20E contributes to increased levels of CD209 and Lys1 during the wandering stage.

CD209 depletion results in a significant increase of Lactobacillus load in the midgut and hemolymph during the wandering stage

Some studies have indicated that CTLs have antibacterial activities and play crucial roles in maintaining gut microbial homeostasis [8, we elucidated the essential role of a CD209 (CTL) − Lys1 (Lys) immune function in maintaining gut homeostasis, thereby minimizing the risk of opportunistic infections during complete metamorphosis. We showed that opportunistic infections caused by gut Lactobacillus result in insufficient energy and 20E, thus delaying insect metamorphosis. Moreover, 20E elevates the expression of CD209 and Lys1, which in turn maintain high 20E titers by suppressing Lactobacillus production in the midgut and hemolymph during the wandering stage. Our results reveal a positive feedback loop of endocrine–immune interactions mediated by Lactobacillus, highlighting the significance of gut homeostasis and local–systemic communication during complete metamorphosis.

Fig. 8
figure 8

Model showing CTL − Lys immune function regulates gut homeostasis and controls insect metamorphosis. CD209 and its downstream Lys1 maintain high expression levels in the midgut during the wandering stage due to the high 20E titers. Lys1 suppresses the proliferation of Lactobacillus in the midgut, resulting in a low abundance of Lactobacillus (and the associated PGN) in the hemocoel. In this condition, fat body cells are mainly involved in the storage of lipids and cholesterol esters in LDs, thus providing sufficient energy and 20E titers for timely pupariation. However, low expression levels of CD209 and Lys1 increase the Lactobacillus load in the midgut. The increased loads of Lactobacillus (and PGN) subsequently translocate into the hemocoel and trigger a shift from lipid storage to phospholipid synthesis. Moreover, cholesterol esters stored in LDs decrease. This leads to insufficient energy and 20E titers, thereby delaying pupariation. The low 20E titers, in turn, fail to induce high levels of CD209 and Lys1 expression in the midgut

Full size image

Endogenous 20E-mediated increase in immunity contributes to effective gut sterilization, thereby minimizing the risk of opportunistic infections during complete metamorphosis

Consistent with reports on species, such as mosquitoes [43], our results indicated that although the diversity of gut bacteria increased, their load largely decreased when larvae entered the wandering stage. Lactobacillus accounted for the gut bacteria with the largest decrease. By contrast, the total bacterial load in the hemolymph increased significantly during metamorphosis, possibly from the metamorphosing midgut, as did Lactobacillus. Because the peritrophic membrane, which lines the midgut lumen and prevents gut microbial diffusion, is purged at the onset of metamorphosis [8, 44], we believe that the fewer gut bacteria there are, the lower the risk of opportunistic infections. The transcript abundance of most of the PRRs and antibacterial effectors increased in the metamorphosing midgut, as reported in Bombyx mori [45]. Endogenous 20E titers contributed to the high level of CD209 and Lys1, and possibly other immunity-related genes, because they were upregulated in the metamorphosing midgut. Considering that loss of immunity-related genes, such as antibacterial effectors, leads to an increased load of gut microbiota [9], 20E-mediated increase in immunity might be an effective gut sterilization mechanism during complete metamorphosis.

CD209 and its downstream Lys1 restrict Lactobacillus during metamorphosis

Lys accumulates in the midgut epithelium and is released into the gut lumen after the larval epithelium has been sloughed off during metamorphosis [46]. In Drosophila, Lys loss results in an increased load of gut microbiota [9]. A similar result was observed in this study, with Lys1 depletion leading to an increased load of gut Lactobacillus. We further showed that Lys1 directly targets L. plantarum to suppress its proliferation, which can be explained by the fact that Lys can cleave PGN on the cell wall of gram-positive bacteria [47]. However, we cannot rule out the possibility that Lys1 contributes to gut immunity by modulating immune signaling, because Lys also cleaves PGN shed from gut microbiota [48]. Similar with the function of Lys1 in restricting gut Lactobacillus during metamorphosis, as a CTL, CD209 was suggested to act upstream of Lys1. Although many CTLs are known to regulate the expression of antibacterial effectors [15, 49, 50], the detailed regulatory mechanisms are unclear. Since injection of dsRNA results in systemic rather than gut specific knockdown of the target gene, the effect of downregulation of CD209 or Lys1 in other tissues cannot be ruled out. However, considering the predominant expression of CD209 and Lys1 in the metamorphosing midgut, our explanation is reasonable.

Increased load of Lactobacillus and its associated PGN translocate from the midgut to hemocoel, delaying metamorphosis in CD209- or Lys1-depleted larvae

L. plantarum, as a commensal bacterium in the gut, exerts its benefits by promoting Drosophila systemic growth [3]. Similar to E. faecalis in M. sexta [14], however, L. plantarum in H. armigera completes its commensal-to-pathogen switch after translocation from the midgut to hemocoel. Opportunistic infections caused by gut L. plantarum did not induce sepsis resulting in H. armigera death, but elicited delayed metamorphosis of the final larval instar. Injection of gut E. mundtii into the hemocoel induced a similar phenotype in H. armigera [15], suggesting that it is not bacteria-specific and a common mechanism may be involved. A “malaise syndrome” associated with delayed metamorphosis was characterized in M. sexta and could be elicited by PGN [41]. Given that PGN is present on the cell wall of bacteria and is shed during bacterial growth and maturation [42], we suggest that the delayed metamorphosis caused by opportunistic infections is triggered by the PGN derived from gut bacteria that invade the hemocoel. Lysine-type PGNs are derived from gram-positive bacteria [42], such as L. plantarum and E. mundtii. Delayed pupariation was observed in both H. armigera and D. melanogaster after Lysine-type PGN challenge, supporting this hypothesis. Gut microbiota-shed PGN could translocate from the gut lumen into systemic circulation [51]. Therefore, it was suggested that the Lysine-type PGN derived from both gut and hemolymph Lactobacillus contributed to delayed metamorphosis in CD209- or Lys1-depleted larvae.

High abundance of Lactobacillus and its associated Lysine-type PGN in hemocoel decreases lipids and cholesterol esters stored in LDs, resulting in insufficient energy and 20E available for metamorphosis

Fighting infection is an energy-consuming process [16], as is committing metamorphosis [52], so there may be a competition for energy between the two. Metabolism, such as lipid metabolism, needs to fulfill the energy requirements of an organism [53]. Opportunistic infections caused by gut L. plantarum and its associated Lysine-type PGN perturb lipid homeostasis, leading to a reduction in fat body lipids, which are the primary source of energy during metamorphosis [52]. This may be one of the causes why opportunistic infections delay metamorphosis. Shifting lipid metabolism from TAG storage to membrane phospholipid synthesis also occurs in Drosophila upon activation of toll signaling. This shift favors endoplasmic reticulum expansion and sustains the immediate demand for AMP secretion during infection [54]. Because fat body cells are major sites for energy metabolism and innate immunity [55], their coordination during metamorphosis is of great interest.

Ecdysone, produced when needed from cholesterol esters stored in LDs [18, 37], is essential for triggering insect metamorphosis [19]. Interestingly, we found that the level of cholesterol esters decreased with the decrease of LD size in L. plantarum-injected larvae. The level of ecdysone correspondingly decreased. Therefore, insufficient cholesterol mobilization and ecdysone production may be another cause of delayed metamorphosis due to opportunistic infections. This was confirmed by the administration of 20E, which partially rescued pupariation in L. plantarum-injected or fed larvae. The expression of Halloween genes involved in ecdysteroid biosynthesis are suppressed in E. mundtii-injected larvae [15], suggesting that endocrinological changes caused by opportunistic infections are multilayered. Given that 20E suppresses TAG storage [56], low 20E levels may facilitate lipid storage to compensate for energy expenditure during the fight against infections.

Lysine-type PGN induced storage defects in lipids and cholesterol esters, suggesting that delayed metamorphosis due to energy and ecdysone deficiency is applicable to opportunistic infections caused by gram-positive bacteria. Similar phenomena, such as delayed pupariation and decreased storage of lipids and cholesterol esters, have been observed in Drosophila larvae challenged by Lysine-type PGN, suggesting that this is a common mechanism.

Conclusions

Opportunistic infections caused by gut symbiotic bacteria occur frequently [57, 58], affecting mammal or insect development through mechanisms that are obscure. Here, we have shown that opportunistic infections caused by gut Lactobacillus lead to a reduction in storage of lipids and cholesterols, thereby reducing the availability of energy and ecdysone for insect pupariation. CD209 and its downstream Lys1, induced by ecdysone, contribute to gut homeostasis, thereby minimizing the risk of opportunistic infections (Fig. 8). Therefore, the endocrine system controls the risk of opportunistic infections by recruiting gut immunity, which in turn ensures normal endocrine function and timely metamorphosis.

Methods

Insect rearing, antibiotic treatment, and PGN challenge

H. armigera larvae were maintained in an insectarium under the photoperiod of 14 h light/10 h dark at 27 ± 1°C. The larvae were fed on an artificial diet comprising soybean powder, wheat germ, compound vitamins, and mineral salt. Axenic H. armigera larvae were obtained by constantly feeding on a diet supplemented with gentamicin (15 μg/mL), penicillin (10 units/mL), and streptomycin (10 μg/mL) from the third instar. The midguts were dissected from sixth-instar larvae at 0 h PE, reared on a diet with or without antibiotics. The efficacy of cultivable bacterial elimination was confirmed by plating gut homogenates onto LB agar plates. PCR analysis against gut homogenates was conducted using universal 16S rRNA gene primers 16S-F and 16S-R (Additional file 1: Table S8). For PGN challenge, sixth-instar larvae at 48 h PE were injected with 5 μL of Lysine-type PGN (500 ng/μL, KALANG, Shanghai, China).

D. melanogaster were raised on a standard cornmeal/yeast diet under the photoperiod of 8 h light/16 h dark at 25 ± 1°C. Drosophila late-third instar larvae (~ 108 h after egg lay) were collected and rinsed in sterile PBS. The larvae were punctured laterally with a needle dipped in Lysine-type PGN or PBS. The punctures were performed in mineral oil, where the punctured larvae were kept for another 10 min before transfer to fresh food, as described [54].

Sequencing and analysis of bacterial 16S rRNA gene from the midgut

Midguts were collected from sixth-instar larvae at 24 and 72 h PE. Before the midguts were dissected, larvae were surface-sterilized with 75% ethanol. Five independent biological replicates were included. A total of 20 midguts from each stage were pooled as a replicate and used for metagenomic DNA extraction. The V3 and V4 hypervariable regions of the 16S rRNA gene were amplified with primers 338F and 806R (Additional file 1: Table S8). The PCR amplicon pools were prepared for sequencing on an Illumina MiSeq platform (San Diego, USA) by Majorbio Bio-Pharm Technology Co., Ltd. (Shanghai, China). The raw sequencing reads were demultiplexed, quality-filtered by Trimmomatic, and merged using FLASH. OTUs with 97% similarity cut-off were clustered using UPARSE (http://drive5.com/uparse/). The taxonomy of each OTU representative sequence was analyzed by RDP Classifier (http://rdp.cme.msu.edu/) against the SILVA database (http://www.arb-silva.de) using a confidence threshold of 0.7.

Isolation, characterization, inoculation, and colonization of L. plantarum

To isolate bacteria from the midgut, sixth-instar larvae at 24 h PE were surface-sterilized in 75% ethanol and rinsed in sterile PBS. The midguts were dissected, homogenized, and diluted (1:1000) in sterile PBS. The diluted homogenates were spread on MRS (Solarbio, Bei**g, China) agar plates. The plates were incubated at 37°C for 24 − 48 h. Morphologically distinct colonies were selected for genomic DNA extraction using a bacterial DNA kit (Omega, GA, USA). DNA fragments were amplified using universal 16S rRNA gene primers 27F and 1492R (Additional file 1: Table S8), followed by sequencing. The closest L. plantarum matching the 16S rRNA gene sequence (99% identity) was obtained after BLAST search (https://blast.ncbi.nlm.nih.gov/Blast.cgi) against the 16S rRNA database.

L. plantarum in mid-logarithmic phase was harvested, washed twice, and resuspended in PBS at 2 × 108 cells/mL. Each sixth-instar larva at 48 h PE was injected with 5 μL of bacterial suspension. L. plantarum was introduced into sixth-instar larvae at 0 h PE by feeding on artificial diet, supplemented with the bacterial suspension. An equal amount of sterile PBS was used as a control. Midguts, hemolymph, and fat bodies from the bacteria-injected or -fed larvae were collected for further assays.

RNAi

To produce dsCD209 and dsLys1, the fragments of these two genes were amplified using the primers listed in Table S8 (Additional file 1). The dsCD209 and dsLys1 were synthesized with their respective PCR products as templates using a MEGAscript kit (Ambion, Austin, TX, USA). The dsGFP was synthesized as a control. The dsRNAs were eluted from the purification column with nuclease-free water and concentrated to 2.0 μg/μL. Each fourth-instar larva at 6 − 12 h PE was injected with 5 μL of dsCD209, dsLys1, or dsGFP. The midguts, hemolymph, and fat bodies were collected from the larvae after dsRNA injection for analysis.

RNA-seq analysis

Total RNA was extracted from the midguts of sixth-instar larvae at 24 h PE (feeding; MgF) and 72 h PE (wandering; MgW), and those at 72 h PE pretreated with dsCD209 or dsGFP using TRIzol (Invitrogen, Carlsbad, CA, USA). The cDNA libraries (MgF-1, -2, -3, -4; MgW-1, -2, -3, -4; dsCD209-1, -2, -3; dsGFP-1, -2, and -3) were constructed as described [59]. Briefly, mRNAs were isolated using oligo(dT) beads and then fragmented. Double-stranded cDNAs were synthesized using EasyScript cDNA synthesis SuperMix (TransGen Bio-tech, Bei**g, China), followed by sequencing on the Illumina Novaseq 6000 platform (San Diego, USA) at Majorbio Bio-Pharm Technology Co., Ltd. Clean reads were obtained after filtration of raw reads and aligned to the H. armigera genome (https://www.ncbi.nlm.nih.gov/genome/13316?genome_assembly_id=319039) using TopHat software (http://tophat.cbcb.umd.edu/). Mapped reads were assembled and all unigenes were run in BLASTx against the non-redundant database. DEGs in the libraries (MgW vs. MgF; dsCD209 vs. dsGFP) were determined using a fold cut-off change value ≥ 2 and adjusted p-value < 0.05.

Measurement and injection of 20E

We determined 20E titers using the double antibody sandwich method with an insect 20E ELISA kit (Kmaels Biotechnology, Shanghai, China), as described [60]. Hemolymph was collected from sixth-instar larvae at 72 and 96 h PE treated with dsRNA or L. plantarum. Hemolymph from five larvae was pooled as a replicate, and three replicates in each treatment were obtained for 20E measurement. For Drosophila, larvae were collected at 3 h post-PGN challenge, homogenized, and centrifuged. The supernatants were used for assays. The samples were added to the microwells precoated with 20E antibodies. Subsequently, horseradish peroxidase-labeled detection antibodies were added to the wells. After thorough washing was complete, 3,3′,5,5′-tetramethylbenzidine was used for color development. Absorbance was measured at 450 nm, and 20E titers were calculated from the standard curve.

To determine gene expression, each sixth-instar larva at 48 h PE was injected with 20E (ABMole, Houston, TX, USA) at 200 ng/5 μL, which was close to the amount of endogenous 20E during the wandering stage [26]. An equal amount of dimethyl sulfoxide (DMSO), used for dissolving 20E, was injected as a solvent control. Three biological replicates were included, with each replicate comprising four larvae. Midguts were collected for total RNA or protein extraction at 12 and 24 h post-injection.

Quantification of total bacteria and Lactobacillus in the midgut and hemolymph

Genomic DNA from the midguts and hemolymph of sixth-instar larvae at 24 or 72 h PE was extracted using a genomic DNA isolation kit (Omega). Genomic DNA was extracted from the midguts and hemolymph of sixth-instar larvae at 72 h PE after treatment with dsCD209, dsLys1, or dsGFP. Moreover, genomic DNA was isolated from the midguts and hemolymph of sixth-instar larvae at 24 and 72 h PE fed with a diet containing L. plantarum or PBS. The qPCR analysis was performed on genomic DNA using universal 16S rRNA gene primers 16S-F and 16S-R or Lactobacillus-specific 16S rRNA primers (Additional file 1: Table S8). The qPCR was performed using TransStart Tip Green qPCR SuperMix (TransGen Bio-tech). H. armigera β-actin gene was used as an endogenous control.

Recombinant protein generation, bacterial-binding assay, and antiserum preparation

The DNA fragment encoding the mature peptide of CD209 or Lys1 was cloned into pET32a plasmids, which were transformed into Escherichia coli BL21 (DE3) competent cells. After addition of 0.3 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) at 16°C overnight, rCD209 was induced in a soluble form, whereas rLys1 was present in inclusion bodies. The inclusion bodies were denatured and renatured, as described [61]. The recombinant proteins were purified using High-Affinity Ni–NTA Resin (Novagen, WI, USA), dialyzed against PBS, and concentrated to desired concentrations. The rTrx was prepared simultaneously and used as control.

L. plantarum (2 × 108 cells/mL) was incubated with rCD209 (1.0 mg/mL) in the presence or absence of 10 mM CaCl2 for 30 min. The bacterial cells were pelleted, washed, and eluted with 7% SDS. The rTrx was used as negative control. The samples were analyzed by western blotting.

After homogenization of rCD209 (200 μg) with complete Freund’s adjuvant, the mixture was hypodermically injected into the back of a rabbit. After 3 weeks, the same amount of recombinant protein, homogenized with incomplete Freund’s adjuvant, was subcutaneously injected. The antiserum against CD209 was prepared after a booster injection of 500 μg antigen 2 weeks later.

Western blotting

Total proteins isolated from various tissues, including the midgut treated with 20E or RNAi, were subjected to western blot analyses. Purified rCD209 or rTrx were loaded as controls. Equal aliquots of protein samples (100 μg/lane) were resolved using 12.5% SDS-PAGE and transferred to nitrocellulose membranes (Millipore, Darmstadt, Germany). The membranes were blocked with 5% skim milk and incubated with the antiserum against rCD209 at a 1:500 dilution. An antibody against β-actin (ABclonal, Cat. AC026) was used as a loading control. Protein samples from bacterial washing and elution were subjected to western blot analyses using anti-His antibody (ABclonal, Cat. AE003). Dosage analyses were performed using ImageJ software (Bethesda, MD, USA), based on three biological replicates.

Analysis of the antibacterial activity of Lys1

To analyze the antibacterial activity of Lys1, rLys1 was incubated with enterokinase (Beyotime Bio-tech, Shanghai, China) to remove Trx (His-tagged). A total of 90 μL of Lys1 (25 μM) was mixed with 10 μL of bacterial suspension at 25°C for 1 h. The mixture was plated on MRS agar plates and cultured at 37°C for 24 h. The number of colonies on each plate was recorded. The same concentration of bovine serum albumin (BSA) was used as a control. Three replicates were used.

RT-qPCR

Total RNA was extracted from various tissues, including the midgut treated with 20E or RNAi, using TRIzol Reagent (Invitrogen). Total RNA was isolated from fat bodies of sixth-instar larvae at 72 and 96 h PE after either dsRNA injection or L. plantarum treatment. First-strand cDNA was synthesized using EasyScript cDNA synthesis SuperMix (TransGen Bio-tech). RT-qPCR analysis was performed using TransStart Tip Green qPCR SuperMix (TransGen Bio-tech). The housekee** gene β-actin was used as an endogenous control. Primers used for RT-qPCR analysis are listed in Table S8 (Additional file 1).

Untargeted metabolomics analysis

Hemolymph samples (100 μL) from sixth-instar larvae at 24 h post-L. plantarum or -PBS injection were mixed with 400 μL extraction solution (acetonitrile:methanol = 1:1) containing 0.02 mg/mL internal standard (L-2-chlorophenylalanine). After vortexing for 30 s and low-temperature ultrasonic extraction for 30 min was complete, the samples were kept at − 20°C for 30 min to precipitate the proteins and centrifuged for 15 min at 13,000 × g. The supernatants were collected and blow-dried under nitrogen. The samples were resolubilized with 100 μL solution (acetonitrile:water = 1:1) and centrifuged at 13,000 × g. The supernatants were obtained for LC–MS/MS analysis.

The metabolite extracts from hemolymph samples were analyzed using a UPLC-Q-Exactive mass spectrometer (Thermo Fisher Scientific, MA, USA). Chromatographic separation was conducted using an ACQUITY HSS T3 column (Waters, Milford, USA). The mobile phases consisted of 0.1% formic acid in water:acetonitrile (95:5; solvent A) and 0.1% formic acid in water:acetonitrile:isopropanol (5:47.5:47.5; solvent B). Raw data were processed with Progenesis QI software (Waters). Internal standard peaks and known false positive peaks were removed. The metabolites were searched against the HMDB database (http://www.hmdb.ca/). The R package “ropls” (version 1.6.2) was used to perform PLS-DA. Metabolites with VIP > 1 and p < 0.05 were considered significantly different metabolites. Differential metabolites between two groups were mapped into their biochemical pathways through metabolic enrichment based on the KEGG database (http://www.genome.jp/kegg/).

LD staining

Fat bodies were collected at 24 h post-injection of L. plantarum or PBS, incubated in Nile red solution (20% glycerol in PBS, with 1:500 dilution of 10% Nile red in DMSO) for 20 min, and imaged under a laser-scanning confocal microscope (Leica, Germany). LD diameter and fluorescence intensity were determined using ImageJ software. Each group consisted of 15 average LD diameters and fluorescence intensities, each of which was based on one image.

TAG and cholesterol assays

Fat bodies were collected at 24 h post-L. plantarum, -PGN (Lysine-type), or -PBS injection to determine TAG and cholesterol levels, as described [62]. Fat bodies were homogenized in isopropanol and centrifuged at 10,000 × g for 10 min. The supernatants were used for TAG and cholesterol quantification using a triglyceride colorimetric assay kit (Elabscience, Wuhan, China) and total cholesterol colorimetric assay kit (Elabscience), respectively. For Drosophila, larvae were collected at 3 h post-PGN challenge, homogenized, and centrifuged, following which the supernatants were used for assays.

Measurement of developmental timing

To determine the mean duration of the fourth-, fifth-, and sixth-instars, we recorded each individual every 1 h from the beginning of each treatment until pupariation. The duration measurement was based on 23 dsCD209- and 24 dsGFP-injected larvae, as well as 28 dsLys1- and 34 dsGFP-injected larvae. A total of 28 L. plantarum- or PBS-injected larvae, 36 L. plantarum-fed larvae, and 41 PBS-fed larvae were used for the assays. A total of 45 PGN (Lysine-type)- and 44 PBS-injected larvae were assayed. For Drosophila, 22 PGN (Lysine-type)- and 30 PBS-challenged larvae were analyzed. The mean duration of larval stages was determined.

To determine whether administration of 20E rescued delayed pupariation, sixth-instar larvae pre-treated with dsCD209 or dsLys1 were divided into two groups. Sixth-instar larvae injected or fed with L. plantarum were divided into two groups. One group was injected with 5 μL of 20E (500 ng/μL) per larva, whereas the other group was injected with an equivalent amount of DMSO as solvent control. Both injections were committed at 72 h PE of sixth-instar. In the dsCD209-injected group, 34 20E- and 29 DMSO-injected larvae were analyzed. In the dsLys1-injected group, 33 20E- and 30 DMSO-injected larvae were analyzed. In the L. plantarum-injected group, 32 20E- and 30 DMSO-injected larvae were used for calculating the mean of the sixth-instar. In the L. plantarum-fed group, 35 20E- and 34 DMSO-injected larvae were analyzed. Each analysis was conducted twice.

Statistical analyses

Statistical analysis was performed using GraphPad Prism 8 (GraphPad Inc., La Jolla, CA, USA). The Mann–Whitney U test was used to evaluate differences in bacterial composition between the two groups. Unless otherwise indicated, Student’s t test was used to determine differences between two groups. One-way analysis of variance (ANOVA) coupled with Tukey’s multiple comparison test was used to evaluate differences among multiple groups. A value of p < 0.05 was considered statistically significant.