Black soldier fly larvae were obtained from Huazhong Agricultural University, Wuhan, China. The Wuhan strain was established from eggs collected at a poultry facility in the Wuhan suburbs in November 2008 (Cai et al., 2018). Colonies were maintained on a standard diet (Wheat bran: wheat middlings =1:1; Sheppard et al., 2002).

We dissected BSFL gut samples on a UV-sterilized ultra-clean bench. The BSFL were removed from the artificial diet with autoclave-sterilized tweezers and rinsed with autoclave-sterilized water. Then, the BSFL was immersed in 75% ethanol for 2–3 min to surface sterilize them. After that, the BSFL surface was rinsed again with autoclaved sterile water. The BSFL gut was accessed using a sterilized dissection needle and placed in a sterilized 5-ml internal rotation cryopreservation tube. We froze the samples in liquid nitrogen for 15 min, and stored them at −80°C.

We collected about 5 g of fresh BSF eggs produced within 6 h and put them into autoclave-sterilized water. The eggs were dispersed with a soft brush, and autoclaved filter paper separated the eggs from water. Then, the eggs were transferred to a 15-mL autoclave-sterilized centrifuge tube, and added 10 ml Sporgon^® (Decon Labs, Inc. Bryn Mawr, PA, United States). The sterilized eggs were poured into a filter paper, followed by three washes with sterile water. The drained eggs were transferred to a sterile standard diet (sealed with 8 layers of sterile gauze), and placed in a 28°C incubator.

Guts from GF BSFL (after rearing for 15 days at 28°C) were dissected to verify their GF status. Gut homogenates were diluted 10⁻² fold with sterile water. The homogenates were then spread on the surface of Nutrient Agar-NA (Solarbio^®, N8290) or Potato Dextrose Agar-PDA (Solarbio^®, P8931). The plates were incubated upside down for 48 h at 28°C. The E.Z.N.A.^® soil DNA kit (Omega Bio-Tek, D5625) was used to extract the DNA from gut homogenates.

We collected 10-day-old BSFL from standard feed and then surface sterilized using 75% ethanol for 2–3 min, followed by three washes with sterile water. Then, we collected 5 BSFL guts for homogenization. Sterile water was added in gut homogenization to 5 ml, then vortexed them with shaking for further use. GB BSFL was prepared by culturing 15-day-old GF BSFL and 1 ml of gut homogenates of 5 natural 10-day-old BSFL (fed on standard diet) together in artificial diet.

We tested whether GF BSFL remained germ-free during conversion and GB BSFL inoculated with the gut microbes. At the end of conversion, DNA was extracted from GF BSFL gut samples and GB BSFL gut samples. PCR verification was performed on 16S and ITS1. The primers used are shown in Table 1.

The artificial diet formula used was taken from the Drosophila melanogaster artificial diet (Piper et al., 2014), with some modifications. See Supplementary Tables S1, S2 for the formula of the artificial diet used. For the sterilization method of the artificial diet, see Supplementary Table S3. The reagents used for the artificial diets are shown in Supplementary Table S4.

We used germ-free BSFL (GF) and gnotobiotic BSFL (GB) to digest a high-protein artificial feed model. The experimental groups tested are shown in Table 2. Each group was prepared in triplicate.

We examined the protein content of the artificial diets referencing the standards of the ministry of agriculture and rural affairs of the People’s Republic of China NY/T 3493–2019. The artificial diet was placed in an electric thermostatic drying oven at 60°C and oven dried to a constant weight. Then, the artificial diet protein reduction rate was determined as follows:

W1 and W2 are the dry weight protein amounts of BSFL artificial diet before and after rearing, respectively.

Three kits were used to detect protein digestive enzyme activities of the of the GF and GB group gut samples. A trypsin activity assay kit (Solarbio^®, BC2310) was used for trypsin and a pepsin activity assay kit (Solarbio^®, BC2320) was used for pepsin. A leucine aminopeptidase (LAP) activity assay kit (Solarbio^®, BC4140) was used for peptidases. Trypsin activity and pepsin activity were measured by UV spectrophotometry, and peptidase activity was measured by visible spectrophotometry. Enzyme activities were calculated according to sample protein concentrations. The detection methods used followed kit instructions.

We collected separate gut samples from the GF and GB groups. DNA was extracted from the samples using the E.Z.N.A.^® Stool Kit (Omega Bio-Tek, Norcross, GA, United States) following manufacturer instructions. The primers 806R (5`-GGACTACHVGGGTWTCTAAT-3`) and 338F (5`-ACTCCTACGGGAGGCAGCAG-3`) were used to amplify the bacterial V3–V4 region of 16S rDNA genes (Liu et al., 2016). The primers ITS1 (5`-CTTGGTCATTTAGAGGAAGTAA-3`) and ITS2 (5`-TGCGTTCTTCATCGATGC-3`) were used to amplify the fungus ITS1-ITS2 region of ITS1 genes (Wang et al., 2018). Purified PCR products were analyzed using the Illumina Miseq PE300 sequencing platform (Illumina, San Diego, CA, United States).

We processed the raw data using the standard procedures of Beijing Allwegene Technology Co., Ltd. The 16S rDNA and ITS gene sequences were processed and demultiplexed through QIIME2 (Version 1.8.0). Quality filtering and trimming, denoising, pair-end read merging, chimeric removal and taxonomy assignments were processed by Vsearch (Version 2.7.1). Analyses were performed using IBM SPSS 20.0 (IBM, Chicago, IL, United States).

Clean tags were used to cluster gene sequences (or reduce noise) to generate operational taxonomic units (OTUs). The clustering method could be selected as uparse, uclust, or ref. reference database, with the default being uparse. The noise reduction method used was the unoise3 method (Edgar, 2010, 2013; Rognes et al., 2016).

We performed three replications for all experiments. Data are presented as the mean ± standard error (SE). We analyzed the results of bio-physiological indicators using one-way analysis of variance (ANOVA) and Tukey’s post hoc comparisons. Student’s t-test was used to analyze the results of artificial diet protein reduction rate. The number of replicate samples for all analyses was three. Shannon rarefaction curves and other richness and diversity indices of bacterial community (Chao1, observed_species and PD_whole_tree) were estimated using the QIIME2 platform. SPSS 20.0 (IBM SPSS Inc., Chicago, IL, United States). The free online platform of Allwegene cloud V2.0 Platform¹ and a free online platform² were used to analyze the statistical differences. Spearman’s rank correlation analysis was used to determine the significant correlation among the relative abundance of dominant microbes. p < 0.05 was considered statistically significant.

We used amplicon (16S rDNA and ITS1) abundances top20 (Supplementary Table S13) to calculate the correlation coefficient among the gut microbes. Co-occurrence network analysis was conducted at the genus level for the top 20 abundances of all samples (16S and ITS).

We used 16S rDNA amplicon OTU abundance (Supplementary Table S14) and 16S rDNA amplicon KEGG enrichment abundance (Supplementary Table S15) to calculate the 16S rDNA amplicon correlation coefficient between GB BSFL gut microbes and KEGG enrichment abundance. For network analysis of the co-occurrence patterns between protein metabolism and 16S rDNA amplicon (genus level) during BSFL conversion to artificial diets.

We used mixed data (16S rDNA amplicon OTU abundance Supplementary Table S14 and ITS1 amplicon OTU abundance Supplementary Table S16) and GB BSFL gut samples’ proteolytic enzyme activity (Supplementary Table S5) to calculate correlation coefficient between GB BSFL gut microbes and protein digestive enzymes. For the network analysis among 16S rDNA amplicon (genus level) abundance, ITS1 amplicon (genus level) abundance, and protein digestive enzyme activity data.

We used the Spearman test method on the calculated results to filter out the p values greater than 0.05 and values of R less than 0.6 (R > 0.6, p < 0.05; de Vries et al., 2018).

The 16S rDNA amplicon was used to discriminate bacteria, and the ITS1 amplicon was used to discriminate fungi. Calculation of the Spearman correlation coefficient was performed on the Allwegene cloud V2.0 Platform (see Footnote 1).

A Venn diagram and flower diagram were plotted by the online platform.⁴ Cytoscape (v3.9.1) was used for network analysis data visual plotting. GraphPad Prism 9 was used for other data visualization plots.

The validation results of GF and GB BSFL are shown in Figures 1A–C. To evaluate protein degradation efficiency of the GF and GB groups, we calculated the protein reduction rate of dry artificial diet after conversion by GF and GB BSFL. Then, the protein reduction rate of the food dry matter was calculated. There was a significant difference in the protein reduction rate between GF and GB artificial food residues (Figure 1D). The protein reduction rate of the GB artificial feed residues was significantly higher than that of the GF artificial food residues. The protein reduction rate of GB increased by 73.44% compared to the GF group (Table 3).

The presence of gut microbiota effected BSFL gut protein digestive capacity. This effect was manifested by changes in the activities of gut protein digestive enzymes. To evaluate this, protein digestive enzyme activities of GF and GB gut samples were determined (Figure 2). At 12 days, the activities of protein digestive enzymes were significantly different between GF and GB. However, the protein digestive enzyme activities of GF and GB were not significant at 0 day and 24 days, and GF was slightly lower than GB. During the whole process, trypsin activity was the highest, followed by peptidase activity and pepsin activity was the lowest (Table 4). During the whole process, the protein digestion enzyme activities of GF and GB both showed an initially increasing and then a decreasing trend. These results suggest that the enhanced gut protein degradation ability of GB BSFL resulted from the enhanced activities of protein digestive enzymes associated with the presence of gut microbes. Trypsin and peptidase play a major role in protein degradation by BSFL.

In order to reveal the change pattern of gut microbiota in the artificial feed converted by BSFL and better explain the reason why BSFL efficiently degrades protein with the help of its gut microbiota. The microbiome (16S rDNA and ITS1) of gnotobiotic BSFL gut was detected at three time points (day 0, day 12, and day 24).

A total of 190 OTUs were generated from the ITS1 sequences and 133 OTUs were generated from the 16S rDNA sequences. Statistics of OTU numbers for individual samples are shown in Table 5.

We explored the interaction between the gut microbiota during the conversion process. The genus levels of the top 20 abundances of all samples were selected for correlation analysis (Supplementary Table S8). Correlation analysis data between16S rDNA bacteria and ITS1 fungi are presented in Supplementary Table S9.

Co-network analysis of 16S rDNA bacteria and ITS1 fungi is shown in Figure 4. Proteobacteria (bacteria) and Ascomycetes (fungi) are the most common microbial types in top 20 genera. Certainly, the presence of uncultured spp. (Proteobacteria) and unidentified spp. (Ascomycetes) in top 20 genera which may disturb the focus of the analysis. In addition to uncultured spp. (Proteobacteria) and unidentified spp. (Ascomycetes), the highest abundant Issatchenkia spp. and other gut microbes were mostly negatively correlated. But, fungal Issatchenkia spp. and bacterial Lactobacillus spp. and Orbus spp. had a positive correlation.

We used PICRUSt 2 (Douglas et al., 2020) to predict the 16S rDNA gene amplification function during BSFL conversion to artificial food. Functional pathways of the GB 16S rDNA gene (Figure 5) showed that carbohydrate metabolism, metabolism of cofactors and vitamins and amino acid metabolism were the top three in abundance of bacteria functional pathways. Adding the bacterial abundance of amino acid (proteinogenic amino acid) metabolism and metabolism of other amino acids (non-protein amino acids) exceeded that of the pathways of carbohydrate metabolism and cofactors and vitamins metabolism. These results show that amino acid and other amino acids metabolism are the most abundant of all functional pathways.

Prediction of 16S rDNA gene amplification functional KEGG pathways (Figure 5) showed that gut microbiota abundance enriched by digestive system pathways was much lower than that enriched by pathways associated with amino acid metabolism. This suggests that in the presence of the BSFL host, protein degradation is dominated by the host itself. This is unlike the situation with microorganisms during composting. The main function of BSFL gut bacteria is amino acid metabolism, followed by protein degradation. We used online annotation tools⁵ for analysis of published genomic data on the black soldier fly (Zhan et al., 2020; Generalovic et al., 2021). We plotted black soldier fly pathways of protein digestion and absorption-map04974 (Figure 6), which provided evidence supporting this inference.