Rhoptroceros cyatheae was obtained from the Guizhou Chishui Alsophila National Nature Reserve (28°25′12″N, 106°01′03″E) in May 2023. All larvae collected from the forest were fed in an artificial climate incubator (SPX-280, Ningbo Jiangnan Instrument Factory, China; 26 ± 1°C, 75% ± 10% relative humidity, and 16L:8D photoperiod). Newly hatched larvae were reared with A. spinulosa and G. metteniana until the fifth instar. These two plants were obtained from the Chishui Alsophila National Nature Reserve. After 24 h of starvation, the larvae were sterilized with 75% ethanol for 1 min and then washed thrice with sterile water. The complete intestinal tract was dissected in sterile phosphate-buffered saline under a stereoscope, and 15–20 intestinal tracts were mixed as a biological replication, with three replicates per treatment (As: feeding on A. spinulosa; Gm: feeding on G. metteniana). The larvae were identified as the fifth instar according to their molting times (Xu et al., 2021). A sterile 15-mm sample puncher was used to collect samples from fresh and intact A. spinulosa leaves (AsL) and G. metteniana leaves (GmL) into a sterile culture plate, and 75% ethanol was used to clean the surface. The samples were mixed and ground under liquid nitrogen for subsequent DNA extraction.

The total DNA of the samples was extracted as for the E.Z.N.A^® soil DNA Kit (Omega Bio-tek, Norcross, GA, USA) instructions. The primers 799F (5′-AACMGGATTAGATACCCKG-3′)/1193R (5′ACGTCATCCCCACCTTCC-3′) and ITS1F (5′-CTTGGTCATTTAGAGGAAGTAA-3′)/ITS2R (5′-GCTGCGTTCTTCATCGATGC-3′) were used for PCR. PCR cycling conditions were as follows: 5 min at 94°C for initialization, 30 cycles of denaturation for 30 s at 95°C, 30 s of annealing at 55°C, and 30 s of extension at 72°C, followed by a 10-min final elongation at 72°C. The PCR product was extracted by performing 2% agarose gel electrophoresis and purification with PCR Clean-Up Kit (YuHua, Shanghai, China) according to the manufacturer’s instructions and quantified using Qubit 4.0 (Thermo Fisher Scientific, USA). NEXTFLEX^® Rapid DNA-Seq Kit (YuHua, Shanghai, China) was used to construct a library of purified PCR products.

Fastp¹ software was used to perform quality control on the double-ended original sequence, and FLASH² software was used for splicing: (i) The reads were truncated at any site receiving an average quality score of <20 over a 50 bp sliding window, the truncated reads shorter than 50 bp were discarded, and the reads containing ambiguous characters were also discarded; (ii) Only overlapping sequences longer than 10 bp were assembled according to their overlapped regions. The maximum mismatch ratio of the overlap region was 0.2. Reads that could not be assembled were discarded; (iii) The samples were distinguished according to their barcodes and primers, and the sequence direction was adjusted, the exact barcode was matched; two nucleotides were mismatched in primer matching. Then, the optimized sequences were clustered into operational taxonomic units (OTUs) using UPARSE (version 7.1) with 97% sequence similarity. The OTU table was manually filtered, and chimeric, chloroplast, and mitochondrion sequences were removed. RDP Classifier (version 2.11) was used to compare each OTU representative sequence with those in the database Silva and Unite, and the confidence threshold was 70%.

Alpha and beta diversity indices were estimated at the OTU level using MOTHUR and UniFrac, respectively. The Kruskal–Wallis test was used to analyze the α-diversity index of bacteria and fungi in the samples. The Wilcoxon signed-rank test and T-test (false discovery rate correction) were used for pairwise comparison. The principal coordinate analysis was performed based on the Bray–Curtis distance, and the Adonis test was used to analyze the difference between the groups, with p < 0.05 considered statistically significant. The linear discriminant analysis (LDA) effect size³ was calculated to identify the significantly abundant taxa (phylum to genera) of bacteria among the different groups (LDA > 4, p < 0.05). To understand the relationships among the genera, Spearman’s correlation coefficients were used for network analyses (Networkx: version 1.11). Network topological properties were calculated using Gephi. PICRUSt2⁴ and FAPROTAX (1.2.1) were used to predict the function of the microbial community in different samples.

Illumina Miseq sequencing was performed to characterize the 16S rDNA and ITS2 regions of the leaves and R. cyatheae gut. A total of 831,798 bacterial reads (Supplementary Table S1) with an average length of 376 bp were obtained. Additionally, 1,074,630 fungal reads with an average length of 251 bp (Supplementary Table S2) were obtained. Rarefaction curves indicated that the sequencing volume and sample depth were saturated (Supplementary Figure S1). In addition, good coverage reflects the integrity of sequencing (Wang et al., 2023). In this study, the coverage of each sample was above 99%, indicating that most of the species in the sample were detected (Supplementary Table S3).

After high-quality filtering, the optimized sequence clustering analysis based on 97% sequence similarity assigned the high-quality bacterial sequences into 721 OTUs (544 intestinal bacterial OTUs and 603 endophytic bacterial OTUs in leaves, shared 426), belonging to 16 phyla, 24 classes, 75 orders, 132 families, 265 genera, and 417 species. Among them, 201 OTUs were observed in As, 75 were observed in Gm, and 268 were shared between groups. The analysis of OTUs in the intestine and host leaves showed that 144 OTUs were observed in As, 171 were observed in AsL, and 325 were shared between groups. Furthermore, 154 OTUs were observed in Gm, 195 were observed in GmL, and 189 were shared between groups (Figure 1A). The high-quality sequences of fungi were assigned into 1,435 OTUs (627 intestinal fungal OTUs and 1,404 endophytic fungal OTUs in leaves, shared 596), belonging to 3 phyla, 21 classes, 75 orders, 215 families, 435 genera, and 637 species. Among them, 147 OTUs were observed in As, 178 were observed in Gm, and 302 were shared between groups. The analysis of OTUs in the intestine and host leaves showed that 144 OTUs were observed in As, 171 were observed in AsL, and 325 were shared between groups. Furthermore, 154 OTUs were observed in Gm, 195 were observed in GmL, and 189 were shared between groups (Figure 1B).

Proteobacteria, Firmicutes, and Actinobacteria were dominant phyla in the As and Gm groups, accounting for 97.44 and 99.74% of the total sequences, respectively (Figure 2A). Proteobacteria was the dominant phylum in the AsL and GmL groups, accounting for 86.60 and 90.41% of the total sequences, respectively. At the genus level, the top five abundant genera in the gut of the two groups of R. cyatheae were different. The dominant genera in the gut of R. cyatheae feeding on A. spinulosa were Allorhizobium-Neorhizobium-Pararhizobium-Rhizobium (31.07%) and Methylobacterium-Methylorubrum (15.55%), which were also found in the host plants and accounted for a high proportion (25.83 and 11.96%, respectively). Enterococcus (59.89%) and unclassified_o_Enterobacterales (23.62%) were the dominant genera in the gut of R. cyatheae feeding on G. metteniana but were absent in GmL (Figure 2B). Collectively, the results showed that Proteobacteria, Firmicutes, and Actinobacteria were the most abundant phyla in the two hosts, whereas the dominant genera changed according to the host. At the phylum and genus levels, the similarity between the intestinal bacteria of R. cyatheae feeding on A. spinulosa and its host plant bacteria was higher than that observed in the Gm group. At the phylum level, the dominant intestinal fungus of the two feeding groups of R. cyatheae was Ascomycota (As: 98.98%; Gm: 96.58%). The dominant fungus in the AsL and GmL groups was also Ascomycota, accounting for 78.16 and 91.74% of the total sequences, respectively (Figure 2C). At the genus level, the dominant genus of intestinal fungi in the two feeding groups of R. cyatheae was Candida (As: 93.49%; Gm: 89.28%). However, Candida was not detected in the AsL and GmL groups (Figure 2D). In terms of fungal genera, the dominant genus in R. cyatheae was not affected by the microbiome of these two host leaves; nonetheless, the fungus exhibited its formation mechanism, which may play a special role in the intestine.

Alpha diversity was used to assess the richness and diversity between groups. The intestinal bacteria of R. cyatheae feeding on different plants showed significant differences in Chao (Figure 3A) and Shannon indices (Figure 3B). Figure 3 shows that bacterial diversity in the intestine of R. cyatheae feeding on G. metteniana decreased significantly. For fungi, no significant difference was observed in bacterial diversity and richness in the intestine of R. cyatheae fed on different plants (Figures 3C,D). It is worth noting that when the Wilcoxon signed-rank test was used for pairwise comparison, we found that at the bacterial level, the Chao and Shannon indices of the gut microbiome of R. cyatheae were close to those of the leaves of the plants they feed on. However, significant differences were observed for fungal communities. The overall difference is explained in the principal coordinate analysis (PCoA) diagram (bacteria: Adonis, R² = 0.897, p = 0.001; fungi: Adonis, R² = 0.938, p = 0.001). The intestinal bacteria of R. cyatheae feeding on different host plants were clustered into two different quadrants. The similarity among fungi was higher than that among bacteria (Figure 4). The similarity among bacteria between the intestinal communities of R. cyatheae and the leaf communities of host plants was higher than that among fungi, showing a closer distance in the PCoA diagram (Figures 4A,C). To further elucidate the effects of the two host plants on the microbiota of R. cyatheae larvae, differential abundances of dominant genera were compared. Notably, Enterococcus, Allorhizobium–Neorhizobium–Pararhizobium–Rhizobium, unclassified_o__Enterobacterales, and Glutamicibacter showed differential abundances between A. spinulosa- and G. metteniana-feeding R. cyatheae. Allorhizobium–Neorhizobium–Pararhizobium–Rhizobium showed no differential abundance between R. cyatheae and host plants. These results indicate that host plants affected the intestinal microbiota of R. cyatheae larvae (Figures 4B,D).

Association networks were constructed to determine the patterns of gut bacterial communities of R. cyatheae fed on A. spinulosa and G. metteniana (Figure 5). The network diagram of intestinal bacteria of the insect fed on A. spinulosa included 27 nodes and 172 edges (88 positive and 84 negative correlations) and that of intestinal bacteria of the insect fed on G. metteniana included 29 nodes and 223 edges (197 positive and 26 negative correlations). There were 53 positive correlations and 65 negative correlations among the fungi of R. cyatheae fed on A. spinulosa, and 159 positive correlations and 49 negative correlations among the fungi of R. cyatheae fed on G. metteniana, indicating that the complexity and pattern of the intestinal communities network structure in R. cyatheae fed on G. metteniana were higher than those in R. cyatheae fed on A. spinulosa. We found that Allorhizobium-Neorhizobium-Pararhizobium-Rhizobium, which showed the highest abundance in the As group, was not closely related to other bacteria in the network diagram (5 degrees), and the second dominant genus Methylobacterium-Methylorubrum (16 degrees) was the most closely related one. The highest abundance of Enterococcus was 2 degrees, and there were 22 genera with more than 16 degrees of abundance in the Gm group. Although no difference was observed in the abundance of the dominant genus Candida between the two groups, the degree of association with other genera in the network diagram was not the same (As: 11 degrees; Gm: 18 degrees). The above results indicated that the dominant intestinal bacteria of R. cyatheae fed on A. spinulosa were less related to other genera in life activities than those in R. cyatheae fed on G. metteniana. Overall, the results indicated that there were more cooperation and exchange events among most bacterial genera during the adaptation of R. cyatheae larvae to different hosts.

The PICRUSt2 results showed that the functional prediction categories of the gut bacteria of R. cyatheae feeding on different hosts were focused on metabolism, environmental information processing, genetic information processing, cell transformation, human diseases, and organismal systems. Additionally, the relative abundance of the metabolic pathway was the highest (Figure 6A). In addition, significant differences were observed in the relative abundance of the secondary classification level of metabolic pathways in each treatment (Figure 6B). Specifically, the relative abundance of carbohydrate metabolism (p < 0.05), amino acid metabolism (p < 0.05), and membrane transport (p < 0.05) in the gut bacteria of R. cyatheae feeding on G. metteniana was significantly higher than that of those feeding on A. spinulosa. The relative abundance of cofactor and vitamin metabolism (p < 0.05), xenobiotics symbiosis and metabolism (p < 0.05), and cell growth and death (p < 0.05) was dominant in the gut bacteria of R. cyatheae feeding on A. spinulosa.

The levels of adenosine triphosphatase, DNA-directed RNA polymerase, and DNA-directed DNA polymerase were high in both groups (Figure 7). No significant difference was observed in the function of the intestinal fungi of R. cyatheae feeding on the two host plants.