Oligosaccharides Reduce the Survival of Apis cerana and Disrupt the Gut Symbiont Gilliamella

TL;DR

Oligosaccharides in Camellia flowers harm the health of Apis cerana bees and disrupt their gut microbes, explaining their reluctance to pollinate these plants.

Contribution

This study reveals how oligosaccharides affect bee survival and gut microbiota, providing a molecular explanation for pollination reluctance.

Findings

Oligosaccharides significantly reduced the survival rate and sucrose consumption of Apis cerana.

Gut microbial communities, especially Gilliamella, were disrupted by oligosaccharide treatments.

Metabolomic changes suggest oligosaccharides alter key metabolic pathways in bees.

Abstract

Pollinating insects play a crucial role in maintaining global biodiversity and enhancing crop yield and quality. As a native bee species in China, studying the health of the Apis cerana is of significant value. The study elucidates the detrimental effects of oligosaccharides (stachyose and raffinose) on the physiology, gut integrity, and microbial homeostasis of Apis cerana, offering a mechanistic explanation for the species’ reluctance to pollinate Camellia reticulata. These results advance the understanding of host-diet-microbiota (especially Gilliamella) interactions in pollinators and underscore the ecological risks associated with specific floral metabolites. The research provides a theoretical foundation for developing oligosaccharide-adapted bee diets and for optimizing pollination management strategies for C. reticulata and related crops, ultimately promoting sustainable apiculture and crop production.

Honeybees are vital pollinators that contribute substantially to global ecosystem stability and agricultural productivity. Camellia reticulata, a cross-pollinated crop species, depends on honeybees for successful reproduction. Apis cerana shows reluctance to pollinate C. reticulata, yet the molecular mechanisms underlying this phenomenon remain unexplored. In this study, we performed controlled feeding experiments in which adult worker A. cerana were supplied with stachyose, raffinose, and their combination. We assessed physiological traits including survival rate, sucrose solution consumption, and body weight gain, alongside histological changes in intestinal cell structures. We conducted RNA-seq of gut tissues as well as 16S rRNA sequencing and metabolomic profiling. Our findings revealed that the mixed oligosaccharide treatment significantly reduced the survival rate of workers, and three oligosaccharide treatments significantly reduced sucrose consumption in A. cerana. Both mixed and single-oligosaccharide treatments caused pronounced intestinal cell damage and disrupted the gut microbial community structure. Among the gut microbes, Gilliamella exhibited the most substantial decline in the stachyose group. Metabolomic analysis further demonstrated that oligosaccharide feeding significantly altered amino acid and galactose metabolism pathways, which may play critical roles in oligosaccharide utilization and directly influence honeybee survival. In summary, this study provides new insights into the molecular mechanisms underlying A. cerana mortality associated with C. reticulata pollination. These findings not only enhance our understanding of host-diet-microbiota interactions in honey bees but also offer a theoretical basis for the integrated management of A. cerana for C. reticulata pollination and the development of oligosaccharide-adapted bee diets.

Figures

Full text

1. Introduction

As one of the most important pollinators, honeybees play a critical role in maintaining global ecosystem diversity and crop yield [1,2]. Camellia is the largest genus within the family Theaceae, comprising 120 known species, of which more than 50 species possess high seed oil content [3,4]. Both Camellia oleifera and Camellia reticulata are two important species belonging to the genus Camellia. Camellia oleifera is an oil-bearing woody plant endemic to China, with a cultivation and utilization history of more than 2300 years [5]. In recent decades, biological studies on C. oleifera have mainly focused on breeding and the development of new cultivars, whereas research on its pollination biology remains limited. Understanding the pollination process of C. oleifera is fundamental not only for improving fruit set and yield but also for promoting the development and utilization of its by-products.

C. oleifera is a highly self-incompatible and obligate outcrossing oilseed crop, thus heavily relying on pollinating insects to enhance reproductive efficiency and increase fruit yield [6]. Among these pollinators, Andrena camellia and Colletes gigas are recognized as the most efficient and frequent floral visitors, making them the key pollinators of C. oleifera [7,8]. The western honeybee (Apis mellifera) and the Eastern honeybee (Apis cerana) are the two most widely managed bee species for commercial use. However, A. mellifera, when used for pollination of C. oleifera, frequently suffers from larval decay and abdominal swelling in adult workers, often leading to death [9]. This phenomenon is mainly attributed to the inability of adult workers to digest specific oligosaccharides in C. oleifera (including stachyose, raffinose, and manninotriose). These oligosaccharides gradually accumulate in the gut over time, severely disrupting the glucose–trehalose biosynthetic pathway in worker bees, and ultimately causing mortality [9]. Furthermore, galactose, a metabolic product of oligosaccharides in camellia nectar and pollen, also exhibits strong toxicity to honeybees [10], thereby limiting the widespread application of managed honeybee colonies in camellia pollination.

Similarly, A. cerana may also encounter such detrimental effects, yet research on its pollination of Camellia reticulata remains scarce. In our previous laboratory work, we analyzed the sugar composition of nectar from C. reticulata and found that it also contains the same oligosaccharides, including stachyose and raffinose (Table S1). We aimed to investigate whether stachyose and raffinose could similarly induce digestive dysfunction or mortality in A. cerana; therefore, we conducted the present study. Several key questions remain unresolved: Do oligosaccharides in camellia nectar significantly impair the survival and health of A. cerana? Do they damage the intestinal tissue of A. cerana? Do oligosaccharides disrupt the gut microbiota community and alter microbial abundance? Which gut microbial taxon is most significantly affected by oligosaccharides? These issues warrant further investigation.

To address these questions, we simulated the natural concentrations of stachyose and raffinose in camellia nectar and fed them to adult A. cerana workers, aiming to elucidate the molecular mechanisms underlying bee abdominal distension and mortality following camellia pollination. Specifically, we investigated:

• (a)The effects of oligosaccharides on honeybee survival curves, sucrose consumption, body weight gain, and intestinal tissue morphology;
• (b)The molecular mechanisms underlying intestinal cell damage induced by oligosaccharides through intestinal transcriptome (RNA-seq) analysis;
• (c)The effects of oligosaccharides on gut microbial community composition and microbial metabolites by using 16S rRNA sequencing and gut microbial metabolomics.

2. Materials and Methods

2.1. Honeybee Sample Preparation

One-day-old worker bees (Apis cerana) were collected from a colony (5 frames) at the Eastern Bee Research Institute of Yunnan Agricultural University and randomly divided into experimental and control groups. Each group had eight replicates with 15 bees per replicate, totaling 120 bees per group. The experimental groups were fed solutions containing stachyose, raffinose, or their mixture for 10 days. The control group was fed sterile sucrose solution (50%, wt/wt) for the same duration.

Bees were fed with three different sugar solutions (stachyose, raffinose, and a mixture of stachyose and raffinose) and one control solution (50% sucrose without oligosaccharides), resulting in four groups: S (stachyose), M (raffinose), H (mixture), and Control. Each group consisted of eight replicates, with 15 bees per replicate, totaling 120 bees per group. Using the S group as an example, after 10 days of feeding, sampling was performed on day 11. During the feeding period, 21 bees died in the S group, leaving 99 bees. These 99 bees were used for survival rate and sugar consumption analyses. Subsequently, gut samples from these bees were collected for transcriptomics, metabolomics, 16S rRNA sequencing, and HE staining. Sampling details were as follows: Transcriptomics: three replicates per group, six gut samples pooled per replicate, resulting in 18 bees per group (not three bees). Metabolomics: three replicates per group, six gut samples pooled per replicate, resulting in 18 bees per group. 16S rRNA sequencing: three replicates per group, twelve gut samples pooled per replicate, resulting in 36 bees per group. HE staining: four replicates per group, one gut sample per replicate, resulting in four samples per group. The M, H, and Control groups had the same sampling scheme as the S group for transcriptomics, metabolomics, and 16S rRNA analyses. Body weight gain analysis was conducted using a second batch of bees, with three replicates per treatment and 20 bees per replicate. Due to mortality during rearing, the final sample sizes were: S group, 50 bees; M group, 48 bees; H group, 37 bees; and Control group, 57 bees.

The feeding experiments were conducted under controlled conditions (35 °C, 70% relative humidity, darkness) using an artificial climate chamber (Ningbo Haishu Saifu Laboratory Instrument Factory). Bees were anesthetized with CO_2_ prior to dissection. Midgut and hindgut tissues were aseptically collected, immediately frozen in liquid nitrogen, and stored at −80 °C until further processing. Animal handling complied with international ethical guidelines.

2.2. Oligosaccharide Preparation and Consumption Measurement

Commercial standards of stachyose and raffinose (Aladdin Biochemical Technology Co., Ltd. Shanghai, China) were dissolved in sterile sucrose syrup (50%, wt/wt) at concentrations of 36 mg/mL stachyose (3.6%), 16 mg/mL raffinose (1.6%), and a mixture of stachyose and raffinose (stachyose: 36 mg/mL, raffinose: 16 mg/mL). Daily solution consumption per bee was recorded and mortality was counted daily at 10:00 AM to plot survival curves. First, the normality of the sugar water consumption data from the four groups was assessed. After confirming that the data met the assumption of normality, a one-way analysis of variance (one-way ANOVA) was performed. For multiple comparisons, Dunnett’s multiple comparisons test was applied, yielding the following adjusted p values: S vs. Control, adjusted p = 0.0020; M vs. Control, adjusted p = 0.0426; and H vs. Control, adjusted p = 0.0021.

2.3. Survival Curves and Body Weight Measurements

The daily sugar solution consumption of worker bees was recorded for the three treatment groups and one control group. Consumption was determined as the difference between the total volume of sugar solution supplied on the previous day and the remaining volume on the current day. This value was then normalized by dividing by the number of worker bees to obtain the per-bee daily sugar solution consumption. Concurrently, the number of dead worker bees in each group was recorded every morning at a fixed time (10:00 AM) to construct survival curves. Survival analysis was performed using the Kaplan–Meier method, and differences in survival probability among groups were assessed by the log-rank test. Both sugar consumption measurements and survival observations were conducted daily over a 10-day period, until the feeding experiment was completed.

2.4. Gut Histological Hematoxylin and Eosin Staining (HE Staining)

Midgut tissues from honeybees were fixed overnight at room temperature with 10% neutral buffered formalin, dehydrated through graded ethanol, and embedded in paraffin. Sections (5 µm) were stained with hematoxylin and eosin (HE, Sigma-Aldrich, Saint Louis, MO, USA). The tissues were processed through the HE staining protocol [11]. Images were captured with an Olympus BX53 microscope (Olympus Corporation, Tokyo, Japan) and analyzed using ImageJ software (Version 1.46r) to quantify morphological changes such as villus height and epithelial integrity. Scale bar, 25 μm. Additionally, four view fields of HE sections from the four groups (S group, M group, H group, and control group) were selected as the figures. Twenty positions were randomly selected from each figure to determine the thickness of intestinal wall tissue, and the mean thickness of intestinal wall tissue was calculated for each figure. Subsequently, analysis of variance between groups was conducted using Dunnett’s multiple comparisons test.

2.5. Gut Transcriptomic Analyses

Total RNA was extracted from gut tissues using TRIzol^®^ reagent (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s instructions. RNA quality and quantity were assessed using a Bioanalyzer 5300 (Agilent Technologies, Hangzhou, China) and a NanoDrop ND-2000 spectrophotometer (NanoDrop Technologies, Suzhou, China). RNA purification, reverse transcription, library preparation, and sequencing were performed following the manufacturer’s protocols; RNA sequencing libraries were prepared using the Illumina^®^ Stranded mRNA Prep (Ligation) kit (Illumina, San Diego, CA, USA). Libraries were quantified by Qubit 4.0 fluorometer and sequenced on an Illumina NovaSeq X Plus platform (paired-end 150 bp).

Raw paired-end reads were trimmed and quality filtered using fastp [12] (v0.19.6) with default parameters. Clean reads were subsequently mapped to the Apis cerana reference genome (NCBI accession: GCF_029169275.1_AcerK_1.0) in a strand-specific manner using HISAT2 [13] (v2.2.1). Transcript assembly for each sample was then performed with StringTie [14] using a reference-guided strategy. Transcript abundance was quantified as transcripts per million (TPM). RSEM [15] was used to quantify gene abundances. Differentially expressed genes (DEGs) between treatment and control groups were identified using DESeq2 [16]. For the identification of significantly differentially expressed genes with applied thresholds of |log_2_Fold Change| ≥ 1 and a false discovery rate (FDR) < 0.05. Functional enrichment analyses, including Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG), were conducted using Goatools (Version 1.5.2) and Python SciPy (Version 0.9.0). Significantly enriched GO terms and KEGG pathways were identified against the whole-transcriptome background with a Bonferroni-corrected p < 0.05.

2.6. Gut Microbial 16S rRNA Sequencing and Analyses

Gut sample (each 200 mg) was aseptically collected, transferred into sterile 5 mL tubes, and immediately frozen at −80 °C. All steps were carried out under sterile conditions, with blank negative controls included to monitor contamination. Microbial DNA was extracted using the TIANMicrobeMagneticEnvir-DNAKit4 (DP713-T14, Tiangen Biochemical Technology Co., Ltd., Beijing, China) following the manufacturer’s instructions. DNA quality was verified using 1% agarose gel electrophoresis, and concentration/purity measured using a NanoDrop2000 spectrophotometer (Thermo Scientific, Wilmington, DE, USA). PCR amplification of the V3–V4 region was performed with barcoded primers and TransStart Fastpfu DNA Polymerase (TransGen AP221-02, TransGen Biotech, Beijing, China) in an ABI GeneAmp^®^ 9700 thermal cycler (Applied Biosystems, Foster City, CA, USA). Each sample was amplified in triplicate, combined, and purified with an AxyPrepDNA gel recovery kit (Axygen, Union City, CA, USA). PCR products were quantified using QuantiFluor™-ST (Promega, Madison, WI, USA), pooled proportionally, and used to construct sequencing libraries (NEXTFLEX Rapid DNA-Seq Kit, PerkinElmer, Austin, TX, USA). Libraries were quality-assessed using an Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA) and Qubit dsDNA HS Assay (Invitrogen, Carlsbad, CA, USA) before high-throughput sequencing on an Illumina NextSeq 2000 platform (2 × 150 bp) (Illumina, San Diego, CA, USA).

2.7. Amplicon Sequence Processing and Analyses

The demultiplexed raw reads were first quality-filtered and merged by fastp [12] and FLASH [17], respectively. The resulting high-quality sequences were then processed for error correction and resolution of amplicon sequence variants (ASVs) using the DADA2 (Version 1.26) [18] algorithm implemented in the QIIME2 [19] pipeline with recommended parameters, which obtains single nucleotide resolution based on error profiles within samples. Taxonomic assignment of ASVs was performed using the Naive bayes consensus taxonomy classifier implemented in QIIME2 and the SILVA 16S rRNA database (v138). The metagenomic function was predicted by PICRUSt2 [20] based on ASV representative sequences.

2.8. Statistical Analyses

Based on the ASVs information, rarefaction curves and alpha diversity indices including observed ASVs, Chao1 richness, Shannon index and Good’s coverage were calculated with Mothur [21]. Microbial community similarities between samples were assessed via principal coordinate analysis (PCoA) based on Bray–Curtis dissimilarity using Vegan v2.5-3 package. Treatment-related variation in community composition was tested for significance using PERMANOVA within the same package. The linear discriminant analysis (LDA) effect size (LEfSe) (http://huttenhower.sph.harvard.edu/LEfSe, accessed on 23 June 2025) [22] was performed to identify the significantly abundant taxa (phylum to genera) of bacteria among the different groups (LDA score > 2, p < 0.05). The distance-based redundancy analysis (db-RDA) was performed using Vegan v2.5-3 package to investigate the effect of Oligosaccharide treatment on the gut bacterial community structure. Based on the db-RDA results, linear regression analysis was performed to examine the relationships between the significant explanatory variables (including Oligosaccharide treatment) and microbial alpha diversity indices.

2.9. Gut and Gut Microbiota Metabolomics

Bee gut sample (100 mg each) was placed into 2 mL centrifuge tubes containing one grinding bead (6 mm diameter) and 800 µL extraction solution (methanol: water = 4:1 v/v), including four internal standards (e.g., L-2-chlorophenylalanine, 0.02 mg/mL). Samples were homogenized using a cryogenic tissue grinder for 6 min (−10 °C, 50 Hz), followed by ultrasonic extraction for 30 min at low temperature (5 °C, 40 kHz). After incubation at −20 °C for 30 min, samples were centrifuged for 15 min (4 °C, 13,000× g). The supernatant was collected into autosampler vials with insert tubes for LC-MS analysis. Metabolomic analysis was performed using an AB SCIEX ultra-performance liquid chromatography tandem time-of-flight mass spectrometry (UPLC-TripleTOF) system.

The raw data matrix was preprocessed as follows. Metabolic features detected in at least 80% of the samples within any group were retained. Missing values were imputed using the minimum value of the data matrix, and each metabolic feature was normalized to the total ion intensity. To reduce systematic variation caused by sample preparation and instrument instability, peak intensities were further normalized by the sum normalization method. Variables from quality control (QC) samples with a relative standard deviation (RSD) > 30% were excluded, and the data were log10-transformed to generate the final normalized matrix for subsequent analyses. Variance analysis was performed on the preprocessed matrix. Principal component analysis (PCA) and orthogonal partial least squares discriminant analysis (OPLS-DA) were conducted using the R package “ropls” (v1.6.2), with seven-fold cross-validation employed to evaluate model stability. Significantly different metabolites between two groups were identified based on the variable importance in projection (VIP) values derived from the OPLS-DA model (VIP > 1) and statistical significance from Student’s t-test (p < 0.05). Differential metabolites were mapped to biochemical pathways through enrichment and pathway analysis using the Kyoto Encyclopedia of Genes and Genomes (KEGG) database (http://www.genome.jp/kegg/, accessed on 25 June 2025). Metabolites were classified according to their associated pathways or biological functions. Enrichment analyses were carried out using the Python package scipy.stats (https://docs.scipy.org/doc/scipy/, Version 1.17.0, accessed on 25 June 2025), yielding the most relevant biological pathways associated with experimental treatments.

3. Results

3.1. Survival Curve, Sugar Solution Consumption Rate, and Body Weight Gain

To assess the effects of stachyose, raffinose, and their combination on the physiological status and gut microbiota of Apis cerana, survival probability, sucrose solution consumption, and body weight gain were compared among the stachyose group (S), raffinose group (M), combined group (H), and the control group (Figure 1). The survival analysis revealed that the 10-day survival probability of the H group was significantly lower than that of the control group (p = 0.011), while no significant differences were observed between either the S or M groups and the control group (p = 0.64 and p = 0.20, respectively) (Figure 1a). With respect to sucrose solution consumption, all treatment groups (S, M, and H) demonstrated significantly reduced consumption rates compared with the control (p = 0.0021, p = 0.0468, and p = 0.0022, respectively) (Figure 1b). Furthermore, body weight gain exhibited significant group-specific differences: the S and M groups showed markedly lower weight gain than the control group (p < 0.0001 and p = 0.0051, respectively), whereas the H group exhibited significantly higher weight gain relative to the control (p = 0.0023) (Figure 1c).

3.2. HE Staining of Apis Cerana Worker Midgut

To evaluate the effects of stachyose (S group), raffinose (M group), and their combination (H group) on the midgut tissue of honeybees, hematoxylin–eosin (HE) stained sections were compared between treatment groups and the control group (Figure 2). Honeybees in all treatment groups (S, M, and H) exhibited varying degrees of midgut epithelial damage, primarily characterized by the loss of cellular integrity within the intestinal wall. Among these, the H group showed the most pronounced structural disruption of midgut cells (Figure 2c), followed by the M group (Figure 2b), while the S group displayed only mild alterations (Figure 2a). In contrast, the control group exhibited relatively intact midgut cellular architecture (Figure 2d). The mean cross-sectional thickness of intestinal wall tissue in the four selected groups (S group, M group, H group, and control group) showed that the H group had the lowest value, followed by the M group. The S group exhibited a higher value than the M group, while three treatment groups all demonstrated lower thickness values compared to the control group (Figure 2e).

3.3. Intestinal Tissue Transcriptome Analyses

To further investigate the regulatory effects of stachyose (S), raffinose (M), and their combination (H) on gene expression in honeybee intestinal tissues, RNA-seq analysis was performed on whole gut samples from the S, M, and H groups. Principal component analysis (PCA), based on TPM values of all expressed genes, revealed distinct transcriptional profiles across treatments. The first two principal components explained 50.64% (PC1) and 12.15% (PC2) of the total variance, respectively. Biological replicates clustered tightly within each group, with the exception of the M group, indicating high experimental reproducibility and overall sample consistency (Figure 3a). Meanwhile, we performed non-metric multidimensional scaling (NMDS) analysis, the result showed the stress value of 0.042. In addition, ANOSIM showed a significant difference among treatment groups (R = 0.454, p = 0.014) (Figure S1). The numbers of the identified differentially expressed genes (DEGs) were 718 for S compared with the control, 88 for M compared with the control, and 266 for H compared with the control (Figure S2). Functional enrichment analysis was then performed to evaluate the potential contribution of DEGs to intestinal structural alterations at the transcriptional regulatory level. In the S group, the top five enriched KEGG pathways included the Wnt signaling pathway, MAPK signaling pathway, insect hormone biosynthesis, biosynthesis of unsaturated fatty acids, and sphingolipid metabolism (Figure 3c; Table S2). The most enriched GO terms were lipid metabolic process, cyclase activity, phosphorus–oxygen lyase activity, guanylate cyclase activity, and phospholipase activity (Table S3). In the M group, DEGs were predominantly enriched in KEGG pathways related to arachidonic acid metabolism, apoptosis, ether lipid metabolism, folate biosynthesis, and nucleotide metabolism (Figure 3d; Table S4). The top enriched GO terms included monatomic ion channel activity, channel activity, passive transmembrane transporter activity, salt transmembrane transporter activity, and monatomic ion transmembrane transporter activity (Table S5). For the H group, KEGG pathways enrichment highlighted ECM–receptor interaction, longevity regulating pathway, pentose phosphate pathway, propanoate metabolism, and sulfur relay system (Figure 3e; Table S6). The top enriched GO terms were structural constituent of chromatin, nucleosome, protein–DNA complex, DNA biosynthetic process, and protein heterodimerization activity (Table S7).

3.4. Intestinal Microbial 16S rRNA

To evaluate the effects of stachyose (S), raffinose (M), and their combination (H) on the gut microbiota of honeybees, 16S rRNA gene amplicon sequencing was performed on gut samples from the S, M, and H groups. Following quality filtering, denoising, and chimera removal using DADA2, a total of 658,012 high-quality reads were retained, with an average of 54,834 reads per sample (Table S8). At 100% sequence similarity, 244 amplicon sequence variants (ASVs) were identified across all samples (Table S9). Alpha diversity indices were calculated to assess microbial richness and diversity within samples. The Chao1 index showed no significant differences among groups (p > 0.05) (Table S10), indicating comparable microbial richness. Similarly, the Shannon index revealed no significant differences among groups (p > 0.05) (Table S11), indicating similar community diversity across treatments.

Beta diversity was assessed using permutational MANOVA based on Bray–Curtis dissimilarity. No significant separation was observed among groups (R^2^ = 0.2796, p = 0.43; Table S12). Consistently, non-metric multidimensional scaling (NMDS) analysis confirmed the absence of significant differences in community composition between groups (ANOSIM, Stress = 0, R = 0.0031, P = 0.464; Figure 4a). Venn diagram analysis revealed differences in genus-level composition across groups. Specifically, 26 genera were detected in the control group, 11 in the S group, 11 in the M group, and 17 in the H group (Figure 4b). Taxonomic classification of ASVs revealed that the honeybee gut microbiota was predominantly composed of Bacillota, Pseudomonadota, Bacteroidota, Actinomycetota, and Cyanobacteriota at the phylum level (Figure S3). No significant differences were observed in the relative abundances of these phyla among groups (p > 0.05; Table S13). Co-occurrence network analysis revealed distinct distribution and interaction patterns among dominant gut bacteria across treatments. Lactobacillus was connected to nine sample nodes spanning all four treatment groups, indicating its presence across all treatments. Snodgrassella exhibited nine edges, including eight associations with samples from all four treatment groups and one association with Gilliamella. Melissococcus showed eight edges, with seven connections to samples from three treatment groups, excluding the S group, and one association with Gilliamella. Gilliamella displayed seven edges, including five connections to samples from two treatment groups, excluding the S and M groups, and two associations with Snodgrassella and Melissococcus, indicating its absence in the S and M groups. Notably, interaction relationships were observed among Snodgrassella, Gilliamella, and Melissococcus within the network (Figure 4c). At the genus level, the most abundant taxa included Lactobacillus, Snodgrassella, Melissococcus, Gilliamella, Apibacter, Bombilactobacillus, Bifidobacterium, and Apilactobacillus. The top five genera by average relative abundance were Lactobacillus (76.87%), Snodgrassella (6.44%), Melissococcus (6.33%), Gilliamella (4.74%), and Apibacter (1.93%) (Figure 4d; Table S14). Notably, Gilliamella was significantly enriched in the control group compared with the treatment groups (S, M, and H), with the most pronounced reduction observed in the S group (p = 0.04148; Figure 5a,b). The average abundance values of Lactobacillus in treatment groups (S, M, and H) were all higher than those in the control group, and the average abundance values in groups S, M, and H decreased successively. The average abundance values of Snodgrassella in groups S and M were higher than those in the control group, whereas those in group H were lower than those in the control group. The average abundance values of Snodgrassella in groups M, S, and H decreased successively (Figure 4c; Table S14). To further explore the functional potential of the gut microbial communities, functional profiling was performed using PICRUSt2. The top five predicted pathways across all groups were metabolic pathways, biosynthesis of secondary metabolites, microbial metabolism in diverse environments, phosphotransferase system, and ribosome (Figure 5c).

3.5. Gut Microbial Metabolomic Analyses

The biological replicates in the four comparisons were well-clustered (Figure 6a; Figure S4). In total, 4269 metabolic features were detected, of which 1242 metabolites were successfully annotated to KEGG compounds. These metabolites spanned diverse categories, including antibiotics, carbohydrates, hormones and transmitters, lipids, nucleic acids, organic acids, peptides, steroids, vitamins, and cofactors. Furthermore, 1389 metabolites were mapped to KEGG pathways, encompassing amino acid metabolism, carbohydrate metabolism, energy metabolism, global and overview maps, lipid metabolism, membrane transport, metabolism of cofactors and vitamins, metabolism of other amino acids, nucleotide metabolism, signaling molecules and interaction, translation, transport and catabolism, and xenobiotics biodegradation and metabolism (Figure 6b; Table S15). Unsupervised PCA revealed a clear trend of separation among groups, suggesting distinct metabolic profiles. This distinction was further confirmed by the supervised PLS-DA model (R = 0.5022, p = 0.001) (Figure 6a; Figure S4). Differential metabolites were identified based on the criteria of variable importance in projection (VIP > 1) and Student’s t-test (p < 0.05). In the S group, 2,145 differential metabolites were detected compared with the control group, among which 362 were successfully annotated with compound names (Table S16). Of these, 163 were up-regulated and 199 were down-regulated (Figure S5). KEGG pathway enrichment analysis revealed that these metabolites were significantly enriched (adjusted p-value < 0.05) in galactose metabolism, ABC transporters, alanine, aspartate and glutamate metabolism, starch and sucrose metabolism, and butanoate metabolism (Figure 6c; Table S17). In the M group, 4779 differential metabolites were identified, including 825 annotated compounds (Table S16). Among them, 435 metabolites were up-regulated and 390 were down-regulated (Figure S5). The significantly enriched KEGG pathways (adjusted p-value < 0.05) were biosynthesis of cofactors, nucleotide metabolism, ABC transporters, glycerophospholipid metabolism, and purine metabolism (Figure 6d; Table S17). In the H group, 2281 differential metabolites were identified relative to the control, of which 477 were annotated with compound names (Table S16). Among these, 157 metabolites were up-regulated and 320 were down-regulated (Figure S5). Enrichment analysis revealed significant associations (adjusted p-value < 0.05) with alanine, aspartate and glutamate metabolism, arginine biosynthesis, biosynthesis of cofactors, purine metabolism, and galactose metabolism (Figure 6e; Table S17). Collectively, these pathways and metabolites are likely to play pivotal roles in oligosaccharide digestion, nutrient absorption, cellular damage, and survival in Apis cerana.

4. Discussion

To evaluate the impact of oligosaccharides (stachyose, raffinose, and their mixture) on the survival of Apis cerana, we compared the 10-day survival curves of bees fed with individual oligosaccharides and with their mixture. Our results demonstrated that the survival rate of bees in the mixture group was significantly lower than that of the control, whereas bees fed with single oligosaccharides (stachyose or raffinose) showed no significant differences from the control, despite a clear downward trend (Figure 1a). These findings indicate that, in the short and medium term, individual oligosaccharides exert only a limited threat to the survival of A. cerana, while their combination imposes a more pronounced risk. These adverse effects may partially explain the reluctance of A. cerana to forage on the nectar and pollen from Camellia reticulata.

Consistent with our observations, Li et al. (2022) reported that supplementation with raffinose, stachyose, or their mixture significantly reduced the lifespan and survival rate of adult worker Apis mellifera [9]. In addition, we found that A. cerana fed with oligosaccharides consumed significantly less sucrose compared with the control. Notably, feeding with single oligosaccharides markedly reduced body weight gain, whereas feeding with the oligosaccharide mixture significantly increased body weight gain (Figure 1b,c). These contrasting effects suggest that oligosaccharides interfere with nutrient assimilation in A. cerana, with distinct physiological consequences depending on whether they are ingested individually or in combination. The precise mechanisms underlying these differences warrant further investigation.

Li et al. (2022) The further research demonstrated that oligosaccharides such as manninotriose, raffinose, and stachyose cannot be effectively digested in the gut of adult worker A. mellifera due to the absence of specific hydrolytic enzymes, and the subsequent accumulation of indigestible oligosaccharides leads to metabolic disorders and ultimately death [9]. By analogy, it is plausible that the core mechanism of oligosaccharide-induced toxicity in A. cerana also involves intestinal dysfunction arising from the lack of oligosaccharide-digesting enzymes. In support of this, we observed pronounced structural damage in the intestinal epithelial cells of A. cerana fed both single and mixed oligosaccharides. We therefore hypothesize that the accumulation of undigested oligosaccharides may cause progressive intestinal cell injury.

To elucidate the molecular mechanisms underlying this damage, we conducted transcriptome profiling of the intestinal tissues. In the stachyose-fed group, differentially expressed genes (DEGs) were significantly enriched in biological processes including the MAPK signaling pathway, lipid metabolic processes, and phospholipase activity. Notably, activation of the MAPK signaling pathway has been implicated in oxidative stress and inflammatory responses. For example, stress-induced MAPK activation has been shown to promote inflammation and oxidative injury [23]. Perturbations in lipid metabolism also play a critical role in inflammation and apoptosis, as evidenced by studies linking lipid metabolic disorders to epithelial cell damage in mastitis [24]. Moreover, phospholipase-mediated signaling has been associated with early endothelial injury, with VEGF-A inducing retinal endothelial damage via phospholipase A2 activation [25]. Taken together, these observations suggest that the enrichment of MAPK signaling, lipid metabolic dysregulation, and phospholipase activity in the stachyose-fed bees likely represents key drivers of the intestinal structural damage observed in A. cerana.

In the raffinose-fed group, DEGs in the gut were predominantly enriched in pathways closely linked to cellular injury, including arachidonic acid metabolism, apoptosis, and ether lipid metabolism. Elevated levels of arachidonic acid (AA) have been reported to induce oxidative stress and acute erythrocyte injury [26]. Ether lipids are also recognized as mediators of cellular damage in neuroblastoma models [27]. Collectively, these findings support the notion that enrichment of these pathways may underlie the structural injury of intestinal cells observed in the raffinose group.

In contrast, in the mixture group (stachyose plus raffinose), DEGs were enriched in pathways and functions potentially associated with membrane and structural damage, including ECM–receptor interaction, longevity regulating pathway, and the pentose phosphate pathway, as well as molecular functions related to chromatin and nuclear integrity, such as structural constituent of chromatin, nucleosome assembly, and protein–DNA complex formation. Prior studies have demonstrated that lipopolysaccharides (LPS) can disrupt ECM organization and actin cytoskeleton architecture, and ultimately compromise ECM–cytoskeleton crosstalk [28]. Similarly, in Drosophila, the Efl21 (K09542) gene within the longevity regulating pathway has been shown to stabilize intermediate filament proteins, preventing their aggregation under stress conditions and thereby preserving cytoskeletal, organelle, and myofilament integrity [29]. Our study revealed a significant upregulation of Efl21 (K09542) in the mixture group, suggesting that intestinal structural damage induced by mixed oligosaccharides may elicit a compensatory response aimed at maintaining cytoskeletal stability. Moreover, excessive histones are known to exert cytotoxic effects through multiple mechanisms, ultimately compromising genome stability and cell viability [30]. In line with this, we detected significant upregulation of histone-related genes (LOC108003504, LOC108003589, LOC108003624) in the intestines of bees fed mixed oligosaccharides. This observation suggests that histone overexpression may represent an additional contributor to oligosaccharide-induced intestinal structural injury.

To further elucidate the effects of oligosaccharides on gut homeostasis, we examined the gut microbial community structure of A. cerana through 16S rRNA sequencing. In our study, we used the co-occurrence network analysis to explore the distribution patterns and potential interactions of dominant gut bacteria under different treatments. The ubiquitous presence of Lactobacillus across all treatment groups suggests that this taxon represents a stable core member of the gut microbiota, likely exhibiting a high tolerance to oligosaccharide treatment-induced perturbations. In contrast, the absence of Gilliamella in the S and M treatment groups indicates a higher sensitivity of this bacterium to oligosaccharide treatments. Notably, the observed associations among Snodgrassella, Gilliamella, and Melissococcus highlight potential interspecific interactions within the gut microbial community. The selective loss of Gilliamella in oligosaccharide treatment groups may therefore disrupt these microbial interactions, potentially leading to alterations in gut community stability and host health. Although co-occurrence network analysis does not imply direct causal relationships, the observed changes in network connectivity and species presence suggest that the treatments reshaped the gut microbial interaction landscape. These findings provide a network-based perspective on how specific bacterial taxa respond to oligosaccharide treatment-induced stress and underscore the importance of microbial interactions in maintaining gut community structure. Previous studies have demonstrated that the worker bee gut is dominated by nine characteristic bacterial phylotypes. Based on 16S rDNA analyses [31,32] and metagenomic profiling [33], these phylotypes account for approximately 95–99% of the total gut microbiota. Among them, five taxa constitute the bee core microbiota, including two Gram-negative bacteria, Snodgrassella alvi (class Betaproteobacteria) and Gilliamella apicola (class Gammaproteobacteria) [34], as well as three Gram-positive bacteria: two abundant and widespread Lactobacillus clades (Firm-4 and Firm-5) [35] and one less abundant Bifidobacterium species [36]. These five core taxa accounted for more than 80% of the gut microbiota in the control group (Figure 4d; Table S14). We further found that oligosaccharide feeding altered the relative abundance of gut microbiota, with the most notable change being a significant reduction in Gilliamella abundance (Figure 4d; Figure 5a,b). Toxicological assays of Camellia oleifera nectar and pollen revealed that manninotriose, raffinose, and stachyose in the nectar were toxic to bees [9]. Time-resolved metabolomic analyses demonstrated that the toxic mechanism of oligosaccharides lies in their inability to be further digested in A. mellifera, leading to accumulation in the body, disruption of trehalose synthesis and metabolism, and ultimately bee death [9]. Earlier, Zheng et al. (2016) reported that Gilliamella apicola in the bee gut can metabolize sugars toxic to bees, such as mannose, arabinose, and xylose [37]. The oligosaccharides upregulate the expression of beta-galactosidase gene (LOC107999782, Table S18), likely activated the lactose metabolic pathway in the treatment groups (S, M and H groups). This activation may have led to the oligosaccharides competitively binding to the sugar metabolism enzymes within Gilliamella. Consequently, sugar could not be metabolized effectively, which may explain the observed reduction in Gilliamella abundance. More recently, Chen et al. (2025) found that A. mellifera colonies living long-term in C. oleifera forests harbored significantly higher abundances of Gilliamella apicola compared with colonies that had not been exposed to pollination, indicating that adaptation of gut microbiota to this ecological niche enhanced the host’s ability to metabolize toxic oligosaccharides [38]. These findings suggest that Gilliamella is most likely the major bacterial taxon involved in oligosaccharide digestion (e.g., stachyose and raffinose). In our study, both the single oligosaccharide and the mixture groups exhibited reduced relative abundances of Gilliamella, with the stachyose group showing a particularly significant reduction (Figure 5a,b). This indicates that oligosaccharides depleted Gilliamella populations, and the consequent reduction in Gilliamella further impaired oligosaccharide digestion, leading to substantial accumulation of oligosaccharides in the gut (Figure 4d; Figure 5a–c).

To further investigate the metabolic alterations in the gut of A. cerana caused by oligosaccharides, we performed metabolomic sequencing of both the intestinal tissue and its associated microbiota.

Both single oligosaccharide groups (stachyose or raffinose) and the mixed oligosaccharide group significantly altered the galactose metabolism pathway which has been related to intestinal physiology. Moreover, comparative analysis of differential metabolites within this pathway revealed that stachyose levels were markedly increased in both the stachyose and mixture groups, suggesting substantial accumulation of stachyose in the gut. In contrast, raffinose levels did not significantly increase in either the raffinose or mixture groups, implying that both the microbiota and host possess stronger metabolic capacity for raffinose than for stachyose (Table S16). This difference may be attributable to the more complex heterocyclic structures of stachyose compared with raffinose. In the raffinose group, most differential metabolites were enriched in amino acid metabolism pathways, suggesting that amino acid metabolic dysregulation may be closely related to intestinal cell structural damage and reduced survival. Glutamine is an essential nutrient for the growth, differentiation, mucosal integrity, and barrier function of intestinal epithelial cells. Multiple cell culture studies have confirmed its role in maintaining tight junction integrity of the intestinal mucosa [39]. In our study, levels of 1-pyrroline-5-carboxylic acid, glutamic acid, and L-glutamine were all significantly reduced in the raffinose group, indicating suppression of the L-glutamine biosynthesis pathway, which greatly lowered intestinal L-glutamine levels and may have impaired the growth, differentiation, and integrity of intestinal epithelial cells (Figure 2b; Tables S16 and S18). Similarly, these metabolites were also significantly reduced in the mixture group, suggesting that the mechanisms of intestinal cell damage in the mixture group were partly similar to those in the raffinose group (Figure 2c; Tables S16 and S18). In addition, a reduction in the microbial metabolite 5-hydroxyindoleacetic acid (5-HIAA) is associated with impaired intestinal motility and function. Reduced 5-HIAA levels may indicate broader gut microbiota dysbiosis, potentially compromising immune function and inducing metabolic imbalance [40]. In our study, differential metabolites in the tryptophan metabolism pathway—including formyl-5-hydroxykynurenamine, 5-hydroxyindoleacetylglycine, and 5-hydroxyindole-3-acetic acid—were all significantly decreased, suggesting that oligosaccharide feeding in the raffinose group impaired intestinal motility and function, which in turn reduced 5-HIAA levels and exacerbated metabolic dysregulation (Tables S16 and S18). Compared with the mixed group, both single oligosaccharide groups (stachyose and raffinose) exhibited significant enrichment of differential metabolites in the ABC transporter pathway. ABC transporters are complex molecular systems that utilize energy to actively transport diverse molecules across cellular membranes against concentration gradients, and they are essential for bacterial metabolism of vital molecules such as vitamins, metal ions, and sugars [41]. These results suggest that in the stachyose or raffinose group, gut microbiota may reduce host energy supply through altered ABC transporter pathways, thereby contributing to reduced body weight gain in honeybees (Figure 1c). In the future, we can attempt to enhance the tolerance of A. cerana to Camellia reticulata honey by supplementing the probiotics of Gilliamella or the degradation enzyme of oligosaccharides. This facilitates the utilization of Camellia reticulata nectar by A. cerana.

5. Conclusions

In conclusion, through controlled feeding experiments, we demonstrated that raffinose and stachyose—particularly in combination—significantly reduced the survival and sucrose consumption of Apis cerana. Both single and mixed oligosaccharide treatments induced intestinal epithelial damage and disrupted the gut microbial balance, with a marked decline in Gilliamella abundance in the stachyose-fed group. Integrative transcriptomic and metabolomic analyses further revealed that oligosaccharide exposure altered the expression of genes related to intestinal integrity and modulated key metabolic pathways. Collectively, these findings provide novel mechanistic insights into the adverse effects of plant-derived oligosaccharides on honeybee physiology and gut homeostasis, shedding light on the molecular basis of bee–plant interactions in pollination ecology.