From nurse bee to queen egg: RNA-seq analysis of Apis mellifera eggs shows dietary protein-dependent gene regulation

Abstract

Honeybees, Apis mellifera, play a vital role as pollinators in global agricultural ecosystems. Nutrition, particularly dietary protein content, profoundly impacts honeybee health and reproduction. Yet, the molecular mechanisms connecting diet composition and gene expression in honeybee eggs remain underexplored. In this study, we investigate the intricate relationship between diet, gene expression, and honeybee egg development. Using RNA-seq analysis, we explore the effects of different protein-to-carbohydrate (P: C) ratios in honeybee diets on differential gene expression in the eggs laid by the queen and potential associated molecular responses. Our research identifies 1007 differentially expressed genes (DEGs) across various dietary conditions, highlighting the pivotal role of nutritional composition in shaping gene expression during egg development.

Cluster analysis revealed two DEG profiles corresponding to low protein diets (LPD) and high protein diets (HPD). LPD conditions upregulate genes linked to protein catabolism, autophagy, and ubiquitin-mediated proteolysis, indicating potential cellular responses to nutritional stress. Conversely, HPD conditions upregulate genes related to RNA processing, spliceosome activity, and the MAPK signalling pathway, suggesting normal cellular development.

Notably, the Hippo signalling pathway exhibits distinct gene regulation patterns under LPD and HPD conditions, potentially influencing cellular growth and differentiation in response to nutrient availability.

Our findings underscore the critical role of nutrition in honeybee health and reproduction, providing insights into optimizing honeybee diets for colony health and resilience. As honeybee populations confront challenges from changing environmental conditions and resource availability, understanding these molecular responses is crucial for their effective management and conservation as essential pollinators. This study establishes a foundation for further investigations into the functional consequences of these molecular responses at the individual level and their broader implications for honeybee colony development and health.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12864-026-13155-2.

Introduction

Honeybees, Apis mellifera, are vital pollinators that play a pivotal role in maintaining global agricultural ecosystems and biodiversity [1]. However, populations in North America and Europe are experiencing alarming declines [2–4] raising urgent concerns about the factors affecting honeybee health [5–7]. Nutrition is one of the key drivers of colony performance. For instance, pollen diversity and abundance strongly influence honeybee egg production and overall colony growth [8], while dietary protein intake, closely linked to landscape composition, significantly affects honeybee health [9–13].

Nurse bees play a central role in this process by transforming pollen and nectar into royal jelly, a protein-rich secretion that sustains both larvae and the queen [14, 15]. Proteins provide the essential amino acids for honeybee growth and reproduction [16], and protein content in diets directly influences colony dynamics through queen egg laying [17–20]. However, colony nutrition depends not only on absolute protein intake but also on macronutrient balance, particularly P: C ratios. Honeybee colonies regulate nutrient intake collectively through social nutritional mechanisms, and macronutrient balance has been shown to influence foraging behaviour, nurse bee physiology, and larval development [21–23]. While the connection between dietary protein intake and honeybee development is well established [24–26], the molecular mechanisms linking nurse bee diet to queen egg remain poorly understood in A. mellifera and eusocial insects more broadly. Most studies have focused on ovarian development or hormonal regulation in reproductives [27], leaving open the question of how maternal diet influences egg gene expression and subsequent development.

Recent work in insects highlights the importance of maternal provisioning and gene regulation in eggs [28–31]. Maternal gene products, including mRNAs and proteins, drive the start of embryonic development in many organisms before the zygotic genome becomes active [31]. This maternal control is conserved across species and plays a crucial role in early development and subsequent growth [29]. In Drosophila, maternal nutritional stress alters egg provisioning [30] and transcript composition [31], reflecting plastic responses to poor conditions.

In honeybees, nutritional conditions have been shown to influence both the number [19] and the quality (development, phenotype, survival) of eggs laid [16], with possible long-term effects on adult phenotype, health, behaviour and survival.

However, the impact of nutrition in eusocial species is constrained by the complexity of social nutrition and the division of labour, primarily between reproductive and non-reproductive castes [32]. Because of the complexity of social nutrition, research has primarily focused on queenless bees [33], royal larvae [34], and virgin queens [35], examining ovarian development, hormonal regulation, and gene expression. Studies in Apis mellifera have been constrained to this point because evaluating queen performance in the field is challenging [36] due to variations in environmental factors that affect insect development [9, 37]. Additionally, obtaining enough eggs while controlling nutrition presents technical challenges to recreate the conditions for egg laying. To bridge this gap, Fine, Shpigler [38] introduced an innovative laboratory system designed to monitor honeybee queen egg laying under controlled conditions. These experiments emphasize the utility of mesocosms adapted to honeybees as a valuable tool for monitoring queen reproduction [8, 39]. Overall, this research not only presents a novel approach for monitoring queen egg laying but also highlights the broader implications of pollen nutrition on honeybee biology. This sets the stage for an investigation into the impact of colony nutrition on egg production. Using a similar device, our research explores how the nutrition of nurse bees feeding the queen impacts gene expression in the eggs.

Such approaches open new avenues of research, raising key questions about how colony size, nutritional status, and egg production interact to shape adult worker bee phenotype, health, behaviour, survival, and ultimately overall colony development and performance. In this study, we specifically tested the hypothesis that nurse-mediated queen nutrition influences gene expression in freshly laid eggs. Adequate nutrition of the reproducing queen is essential to support embryonic development and may have consequences for both individual and colony growth, health and survival. By exploring the molecular responses to varying P: C ratios, we aim to provide a comprehensive understanding of how diet composition influences gene expression in the egg and potential consequences in embryo development. We employ RNA-seq analysis to study the differentially expressed genes under various dietary conditions. These insights hold implications for optimizing honeybee diets to promote colony health and resilience, especially in the context of changing environmental conditions and resource availability.

Materials and methods

Egg production under Honeybee diet laboratory experiment

Adaptation of FECS Cage for egg-laying monitoring and egg collection under laboratory conditions

The FECS (Film-Assisted Honeybee Egg Collection System) cage [40] was selected for its suitability to monitor egg-laying and collect eggs under controlled conditions. The blueprint was obtained from the electronic supplementary materials of [40] and 3D-printed using PLA+ filament (1.75 mm diameter, eSun^®) at the Biochemistry Department, University of Otago (Supplementary Fig. 1). The cage features 256 cells, each with a hole for easy egg collection. To enhance adhesion, the frame was coated with silicon. An adapted version was created by adding a wooden frame and wire mesh to confine the bees. The internal volume of each cage was approximately 400 mL, accommodating around 100 bees per cage (Supplementary Fig. 2).

Preparation of bee food and queen diets

Pollen patties were formulated using pollen granules coming from a variety of plants and different hives sites (Happy Valley Honey Ltd.) (Supplementary Table 1), sucrose (graded white cane sugar, Chelsea^®, NZ Sugar Co), brewer’s yeast (Sunray Apothecary Health Store, NZ), lactalbumin enzymatic hydrolysate powder (Sigma-Aldrich GmbH), and unpasteurized liquid honey, mainly clover honey (Arataki Honey Ltd.) (Supplementary Table 2). Pollen granules were ground into a fine powder using a coffee grinder. The powdered mixture of sugar, brewer’s yeast, and lactalbumin was combined in a mortar. Honey was slowly added to create a thick, sticky dough. The final amount of honey was adjusted slightly to obtain a consistent dough texture suitable for bee feeding, as honey viscosity varies with temperature and humidity. These minor adjustments (typically a few drops) did not significantly affect the intended protein-to-carbohydrate ratios. The doughs with P: C ratios of 1:3, 1:5 and 1:9 were sent to the Nutrition Laboratory of Massey University for macronutrient composition analysis. The intermediate diets (1:7 and 1:10) were formulated proportionally from the analysed diets using identical ingredients and preparation procedures, but were not chemically re-analysed. Because all diets were prepared from the same ingredient sources and protocols, any deviation between nominal and actual macronutrient ratios is expected to be minimal. Six diets were formulated: 0:1, 1:10, 1:9, 1:7, 1:5 and 1:3 (Table 1). For comparative analyses, diets were grouped according to their P: C ratios. Diets 0:1, 1:10, and 1:9 were classified as low-protein diets (LPD), whereas diets 1:7, 1:5, and 1:3 were classified as high-protein diets (HPD). This classification was primarily based on relative dietary protein content and was further supported by transcriptomic clustering analyses showing consistent expression patterns across these nutritional categories. The dough of each diet was divided into small patties (~ 5 g), placed in a 15 mL centrifuge tube, and sealed. Diet patties were freshly prepared and provided to the bees on the same day to avoid any compositional change.

Table 1.

P:C ratio, ingredient quantity (g) and protein, carbohydrate and lipid percentage in diet

	Low Protein Diet (LPD)			High Protein Diet (HPD)
P:C Ratio	0:01	1:10	1:09	1:07	1:05	1:03
Ingredient:
Pollen granules	0	11.18	12.5	15	20	30
Sucrose	60	72.94	70	64	52	28
Brewer's Yeast	0	10.6	11.8	14.2	19	28
Lactalbumin	0	5.1	5.63	6.75	9	13.5
Honey	40	30	30	30	30	30
Percentage of Macronutrients:
% of Protein	0	7.94	8.78	10.45	13.8	17.5
% of Carbohydrate	100	80.5	78.9	75.7	69.3	62.1
% of Lipid	0	1.14	1.28	1.55	2.1	2.2

Honeybee eggs production under controlled environment

Frames with capped brood were sourced from colonies provided by Otto’s Bees Ltd from the same beeyard in Dunedin, NZ, under the same beekeeping management from December to January during the NZ summer. The bees originated from commercial New Zealand apiaries and likely represent admixed Apis mellifera stocks primarily derived from A. m. ligustica and A. m. carnica, although precise subspecies identification was not available. A total of 33 sister queens originating from the same apiary were free-mated at the same site and introduced into small cages with 7–9 emerging bees with candy for 2 days. Capped brood frames were placed in a nucleus box and stored in an incubator at 34 °C for 24 h. Relative humidity in the incubator was controlled and maintained between approximately 70% and 80%. The queens and 100 newly merging workers (0–24 h old) were introduced simultaneously into the laying cages. Workers in each cage were fed one diet for one week (Table 1), with food replaced daily. No significant differences in consumption were observed among the pollen patty diets, indicating comparable intake across these nutritional treatments. However, food consumption under the pure carbohydrate (candy) diet differed significantly from pollen patty diets. Direct quantification of nurse bee food processing and queen intake was not feasible within this experimental system. Daily pooled egg batches were collected and stored at -80 °C. Each RNA-seq sample consisted of eggs pooled from a single queen during one day of laying, typically comprising approximately 80–150 eggs. Pooling was performed to ensure sufficient RNA yield while minimizing technical variability associated with individual egg extraction.

After a 5-day trial, the workers were released from the cages and the queens were stored individually in small cages with newly emerging bees under a candy diet for two days before being sequentially re-exposed to a different diet with a new cohort of emerging workers. For transcriptomic analysis, each RNA-seq sample consisted of pooled eggs collected during a single day from one queen under a defined dietary treatment. A total of 24 RNA-seq samples were analysed, representing 18 unique queens across multiple dietary conditions (Supplementary Table 3 A). Individual queens contributed one or more samples across different dietary treatments using distinct worker cohorts, resulting in repeated measures across queens. Each RNA-seq sample (pooled eggs collected from a single queen on a given day; Supplementary Table S3B) was treated as an independent biological replicate corresponding to a specific queen and sampling day. Non-independence due to repeated sampling of the same queens was accounted for by including queen identity in the statistical model.

RNA extraction and library preparation for RNA-seq

Total RNA from a group of eggs was extracted using a hybrid Trizol (Ambion) and RNeasy mini kit (Qiagen) protocol [41]. RNA samples were quantified, and their quality was assessed using a Nanodrop 2000. Libraries were constructed using the TruSeq Stranded mRNA Library Prep kit and uniquely indexed. Libraries were pooled and sequenced using NextSeq 2000 P2 reagents, producing 100 bp paired-end reads, by Otago Genomics Facility.

RNA-SEQ analysis

Expression analysis

RNA sequencing data processing was performed using the nf-core/rnaseq pipeline v3.9, Nextflow version available Dec 16 2022 [42], using the Apis mellifera reference genome Amel_HAv3.1 [43]. Quality control steps included Read QC via FastQC [44], UMI extraction using UMI-tools [45], and Adapter and quality trimming with Trim Galore [46]! . Star [47] and Salmon [48] were used for alignment and quantification. DESeq2 was employed to generate the DESeqDataSet (DDS) object [49]. This was achieved by importing Salmon output files via tximport [50], as part of an analysis conducted in R version 4.2.2 [51].

Power calculation

To estimate the statistical power of our RNA-seq experiment, we employed the RnaSeqSampleSize v1.22.0 package [38]. This package allows for the calculation of statistical power based on count data, enabling a more biologically relevant assessment. Count dispersions were estimated from the count matrix within the DESeqDataSet [49]. Using a repNumber of 2,000 and considering LFC thresholds of 2, 5, and 10, we generated power distributions to determine the proportion of genes showing specific Log2 fold changes (LFC) that can be detected with our sequencing depth and sample replication.

Principal component analysis (PCA)

PCA was conducted to explore the relationships among samples based on gene expression profiles. Raw read counts were obtained by nf-core, and then processed by DESeq2, filtering out genes with low expression. The remaining genes were log-transformed, scaled, and centred to account for technical effects. PCA was performed to visualize the variation between samples along the principal components.

Likelihood ratio test (LRT)

Differential gene expression analysis was conducted using the likelihood ratio test (LRT) implemented in the DESeq2 package (Love, Huber, & Anders, 2014). The primary model tested the effect of diet on gene expression while accounting for egg collection day (~ day + diet vs. ~ day). Additional LRT analyses were performed to evaluate the effects of queen identity and worker cohort. These analyses indicated that queen identity explained the largest proportion of transcriptional variation, while diet also showed a substantial effect, whereas worker cohort and egg collection day contributed comparatively little to gene expression variation (Supplementary Table 4). Because some queens were sequentially exposed to multiple dietary treatments, queen identity was explicitly included in the statistical modelling to account for potential non-independence among samples. The limited contribution of egg collection day suggests that temporal effects had only a minor influence on transcriptomic variation under the controlled experimental conditions. Genes with an adjusted p-value (Padj) < 0.05 after Benjamini–Hochberg correction were considered differentially expressed for the global analysis. No fold-change threshold was initially applied in order to capture broad transcriptional responses. Additional filtering using |log2FC| ≥ 1 and |log2FC| ≥ 2 thresholds was subsequently performed for visualization and interpretation of high-effect-size genes.

Functional Enrichment Analysis (FGSEA and GO Term Heat-map):

• Gene Ontology (GO) term overrepresentation in the diet DEG list was investigated with gene set enrichment analysis using Fast Gene Set Enrichment Analysis (FGSEA) [52], with Gene Ontology (GO) annotations sourced from the Hymenoptera Genome Database [53].

To visualize the overlap across enriched pathways, a GO terms heat-map was generated. GO annotations were loaded, and the FGSEA analysis was conducted on DEGs. The overlap between each pair of pathways was calculated, and hierarchical clustering organized pathways based on the similarity of gene overlap. The heatmap, created using the gplots package [54], visually represented shared and distinct patterns of gene expression changes across diets.

Standard GO Overrepresentation Analysis:

• To further uncover enriched biological functions, we conducted standard gene ontology overrepresentation analysis using the ClusterProfiler v4.6.2 package [55]. GO annotations from the study by Walsh et al. (2021) were used to annotate differentially expressed genes (DEGs) with adjusted p-value < 0.05. This analysis provided insights into functional categories enriched among the DEGs.

KEGG Pathway Analysis:

• Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis was performed using the DAVID Bioinformatics Resources 6.8 platform [56]. Gene names derived from model organisms refer to predicted orthologues in the Apis mellifera genome annotation and do not imply direct functional validation in honeybees. DEGs were analyzed for their associations with KEGG pathways, with significance thresholds set at adjusted p-value < 0.05. The analysis utilized the whole gene list of the Apis mellifera genome (Amel_HAv3.1) as the background against the DEGs.

Gene Expression Patterns:

• To investigate expression patterns of DEGs across multiple diets, we performed clustering analysis using the Degreport v1.34.0 package [57]. Six distinct clusters were initially identified, which were further grouped into two merged clusters based on opposite expression patterns to the P: C ratio diet.

Differential Gene Expression between LPD and HPD:

• In order to explore differential gene expression between the LPD and HPD groups, multiple analyses were performed comparing the two groups. This included DESeq2 Wald test comparison, FGSEA, and KEGG analysis as described above. These analyses provided an understanding of the gene expression changes associated with different P: C ratios. To visually assess the significance and magnitude of differential expression between the LPD and HPD groups, we created volcano plots using EnhancedVolcano package in R [58].

Results

RNA-seq data quality overview

Sequencing quality metrics indicated overall high-quality RNA-seq data across all samples. Mean Phred quality scores were consistently above Q30, indicating high base-calling accuracy (Supplementary Figs. 3 and 4). Alignment rates to the Apis mellifera reference genome were high across samples, with the majority of reads uniquely mapped (Supplementary Fig. 5). Sequencing depth was comparable among libraries, with no evidence of major technical bias (Supplementary Fig. 6). Together, these quality metrics confirm the robustness of the dataset for downstream differential expression analysis.

Differential gene expression analysis

Differential gene expression analysis was conducted to investigate the effects of different P: C ratios on honeybee gene expression. To assess the relative contribution of experimental factors to transcriptomic variability, likelihood ratio tests including queen identity, diet, worker cohort, and egg collection day were performed. Queen identity accounted for the largest proportion of transcriptional variation, with 2193 differentially expressed genes (18.95%), compared with 1007 DEGs associated with diet (8.70%). Worker cohort and egg collection day contributed comparatively less variation (79 DEGs, 0.68%, and 6 DEGs, 0.05%, respectively), indicating substantial intrinsic inter-queen variability while confirming a significant dietary effect on egg gene expression. Differential expression analysis focusing specifically on diet, identified a total of 1007 DEGs across the different diets.

PCA of transcriptomic profiles

Principal component analysis (PCA) was performed to visualize global transcriptomic variation across dietary P: C ratios (Supplementary Fig. 7). PC1 explained 53% of the total variance, and PC2 explained 8%. Samples did not form strictly discrete clusters according to diet, and substantial dispersion was observed within most dietary groups, consistent with significant biological variability among queens. However, samples from the highest protein diet (1:3) showed a tendency to cluster together, suggesting a more consistent transcriptional response under strongly protein-rich conditions. In contrast, intermediate diets such as 1:9 and 1:7 displayed partial overlap and occupied transitional positions along the PCA axes, supporting the interpretation of a largely continuous nutritional effect rather than sharply discrete categories.

Clustering and expression patterns

Supplementary Fig. 8 indicates that there are two main DEG profiles. The first profile concerns the samples under the diets 0:1, 1:10 and 1:9, which present a lower protein level than the diets 1:7, 1:5 and 1:3, which present a higher protein level. These two groups are confirmed by the clustering analysis that grouped DEGs with similar expression patterns (Supplementary Fig. 9): groups 1,3 and 8 have similar differential expression pattern profiles with upregulated genes for the 0:1, 1:10 and 1:9 diets and downregulated genes for the 1:7, 15 and 1:3 diets. Conversely, groups 2,4 and 5 display the opposite patterns with down-regulated genes for the 0:1, 1:10 and 1:9 diets and up-regulated genes for the 1:7, 15 and 1:3 diets. Therefore, DEGs are grouped according to none or low protein content in the diet and medium/high protein content in the diet. For convenience, the two groups of diets are named low protein diets (LPD), comprising the 0:1, 1:10 and 1:9 diets, and HPD, comprising the 1:7, 1:5 and 1:3 diets. DEGs from groups 1, 3 and 8 were merged in cluster 138, referring to upregulated genes on LPD and group 2,4,5 in cluster 245, referring to upregulated genes on HPD (Supplementary Fig. 10). Clusters 138 and 245 contain 403 and 451 DEGs, respectively, showing distinct expression profiles related to protein level in the diet. Clustering analyses captured coordinated expression patterns across the dietary gradient rather than strictly binary differences between diet groups. LPD conditions resulted in the upregulation of genes associated with protein catabolic processes, autophagy, and ubiquitin-mediated proteolysis, likely reflecting cellular responses to nutritional stress. HPD conditions, on the other hand, showed upregulation of genes involved in RNA processing, spliceosome activity, and the MAPK signalling pathway, indicating normal cellular development and functioning.

Standard GO overrepresentation analysis and KEGG

To better interpret transcriptional responses to dietary protein levels, Gene Ontology (GO) overrepresentation and KEGG pathway analyses were performed separately on the two main DEG clusters identified according to their expression patterns across diets (clusters 138 and 245).

Gene Ontology Analysis / KEGG on Cluster 138 (LPD-associated upregulated genes)

Standard GO overrepresentation analysis with DEGs in cluster 138 identified significant enrichment of 403 overrepresented GO terms (Figure 1). Several GO terms correspond to pathways also confirmed by the KEGG analysis (Table 2). Notably, enrichment was observed for ubiquitin-dependent protein catabolic processes, lysosome activity, mitophagy and autophagy pathways (Figure 1 and Table 2). The upregulation of these genes suggests increased proteolysis and recycling cellular components under conditions of protein limitation in eggs, consistent with a molecular response to nutritional stress.

Table 2.

KEGG pathways significantly enriched in DEGs from cluster 138. Gene count refers to the number of differentially expressed genes associated with each KEGG pathway. The percentage corresponds to the proportion of DEGs assigned to each pathway relative to the total number of genes annotated in that pathway. p-values represent enrichment significance obtained from pathway overrepresentation analysis

GO term	Gene count	%	p-value
Ubiquitin-mediated proteolysis	17	4	1.8E-05
Nucleocytoplasmic transport	14	3	7.0E-04
Lysosome	11	3	1.3E-02
Mitophagy	6	2	5.1E-02
Biosynthesis of amino acids	7	2	5.4E-02
Autophagy	9	2	6.4E-02
Endocytosis	11	3	8.9E-02

Fig. 1.

GO terms overrepresented in cluster 138

Gene Ontology Analysis / KEGG on Cluster 245 (HPD-associated upregulated genes)

Significant enrichment of GO terms among the 451 DEGs of cluster 245 is displayed in Fig. 2. These GO terms are mainly associated with RNA processing, protein metabolism, and developmental pathways. Some terms are related to nervous system development, which may influence honeybee behaviour, learning and memory with a potential impact on the whole colony organisation [59, 60].

Fig. 2.

GO terms overrepresented in cluster 245

The KEGG pathways analysis confirms the findings presented above (Fig. 2; Table 3). Under nutritional treatment with a protein percentage above 10.45%, RNA processing pathways via the spliceosome and RNA polymerase are preferentially activated. The SNARE interactions in vesicular transport are involved in the regulation of docking granules and vesicles to target membranes, including the plasma membrane and membrane fusion [61]. This pathway plays a role in the transport of synthesised proteins in the endoplasmic reticulum and may be involved in developmental pathways such as the Hippo signalling pathway [62].

Table 3.

KEGG pathways significantly enriched in DEGs from cluster 245

Pathway	Gene count	%	p-value
SNARE interactions in vesicular transport	18	4	4.5E-07
RNA polymerase	5	1	8.3E-03
Hippo signalling pathway	5	1	1.3E-02
Spliceosome	5	1	8.7E-02

Protein levels in nurse bee diet influence gene expression in eggs laid by the queen: Upregulation of autophagy under LPD and enhanced RNA processing under HPD

In the following section, DEGs were categorised based on their association with either the LPD (0:1, 1:9, 1:10) or HPD (1:7, 1:5, 1:3) groups. A volcano plot was generated using a Padj < 0.05 and a cutoff of |log2FC| ≥ 2 (Fig. 3A), using HPD as a reference, meaning that genes are upregulated or downregulated under this treatment. A total of 3 genes were identified (Table 4). Among these genes, one specific gene called anaphase-promoting complex subunit 13 (LOC102653956) was downregulated, with the other 2 genes upregulated, one of which is associated with 5.8 S ribosomal DNA (LOC113219354) and the other one remains uncharacterized (LOC113218980). Figure 3B shows the broader DEG set using Padj < 0.05 and a permissive threshold of |log2FC| ≥ 0.4, resulting in 248 significantly downregulated and 287 significantly upregulated genes under HPD conditions. Applying a more stringent threshold of |log2FC| ≥ 1 reduced the number of DEGs from 535 to 153 genes (Supplementary Fig. 11).

Table 4.

FGSEA graphic representation with HPD as reference. The leading edge size in each plot represents the subset of genes within the gene set that contributes most significantly to the observed enrichment score. (Supplementary Table : sig_fgsea_res_annot_GO.csv)downregulated (<0) and upregulated (>0) genes under HPD, Padj < 0.05

Gene ID	Annotation	baseMean	log2FoldChange	Padj
LOC102653956	anaphase-promoting complex subunit 13	125.2617	-9.9846376	2.30E-15
LOC113219354	5.8S ribosomal RNA	127.2008	20.9904175	7.04E-05
LOC113218980	uncharacterised	147.1378	2.294097	1.24E-05

FGSEA was performed as a ranked gene-set analysis using the DESeq2 test statistic across all genes, rather than only the fold-change-filtered DEG set. The FGSEA plot shows enrichment of gene sets taking HPD as reference. Then, the enrichment of gene sets presents the opposite pattern under the LPD treatment (Fig. 4). Based on FGSEA (Fig. 4), it can be deduced that under LPD conditions, there is an upregulation of the ubiquitin-dependent catabolic process, lysosomes, mitophagy, and autophagy pathways. Conversely, under HPD, normal cell development through RNA processing and spliceosome activities appears to be occurring.

Fig. 3.

A:Volcano plots of the DEGs obtained by HPD and LPD clustering, using HPD as reference (Padj cutoff 0.05; Log₂FC cutoff 2). B: Volcano plots of the DEGs obtained by HPD and LPD clustering, using HPD as reference (Padj cutoff 0.05; Log₂FC cutoff 0.4). 248 genes are significantly downregulated, and 287 genes are upregulated under HPD. (See Supplementary Table 8)

Fig. 4.

FGSEA graphic representation with HPD as reference. The leading edge size in each plot represents the subset of genes within the gene set that contributes most significantly to the observed enrichment score. (Supplementary Table : sig_fgsea_res_annot_GO.csv) downregulated (< 0) and upregulated (> 0) genes under HPD, Padj < 0.05

KEGG on LPD and HPD clusters

KEGG pathway analysis revealed contrasting molecular responses associated with low-protein (LPD) and high-protein (HPD) dietary conditions in honeybee eggs.

Under LPD conditions, enriched pathways were primarily associated with ubiquitin-mediated proteolysis, autophagy, lysosome activity, endocytosis, and nucleocytoplasmic transport. Together, these pathways suggest coordinated cellular responses linked to protein recycling, intracellular degradation, and resource conservation under reduced protein availability. Notably, the ubiquitin-mediated proteolysis pathway involved genes, including anaphase-promoting complex subunit 13 (LOC102653956). Overall, these transcriptional signatures are consistent with a cellular response to nutritional stress, although their precise developmental consequences remain to be determined [63–70].

In contrast, HPD-associated genes showed enrichment of pathways related to spliceosome activity, RNA processing, MAPK signalling, and Notch signalling. These pathways are broadly associated with active cellular regulation, developmental processes, and transcriptional activity, suggesting conditions compatible with active embryonic development under higher protein availability.

KEGG analysis also indicated differential representation of genes associated with the Hippo signalling pathway across dietary conditions. Under HPD, genes annotated as putative orthologues of the Drosophila genes Ed, Ds, and 14-3-3 showed increased expression, whereas under LPD conditions, downregulation of Mad as well as core components such as Hippo and Mats was observed (Supplementary Fig. 12). These patterns suggest diet-associated modulation of growth-regulatory signalling pathways in honeybee eggs.

Discussion

The present study aimed to investigate the effects of different P: C ratios in honeybee diets on egg gene expression and associated molecular responses. The RNA-seq analysis provides insights into gene expression patterns and molecular pathways influenced by dietary compositions, shedding light on how colony nutrition may affect early developmental processes in honeybees.

One of the key findings of this study is that queen identity explained a larger proportion of transcriptional variation than nurse-mediated queen nutrition. This highlights substantial biological variability among queens, likely reflecting differences in genetic background, physiological state, and reproductive dynamics. However, despite this dominant source of variation, diet still exerted a significant and biologically meaningful effect on gene expression. Differential gene expression analysis identified 1007 DEGs across the different diets, indicating that nutritional composition plays a role in regulating transcriptional profiles in freshly laid eggs and may have implications for egg development. Distinct clustering patterns observed in the expression profiles of these DEGs under different diets (Clusters 138 and 245) highlight the importance of the P: C ratio in shaping the molecular responses during embryonic development.

Functional enrichment analysis further revealed biological processes and pathways associated with contrasting nutritional conditions. The upregulation of genes associated with protein catabolic processes, autophagy, ubiquitin-mediated proteolysis, and lysosomal pathways under LPD indicates a cellular response consistent with nutritional stress. Ubiquitin-mediated proteolysis, including genes such as anaphase-promoting complex subunit 13 (LOC102653956), may contribute to selective protein turnover and cell-cycle regulation during early development. In parallel, autophagy and lysosomal pathways may support the degradation and recycling of damaged organelles and macromolecules under protein-limited conditions. The enrichment of endocytosis and nucleocytoplasmic transport pathways under LPD further suggests that protein limitation is also associated with changes in intracellular trafficking and macromolecular exchange between the nucleus and cytoplasm. Together, these transcriptional responses are consistent with mechanisms that may help maintain cellular homeostasis and conserve resources under reduced protein availability. While such mechanisms may help sustain early development, they could also reflect resource constraints that compromise normal embryonic growth and physiological function, although the developmental consequences remain to be experimentally validated.

Conversely, under HPD conditions, enrichment of pathways related to RNA processing, spliceosome activity, MAPK signalling, Notch signalling, and developmental regulation suggests conditions compatible with active embryonic gene regulation and cellular differentiation. Such pathways are consistent with the high metabolic and transcriptional activity required during early embryogenesis, and suggest that higher protein availability supports a molecular environment conducive to normal developmental progression. MAPK signalling is involved in cell proliferation, differentiation, stress responses, and developmental regulation, whereas Notch signalling plays a central role in cell fate determination and tissue patterning. Their enrichment under HPD conditions suggests that increased protein availability may support regulatory processes involved in coordinated embryonic development.

From a mechanistic perspective, honeybee larval and queen nutrition is mediated by nurse bees through royal jelly production, which reflects both dietary inputs and colony nutritional status. Variations in nurse bee diet composition may therefore indirectly influence maternal provisioning of eggs, potentially affecting transcript abundance through nutrient-sensitive regulatory pathways. However, the physiological dynamics of nutrient allocation in actively laying honeybee queens remain incompletely understood. Although oogenesis and vitellogenin deposition are continuous processes, the extent to which queens accumulate persistent nutritional reserves specifically for egg production has not been clearly established. In our experimental design, the two-day carbohydrate-only feeding period between dietary treatments was intended primarily to standardize feeding conditions rather than to achieve a complete metabolic reset; consequently, some residual nutritional effects from previous diets cannot be entirely excluded. In insects more broadly, nutritional signalling pathways such as insulin/insulin-like signalling (IIS), TOR signalling, and vitellogenin-associated pathways have been shown to regulate reproductive physiology, oocyte provisioning, and early developmental gene expression [71–73]. While these mechanisms were not directly investigated here, they provide a plausible biological framework linking colony nutrition to egg transcriptomic profiles.

In the present study, the queen’s nutritional intake was not directly manipulated or quantified, as dietary treatments were applied to worker bees. Therefore, the observed effects should be interpreted as worker-mediated nutritional effects rather than direct maternal dietary effects. Variability in worker feeding behaviour and nutrient transfer to the queen may contribute to the observed variation and cannot be fully controlled in this experimental system. However, this indirect pathway reflects the biological reality of eusocial insects, where queen nutrition is inherently mediated by workers.

The involvement of conserved developmental signalling pathways, including Hippo, MAPK, and Notch, suggests that dietary conditions may influence regulatory networks controlling growth, cell differentiation, and stress responses. The Hippo pathway plays a fundamental role in regulating organ size and tissue growth during development [74]. In our dataset, genes associated with this pathway were differentially represented under both LPD and HPD conditions. Under HPD, putative orthologues of the Drosophila genes Ed, Ds, and 14-3-3 showed increased expression, whereas under LPD, reduced expression of Mad, as well as core pathway components such as Hippo and Mats, was observed. These patterns suggest that nutritional status may modulate growth-regulatory signalling through distinct mechanisms. Under protein-limited conditions, altered expression of core Hippo pathway components may reflect a reduction in proliferative activity aimed at conserving resources, whereas under protein-rich conditions, increased expression of upstream regulators may indicate developmental conditions more compatible with normal tissue growth and differentiation. The differential expression of genes involved in these pathways may represent an adaptive response to the dietary constraints faced by the developing embryos. However, caution is warranted as pathway enrichment does not establish causality, and further experimental validation will be required.

In Drosophila, it was demonstrated that autophagy deficiency led to altered expression patterns of downstream target genes of the Notch signalling pathway [75]. The similarity in Notch pathway modulation across different species [76], including honeybees [77], in response to varying environmental conditions, suggests the conservation of molecular mechanisms governing cellular responses to dietary changes.

Furthermore, previous studies have suggested potential interactions between autophagy and the MAPK pathway [75]. Given the well-established conservation of cellular signalling networks across species, the notable upregulation of the MAPK pathway in honeybee eggs under HPD may imply its involvement in mediating responses to dietary alterations. This suggests that the MAPK pathway may play a role in regulating processes such as cell growth, proliferation, differentiation, and responses to external stimuli in honeybee eggs exposed to the HPD.

The findings of our study also have broader implications for honeybee health and colony survival. Adequate nutrition, particularly protein intake, is crucial for honeybee development and overall colony health [1, 16]. Protein-rich pollen serves as the primary source of amino acids and nutrients for larval and adult honeybees [16]. Molecular responses to different P: C ratios in honeybee eggs may provide a useful starting point for exploring how dietary composition influences gene expression and developmental outcomes. Lipids also play a crucial role in colony health [78], and further research is needed to explore their impact on vitellogenesis, egg quality, survival, embryo development, and subsequent colony growth, for example through controlled experiments using pollen sources with contrasted lipid profiles.

Furthermore, these results suggest that a protein-deprived diet can have implications for the health and survival of honeybee colonies, ultimately affecting their overall health and survival. When nurse bees are subjected to a protein-deprived diet, their nutritional status may be compromised. Hence, the production of royal jelly by nurse bees, whether in terms of quantity, quality, or both, may influence queen nutrition and consequently the provisioning of eggs. Further research is needed to determine whether eggs laid by a poorly nourished queen may result in larvae and adult bees with reduced immunity and developmental deficiencies. This scenario can set off a vicious cycle: weaker and less productive queens produce offspring with compromised immunity and development. This, in turn, can lead to a decline in the overall health of the colony.

Conclusion

In conclusion, this present study demonstrates that the nutritional composition of nurse-mediated queen diet significantly influences gene expression patterns in freshly laid eggs, with potential implications for embryonic development. Differential expression and pathway enrichment analyses revealed that contrasting P: C ratios are associated with distinct molecular responses, including conserved developmental signalling pathways such as Hippo, MAPK, and Notch, which are involved in cellular growth, differentiation, and developmental regulation [79]. These findings contribute to our understanding of the complex interplay between diet, gene expression, and developmental outcomes in honeybees. Similar molecular mechanisms may also operate in other organisms and may help explain how nutritional conditions influence developmental responses across species [80]. Further investigations into the functional consequences of these molecular responses and their implications for honeybee health are required, with potential implications for the management and conservation of these crucial pollinators.

Authors' contributions

DF designed the study, performed experiments, analysed and interpreted data, and wrote the original draft.SI performed experiments and data curation, contributed to data analysis, and critically revised the manuscript.JG contributed to data processing, analysis, and critically revised the manuscript.PKD contributed to study conception and supervision, provided resources and funding, and critically revised the manuscript.All authors read and approved the final manuscript and agree to be accountable for their contributions and for the integrity of the work.

Funding

This research was supported by a Callaghan Innovation scholarship, and the Ministry of Business, Innovation and Employment Future Bees programme. A Publishing Bursary from the University of Otago supported the writing of this article.

Data availability

The data that support the findings of this study are available in the supplementary material of this article. All sequencing data is available at NCBI at Bioproject Accession PRJNA1471325 (https://www.ncbi.nlm.nih.gov/bioproject/1471325)

Declarations

Ethics approval and consent to participate

Honeybees (Apis mellifera) are invertebrates and are not subject to ethical approval requirements under current animal research regulations. All experimental manipulations were performed in accordance with institutional biosafety and beekeeping guidelines, and no procedures causing unnecessary suffering were applied.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.