Dopamine enhances surface glial glycolysis via acetylcholine and insulin/insulin-like growth factor signaling in the honeybee

Summary

Glycolysis plays a crucial role in neuronal homeostasis and is regulated by neuroendocrine signals, particularly dopamine, yet the underlying mechanisms remain unclear. Metabolomics showed that bromocriptine (BRC), a dopamine receptor agonist, increased glycolytic flux in forager bees, as indicated by higher glucose levels, accumulation of glycolytic intermediates, and reduced pentose-phosphate pathway metabolites. Single-cell and single-nucleus RNA-seq further indicated that acetylcholine (ACh) released by Kenyon cells may mediate dopamine-regulated glycolysis in surface glia (SG). We also observed coordinated insulin/insulin-like growth factor signaling (IIS) between cortex glia (CG) and SG. Mechanistically, co-administration of a nicotinic ACh receptor agonist with BRC attenuated the BRC-induced upregulation of SLC2A1, HK, and IGF transcripts, consistent with the proposed pathway. Together, these findings are consistent with a dual regulatory scheme in which a dopamine-ACh axis operates in concert with IIS to shape glial metabolic plasticity with cell-type specificity.

Graphical abstract

Highlights

• •Dopamine increases glycolysis in the honeybee brain, particularly in surface glia
• •Reduced acetylcholine release may mediate this dopamine-induced metabolic shift
• •Bromocriptine raises SLC2A1/HK/IGF; nAChR agonist co-administration attenuates
• •Cholinergic suppression in cortex glia is associated with increased IGF expression

Entomology; Molecular biology; Neuroscience

Introduction

Glycolysis maintains neuronal homeostasis by supporting rapid energy supply, metabolic adaptation, synaptic regulation, and neuroprotection.^1,2,3,4,5,6 It is regulated by neuroendocrine signals and is conserved across both invertebrates and vertebrates.⁷ Neurotransmitters such as dopamine are key modulators of the neuroendocrine control of glycolysis. In mice, increasing dopamine levels has been shown to elevate phosphoglycerate kinase-1 (PGK1) activity and enhance glycolysis.⁸ A Drosophila melanogaster study reported that L-DOPA administration reduces systemic glucose levels, suggesting that dopamine enhances glycolytic flux.⁹ However, the mechanisms through which dopamine regulates glycolysis remain unclear and warrant further investigation. As a metabolic regulator, insulin plays a critical role in the control of cerebral glycolysis. In Drosophila melanogaster, glial cells participate in glycolysis, with cortex glia (CGs) regulating their own glucose uptake via insulin secretion.¹⁰ In humans, insulin facilitates the transport of glucose across the blood-brain barrier into specific glial cells, including astrocytes (ASTs), microglia, and oligodendrocytes, which then perform glycolysis.^11,12,13 Additionally, a Drosophila melanogaster study demonstrated that dopamine modulates insulin-like peptide (ILP) release through the activation of insulin-producing cells (IPCs).¹⁴ These findings collectively suggest that the insulin/insulin-like growth factor (IIS) signaling pathway may be involved in dopamine-mediated glycolytic regulation. However, direct experimental evidence for this link remains lacking.

The honeybee (Apis mellifera), a eusocial insect, serves as an ideal model organism for investigating neuroendocrine regulation. Its broad biological significance—as a key pollinator essential to ecological balance and agriculture, and as a species exhibiting highly organized social behavior—makes it a widely used model in entomology, behavioral biology, and evolutionary biology.^15,16,17,18 This utility is further enhanced by its genomic characteristics: despite limited divergence in gene number compared with humans, its smaller genome size reduces sequencing costs while retaining social complexity.¹⁹ Foraging, an energy-intensive social behavior in honeybees, is associated with elevated brain dopamine levels.²⁰ Thus, forager bees provide a unique model for dissecting socially modulated glycolysis regulated by dopamine—a phenomenon poorly characterized in solitary insects such as Drosophila melanogaster—while offering greater experimental efficiency than mammalian models.

Glycolysis in the honeybee not only sustains basal energy requirements but is also mechanistically linked to physiological behaviors such as reproduction and aggression.^21,22 This process is regulated by signals, including insulin and neurotransmitters such as octopamine.^23,24 As a primary “fight-or-flight” neuromodulator in insects, octopamine regulates aggression and stress by mobilizing energy reserves.²⁵ In contrast, dopamine often plays functionally antagonistic or distinct modulatory roles. For instance, in avoidance learning paradigms, octopamine enhances punishment perception, whereas dopamine reduces it.²⁶ This functional antagonism is a key principle of biogenic amine signaling in insects. Despite these behavioral contrasts, whether dopamine directly modulates glycolysis in the honeybee remains unclear. Given that neurotransmitter systems frequently couple to cellular metabolism, one plausible route for such modulation could involve crosstalk with the cholinergic system. Anatomical observations in the honeybee suggest such points of interaction, including the co-localization of acetylcholine (ACh) transporters and dopamine receptors in mushroom bodies,²⁷ as well as the reciprocal expression of their receptors in both neuron types.²⁸ In the vertebrate striatum, dopamine directly suppresses the firing of cholinergic interneurons and reduces ACh release through the activation of D2-like receptors (D2Rs).²⁹ Conversely, ACh can modulate dopamine release via nicotinic acetylcholine receptors (nAChRs), with effects depending on receptor subtype and neural circuit context.³⁰ Therefore, we propose that a dopamine-ACh axis may influence glial metabolic programs.

In this study, we investigated the dopamine’s regulation of glycolysis in the honeybee using targeted metabolomics and single-cell transcriptomics. Our analyses revealed dopamine-induced enhancement of glycolytic flux in brain cells, with ACh acting as a pivotal mediator. These findings suggest cell type-specific neuroendocrine mechanisms that integrate neurotransmitter signaling with metabolic pathways in eusocial insects and establish the honeybee as a valuable model for studying neuroendocrine regulation.

Results

Monoaminergic receptor activation enhances glycolysis in forager bees

To investigate the regulatory role of dopamine in energy metabolism, bromocriptine (BRC) was administered to forager bees, with ddH₂O serving as the control (Figure 1A). Successful drug delivery was first confirmed by untargeted metabolomics, which showed significantly higher BRC levels in the treated group than in controls (Figures S1A and S1B). Quantitative PCR (qPCR) validation indicated the activation of dopamine signaling, as evidenced by the upregulation of Dop3 (encoding D2R) and downregulation of protein kinase A catalytic subunit 1 (PKA-c1), a canonical downstream effector whose suppression reflects the D2R-mediated inhibition of cAMP signaling (Figures 1B and 1C).

Figure 1.

Dopamine D2 receptor activation enhances glycolysis and alters metabolic pathways in forager bees

(A) Schematic of the experimental workflow, including BRC and control (ddH₂O) treatments of forager bees, followed by qPCR and targeted metabolomics analyses of brain tissue. BRC, Bromocriptine. qPCR, Quantitative PCR.

(B and C) Relative mRNA expression of (B) Dop3 and (C) PKA-c1 in whole brains from BRC- and ddH₂O-treated bees, as measured by qPCR.

(D) Heatmap shows hierarchical clustering of differentially abundant metabolites. Rows represent individual biological replicates (B1–5 for BRC, H1–6 for ddH₂O) and columns represent metabolites. The color scale indicates the z-score-normalized relative abundance of each metabolite (red, high; blue, low).

(E) Volcano plot displays metabolic changes between BRC and ddH₂O groups. The x axis represents the log₂ fold change, and the y axis represents the –log₁₀ of the FDR-adjusted p-value. Significantly upregulated (red) and downregulated (blue) metabolites (p < 0.05, FDR-adjusted) are highlighted.

(F) Pathway enrichment analysis of the differential metabolites. The x axis indicates the enrichment ratio. Upregulated pathways (e.g., glycolysis) are shown in red, and downregulated pathways (e.g., pentose phosphate pathway) are shown in blue.

(G) D-glucose concentration in whole brains from BRC- and ddH₂O-treated bees.

(H) Ratios of key glycolytic intermediates to glucose-6-phosphate (G6P), used as a proxy for glycolytic flux. F6P, fructose-6-phosphate; DHAP, dihydroxyacetone phosphate; 3-PG, 3-phosphoglyceric acid. Data in (B, C) are presented as mean ± SEM. Data in (G, H) are presented as boxplots showing the median, interquartile range, and data distribution. Statistical significance was determined using a two-tailed Student’s t test. ∗p < 0.05, ∗∗p < 0.01.

Targeted metabolomics revealed marked metabolic shifts. Principal component analysis (PCA) showed a clear separation between groups (Figure S1C), and hierarchical clustering highlighted the accumulation of glycolytic intermediates—fructose-6-phosphate, 3-phosphoglyceric acid, and pyruvate—alongside suppression of pentose phosphate pathway (PPP) metabolites, including 6-phosphogluconic acid and D-ribulose 5-phosphate (Figure 1D). Eight metabolites were significantly altered (FDR-adjusted p < 0.05). Specifically, D-glucose, pyruvic acid, and D-ribose-5-phosphate were upregulated, whereas fructose-1-phosphate, dihydroxyacetone phosphate, 6-phosphogluconic acid, glucose-6-phosphate, and D-ribulose-5-phosphate were significantly downregulated (Figure 1E). Notably, upregulated metabolites were predominantly enriched in the glycolysis pathway (FDR-adjusted p = 0.03), whereas downregulated metabolites showed strong enrichment in the PPP (FDR-adjusted p = 2.7 × 10⁻⁴; Figure 1F).

To further assess glycolytic activation, we evaluated glucose levels and ratios of glycolytic intermediates relative to glucose-6-phosphate (G6P). BRC-treated bees exhibited an increase in glucose levels (log₂FC = 5.60, p = 0.002; Figure 1G), along with significant increases in F6P/G6P (log₂FC = 2.13, p = 0.015), 3-PG/G6P (log₂FC = 2.65, p = 0.007), and pyruvate/G6P (log₂FC = 2.31, p = 0.007), collectively indicating accelerated glycolytic flux (Figure 1H).

Together, these findings demonstrate that the activation of monoaminergic receptors enhances glycolysis in forager bees.

Kenyon cells mediate dopamine-dependent glycolytic regulation through acetylcholine secretion

To elucidate the cellular basis of dopamine-induced metabolic alterations, single-cell RNA sequencing (scRNA-seq) data from 43,378 forager-bee brain cells (GSE184507) were analyzed. Clustering identified 43 distinct neuronal and glial populations (Figure 2A). Among neuronal subsets, we primarily detected Kenyon cells (KCs), olfactory-lobe cells (OLCs), and olfactory projection neurons (OPNs). KCs showed high expression of the vesicular acetylcholine transporter (VAChT), a marker of ACh secretion (Figures 2B, 2C, and S1D). A joint kernel density plot, supported by individual feature plots, revealed that Dop3 expression was strongly enriched in KCs, co-expressing with PKA-c1, PLCε, and VAChT (Figures 2D and 2E). Enrichment analysis of KC marker genes indicated a significant association with pathways related to neurotransmitter transport and G-protein-coupled receptor signaling (Figure 2F). These observations suggest that KCs may serve as key mediators of dopamine signaling while concurrently secreting ACh.

Figure 2.

Single-cell analysis identifies Kenyon cells as a primary hub for dopamine and acetylcholine-mediated signaling

(A) Uniform manifold approximation and projection (UMAP) projection of all sequenced cells from the forager bee brain, colored by unbiased Seurat clusters.

(B) Dot plot shows the expression of canonical marker genes across the Seurat clusters, which was used for cell type annotation. The y axis represents the clusters shown in (A). Dot size corresponds to the percentage of cells in a cluster expressing the gene, while color intensity reflects the average normalized expression level. KC, Kenyon cells; OPNs, olfactory projection neurons; OLCs, olfactory-lobe cells; hml, hemocytes; epi, epithelial cells.

(C) UMAP projection of all cells, colored by the final cell type annotations derived from the marker gene expression shown in (B). “Vmat+KC” represents a subset of Kenyon cells that also express the vesicular monoamine transporter (Vmat), suggesting a monoaminergic phenotype. KC, Kenyon cells; OPNs, olfactory projection neurons; OLCs, olfactory-lobe cells.

(D) Co-expression analysis of the dopamine receptor Dop3, KC markers (PLCε and PKA-c1), and VAChT. Left: joint kernel density plot shows regions of high co-expression. The color gradient reflects normalized density from low (blue) to high (red). Right: individual feature plots show the normalized expression pattern of each gene across the UMAP. The color gradient reflects normalized gene expression from low (blue) to high (red).

(E) Heatmap shows the Pearson correlation coefficient between the expression of Dop3, VAChT, PKA-c1, and PLCε within the identified KC population, where darker shades indicate a stronger correlation, with blue and red representing negative and positive correlations, respectively.

(F) Gene Ontology (GO) pathway enrichment analysis for genes highly expressed in the KC clusters. Nodes represent enriched biological pathways grouped by related functions (colors).

An acetylcholine-responsive insulin/insulin-like growth factor signaling signaling axis connects cortex and surface glia

To explore inter-glial interactions, we re-clustered the glial populations and identified four primary subtypes: CG, surface glia (SG), astrocyte-like glia (AST), and ensheathing glia (EG) (Figures 3A, 3B, S1E, and S1F). We then assessed their responsiveness to ACh. The signature score for the ACh-receptor gene set—comprising nAChRα1, nAChRα2, nAChRα3, and nAChRα6—was high in both CG and SG cells (Figure 3C). This composite score integrates relative gene expression and enrichment metrics to quantify receptor-set activity.

Figure 3.

scRNA-seq reveals a potential acetylcholine-driven IIS signaling axis from cortex glia (CG) to surface glia (SG)

(A) Uniform manifold approximation and projection (UMAP) projection of all identified glial cell subtypes from the scRNA-seq data.

(B) Dot plot shows the expression of key marker genes used to annotate the glial cell subtypes. Dot size corresponds to the percentage of cells in a cluster expressing the gene, while color intensity reflects the average normalized expression level. AST, astrocyte-like glia; CG, cortex glia; EG, ensheathing glia; SG, surface glia.

(C and D) UMAP plots show the density of signature scores for (C) acetylcholine receptor (AChR) genes (including nAChRα1, nAChRα2, nAChRα3, and nAChRα6) and (D) glycolysis-related genes (including LOC409424, LOC551005, LOC408818, and LOC724724) across all cells. The color scale indicates the signature score density, from low (blue) to high (yellow).

(E) Co-expression analysis of CG markers (wrapper, zyd) and IIS signaling ligands (IGF, ILP-2). Left: joint kernel density plot shows co-expression patterns. The color gradient reflects normalized density from low (blue) to high (red). Right: Individual feature plots show the normalized expression of each gene. The color gradient reflects normalized gene expression from low (blue) to high (red).

(F) Co-expression analysis of SG markers (Tret1, Mdr49) and IIS signaling receptors (ILPR, Imp). Left: Joint kernel density plot. The color gradient reflects normalized density from low (blue) to high (red). Right: Individual feature plots for each gene. The color gradient reflects normalized gene expression from low (blue) to high (red).

Spatial expression analysis of key metabolic and signaling pathways revealed that glycolysis-related genes, including SLC2A1 and HK, were predominantly enriched in SG cells (Figures 3D and S1G). In contrast, insulin/insulin-like signaling (IIS) ligands (IGF and ILP-2) were mainly expressed in CG cells (Figure 3E), whereas IIS receptors (ILPRs) and binding proteins (Imp) were enriched in SG cells (Figure 3F). Gene expression within the glial population showed strong correlations (Figure S1H). This compartmentalized pattern—ligand expression in CG and receptor expression in SG—suggests a directional IIS-mediated signaling axis between these two glial subtypes.

Activation of monoaminergic receptors suppresses acetylcholine release in Kenyon cells

To further elucidate the roles of KCs and glial cells in glycolysis enhancement mediated by monoaminergic receptor activation, we performed single-nucleus RNA sequencing (snRNA-seq). This approach enabled comprehensive and unbiased profiling of all cell types within the forager bee brain, comparing samples from BRC-treated bees with a ddH₂O control group. After quality control, 34,232 high-quality nuclei were retained. Among glial cells, four distinct subtypes were identified: SG, CG, ensheathing glia (EG), and astrocyte-like glia (AST) (Figures 4A and 4B). Re-clustering of neuronal populations identified KCs, OLCs, and OPNs (Figures 4C and S2A). Cell proportion analysis revealed significant increases in the numbers of OPNs, CG, SG, and KCs following monoaminergic receptor activation, while the proportion of AST cells significantly decreased (Figure 4D). A joint kernel density plot, supported by individual feature plots, showed Dop3 co-enrichment with PKA-c1, PLCε, and VAChT, which was markedly upregulated in BRC-treated bees (Figures 4E and S2B). Pathway enrichment analysis of downregulated genes in KCs indicated broad suppression of neuronal activity, as evidenced by the downregulation of pathways related to neurotransmitter transport and ion channel activity (Figure 4F), suggesting reduced synaptic transmission. Guided by the “ion channel activity” findings, we further examined voltage-dependent calcium channel (VDCC) genes, which are essential for neurotransmitter release. Consistent with this, key VDCC genes were significantly downregulated in KCs from BRC-treated bees (Figure 4G). At the single-cell level, VAChT expression in KCs was reduced (log₂FC = −0.57, p < 0.05), a finding validated by qPCR (p < 0.05) (Figures 4H and 4I). Enzyme-linked immunosorbent assay (ELISA) confirmed a significant reduction in ACh secretion in BRC-treated bees (p = 2.4 × 10⁻⁵) (Figure 4J). These results demonstrate that the activation of monoaminergic receptors downregulates neuronal activity pathways in KCs, leading to reduced ACh release.

Figure 4.

Single-nucleus RNA-seq reveals that the activation of monoaminergic receptors downregulates acetylcholine-related pathways in Kenyon cells

(A) Dot plot shows the expression of key marker genes across all identified cell clusters from both BRC and ddH₂O-treated groups. Dot size corresponds to the percentage of cells in a cluster expressing the gene, while color intensity reflects the average normalized expression level.

(B) Uniform manifold approximation and projection (UMAP) projection of all sequenced cells from both treatment groups, colored by major cell type. AST, astrocyte-like glia; CG, cortex glia; EG, ensheathing glia; SG, surface glia.

(C) UMAP projection of the neuronal subset, colored by identified neuronal subtypes. KC, Kenyon cells; OLCs, olfactory-lobe cells; OPNs, olfactory projection neurons.

(D) Differential analysis of cell type proportions between BRC and ddH₂O groups. The x axis shows the observed log₂ fold difference (Obs log2FD), where negative values indicate a higher proportion in the BRC group. Error bars represent the 95% confidence interval. Significant changes (p < 0.05 FDR-adjusted) are highlighted. n.s., not significant.

(E) Co-expression analysis of KC markers (PLCε, PKA-c1), the dopamine receptor Dop3, and VAChT in neuronal cells. Left: joint kernel density plot where the color gradient reflects normalized density from low (blue) to high (red). Right: individual feature plots showing the normalized expression level of each gene.

(F) Gene Ontology (GO) pathway enrichment analysis of genes significantly downregulated in KCs of the BRC-treated group. Nodes represent enriched pathways, grouped by functional modules (colors).

(G) For voltage-dependent calcium channel genes in KCs, violin plots (left) show the distribution of normalized expression in cells with non-zero expression (cells with expression = 0 were omitted from the violin visualization). For the genes LOC408430 and LOC408841, expression log₂ fold changes (BRC vs. ddH₂O) were −0.48 and −0.73, respectively, as estimated by MAST using all KC cells, including both expressing and non-expressing (zero-expression) cells, with Benjamini-Hochberg correction for multiple testing. The accompanying bar plots (right) display the percentage of expressing cells in each condition, with log₂ fold changes in the proportion of expressing cells of −0.25 for LOC408430 and -1.16 for LOC408841, as estimated using the scProportionTest R package.

(H) For VAChT in KCs, the violin plot (left) shows the distribution of normalized expression in cells with non-zero expression (cells with expression = 0 were omitted from the violin visualization). The expression log₂ fold change (BRC vs. ddH₂O) for VAChT was −0.57, as estimated by MAST using all KC cells, including both expressing and non-expressing (zero-expression) cells, with Benjamini-Hochberg FDR correction. The accompanying bar plot (right) displays the percentage of VAChT-expressing cells in each condition, with a log₂ fold change in the proportion of expressing cells of −1.28, as estimated using the scProportionTest R package.

(I) Relative mRNA expression of VAChT in whole brains, measured by qPCR.

(J) Acetylcholine (ACh) concentration in whole brains, measured by ELISA. For violin plots (G and H), significance was determined using MAST with Benjamini-Hochberg FDR correction. For bar plots (G and H), significance was calculated using the scProportionTest R package. For qPCR and ELISA data (I and J), significance was determined using a two-tailed Student’s t test. Significance levels: ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

Dopamine enhances glycolytic gene expression in surface glia by suppressing acetylcholine signaling

To investigate signaling mechanisms in SG, we first assessed their responsiveness to cholinergic input. A joint kernel density plot, supported by individual feature plots, showed partial co-expression of nAChRα6 with SG markers Tret1 and Mdr49 (Figure 5A). Consistent with the dopamine-mediated inhibition of cholinergic signaling, snRNA-seq analysis revealed significant downregulation of nAChRα6 in SG cells following BRC treatment, accompanied by the upregulation of the glycolytic genes SLC2A1 and HK (Figures 5B–5D). To determine whether this effect was mediated by cholinergic signaling, we performed qPCR across three experimental groups: control (ddH₂O), BRC, and co-treatment (BRC + imidacloprid [IMI]). As a positive control for nAChR activation, imidacloprid treatment alone significantly upregulated its target receptor nAChRα1 compared with controls (Figure S2C). qPCR of the main experimental groups confirmed the downregulation of nAChRα6 and upregulation of SLC2A1 and HK in the BRC group relative to controls. Importantly, co-treatment with imidacloprid attenuated the BRC-induced upregulation of SLC2A1 and HK, indicating that this effect is mediated through suppressed nAChR signaling (Figure 5E). Spatial validation using RNA fluorescence in situ hybridization (FISH) further confirmed reduced nAChRα6 expression and increased SLC2A1 and HK signals in the SG layer of BRC-treated bees (Figures 5F, 5G, and S2D). Glycolytic gene scores were significantly elevated in SG from BRC-treated bees (Figure S3A). Similarly, other carbohydrate kinase genes showed increased expression (Figurer S3B and S3C). Collectively, these findings indicate that the activation of monoaminergic receptors enhances glycolytic activity in SG as a result of dopamine-induced suppression of ACh signaling.

Figure 5.

Monoaminergic receptors activation enhances glycolysis in surface glia (SG) by suppressing acetylcholine signaling

(A) Uniform manifold approximation and projection (UMAP) projections of glial cell subsets shows co-expression of SG markers Tret1 and Mdr49 with the nicotinic acetylcholine receptor subunit nAChRα6. Left: joint kernel density plot of all three genes across both BRC and ddH₂O-treated groups. The color gradient reflects normalized density from low (blue) to high (red). Right: individual feature plots show the normalized expression of each gene. The color gradient reflects normalized gene expression from low (blue) to high (red).

(B) For nAChRα6 in SG, the violin plot (left) shows the distribution of normalized expression in cells with non-zero expression (cells with expression = 0 were omitted from the violin visualization). The expression log₂ fold change (BRC vs. ddH₂O) for nAChRα6 was −4.23, as estimated by MAST using all SG cells, including both expressing and non-expressing (zero-expression) cells, with Benjamini-Hochberg FDR correction. The accompanying bar plot (right) displays the percentage of nAChRα6-expressing cells in each condition, with a log₂ fold change in the proportion of expressing cells of −2.3, as estimated using the scProportionTest R package.

(C) For SLC2A1 in SG, the violin plot (left) shows the distribution of normalized expression in cells with non-zero expression (cells with expression = 0 were omitted from the violin visualization). The expression log₂ fold change (BRC vs. ddH₂O) for SLC2A1 was 1.19, as estimated by MAST using all SG cells, including both expressing and non-expressing (zero-expression) cells, with Benjamini-Hochberg FDR correction. The accompanying bar plot (right) displays the percentage of SLC2A1-expressing cells in each condition, with a log₂ fold change in the proportion of expressing cells of −2.30, as estimated using the scProportionTest R package.

(D) For HK in SG, the violin plot (left) shows the distribution of normalized expression in cells with non-zero expression (cells with expression = 0 were omitted from the violin visualization). The expression log₂ fold change (BRC vs. ddH₂O) for HK was 0.77, as estimated by MAST using all SG cells, including both expressing and non-expressing (zero-expression) cells, with Benjamini-Hochberg FDR correction. The accompanying bar plot (right) displays the percentage of HK-expressing cells in each condition with no significant differences, as estimated using the scProportionTest R package.

(E) qPCR validation and rescue experiments for nAChRα6, SLC2A1, and HK expression in whole brains. Bars represent relative mRNA expression in control (ddH₂O), bromocriptine (BRC)-treated, and co-treatment (BRC+IMI) groups, with the numbers above the bars indicating the group means.

(F) Representative RNA FISH images of the SG layer in brains from BRC- (top) and ddH₂O-treated (bottom) bees. Colors are as follows: DAPI (nuclei, blue), Tret1 (red), nAChRα6 (green), SLC2A1 (white), and HK (yellow). Scale bars, 200 μm (main) and 50 μm (inset).

(G) Boxplots show the quantification of fluorescence intensity for each gene from the FISH images in (F), comparing BRC and ddH₂O groups.

For violin plots (B–D), significance was determined using MAST with Benjamini-Hochberg FDR correction. For bar plots showing cell percentages (B–D), significance was calculated using scProportionTest. For qPCR and FISH quantification data (E and G), significance was determined using a two-tailed Student’s t test. Significance levels: ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001; n.s., not significant.

Suppression of cholinergic signaling in cortex glia associates with increased IGF expression

To explore the role of the insulin/insulin-like growth factor signaling (IIS) pathway in monoaminergic receptor-induced enhancement of glycolytic activity, we examined the expression of IIS-related genes in CG and SG using snRNA-seq data. As expected, nAChR expression (AChR_score) was enriched in both CG and SG cells (Figure S3D). We observed a spatially segregated expression pattern of IIS components: the ligands IGF and ILP-2 were predominantly expressed in CG cells, together with nAChRα6, whereas their receptors (ILPRs) and binding proteins (Imp) were enriched in SG cells (Figures 6A and 6B). Expression levels of these components were significantly correlated, suggesting a functional signaling axis between CG and SG (Figure S3E). Differential expression analysis revealed that BRC treatment led to significant downregulation of nAChRα6 and concurrent upregulation of IGF in CG cells (Figures 6C and 6D). qPCR validation confirmed IGF upregulation in the BRC group and showed that co-treatment with imidacloprid reduced IGF expression relative to BRC alone (Figure 6E). Collectively, these findings suggest that the dopamine-induced suppression of cholinergic signaling enhances IGF expression in CG cells, supporting increased IIS signaling from CG to SG.

Figure 6.

Suppression of acetylcholine signaling in cortex glia (CG) enhances the expression of IGF

(A) Co-expression analysis of CG markers (zyd, wrapper), the nAChR subunit nAChRα6, and IIS ligands (IGF and ILP-2) in the glial cell population. The left panel shows the joint kernel density plot, while the right panels show individual feature plots for each gene.

(B) Co-expression analysis of SG markers (Tret1, Mdr49) and IIS receptors (ILPR, Imp) in glial cells, with the joint kernel density plot on the left and individual feature plots on the right.

(C) For nAChRα6 in CG, the violin plot (left) shows the distribution of normalized expression in cells with non-zero expression (cells with expression = 0 were omitted from the violin visualization). The expression log₂ fold change (BRC vs. ddH₂O) for nAChRα6 was −0.77, as estimated by MAST using all CG cells, including both expressing and non-expressing (zero-expression) cells, with Benjamini-Hochberg FDR correction. The accompanying bar plot (right) displays the percentage of nAChRα6-expressing cells in each condition, with a log₂ fold change in the proportion of expressing cells of −0.51, as estimated using the scProportionTest R package.

(D) For IGF in CG, the violin plot (left) shows the distribution of normalized expression in cells with non-zero expression (cells with expression = 0 were omitted from the violin visualization). The expression log₂ fold change (BRC vs. ddH₂O) for IGF was 1.01, as estimated by MAST using all CG cells, including both expressing and non-expressing (zero-expression) cells, with Benjamini-Hochberg FDR correction. The accompanying bar plot (right) displays the percentage of IGF-expressing cells in each condition with no significant differences, as estimated using the scProportionTest R package.

(E) qPCR analysis of relative IGF mRNA expression in whole brains from control (ddH₂O), BRC-treated, and co-treatment (BRC+IMI) groups. For violin plots (C, D), significance was determined using MAST with Benjamini-Hochberg FDR correction. For bar plots (C, D), significance was calculated using scProportionTest. For qPCR data (E), significance between groups was determined using pairwise two-tailed Student’s t-tests. Significance levels: ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

Discussion

This study used metabolomics, single-cell RNA sequencing (scRNA-seq), and single-nucleus RNA sequencing (snRNA-seq) to investigate the dopamine regulation of glial glycolysis. Our findings support a dual regulatory mechanism in which the dopamine-ACh axis coordinates with IIS to orchestrate metabolic plasticity in glia with cell type-specific precision.

Further evidence supports the role of KCs as primary targets of dopamine signaling and the main source of ACh release. Dopaminergic projections have been shown to overlap with the presynaptic zones of KCs in the honeybee brain,³¹ and dopamine receptor expression in these cells has been confirmed by FISH.³² In Drosophila melanogaster, ACh secretion from KCs has also been reported.^33,34 Additionally, FISH has verified the expression of the VAChT gene in KCs, further supporting their cholinergic nature.²⁷ These observations suggest that KCs are both recipients of dopaminergic input and sources of ACh secretion.

Reduction in ACh release appears to represent a key regulatory node through which the activation of monoaminergic receptors accelerates glycolysis. Prior studies have described a reciprocal interaction between dopamine and ACh, in which dopamine suppresses presynaptic ACh release via D2-like receptors (D2Rs) and, conversely, ACh can modulate dopamine signaling.^{31,35,36,37,38,39} Our data build on this paradigm by showing that the activation of monoaminergic receptors in KCs inhibits ACh release through the downregulation of VDCCs. This cascade likely lowers intracellular calcium levels, diminishes cellular excitability, and ultimately suppresses ACh secretion.^{40,41,42,43,44,45} However, the regulatory role of ACh in glycolysis has remained poorly understood, with limited experimental evidence—primarily observations that elevated ACh concentrations suppress glycolytic activity in the murine cerebral cortex.⁴⁶ Our results extend this understanding by demonstrating that reduced ACh secretion enhances glycolytic flux in SG, thereby identifying ACh as a mediator of dopamine-regulated glycolysis.

The regulation of cerebral glycolysis by insulin has been extensively characterized across species. In insects, ILPs are synthesized and secreted by IPCs in the brain, exerting systemic metabolic effects.^47,48,49 In Drosophila melanogaster, these IPCs are regulated by neurotransmitters such as ACh and modulate both insulin secretion and glucose homeostasis.^10,50,51,52 In the honeybee, our results indicate that ACh-responsive CG express ILP/IGF ligands, which act on SG cells to promote glycolytic activity—a mechanism distinct from the self-regulatory IPCs of solitary Drosophila. This ACh-mediated regulation of insulin secretion is evolutionarily conserved: in humans, ACh enhances glucose-stimulated insulin secretion by activating protein kinase C and phospholipid-derived secondary messengers.^53,54

While our findings provide insight into dopamine’s downstream regulation of ACh and glycolysis, they also highlight several directions for future research. Notably, transcriptional heterogeneity was observed among major cell types, including KCs and astrocytes, likely reflecting functional subpopulations—a phenomenon consistent with findings in other insect models.^55,56 Although detailed subtype classification was beyond the scope of this study, identifying their specific molecular identities and roles represents a promising avenue for future research. Additionally, although BRC exhibits higher affinity for D2Rs than for D1Rs or 5-HT receptors,⁵⁷ it can act as a D1R antagonist at higher concentrations.⁵⁸ Our observation of PKA-c1 downregulation, however, is consistent with predominant D2R-mediated signaling, as D1R and most 5-HT receptor pathways typically converge on PKA-c1 upregulation.^59,60,61 While this evidence implicates D2R as the primary mediator, a full delineation of upstream receptor contributions remains beyond the present scope, which focused on the downstream ACh-glycolysis pathway. Future studies are warranted to clarify the potential roles of D1R and serotonin receptor signaling in this process. Complementary behavioral evidence also supports a functional role for dopaminergic signaling in foraging-related activity. Studies using dopamine receptor antagonists, such as fluphenazine, have shown impaired foraging motivation and associative learning.^20,26 However, these behavioral studies did not examine the underlying metabolic mechanisms. Whether the observed behavioral inhibition involves the dopamine-ACh-glycolysis axis identified here remains unexplored and represents an important knowledge gap for future research.

Limitations of the study

Due to the lack of specific antibodies available for the honeybee, changes in key proteins involved in this dual regulatory mechanism remain unidentified. For example, whether CG regulates SG through upregulated IIS and downregulation of VDCCs requires further validation. In addition, because of the challenges associated with capturing electrical signal alterations in KCs, our study did not include neural electrophysiological experiments, such as the patch-clamp technique, which would further validate changes in neuronal excitability.

In conclusion, our findings suggest a dual mechanism of neuroendocrine regulation in the honeybee, addressing a critical gap in understanding how dopamine modulates glycolysis in social insects.

Resource availability

Lead contact

Further information and requests for resources or reagents should be directed to and will be fulfilled by the Lead Contact, Xingan Li (lxingan@sina.com).

Materials availability

This study did not generate new unique reagents.

Data and code availability

Processed and raw snRNA-seq data have been deposited in GEO under accession number GSE295726. The published scRNA-seq dataset is available under GSE184507. The DOI is listed in the key resources table. Microscopy data reported in this article will be shared by the lead contact upon request.

All analysis code and environment files are available on Figshare (DOI: https://doi.org/10.6084/m9.figshare.28846307). Any additional information required to reanalyze the data reported in this article is available from the lead contact upon request.

Acknowledgments

This work was supported by the National Science Foundation of China (No. 31350006), the Jilin Science and Technology Program (No. 20210506044 ZP), and the Natural Science Foundation of Jilin Province, China (20240101272JC).

Author contributions

Conceptualization, methodology, and funding acquisition, X.L. and P.C.; investigation and writing - original draft, M.Z., Y.Z., and H.Z.; visualization, W.F. and A.T.; resources, data curation, validation, and writing - review and editing, M.X., J.L., and Q.N. All co-authors have read and approved the final version of the article.

Declaration of interests

The authors declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Biological samples
Honeybee (Apis mellifera carnica)	Jilin Academy of Apiculture Science	–
Chemicals, peptides, and recombinant proteins
Bromocriptine mesylate tablets (2.5 mg)	Gedeon Richter Plc.	–
Imidacloprid	Macklin	A859444
Fixative solution	Servicebio	Cat# G1113
DAPI-containing antifade mounting medium	Absin	Cat# 9235
Bovine Serum Albumin (BSA)	NEB	Cat# B9000S
Phosphate Buffered Saline (PBS)	Thermo Fisher	Cat# 10010023
Trizol Reagent	Invitrogen	Cat# 15596026
Chloroform	Sigma-Aldrich	Cat# C2432
Anhydrous ethanol	Sigma-Aldrich	Cat# E7023
DEPC-treated water	Thermo Fisher	Cat# AM9920
Methanol (LC-MS Grade)	Fisher Scientific	Cat# A456-500
Formic acid (LC-MS Grade)	Fisher Scientific	CAS 64-18-6
Ammonium acetate (LC-MS Grade)	Fisher Scientific	CAS 631-61-8
3-Nitrophenylhydrazine (3-NPH)	Sigma-Aldrich	CAS 636-95-3
EDC hydrochloride	Sigma-Aldrich	25952-53-8
Critical commercial assays
ELISA kit	COIBO BIO	CB10243-In
RNA extraction kit	SangonBiotech	Cat# B511321
cDNA reverse transcription kit	BBI	Cat# B639251
qPCR reagent kit	Promega	Cat# A6001
Single Cell 3′ Library & Gel Bead Kit V3.1	10x Genomics	Cat# 1000121
Chromium Single Cell G Chip Kit	10x Genomics	Cat# 1000120
Deposited data
snRNA-seq	This paper	GSE295726
Published scRNA-seq data	GEO	GSE184507
Oligonucleotides
Primers for qRT-PCR, See Table S2	–	–
Software and algorithms
Seurat (v5.1.0)	Satija Lab62	RRID:SCR_016341; v5.1.0
scDblFinder (v1.11.4)	Lun et al.63	–
CellRanger (v7.0.1)	10x Genomics	–
Cytoscape (v3.10.3)	Cytoscape Consortium64	–
ClueGO (v2.5.9)	Bindea et al.65	–
scProportionTest	Squair et al.66	–
Nebulosa (v0.99.92)	Alquicira et al.67	–
RStudio (Posit)	https://posit.co/products/open-source/rstudio/	–
MassLynx	Waters	–
Compound Discoverer (v3.0)	Thermo Fisher	–
CellProfiler	Broad Institute	–
Code for analysis	This paper	10.6084/m9.figshare.28846307

Experimental model and study participant details

Honeybees and brain sample collection

Samples used in this study were obtained from honeybees raised at the Jilin Academy of Apiculture Science (Jilin City, Jilin Province, China). The colonies were maintained outdoors under natural environmental conditions, and all specimens were from a single species, Apis mellifera carnica. Forager bees were collected at hive entrances while carrying pollen baskets and returning to the nest. All brain samples were collected during daytime hours (9:00–16:00).

Method details

Drug treatments

Drug treatments were administered to honeybee colonies (approximately 15,000 bees per colony) via uniform spray application—a method previously used for colony-wide pharmacological exposure in eusocial insects.⁶⁸ Bromocriptine (Gedeon Richter Plc.), a dopamine receptor agonist, was used at a dosage extrapolated from the maximum recommended human therapeutic dose, scaled to the average worker bee body weight (0.1 g) relative to a 60-kg human.⁶⁹ This corresponded to an approximate exposure of 2 μg per bee (30 mg total per colony). Imidacloprid (Macklin, A859444), a nicotinic acetylcholine receptor (nAChR) agonist, was administered at a dose of 30 ng per bee (0.45 mg total per colony). This dosage was selected based on prior studies demonstrating its effects on forager gene expression and is well below the reported LD₅₀ for honeybees.⁶⁸ The main experiment included three treatment groups observed over five consecutive days. The first group, the Bromocriptine (BRC) group, received a daily treatment of 12 bromocriptine tablets (2.5 mg each) dissolved in 100 mL of ddH₂O. The Co-treatment (BRC+IMI) group followed a staged regimen: bees were treated with bromocriptine alone for the first two days, followed by three days of combined treatment with bromocriptine (12 tablets) and imidacloprid (0.45 mg) dissolved together in 100 mL of ddH₂O. The Control group received a daily treatment of 100 mL of ddH₂O alone. To validate the efficacy and transcriptional effects of the imidacloprid treatment, a separate Imidacloprid-only group was included. This group was treated with imidacloprid alone (0.45 mg in 100 mL of ddH₂O) for three consecutive days.

For all groups, solutions were sprayed evenly across the hive once daily under standardized outdoor conditions to ensure consistent exposure.⁶⁸

Brain dissection, nuclei isolation

Honeybees were anesthetized on ice in a dissection dish and washed twice with ethanol and phosphate-buffered saline (PBS; Thermo Fisher, Cat# 10010023). Brains were dissected in PBS on ice under a stereomicroscope (OLYMPUS EP50). The surrounding tracheae, ocelli, and compound eyes (absent in drones) were carefully removed, and the brains were rinsed with 1 mL of PBS. Single nuclei were prepared by mechanical extraction.

Cell capture and cDNA synthesis

Single nuclei were processed using the Chromium Single Cell 3′ Library and Gel Bead Kit v3.1 (10x Genomics, Cat# 1000121) and the Chromium Single Cell G Chip Kit (10x Genomics, Cat# 1000120). Cell suspensions (300–600 live cells per μL, determined by CountStar) were loaded onto the Chromium Controller (10x Genomics) to generate single-cell gel beads in emulsion (GEMs) following the manufacturer’s protocol. In brief, single cells were suspended in PBS containing 0.04% bovine serum albumin (BSA; NEB, Cat# B9000S). Approximately 40,000 cells were added per channel, targeting a recovery of ∼10,000 cells per channel. Captured cells were lysed, and released RNA was barcoded through reverse transcription within individual GEMs. Reverse transcription was performed on an S1000TM Touch Thermal Cycler (Bio-Rad) at 53°C for 45 min, followed by 85°C for 5 min and a 4°C hold. The resulting cDNA was amplified and assessed for quality using an Agilent 4200 TapeStation.

Processing and preliminary analyses of scRNA-seq data

scRNA-seq data were obtained from GSE184507.⁵⁶ Processed gene expression matrices were imported into Seurat (v5.1.0)⁶² to create individual Seurat objects for forager samples. Cells were filtered based on the following criteria: nCount_RNA between 200 and 20,000 and nFeature_RNA between 500 and 8,000. To address potential technical artifacts, the scDblFinder package (v1.11.4)⁶³ was used (parameter dbr.sd = 1) to identify and remove putative doublets arising from multiple cells sharing the same barcode during droplet-based capture. Data integration was performed using Canonical Correlation Analysis implemented in Seurat to correct for batch effects. Dimensionality reduction and visualization were conducted through PCA (first 30 PCs) followed by Uniform Manifold Approximation and Projection (UMAP). Cell-type annotation was performed through manual curation based on reported marker genes (Table S1).^{56,70,71,72,73,74} All computational analyses were conducted in R (v4.4.0). Additional details are provided in the accompanying analysis code.

snRNA-seq library preparation

According to the manufacturer’s instructions, snRNA-seq libraries were constructed using the Single Cell 3′ Library and Gel Bead Kit v3.1. The libraries were sequenced on an Illumina NovaSeq 6000 platform, achieving a sequencing depth of at least 100,000 reads per cell with a paired-end 150 bp (PE150) strategy.

Processing and preliminary analyses of snRNA-seq data

CellRanger (v7.0.1) was used to map sequenced reads to the Amel_HAv3.1 reference genome. The resulting gene count matrices were imported into Seurat to generate a Seurat object for each group. This step organized the gene expression data, enabling comprehensive downstream analysis in the Seurat framework. Before integrating cells across groups, filtering was applied using the following criteria: log10GenesPerUMI >0.85, nFeature_RNA >200, and nFeature_RNA <5000. The log10GenesPerUMI value, calculated as log10(nFeature_RNA)/log10(nCount_RNA), reflects cell complexity. To identify potential doublets—instances where multiple cells share the same barcode—the R package scDblFinder (v1.11.4) was used (dbr.sd = 1). This approach ensured accurate detection and removal of doublets, improving data reliability. To eliminate batch effects, the CCAIntegration algorithm in Seurat was employed. Unsupervised clustering was performed using the first 50 principal components, and the results were visualized in a UMAP plot. Cell types were manually annotated based on known marker genes, and photoreceptors were excluded from further analysis. Detailed analytical procedures are available in the provided code. All data analyses were conducted in R (v4.4.0).

Pathway enrichment

Differentially expressed genes (log₂FC > 0.8) for each cell type or comparison group were used as query genes for pathway enrichment analysis. Cytoscape software (v3.10.3)⁶⁴ and the ClueGO plugin (v2.5.9) were used⁶⁵ for analysis, with Apis mellifera [7460] selected as the reference species. A network graph illustrating the enriched pathways was subsequently generated.

Gene set scoring

To assess glycolysis and glycogen synthesis pathways, we evaluated the expression activity of key enzyme genes (LOC409424, LOC551005, LOC408818, LOC724724) using the JASMINE algorithm (https://github.com/NNoureen/JASMINE). Two scores were computed for each gene set: (1) the approximate mean rank of genes among expressed genes, and (2) the enrichment score of the signature within expressed genes. These scores were scaled to a range of 0–1 and averaged to yield the final activity score.

Cell proportions analysis

We employed the scProportionTest⁶⁶ R package to assess differences in cell type proportions between bromocriptine-treated and control bees. This package is specifically designed to test the statistical significance of changes in cell cluster proportions under different conditions. For each cell cluster, a permutation test was conducted to calculate the p-value, determining whether the observed differences were statistically significant. Bootstrapping was then used to estimate confidence intervals for effect size, providing a measure of uncertainty.

Visualization of joint marker gene expression

Joint density plots for marker genes were generated using the R package Nebulosa (v0.99.92).⁶⁷ Nebulosa performs a gene-weighted two-dimensional kernel density estimation on the selected embedding to obtain a smooth density value for each gene at every cell, using local gene expression levels as weights for neighboring cells in the embedding space. For each set of marker genes, we used the function plot_density with the argument joint = TRUE. This procedure first rescales the single-gene density estimates to a common range and then calculates their pointwise product to obtain a joint density that captures the spatial overlap of the markers’ expression. As a result, cells with concordant, even if moderate, expression of all selected markers exhibit high joint density values, whereas cells expressing only a subset of the markers have joint density values close to zero. All single-gene and joint density maps were computed on the same UMAP embedding using the default kernel density parameters implemented in Nebulosa, and no additional manual thresholding, masking, or spatial constraints were applied.

Quantitative polymerase chain reaction

RNA was extracted using an RNA extraction kit (Sangon Biotech, B511321). All dissecting tools and laboratory materials were treated with DEPC-treated water for at least 30 min. Bee samples (eight pooled per group) were collected from the hive and cooled at 4°C for 20 min to induce anesthesia. Brain tissues were dissected and mixed with 200 μL of TRIzol reagent (Invitrogen, Cat# 15596026). After 1 min of homogenization, 0.2 mL chloroform (Sigma-Aldrich, Cat# C2432) was added and vigorously shaken. The mixture was left at room temperature for 3 min, then centrifuged at 12,000 rpm for 10 min at 4°C. The upper aqueous phase was collected, and 100 μL of anhydrous ethanol was added. The mixture was transferred to an adsorption column and centrifuged at 12,000 rpm for 3 min, then repeated using a new collection tube. The column was centrifuged again at 10,000 rpm for 2 min. Finally, 30 μL of DEPC-treated water was added to the column, incubated for 5 min, and centrifuged at 12,000 rpm for 2 min to obtain purified RNA.

The cDNA was synthesized using a reverse transcription kit (BBI, B639251) with the following thermal program: 25°C for 5 min, 42°C for 30 min, and 85°C for 5 min.

qPCR reactions were performed with the GoTaq qPCR reagent kit (Promega, A6001). Primer sequences are listed in Table S2. Reaction mixtures were prepared according to the manufacturer’s protocol and run on an Applied Biosystems 7500 qPCR system. Cycling conditions were: 95°C for 2 min for enzyme activation, followed by 40 cycles of 95°C for 15 s and 60°C for 1 min.

Relative expression levels were calculated using the 2−ΔΔCT method with normalization to the housekeeping gene GAPDH. Statistical significance between groups (n = 3–5 biological replicates) was evaluated using an unpaired two-tailed Student’s t test in R (v4.4.0).

Targeted metabolomics analysis

To minimize degradation, samples were thawed on ice. For each sample, approximately 15 mg of wet brain tissue was prepared by pooling dissected brains from 10 individual bees, with slight variation in brain mass among individuals. Each pooled sample was precisely weighed and mixed with 20 μL of deionized water and 10 ceramic beads. The mixture was homogenized for 3 min using a Bullet Blender (BB24, Next Advance Inc., USA). Subsequently, 120 μL of pre-chilled methanol solution containing internal standards (Glutamic acid-¹³C₅, Pyruvic acid-¹³C₃, Succinic acid-d₄, Malic acid-d₃, Aspartic acid-d₃, Lactic acid-¹³C₃, Serine-d₃, Glutamine-¹³C₅, D-Glucose-d₇, Fructose-¹³C₆, Citric acid-d₄, α-Ketoglutaric acid-¹³C₅, Fructose-6-phosphate-¹³C₆) was added, followed by another 3-min homogenization. The samples were centrifuged at 18,000 × g for 15 min at 4°C (Microfuge 20R, Beckman Coulter, Indianapolis, IN, USA). Twenty microliters of the supernatant were transferred to a 96-well plate. Each well then received 20 μL of 200 mM 3-nitrophenylhydrazine and 20 μL of 120 mM EDC hydrochloride. The reaction proceeded at 30°C and 1,450 rpm for 60 min (MSC-100, Hangzhou Allsheng Instruments Co., Ltd, China). The reaction was quenched by adding 350 μL of ice-cold methanol solution. After centrifugation at 4,000 × g for 20 min at 4°C (Allegra X-15R, Beckman Coulter, Indianapolis, IN, USA), 150 μL of the supernatant was transferred to a new 96-well plate for subsequent analysis.

Metabolite profiling was performed using an Acquity I-Class UPLC coupled with a Xevo TQ-S mass spectrometer (Waters Corp., Milford, MA, USA). To minimize potential order effects, sample injections were randomized according to group information. Quality control (QC) and blank samples were interspersed throughout the analytical sequence. Raw UPLC–MS/MS data were processed using MassLynx software (v4.1, Waters) for peak detection, integration, calibration, and quantification. Statistical analyses—including principal component analysis (PCA), hierarchical clustering, and differential metabolite analysis—were conducted in R (v4.4.0). Metabolite pathway enrichment was performed using the Small Molecule Pathway Database (SMPDB)⁷⁵ and the MetaboAnalyst online platform.⁷⁶

Non-targeted metabolomics

For untargeted metabolomics, six biological replicates were prepared per treatment group. Each replicate consisted of approximately 15 mg of pooled brain tissue from 10 honeybees. Metabolites were extracted by homogenization in liquid nitrogen and resuspension in 500 μL of pre-chilled 80% methanol. After a 5-min incubation on ice, samples were centrifuged at 15,000 × g and 4°C for 20 min. The resulting supernatant was diluted with LC-MS–grade water to achieve a final methanol concentration of 53%. Following a second centrifugation, the cleared supernatant was subjected to UHPLC-MS/MS analysis. The analysis was performed on a Vanquish UHPLC system coupled to an Orbitrap Q Exactive HF mass spectrometer (Thermo Fisher). Samples were separated on a Hypersil Gold column (100 × 2.1 mm, 1.9 μm) using a 12-min gradient at a flow rate of 0.2 mL/min. Data were acquired in both positive and negative ion modes: 0.1% formic acid in water and methanol for the positive mode, and 5 mM ammonium acetate (pH 9.0) and methanol for the negative mode.

Raw data were processed using Compound Discoverer 3.3 software (Thermo Fisher) for peak picking, alignment, and quantification. Metabolite identification was performed by matching features against the mzCloud, mzVault, and MassList databases with a mass tolerance of 5 ppm. Peak intensities were normalized to total spectral intensity, and features with a coefficient of variation (CV) > 30% in QC samples were excluded. Subsequent statistical analyses were conducted in R, including PCA and Partial Least Squares Discriminant Analysis (PLS-DA) to visualize sample distribution. Differential metabolites were defined as those with a VIP score >1 in the PLS-DA model, a p-value <0.05 (Student’s t test), and a fold change ≥2 or ≤0.5.

RNA fluorescence in situ hybridization

Two groups of bees (the bromocriptine-treated group and the ddH₂O-treated group) were used for brain extraction. Brains were dissected and placed into centrifuge tubes containing fixative solution (Servicebio, Catalog No. G1113) for 24 h. The tissues then underwent dehydration and clearing through the following sequence: 4 h in 75% ethanol, 2 h in 85% ethanol, 2 h in 90% ethanol, 1 h in 95% ethanol, 30 min in absolute ethanol I, 30 min in absolute ethanol II, 5 min in alcohol–benzene, 10 min in xylene I, and 10 min in xylene II. After dehydration and clearing, tissues were infiltrated with paraffin as follows: 1 h in melted paraffin I at 65°C, 1 h in melted paraffin II at 65°C, and 1 h in melted paraffin III at 65°C. Following infiltration, the tissues were embedded by placing them in the center of metal molds, immersing them in paraffin, and refrigerating until the paraffin blocks solidified. The blocks were sectioned using a microtome to obtain slices approximately 5 μm thick. Sections were floated in a 65°C water bath for flattening and then carefully mounted onto adhesive microscope slides.

Slides were dried in an oven at 45°C for at least 4 h, air-dried, and baked at 60°C for 60 min. Deparaffinization was performed by sequential immersion in xylene baths, followed by 100% ethanol and ultrapure water. Antigen retrieval was conducted by immersing the sections in preheated 1× target retrieval solution at 97°C–100°C for 10 min. After rinsing briefly in water and air-drying, a hydrophobic barrier was drawn around each tissue section using an ImmEdge Pen. Sections were then subjected to acid treatment with 0.1 M HCl and washed with wash buffer. Permeabilization was achieved using a solution containing pepsin in 0.1 M HCl, followed by incubation in a humidified chamber at 37°C for 30 min. After additional washes with wash buffer, the sections were prepared for downstream hybridization steps. The following procedures were then performed sequentially: probe hybridization, post-hybridization washing, probe ligation, probe circularization, rolling circle amplification, and fluorescence probe hybridization. For each step, the appropriate reaction solution was applied as droplets onto the tissue, followed by incubation at 37°C for the specified time. After the final reaction, the slides were air-dried, and 20 μL of DAPI-containing antifade mounting medium (Absin, Catalog No. 9235) was applied before covering with a coverslip. The gene probes used for FISH are listed in Table S3.

For quantitative analysis, one brain section from each treatment group was analyzed. From each brain, three representative fields of view were imaged from a single section. Images were analyzed using CellProfilerTM (RRID:SCR_007358 v4.2.6),⁷⁷ which enabled automated detection and quantification of DAPI-stained nuclei and fluorescent RNA probe signals. The probes targeted SLC2A1 (encoding a key glucose transporter), HK (encoding hexokinase 1, a rate-limiting enzyme in glycolysis), nAChRα6 (encoding the α6 subunit of the nicotinic acetylcholine receptor), and Tret1 (encoding a trehalose transporter, the primary circulating sugar in insects). CellProfiler pipelines were configured to use the Otsu method for thresholding, prioritizing intensity over shape to improve accuracy in distinguishing and separating clumped objects. The lower threshold bound was set to approximately 0.1 to minimize false object detection. These settings enabled precise segmentation and quantification of fluorescence signals from individual cells. Cells expressing Tret1 were first identified by the presence of Tret1 probe signals, and these Tret1-positive cells were then evaluated for co-expression with SLC2A1, HK, and nAChRα6 probes to determine their co-localization patterns.

Enzyme-linked immunosorbent assay

To determine the effect of bromocriptine on acetylcholine secretion, an ELISA was performed to quantify ACh levels. For each treatment group, nine forager bees were dissected to extract and weigh their brains. Brain tissues were homogenized in PBS using a pestle, and the homogenates were centrifuged at 5,000 × g for 10 min before being equilibrated to room temperature over 30 min.

In pre-coated microtiter plates (CB10243-In), 50 μL of standards, samples, and blanks were added to individual wells. Except for the blank wells, 100 μL of horseradish peroxidase (HRP)-conjugated detection antigen was added to both standard and sample wells. Plates were sealed with adhesive film and incubated at 37°C for 60 min in a water bath or incubator.

After incubation, the adhesive film was removed, and liquids were discarded. Wells were tapped dry on absorbent paper and washed five times with wash buffer. During each wash, wells were filled with wash buffer, left to stand for 20 s, emptied, and tapped dry on absorbent paper. After the final wash, plates were thoroughly dried on lint-free paper.

Substrate solutions A and B were mixed at a 1:1 ratio, and 100 μL of the mixture was added to each well. Plates were resealed with adhesive film and incubated at 37°C for 15 min. Following incubation, 50 μL of stop solution was added to each well, and absorbance (OD values) was measured at 450 nm using a microplate reader.

Quantification and statistical analysis

All statistical analyses were performed using R (v4.4.0) unless otherwise specified. For all tests, a p-value or FDR-adjusted p-value <0.05 was considered statistically significant. Sample sizes were determined based on established standards in the field and were not pre-specified through power analysis. Data were assumed to satisfy the assumptions of the statistical tests used (e.g., normality). Specific details for each experiment—including the exact n, its definition, and corresponding statistical results—are provided in the relevant figure legends.

scRNA-seq and snRNA-seq data analysis

Inclusion and exclusion criteria for single-cell analyses were based on standard QC metrics. After initial processing in Seurat (v5.1.0), cells and nuclei were filtered according to library size and complexity (e.g., nCount_RNA, nFeature_RNA, log10GenesPerUMI). Putative doublets were computationally identified and removed using the scDblFinder package. The specific filtering thresholds for each dataset are described in the method details section.

Metabolomics data analysis

For targeted metabolomics, samples were randomized before UPLC–MS/MS runs to minimize order bias. For untargeted metabolomics, features with a coefficient of variation (CV) > 30% in quality control (QC) samples were excluded from downstream analysis. Differential metabolites between groups were identified using unpaired Student’s t-tests (p < 0.05), Partial Least Squares Discriminant Analysis (PLS-DA; VIP score >1), and fold change thresholds (FC ≥ 2 or ≤0.5).

qPCR data analysis

Relative gene expression was calculated using the 2−ΔΔCT method, normalized to the housekeeping gene GAPDH. Statistical significance between two experimental groups was assessed using an unpaired, two-tailed Student’s t test. Each group comprised n = 3–5 biological replicates, with each replicate representing a pool of 6–8 individual bee brains. Data in bar plots are presented as mean (center) and SEM (dispersion).

RNA FISH image and ELISA data analysis

For RNA FISH, fluorescence intensity was quantified from three fields of view per brain section (n = 1 brain per group). For ELISA, ACh concentration was measured from n = 9 pooled brains per group. For both datasets, comparisons between the two groups were performed using an unpaired, two-tailed Student’s t test. Data are presented as boxplots showing the median (center) and interquartile range (dispersion) or as bar plots showing mean ± SEM.

Pathway enrichment and gene set scoring

Pathway enrichment for differentially expressed genes was conducted using the ClueGO plugin in Cytoscape. Gene set scores were calculated using the JASMINE algorithm, which integrates the mean rank and enrichment of gene expression within the set.

Cell proportion analysis

Changes in cell type proportions were assessed using the scProportionTest R package. The package applies permutation testing to calculate p-values and bootstrapping to estimate 95% confidence intervals.

Published: February 9, 2026