Differential expression of neuronal function genes follows a tissue-specific temporal dynamic during Deformed Wing Virus infection in honey bees Authors: Diego Silva, Lucila Traverso, Gonzalo Dominguez, Felipe Becerra, Yazmin Fuentes, Sheila Ons, Carolina Gomez-Diaz, Marisol Vargas Published: 2026 ABSTRACT Deformed Wing Virus type A (DWV-A) is one of the primary threats to honeybees (Apis mellifera), significantly impacting their nervous system physiology and behavior. While its neurotropic nature is well-recognized, the temporal dynamics of the neuronal transcriptomic response following oral infection, the natural transmission route, remains poorly understood. In this study, we analyzed gene expression in the heads of worker bees orally inoculated with DWV-A over a 16-day time course (1, 4, 7, 10, 13, and 16 days post-inoculation). RNA-seq analysis at day 10 identified 147 differentially expressed genes associated with different biological processes that are critical to the organism, including cellular metabolism and neuronal activity. RT-qPCR validation revealed a persistent downregulation of key genes related to glutamatergic system (eaat-2, neto, and kainate) and sensory perception-related genes in the antennae. Notably, the simultaneous co-expression of nurse-associated and forager-associated marker genes suggests that DWV-A infection induces an asynchrony in behavioral maturation. Our findings demonstrate that DWV-A disrupts neuronal homeostasis and peripheral sensory perception in a tissue-specific and time-dependent manner, providing a molecular framework to understand the behavioral impairment and the loss of coordination at the colony level.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1038/s41598-026-57691-z. Introduction Pollination is a key ecosystem service that sustains biodiversity and agricultural productivity. Among pollinating insects, the honeybee (Apis mellifera L.) plays a central role, being responsible for the pollination of many crop species worldwide, particularly in extensive monoculture systems1–3. This ecological and economic relevance is closely linked to its polylectic foraging behavior and its ability to detect, discriminate, and memorize a wide diversity of floral odors4–6. Such behavioral complexity relies on the integration of olfactory and visual information through learning and memory processes supported by a highly specialized nervous system6–8. The honeybee olfactory system is organized as an extensive neural network connecting peripheral sensory structures in the antennae with central processing centers in the brain. Odorant signals are transduced by olfactory receptor neurons and transmitted to the antennal lobes, where information is encoded within glomerular circuits before being relayed to higher brain regions6,9,10. This architecture enables the precise processing of complex odor landscapes underlying orientation, learning, and foraging decisions11. Olfactory performance and neural function in A. mellifera, as in other insects such as Drosophila melanogaster, are regulated by dynamic cellular and molecular interactions, including glial neuronal coupling, neuromodulatory signaling, and gene regulatory programs. Glial cells contribute to the maintenance and modulation of olfactory networks12,13. These regulatory mechanisms include pathways associated with neurotransmitter dynamics, such as glutamatergic signaling, supporting the functional plasticity required for adaptive behavior14. The integrity of neural system is compromised by the Deformed Wing Virus (DWV), one of the most prevalent and damaging pathogens of honeybees worldwide15,16. Mainly transmitted by the ectoparasitic mite Varroa destructor, DWV comprises several epidemiologically relevant variants, among which DWV-A is the most prevalent in Chile and is frequently associated with colony losses17,18. Deformed Wing Virus (DWV) is neurotropic and has been shown to replicate in both the central nervous system (the brain) and peripheral neural tissues (antennal sensory structures), where it is associated with alterations in neuronal ultrastructure, synaptic gene expression, and olfactory sensitivity19–21. High viral loads are associated with transcriptomic alterations affecting synaptic function and neurotransmission, impairments in memory, and reduced behavioral performance14,22. Furthermore, DWV-A alters the expression of odorant-binding proteins and pre- and postsynaptic genes in the antennae, directly compromising peripheral olfactory processing22,23. Despite recent advances, how DWV-A induced molecular alterations affect information flow, and their functional consequences on odor-guided behaviors remain poorly understood. Here, we combined transcriptomic and targeted expression analyses to test whether DWV-A infection induces coordinated metabolic, proteostatic, and glutamatergic alterations in honeybee heads and antennae, potentially underlying previously reported olfactory and cognitive impairments. Materials and methods Preparation of the viral inoculum The viral inoculum used for all bioassays was prepared as described by Gusachenko et al.24 and Silva et al.22. The viral inoculum was obtained from field honey bee colonies maintained in an experimental apiary. Colonies were previously screened by quantitative PCR (qPCR) to determine the presence and viral load of Deformed Wing Virus variant A (DWV-A). In parallel, the presence of other pathogens, Nosema spp., Lotmaria spp., DWV-B, Acute Bee Paralysis Virus (ABPV), and Chronic Bee Paralysis Virus (CBPV), were also assessed. The selection of the colony used for viral inoculum preparation was based on viral quantification criteria. Specifically, lower Cq values indicate higher viral abundance, while higher Cq values represent lower viral loads. The selected colony exhibited a Cq value > 15 for DWV-A and Ct values < 32 for the other pathogens described above. A group of 20 symptomatic bees was homogenized in phosphate-buffered saline (1X PBS) for 90 s at high-speed using a Stomacher 80 laboratory blender (Seward, London, UK). The homogenate was then centrifuged twice: first at 1500× g for 10 min, followed by 10,000× g for 10 min, both at 4 °C. The supernatant was purified by filtration through a 0.22 µM filter (PES, Merck Millipore, Darmstadt, Germany).To determine the exact concentration of viral particles present in the inoculum, a 200 µL aliquot of the supernatant was collected for RNA extraction, cDNA synthesis, and subsequent quantification of the viral load via qPCR, following the methodology detailed in the RNA extraction section. The remaining extracted inoculum was stored at − 80 °C until further use. Viral inoculation A. mellifera workers used in all experimental procedures were obtained from the apiary of the experimental station “El Nogal” (36°35′58.25″ S, 72°04′51.77″ W), University of Concepción, Chillán, Chile. To ensure uniformity in age, brood frames were extracted directly from healthy hives and maintained under controlled conditions (30 ± 1 °C, 60 ± 3% RH) until adult emergence. Donor colonies were pre-selected based on pathogen screening, prioritizing those free from Nosema ceranae (Cq < 32), Lotmaria passim (Cq < 32), and exhibiting either undetectable or minimal viral loads of acute bee paralysis virus (ABPV, Cq < 32), chronic bee paralysis virus (CBPV, < 32), and DWV (< 32). Bees were obtained from six colonies, and newly emerged individuals were randomly mixed prior to experimental assignment to minimize colony-specific variation. Newly emerged bees were randomly collected and transferred to cylindrical plastic cages (base diameter: 8 cm; top diameter: 10 cm; height: 15 cm). Viral infection procedures followed previously established protocols23, with modifications. Briefly, bees in the inoculated group (Treatment i-DWV) were orally inoculated with 5 µL of a viral suspension containing 1.0 × 10⁹ genome copies per bee, prepared in 60% (w/v) sucrose solution. Individuals that did not ingest the full inoculum were excluded. The control group (Treatment n-DWV) received only sucrose solution (60% w/v), without viral particles. A total of 600 worker bees were used, distributed into 8 independent plastic cages (replicates), each containing 75 bees. Four cages were assigned to the inoculated group (n = 300) and four to the control group (n = 300). All bees received ad libitum access to a 60% sucrose syrup and 3 g of a standardized pollen substitute composed of soybean flour, brewer’s yeast, cornstarch, wheat flour, corbicular pollen, and canola oil, following Arismendi et al.25. Sampling was performed every 3 days, starting 24 h after viral inoculation. At each time point, a pool of five bees was collected per replicate cage. Sampling days included 4, 7, 10, 13, and 16 days post-inoculation (dpi). Additionally, a specific sample was taken at 10 dpi for subsequent RNA-seq analysis. All collected bees were immediately frozen at − 80 °C until use. RNA extraction and quality assessment Total RNA was extracted from bee heads with antennae collected at each bioassay time point, following the protocols described by Kim et al.20, Mondet et al.26, and Silva et al.22, with minor modifications. Briefly, pools of dissected heads including antenna from each replicate and sampling day were homogenized in 500 µL of TRIzol™ Reagent (Invitrogen, Thermo Fisher Scientific, Waltham, MA, USA) using a Retsch homogenizer at maximum speed. Following homogenization, 5 µL of RNA Carrier (Invitrogen) was added according to the manufacturer’s instructions. The homogenate was incubated on ice for 5 min, followed by the addition of 200 µL of chloroform. Samples were then vortexed briefly, incubated on ice for an additional 3 min, and centrifuged at 10,000 × g for 15 min at 4 °C. The aqueous phase was carefully transferred to a new tube for RNA purification. For most samples, RNA was extracted using the E.Z.N.A.® Total RNA I Kit (Omega Bio-Tek, Norcross, GA, USA), following the manufacturer’s protocol. In the case of samples collected at 10 dpi, used for transcriptomic analysis, total RNA was isolated using the PureLink™ RNA Mini Kit (Thermo Fisher Scientific, CA, USA), according to the manufacturer’s instructions. RNA concentration and purity were determined spectrophotometrically using an Infinite® 200 PRO NanoQuant microplate reader (Tecan Group, Männedorf, Switzerland) requiring A260/A280 ration between 1.8 and 2.1. RNA integrity was evaluated by electrophoresis on 1% agarose gels. All RNA extracts were stored at − 80 °C until downstream applications. Sequencing Library construction and high-throughput sequencing services were outsourced to Austral-omics (Valdivia, Chile). A total of six cDNA libraries were generated (three per experimental condition). For mRNA library preparation, 300 ng of total RNA per sample was used as input, and the RNA Directional Library Prep Set (MGI Tech) was employed, following the manufacturer’s instructions. Insert size and final library integrity were verified using capillary electrophoresis on a Fragment Analyzer system (Agilent Technologies) with the DNF-474 High Sensitivity NGS Fragment Analysis Kit. Library quantification was performed using the Qubit™ DNA High Sensitivity Assay Kit (Thermo Fisher Scientific), and the volume required for a final concentration of 1 pM was calculated accordingly. cDNA strand separation and circularization were carried out, followed by DNA nanoball (DNB) synthesis, loading of the flow cell, and sequencing. Short-read sequencing was performed using the DNBSeq technology (MGI Tech) on the DNBSeq-G400 platform, generating paired-end reads of 150 nucleotides. Data preprocessing and identification of differentially expressed genes Raw read quality and the presence of adapter sequences were assessed using FastQC (available at www.bioinformatics.babraham.ac.uk/projects/fastqc). To remove adapter sequences and to trim low-quality bases from the raw reads, we used Trimmomatic (v0.39) in paired-end mode with the following parameters: TRAILING:10, LEADING:5, SLIDINGWINDOW:4:20, ILLUMINACLIP: DNBSEQ_adapter.fa:2:30:10. Reads shorter than 50 bp were filtered out, using MINLEN:50 parameter in the same step. The filtered reads obtained in the previous step were aligned to the most recent A. mellifera genome assembly, Amel_HAv3.1 along with its corresponding gtf file (NCBI accession number: GCF_003254395)27, using STAR software (v2.7.6a)28 with the option --outFilterMultimapNmax set at 1. To improve statistical power, a filtering step to remove low-expressed genes was performed using the HTSFilter package29 with a modeling length of s.len = 100. Differential expression analysis was subsequently conducted on the filtered genes using the DESeq2 package v1.42.030 through the DESeqfunction, employing a Wald test with parametric dispersion fitting. To control the False Discovery Rate (FDR), Storey’s q-values31 were calculated from the raw p-values using the qvalue package. Throughout this study, these significance values are referred to as ‘s-values’ to maintain consistency with the bioinformatics pipeline outputs. Genes with an s-value (q-value) < 0.05 were considered statistically significant. Subsequently, the sequences corresponding to all RNA-seq analysis and all DEGs, were analyzed for functional enrichment in terms of Gene Ontology (GO) categories. To assess the functional enrichment of GO terms, a Gene Score Resampling (GSR) analysis was performed using ErmineJ (V.3.1.2 desktop version for macOS). The analysis utilized the s−values as input, with the ‘big is better’ option enabled to prioritize higher values. The ‘Mean’ class scoring method was applied, and statistical significance was determined through 200,000 iterations to ensure robust empirical p-value estimation. Gene set size constraints were set to a minimum of 10 and a maximum of 200 genes. GO terms with a corrected p-value < 0.05 were considered statistically significant. Pfam protein domains associated with each DEG were identified using InterProScan (v 5.73–104.0), considering the longest transcript as the representative for each gene. Significantly overrepresented domains were identified through a hypergeometric test (p < 0.05). The background set consisted of all domains identified in the longest transcripts of the total gene set encoded in the genome. cDNA syntesis and qPCR cDNA synthesis and subsequent gene expression analysis by qPCR were performed following the standards regulated by MIQE 2.032 to ensure reproducibility. For qPCR analyses, each biological sample consisted of a pool of five bees (n = 4 biological replicates per treatment). For viral load quantification, a standard curve was generated using purified PCR products (Wizard® SV Gel and PCR Clean-Up System, Promega, Madison, WI, USA), quantified by spectrophotometry (Epoch™ Microplate Spectrophotometer, BioTek, Winooski, VT, USA). Copy numbers were calculated following Wu et al.33, and standard curves were generated using serial dilutions (1.0 × 10¹ to 1.0 × 10⁹). Cq values were plotted against log10 copy numbers, and viral loads were expressed as DWV-A copies per bee after normalization and dilution correction34. cDNA derived from head and antenna samples collected at 10 dpi, the same material used for RNA-seq, was included. In parallel, independent samples collected at 4, 7, 13 and 16 dpi were dissected into heads and antennae, and processed for cDNA synthesis and qPCR. For cDNA synthesis, the High-Capacity Reverse Transcription Kit (Applied Biosystems; Invitrogen, Life Technologies, Carlsbad, CA, USA) was used according to the manufacturer’s instructions. To strictly exclude the contribution of residual genomic DNA (gDNA) to the signal, no-reverse transcriptase controls (NRTC) were performed for each biological group. Viral quantification and host gene expression analyses were performed using cDNA corresponding to samples collected at each point, with β-actin and 18 rps used as reference genes (Table 1). qPCR reactions were performed using Applied Biosystems MasterMix (Thermo Fisher Scientific, Wilmington, MA, USA) in a final volume of 10 µL containing 10 ng of cDNA and 530 nM of each primer. Thermal cycling conditions were optimized for each primer pair. Primer efficiencies (E) were determined using standard curves and ranged between 95% and 100% (R2 > 0.99) for all primer pairs. To ensure technical transparency, no-template controls (NTC) were included in every run to monitor for environmental contamination, and a melting curve analysis was conducted at the end of each cycle to confirm single-product specificity. Assay sensitivity and dynamic range were evaluated using serial dilution standard curves for all primers used in qPCR experiment after normalization and dilution correction34. Reactions were run on a StepOne Real Time PCR System de Applied Biosystems (Waltham, MA 02451, USA), and data were analyzed using StepOne software (v 2.3). Table 1Primer sequences designed for quantitative real-time PCR (qPCR) analysis of gene expression, including DWV, housekeeping genes, and genes of interest (Neto, Kainate, and Eaat-2).GeneF/RSequenceTargetDWV-AFTATCTTCATTAAAGCCACCTGGAAVirusRTTTCCTCATTAACTGTGTCGTTGAT Arp1 FATGCCAACACTGTCCTTTCTGGhousekeeping 1RGACCCACCAATCCATACGGA 18rps FGATTCCCGATTGGTTTTTGAATAGhousekeeping 2RAACCCCAATAATGACGCAAACC Neto FATCAGAGGCGACCTTTCAGCNeto (LOC411019)RTGACAGTTTGACGGGGATGG Kainate FTGCTTCCTCCCGACAATGACKainate (LOC102653800)RAGCAATCCCGTAACCCTTGG Eaat-2 FTTATCGCGGCGCTCATTCTTEaat-2 (LOC406113)RTCCTCTTACGTGCAACACAGG Table 1Primer sequences designed for quantitative real-time PCR (qPCR) analysis of gene expression, including DWV, housekeeping genes, and genes of interest (Neto, Kainate, and Eaat-2).GeneF/RSequenceTargetDWV-AFTATCTTCATTAAAGCCACCTGGAAVirusRTTTCCTCATTAACTGTGTCGTTGAT Arp1 FATGCCAACACTGTCCTTTCTGGhousekeeping 1RGACCCACCAATCCATACGGA 18rps FGATTCCCGATTGGTTTTTGAATAGhousekeeping 2RAACCCCAATAATGACGCAAACC Neto FATCAGAGGCGACCTTTCAGCNeto (LOC411019)RTGACAGTTTGACGGGGATGG Kainate FTGCTTCCTCCCGACAATGACKainate (LOC102653800)RAGCAATCCCGTAACCCTTGG Eaat-2 FTTATCGCGGCGCTCATTCTTEaat-2 (LOC406113)RTCCTCTTACGTGCAACACAGG Primer sequences designed for quantitative real-time PCR (qPCR) analysis of gene expression, including DWV, housekeeping genes, and genes of interest (Neto, Kainate, and Eaat-2). Data analysis qPCR data were analyzed to evaluate the effects of viral inoculation and time post infection on gene expression and viral load. For each gene, relative expression levels were calculated using the 2⁻ΔΔCq method32. Normalization was performed using the geometric mean of two validated reference genes, β-actin and 18 Rps (Table 1), to ensure a stable baseline for transcriptomic comparison. Genes were analyzed independently, and at each dpi, inoculated (i-DWV) and non-inoculated (n-DWV) groups were compared. Data normality was assessed using the Shapiro–Wilk test and Q–Q plots. As the data did not consistently meet parametric assumptions, pairwise comparisons were performed using the Wilcoxon rank-sum test (Mann–Whitney U test). p-values were adjusted for multiple comparisons using the Benjamini–Hochberg false discovery rate correction. Viral load data are presented as mean ± standard error (SE). All analyses and graphical visualizations were performed in R (RStudio 2025.09.1). Statistical significance was set at adjusted p < 0.05. Results Sequencing and mapping metrics The RNA-seq analysis yielded over 145 million paired-end reads in total, with an average of 24.29 millions per experimental replicate. After preprocessing, more than 91.5% of the raw reads were retained. These high-quality reads were subsequently aligned to the A. mellifera reference genome, resulting in an average of 20 million mapped reads per replicate, corresponding to 90.23% of the processed reads (see Supplementary Table S1). DWV infection induces immune and glutamatergic genes while suppressing cellular proteostasis Differential expression analysis was conducted on 9,113 genes, representing 74% of the total genes encoded in the genome. A total of 147 genes (1.61% of the analyzed genes) were identified as differentially expressed (DEGs) between the i-DWV and n-DWV treatments (s-value < 0.05) (Fig. 1, Supplementary Table S2). Among these DEGs, 109 genes were up-regulated and 38 were down-regulated in the i-DWV treatment compared to the n-DWV condition (Fig. 2, Supplementary Table S2). Fig. 1Heatmap of differentially expressed genes (DEGs) in bees inoculated with Deformed Wing Virus variant A (DWV-A) at 10 dpi. Inoculated (m_3, m_15, m_17) and non-inoculated (m_18, m_20, m_21) samples are shown. Columns correspond to biological replicates and rows to individual DEGs. The dendrogram above the heatmap depicts hierarchical clustering of samples. Red denotes upregulated genes and blue denotes downregulated genes. Heatmap of differentially expressed genes (DEGs) in bees inoculated with Deformed Wing Virus variant A (DWV-A) at 10 dpi. Inoculated (m_3, m_15, m_17) and non-inoculated (m_18, m_20, m_21) samples are shown. Columns correspond to biological replicates and rows to individual DEGs. The dendrogram above the heatmap depicts hierarchical clustering of samples. Red denotes upregulated genes and blue denotes downregulated genes. Fig. 2Volcano plot summarizing the differentially expressed sequences in i-DWV vs. n-DWV. Transcript abundance is represented as Log10 (TPM + 1). The x-axis and y-axis represent Log2 fold-change (Log2FC) differences between the contrasted groups and statistical significance as the negative Log10 P-value (s-value), respectively. A total of 9,113 genes were included in the analysis, yielding 109 and 38 transcripts, up- and downregulated. Volcano plot summarizing the differentially expressed sequences in i-DWV vs. n-DWV. Transcript abundance is represented as Log10 (TPM + 1). The x-axis and y-axis represent Log2 fold-change (Log2FC) differences between the contrasted groups and statistical significance as the negative Log10 P-value (s-value), respectively. A total of 9,113 genes were included in the analysis, yielding 109 and 38 transcripts, up- and downregulated. Notably, several immune and metabolic genes were among the most highly up-regulated transcripts, including the antimicrobial peptides hymenoptaecin, abaecin, and apidaecin, as well as L-lactate dehydrogenase (LDH; LOC411188), each exhibiting a log2FoldChange > 1.5. Additionally, our analysis identified the up-regulation of three key genes representing complementary regulatory layers of the glutamatergic synapse: Neto (LOC411019), involved in synaptic scaffolding; Eaat-2 (LOC406113), responsible for extracellular glutamate clearance; and the Kainate receptor (LOC102653800), central to synaptic signal reception. We also detected the induction of transcription factors associated with behavioral maturation, specifically cwo (LOC726204) and nau (LOC724636). Conversely, critical components of the cellular proteostasis network were significantly down-regulated (log2FoldChange < − 1). These included Stip1 (Stress-induced phosphoprotein 1, LOC102656934) and Hsp70 (Heat shock protein 70, Gene ID: 409587). (see Supplementary Table S2 for details) Functional enrichment highlights alterations in cell communication and transmembrane transport To provide a functional context, a GO enrichment analysis was first performed using the entire set of 9,113 genes as a background. This identified 176 significantly enriched terms (p < 0.05), primarily associated with essential constitutive processes such as protein synthesis, tRNA processing, and peptide metabolism (Fig. 3). This baseline profile reflects the high metabolic and translational activity required for cellular maintenance under the studied conditions. Fig. 3Gene Ontology (GO) enrichment analysis using a Gene Score Resampling (GSR) approach. The results display the top 20 most significant GO terms enriched in inoculated bees (i-DWV) compared to non-inoculated controls (n-DWV). The Y-axis lists the enriched GO terms and their access codes, while the X-axis represents the statistical significance expressed as −log10 (p-value). The color intensity of the bars indicates the number of genes associated with each specific GO term (count). A total of 9,113 genes were used as the background set for the analysis (corrected p < 0.05). Gene Ontology (GO) enrichment analysis using a Gene Score Resampling (GSR) approach. The results display the top 20 most significant GO terms enriched in inoculated bees (i-DWV) compared to non-inoculated controls (n-DWV). The Y-axis lists the enriched GO terms and their access codes, while the X-axis represents the statistical significance expressed as −log10 (p-value). The color intensity of the bars indicates the number of genes associated with each specific GO term (count). A total of 9,113 genes were used as the background set for the analysis (corrected p < 0.05). Subsequently, a Gene Score Resampling (GSR) analysis was conducted to identify biological functions associated with genes showing coordinated expression changes in response to the i-DWV treatment. This approach considers all genes ranked by their statistical significance (s-values), rather than relying on a predefined DEG list. This analysis revealed 23 significantly enriched terms (p < 0.05) (Fig. 4). The most represented categories included cellular anatomical structure (GO:0110165; 68 genes), cellular process (GO:0009987; 56 genes), and catalytic activity (GO:0003824; 39 genes), spanning the three main GO domains. Fig. 4Gene Ontology (GO) enrichment analysis using a Gene Score Resampling (GSR) approach. The results display the top 20 most significant GO terms enriched in inoculated bees (i-DWV) compared to non-inoculated controls (n-DWV). The Y-axis lists the enriched GO terms and their access codes, while the X-axis represents the statistical significance expressed as −log10 (p-value). The color intensity of the bars indicates the number of genes associated with each specific GO term (count). A total of 147 genes (DEGs) were used as the background set for the analysis (corrected p < 0.05). Gene Ontology (GO) enrichment analysis using a Gene Score Resampling (GSR) approach. The results display the top 20 most significant GO terms enriched in inoculated bees (i-DWV) compared to non-inoculated controls (n-DWV). The Y-axis lists the enriched GO terms and their access codes, while the X-axis represents the statistical significance expressed as −log10 (p-value). The color intensity of the bars indicates the number of genes associated with each specific GO term (count). A total of 147 genes (DEGs) were used as the background set for the analysis (corrected p < 0.05). Focusing on the cellular process category, the increased viral load in the heads of i-DWV bees specifically modulated genes associated with cell communication and transmembrane transport. Within this group, several genes related to neurotransmitter signaling were up-regulated, suggesting an alteration in synaptic excitability and amino acid homeostasis. For instance, we identified the up-regulation of Glutamate receptor ionotropic, kainate (LOC102653800) and Excitatory amino acid transporter 2 (LOC406113), both key players in glutamatergic neurotransmission. Additionally, Neuropilin and tolloid-like (LOC411019), involved in axon guidance, and the Sodium-dependent nutrient amino acid transporter 1 (LOC725346) showed increased expression. Regarding the protein domain analysis, a total of 77 Pfam domains were significantly enriched within the DEG set. The 13 most significantly overrepresented domains (those with the lowest p-values, p < 0.05) are shown in Fig. 5, while the complete list is provided in Supplementary Table S3. Fig. 5Hypergeometric test results for enrichment analysis of DEGs. Horizontal bars indicate total count gene in the genome of A. mellifera (Amel_HAv3.1) and in DEGs dataset (i-DWV vs. n-DWV). P-values for each Pfam domain are indicated. Hypergeometric test results for enrichment analysis of DEGs. Horizontal bars indicate total count gene in the genome of A. mellifera (Amel_HAv3.1) and in DEGs dataset (i-DWV vs. n-DWV). P-values for each Pfam domain are indicated. Notably, the Chitin binding Peritrophin-A domain (PF01607) and the Polysaccharide deacetylase domain (PF01522) were among the enriched categories; these domains are present in the genes cda5 (LOC552276) and serp (LOC408365), respectively, which are known to modulate structural architecture and are involved in the regulation of structural components of extracellular and epithelial tissues. Additionally, we found significant enrichment of the Helix-loop-helix DNA-binding domain (PF00010), a feature of transcription factors such as Cwo (LOC726204) and Nau (LOC724636), previously described as regulators of the nurse-to-forager transition. The analysis also highlighted the enrichment of the CUB domain (PF00431) and the Receptor family ligand binding region domain (PF01094). These domains are structural components of the synaptic gene Neuropilin and tolloid-like (LOC411019) and the Glutamate receptor ionotropic, kainate (LOC102653800), respectively. These findings reinforce the relevance of the glutamate-related regulatory genes identified in the DEG analysis, all of which were up-regulated in i-DWV bees. Conversely, the Tetratricopeptide repeat domain (PF00515) and the Hsp70 protein domain (PF00012) were also significantly enriched. These domains are characteristic of Stress-induced phosphoprotein 1 (LOC102656934) and Heat shock protein 70 (Gene ID: 409587), both of which were down-regulated in i-DWV bee heads. DWV triggers a biphasic response in heads and a progressive decline of synaptic genes in antennae To validate the transcriptional profile observed at 10 dpi and explore the temporal dynamics of the infection, we focused on the glutamatergic triad: Neto, Eaat-2, and Kainate. In head samples, viral loads were significantly higher in the i-DWV group compared to the control (n-DWV) across all time points (p < 0.0001; Fig. 6). Notably, the qPCR results at 10 dpi confirmed the up-regulation of all three genes observed in the RNA-seq analysis, reinforcing the reliability of the transcriptomic snapshot (Fig. 7). Fig. 6Temporal dynamics of DWV-A viral loads in honey bee heads. Viral titers in inoculated bees (i-DWV) ranged from to copies/bee, significantly exceeding non-inoculated controls (n-DWV) at all time points (4–16 dpi). Values at 10 dpi represent the pooled material used for RNA-seq. The horizontal dashed line indicates the basal viral load detected in field-sampled bees, representing the natural background levels of the pathogen in the apiaries Asterisks (*) above the means indicate significant differences between treatments at the same dpi, according to the Wilcoxon test (p < 0.05). Temporal dynamics of DWV-A viral loads in honey bee heads. Viral titers in inoculated bees (i-DWV) ranged from to copies/bee, significantly exceeding non-inoculated controls (n-DWV) at all time points (4–16 dpi). Values at 10 dpi represent the pooled material used for RNA-seq. The horizontal dashed line indicates the basal viral load detected in field-sampled bees, representing the natural background levels of the pathogen in the apiaries Asterisks (*) above the means indicate significant differences between treatments at the same dpi, according to the Wilcoxon test (p < 0.05). Fig. 7Temporal gene expression patterns of the nervous system genes Neto, Eaat-2, and Kainate in head samples of honeybees inoculated (i-DWV) and non-inoculated (n-DWV) with a viral suspension. Values at 10 dpi represent the pooled material (head with antennae) used for RNA-seq. Asterisks (*) above the means indicate significant differences between treatments at the same dpi, according to the Wilcoxon test (p < 0.05). Temporal gene expression patterns of the nervous system genes Neto, Eaat-2, and Kainate in head samples of honeybees inoculated (i-DWV) and non-inoculated (n-DWV) with a viral suspension. Values at 10 dpi represent the pooled material (head with antennae) used for RNA-seq. Asterisks (*) above the means indicate significant differences between treatments at the same dpi, according to the Wilcoxon test (p < 0.05). The temporal analysis in heads revealed a biphasic response: Neto showed early up-regulation at 7 and 10 dpi (p < 0.01), followed by a significant reduction at 13 and 16 dpi (p < 0.01). Similarly, Eaat-2 and Kainate peaked at 10 dpi (p < 0.001 and p < 0.05, respectively) before declining significantly by 16 dpi (p < 0.0001) (Fig. 7). The antennal dynamics showed a progressive down-regulation of the triad, which became statistically significant at the later stages of infection. Specifically, Neto expression decreased at 13 and 16 dpi (p < 0.0001), while Eaat-2 and Kainate reached significant down-regulation at 16 dpi (p < 0.001) (Fig. 8). This suggests that while the initial viral impact triggers a compensatory response in the brain (heads), the sensory tissues (antennae) exhibit a steady decline in the expression of these synaptic integrity genes. For the load virus and expression analysis in antennae, the 10 dpi time point was integrated into the head analysis, as the head capsule constitutes most of the tissue mass and RNA yield at this stage. In antennal samples, viral loads were also consistently higher in inoculated bees (p < 0.001; Fig. 9). Fig. 8Temporal gene expression patterns of the nervous system genes Neto, Eaat-2, and Kainate in antennae samples of honeybees inoculated (i-DWV) and non-inoculated (n-DWV) with a viral suspension. Asterisks (*) above the means indicate significant differences between treatments at the same dpi, according to the Wilcoxon test (p < 0.05). Temporal gene expression patterns of the nervous system genes Neto, Eaat-2, and Kainate in antennae samples of honeybees inoculated (i-DWV) and non-inoculated (n-DWV) with a viral suspension. Asterisks (*) above the means indicate significant differences between treatments at the same dpi, according to the Wilcoxon test (p < 0.05). Fig. 9Temporal dynamics of DWV-A viral loads in honey bee antennae. Viral titers in inoculated bees (i-DWV) ranged from to copies/bee, significantly exceeding non-inoculated controls (n-DWV) at all time points (4–16 dpi). The horizontal dashed line indicates the basal viral load detected in field-sampled bees, representing the natural background levels of the pathogen in the apiaries Asterisks (*) above the means indicate significant differences between treatments at the same dpi, according to the Wilcoxon test (p < 0.05). Temporal dynamics of DWV-A viral loads in honey bee antennae. Viral titers in inoculated bees (i-DWV) ranged from to copies/bee, significantly exceeding non-inoculated controls (n-DWV) at all time points (4–16 dpi). The horizontal dashed line indicates the basal viral load detected in field-sampled bees, representing the natural background levels of the pathogen in the apiaries Asterisks (*) above the means indicate significant differences between treatments at the same dpi, according to the Wilcoxon test (p < 0.05). Discussion Numerous transcriptomic studies in A. mellifera have addressed the consequences of infections by different pathogens, generally analyzing body segments such as the thorax, abdomen, or whole individuals35–39. In contrast, only a few studies have focused on specific tissues such as the brain and/or other specialized tissues21,40,41. In this study, although we included a complete head segment (comprising the brain, antennae, and other sensory organs), our DEG analysis aimed to identify global changes that could encompass both sensory receptor cells and those responsible for processing stimuli in higher-order organs. In our analysis, the increase in viral loads in bee heads was primarily associated with an overexpression of most DEGs. We observed a consistent induction of three antimicrobial peptides (AMPs), hymenoptaecin, abaecin, and apidaecin, under conditions of high viral load. Although AMPs are typically induced by bacterial infections42, they have also been frequently associated with viral challenges36. This observation is further supported by a meta-analysis of transcriptomes from DWV-infected heads, which identified apidaecin as a gene consistently overexpressed in correlation with viral increase43. These consistencies across studies highlight that the response to viral infection involves the activation of immune pathways; however, it is important to note that the AMP expression detected in our head samples might not be restricted to cephalic structures but could also reflect the immune activity of circulating hemocytes present in the cephalic hemolymph. Our findings regarding gene up-regulation contrast with the results reported by Pizzorno et al.21 in bee brains, where a general repression of gene expression predominated. This divergence suggests that DWV exerts a dramatic and distinct effect on the physiology of the head compared to isolated brain tissue. Furthermore, the route of infection appears to be a decisive factor in these differences. While abdominal injection of DWV has been widely used to mimic transmission by V. destructor, generating pathogenic infections in the brain19,21,44,45, our oral inoculation method represents a non-invasive and less stressful approach. Oral infection has been shown to trigger specific behavioral and physiological repercussions44, as well as changes in gene expression related to memory14. Therefore, despite the overlaps in immune markers like AMPs with the study by Pizzorno et al.21, the overall transcriptional landscape seems to be strongly shaped by the mode of viral entry and the specific tissues analyzed. The GO term analysis of our DEGs revealed a pattern centered on metabolic processes, protein synthesis, macromolecule modification, transcriptional regulation, and nucleotide metabolism. Taken together, these results suggest that DWV may reprograms the cellular machinery to favor its replication, enhancing translation and adjusting key metabolic pathways without markedly activating stress responses or apoptosis. This is consistent with the idea that the virus manipulates the host cell rather than inducing immediate damage detectable solely at the transcriptional level. In this context, recent studies have highlighted key metabolic genes such as kynurenine/α-aminoadipate aminotransferase (LOC724239) and l-lactate dehydrogenase, LDH (LOC411188), which emerge as central DEGs and ‘hub genes’ in responses to viral infections, including DWV, along with Toll pathway genes and AMPs46. In agreement with these reports, our analysis identified the overexpression of AMPs and LDH (LOC411188), whose function of catalyzing the reversible conversion of pyruvate to lactate is directly linked to aerobic glycolysis (AG). This process, described in cancer cells as the ‘Warburg effect’47, enables the rapid generation of ATP and NAD⁺ from glucose, even in the presence of oxygen. Such metabolic reprogramming has been proposed as a mechanism that benefits both viral replication and the availability of cellular metabolites48. The increase in LDH and AG in DWV-infected bee heads could therefore have a dual impact: metabolic, by facilitating favorable conditions for viral replication, and behavioral/physiological, by modulating processes associated with the nurse to forager transition. Supporting this, AG in bee heads has been reported to be associated with increased aggressiveness49, cellular proliferation47, and higher glutamate production50. Collectively, these factors could stimulate an accelerated maturation of infected bees. Our temporal expression data further support this neurometabolic disruption, suggesting that Eaat-2 acts as a metabolic sensor within this pathological process. As Eaat-2 is the primary transporter responsible for removing glutamate from the synaptic cleft through an active, ATP-dependent mechanism51,52, its regulation is intrinsically tied to the cell’s energetic status. While increased AG may lead to higher glutamate production, the observed downregulation of Eaat-2, particularly at late stages of infection, points to a critical failure in glutamate clearance. In the context of a ‘Warburg-like’ metabolic shift, the high energetic cost of maintaining Eaat-2 activity likely becomes unsustainable, creating a ‘double hit’ effect: the convergence of increased neurotransmitter production and impaired removal. This imbalance leads to synaptic glutamate accumulation and potential excitotoxicity, providing a mechanistic link between viral-induced metabolic reprogramming and the subsequent cognitive and behavioral decline observed in infected workers. This breakdown coincides with the attainment of peak viral loads (~ 109 copies/bee), suggesting that the high viral replication pressure eventually exhausts the energetic and homeostatic capacity of the neuron. However, contrary to the effects proposed because of a possible increase in AG, Chen et al.14 determined that the increase in DWV viral loads reduced glutamate concentration due to decreased expression of genes involved in glycolysis, while still observing symptoms of both short- and long-term memory loss. In summary, our results and previous studies converge on the idea that DWV, in addition to activating immune responses (AMPs), induces profound metabolic reprogramming centered on aerobic glycolysis. This shift could not only favor the virus but also accelerate maturation processes and alter cognitive and behavioral functions in bees. Traniello et al.43 proposed the hypothesis of genetic markers explaining this accelerated maturation, identifying the increased expression of the transcription factor cwo, a known regulator of the nurse-to-forager transition, as a consistent feature across multiple datasets53,54. The relevance of cwo lies in its role as a key coordinator of circadian rhythms55. In healthy honey bees, circadian regulation intensifies as they progress toward the foraging state, transitioning from the arrhythmic behavior of newly emerged bees to the robust circadian gene expression required for foraging56. This rhythmicity is essential for the spatiotemporal memory needed to anticipate floral resource availability57, which explains why cwo expression is naturally higher in foragers53. Our findings reinforce this link between viral infection and disrupted maturation. The enrichment analysis of Pfam domains in our DEGs highlighted the Helix-loop-helix DNA-binding domain (PF00010), associated with both cwo (LOC726204) and nau (LOC724636). While these transcription factors typically act as markers for different life stages, cwo for foragers and nau for nurses53,54, our results show that both were simultaneously up-regulated in the heads of bees with high viral loads. This suggests that DWV-A does not simply push the bee into a forager state but rather triggers a lack of coordination in the developmental process. The simultaneous increase of forager-like (cwo) and nurse-like (nau) signals could lead to a physiological noise that explains the behavioral deficiencies and the breakdown of normal colony transitions observed in DWV-infected bees. Although some effects of DWV-A are not lethal at the individual level, they can still threaten colony fitness by disrupting organizational cohesion41,45. A striking finding in our study was the significant down-regulation of genes containing the Tetratricopeptide repeat (PF00515) and Hsp70 (PF00012) domains, such as Stress-induced phosphoprotein 1 and Heat shock protein 70. In A. mellifera, the expression of HSPs typically serves as a stress indicator; for instance, their up-regulation is commonly correlated with high infestation rates of pathogens like V. destructor57. Similarly, increases in Hsp90, which in our study also showed a downward trend (log2FoldChange 0.9,s−value0.07), have been associated with neurotoxic damage in bees exposed to glyphosate58. These precedents reinforce the idea that HSP up-regulation is a standard defense mechanism under stress. However, our results show a contrary pattern, with DWV-A inducing a down-regulation of these chaperones, opposing the behavior observed for AMPs. A possible explanation for this relates to the essential role of these proteins in proteostasis, where they act as critical regulators of proper protein folding and functional homeostasis59. We hypothesize that the viral-induced suppression of these folding machineries, combined with high viral replication in the head, could be a key driver of the cognitive impairment reported in infected bees. This link between viral load and the breakdown of cellular homeostasis may ultimately manifest as the memory deficits and learning alterations observed in previous studies14,60,61. Heat shock proteins (HSPs), particularly Hsp70, are widely recognized as key regulators of proteostasis and neuronal function. In model systems such as Drosophila melanogaster, Hsp70 has been shown to exert neuroprotective effects by reducing protein aggregation and supporting neuronal integrity62–64. Additionally, HSPs are dynamically regulated under stress conditions65,66 and respond to a wide array of environmental stressors67. In this context, the downregulation of Hsp70 observed in our study suggests a potential disruption of cellular stress-response mechanisms in DWV-A–infected bee heads. This effect is further supported by the significant enrichment of the Tetratricopeptide repeat domain (PF00515), characteristic of Stress-induced phosphoprotein 1 (Stip1/Hop), a co-chaperone that coordinates Hsp70 and Hsp90 activity to maintain proteostasis68,69. Although the role of Stip1 remains poorly characterized in honey bees, studies in other systems indicate that its homologs contribute to the clearance of misfolded proteins and cellular communication processes70,71. Therefore, the simultaneous downregulation of Hsp70 and Stip1/Hop may reflect a virus-associated impairment of the proteostatic network, potentially increasing neuronal vulnerability and contributing to the cognitive and behavioral deficits reported in DWV-infected bees14,50,61. Our Pfam domain analysis revealed significant enrichment of the CUB domain associated with the synaptic gene Neuropilin and tolloid-like (Neto), as well as the ligand-binding domain of kainate-type ionotropic glutamate receptors (Kainate), together with the altered expression of the glutamate transporter Excitatory amino acid transporter 2 (Eaat-2). These findings suggest that changes in genes associated with synaptic plasticity may not represent isolated transcriptional events, but could instead reflect the combined influence of several molecular processes associated with DWV-A infection. In this context, the deregulation of the transcription factors cwo and nau may indicate alterations in neuronal maturation and physiological homeostasis. At the same time, the enrichment of genes associated with metabolic processes and the downregulation of Hsp70 and Stip1/Hop suggest that infected tissues may experience metabolic and proteostatic stress. Together, these alterations could contribute to changes in synaptic regulation and neuronal functionality during infection. The dynamic expression patterns of Neto, Eaat-2, and Kainate observed during infection are consistent with a possible biphasic response. The transient overexpression detected at 10 dpi may reflect a compensatory or homeostatic response aimed at maintaining synaptic activity under viral stress. In contrast, the reduced expression observed at 16 dpi in bees with high viral loads suggests that this response may not be sustained over time. Such dysregulation could potentially affect glutamatergic signaling, synaptic plasticity, and higher-order processes such as associative learning and olfactory memory, which have previously been reported to be altered in DWV-infected bees. In particular, the loss of synaptic plasticity may be strongly influenced by the changed expression of the Neto gene. Although poorly studied in A. mellifera, its ortholog in D. melanogaster has been shown to act as an essential co-regulator of ionotropic glutamate receptors (iGluRs) at the neuromuscular junction (NMJ), enabling iGluR clustering in postsynaptic densities and ensuring stable and functional synaptic transmission72,73. Neto is indispensable for invertebrate development: null mutations result in paralyzed embryos that never hatch, whereas hypomorphic alleles generate locomotor defects74. It localizes mainly in type I NMJs, together with iGluRs at postsynaptic densities74. Interesting.ly, in the absence of Neto, synapses form normally during the prepatterning stage, but iGluRs fail to cluster at postsynaptic regions, disrupting synapse maintenance and producing electrophysiological defects74. Moreover, in iGluR mutants, Neto aggregates do not form, highlighting a codependent relationship between the two75. Conversely, Neto overexpression, as identified in our study, causes iGluRs in adult D. melanogaster to accumulate at extrajunctional sites, leading to synaptic defects likely due to the trapping of iGluRs outside the synapse76. The iGluRs comprise three major subtypes of glutamatergic receptors: AMPARs, NMDARs, and kainate receptors77. Regulatory changes in these receptors, particularly kainate-type receptors, have been directly linked to synaptic plasticity and memory capacity, processes that have also been associated with neurodegenerative diseases such as Alzheimer’s disease in mammalian models78. Additionally, changes in Eaat-2 expression may also be involved, together with Neto and kainate receptors, in behavioral alterations, synaptic imbalance, and the loss of synaptic plasticity. The taurine/aspartate transporter Eaat-2 is the ortholog of vertebrate glutamate transporters51, and in D. melanogaster it has been described in glial cells of the central nervous system during larval stages52. More recently, its functional relevance in peripheral sensory processing and behavior has been reported: adults with Eaat-2 silenced in perineurial olfactory glial cells exhibited behavioral alterations, showing increased olfactory sensitivity when exposed to ethyl acetate (EA). This led to the proposal that Eaat-2 plays a key role in modulating OSN activity by regulating neuronal signal intensity and preventing excessive responses13. Altogether, our results support the idea that DWV-A infection profoundly disrupts glutamatergic homeostasis in honey bees through coordinated and dynamic alterations in Neto, Eaat-2 and ionotropic glutamate receptors, Kainate. The early induction and subsequent collapse of this regulatory network suggest a transient compensatory attempt that ultimately fails as viral loads increase, giving rise to a state of synaptic dysregulation. This imbalance is likely to compromise the structural and functional integrity of synapses, reduce neuronal plasticity, and contribute to the sensory and cognitive impairments reported in DWV-infected bees. As recently reported by Ferreira et al.79, DWV-A infection triggers a cascade of behavioral changes in honey bees, including precocious foraging and impaired resource collection, that directly impact pollination services and agronomic stability. Our results suggest that the neurophysiological alterations identified in this work provide a plausible mechanistic explanation for such phenomena, acting as the bridge between viral infection and impaired colony performance. Within this context, Neto, Eaat-2 and Kainate emerge not only as molecular targets of DWV-A infection, but also as potential nodes linking viral replication to synaptic dysfunction and, ultimately, may contribute to colony-level impairments such as reduced foraging efficiency, altered behavior, and decreased productivity. Concluding remarks In conclusion, oral DWV-A infection triggers a sequential transcriptomic reprogramming in Apis mellifera, characterized by a biphasic CNS response, involving Warburg-like metabolic activation at 10 dpi followed by homeostatic collapse and terminal sensory impairment in the antennae at 16 dpi. The desynchronization of maturation markers (nau and cwo) demonstrates that the virus dismantles the behavioral genetic program even under moderate viral loads. These findings reveal that subclinical neurobiological damage, rather than overt physical symptoms, is a key driver of colony functional decline, providing a novel molecular framework to understand hive collapse during persistent, asymptomatic infections. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1