Insulin signaling engages divergent transcriptional mechanisms in neural and metabolic tissues

Abstract

Background

Insulin signaling is a conserved regulator of growth, metabolism, and lifespan across metazoans. While its systemic roles are well established, the mechanisms by which insulin coordinates tissue-specific transcriptional programs that underlie distinct functional demands remain incompletely understood. In particular, the differential impact of reduced insulin signaling on different tissues has not been systematically explored.

Results

We performed a comparative transcriptomic analysis of Drosophila melanogaster olfactory sensory neurons (OSNs) and fat body (Fb) to investigate how reduced insulin signaling remodels gene expression in neural and metabolic tissues. Across both tissue types, insulin reduction suppressed key pathways involved in protein synthesis and mRNA surveillance, indicating shared regulatory responses. However, distinct tissue-specific transcriptional adaptations were also observed. In OSNs, insulin reduction led to the upregulation of synaptic and signaling genes, alongside the downregulation of proteostasis-related factors, suggesting enhanced neural plasticity that may come at the cost of long-term neuronal maintenance. In contrast, the Fb exhibited widespread metabolic suppression accompanied by feedback activation of stress-responsive insulin-like peptide genes, consistent with a shift toward hypometabolic adaptation. Network and pathway analyses revealed that these tissue-specific responses involved distinct regulatory architectures affecting core insulin pathway components and gene families.

Conclusions

Our findings demonstrate that reduced insulin signaling elicits both shared and divergent transcriptional programs in neural and metabolic tissues of Drosophila melanogaster. These findings reveal how insulin signaling orchestrates tissue-specific transcriptional landscapes that may underlie differential resilience or vulnerability to cognitive and metabolic decline.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12860-026-00577-9.

Background

Insulin signaling plays a central role in regulating growth, metabolism, stress responses, and lifespan across metazoans [1, 2]. This signaling pathway integrates nutritional, hormonal, and environmental cues to coordinate cellular and systemic functions that are essential for maintaining homeostasis [3, 4]. Dysregulation of insulin signaling is implicated in numerous human diseases, including diabetes, obesity, cancer, and neurodegeneration [5, 6], underscoring the importance of understanding the molecular mechanisms that govern insulin signaling. However, dissecting the intricacies of insulin signaling mechanisms and their differential impact on different tissues has been challenging.

Drosophila melanogaster has emerged as a powerful model for dissecting these mechanisms due to its conserved insulin-like signaling components, sophisticated genetic tools, and the ability to study tissue-specific and systemic regulation in vivo [3, 7–11]. The simplicity of the Drosophila insulin system—comprising a relatively small set of insulin-like peptides and well-characterized signaling effectors—provides an advantageous framework for uncovering fundamental principles of insulin signaling that are often obscured in more complex organisms [1, 7].

Although insulin signaling is often conceptualized as a systemic regulator coordinating organism-wide responses, increasing evidence suggests that it operates through distinct, tissue-specific mechanisms to meet the unique functional demands of different cell types [2, 7]. In Drosophila, as in other organisms, insulin-like peptides and their downstream effectors elicit diverse outcomes across tissues, balancing growth, metabolism, stress tolerance, and lifespan regulation in a context-dependent manner [4, 5, 7, 12]. However, the molecular basis of this tissue-specific modulation remains incompletely understood. We asked whether insulin signaling engages distinct regulatory mechanisms in different tissues, shaping specialized transcriptional programs to support the physiological roles of each tissue.

To answer this question, we leveraged available tissue-specific transcriptomic data from Drosophila olfactory sensory neurons (OSNs) and fat body (Fb) tissues subjected to reduced insulin signaling (Jain et al., unpublished http://GEO Accession [13]; http://GEO Accession. We selected OSNs and fat body because they represent archetypal neural [8, 9, 14] and metabolic [15] tissues, respectively, whose well-characterized roles in sensory modulation and nutrient homeostasis make them ideal for revealing how reduced insulin signaling elicits tissue-specific transcriptional programs of insulin action. By integrating differential gene expression analysis, pathway enrichment, and network-level approaches, we asked whether reducing insulin signaling has a similar or differential effect on gene expression and cellular pathways in these functionally divergent contexts. Our findings offer new insights into how insulin signaling coordinates neural and metabolic adaptation, laying the groundwork for understanding the broader physiological and pathological consequences of insulin pathway dysregulation.

Results

Differential expression analysis of tissue-specific datasets reveals both common and tissue-specific transcriptional responses to reduced insulin signaling

In the olfactory sensory neuron (OSN) dataset, differential expression analysis (padj < 0.05; fold-change threshold of |log₂FC| ≥ 1) identified 2410 differentially expressed genes (DEGs), which included 851 upregulated and 1559 downregulated genes (Fig. 1A). In the fat body (Fb) dataset, 873 DEGs were identified, which included 529 upregulated and 344 downregulated (Fig. 1B). However, when all genes with padj < 0.05 (no fold change cutoff) were considered, 7498 were significant in OSNs and 1452 in fat body (Fig. 1A, B). This gene set, defined by statistical significance rather than magnitude of change, was used consistently across all downstream analyses, including pathway enrichment and quadrant classification. Among these, 754 genes were differentially expressed in both tissues, while 6744 and 698 genes were uniquely expressed in OSNs and the fat body, respectively, upon InR knockdown in these tissues (Fig. 1C).

Fig. 1.

Distinct transcriptomic responses to reduced insulin signaling in Drosophila OSNs and fat body. Volcano plots showing differentially expressed genes (DEGs) in olfactory sensory neurons (A) and fat body (B). Genes with adjusted p < 0.05 and |log₂ Fold change| ≥ 1 are shown as upregulated (red) or downregulated (blue). Genes with adjusted p < 0.05 but 0 < |log₂ Fold change| < 1 are also highlighted and plotted separately (coral for positive Fold change; soft blue for negative Fold change). Only genes with adjusted p ≥ 0.05 are shown in grey. (C) Venn diagram of DEGs (adjusted p < 0.05), showing OSN-specific (pink), fat body–specific (blue), and shared (purple) genes. (D) Scatterplot of log₂ Fold changes for 754 shared DEGs. Colors indicate regulation patterns: down–down (blue; 257 DEGs), up–up (red; 90 DEGs), OSNs–fb/down–up (purple; 357 DEGs), OSNs–fb/up–down (orange; 50 DEGs); Pearson r = 0.22, p = 9.7 × 10⁻¹⁰

To further investigate the variability of transcriptional response, we examined the 754 DEGs common to both OSNs and Fb. Directional categorization of the shared 754 DEGs showed that 257 genes were downregulated in both tissues (Down–Down (blue)), 90 were upregulated in both (Up–Up (red)), 357 were downregulated in OSNs but upregulated in the fat body (Down–Up (purple)), and 50 were upregulated in OSNs but downregulated in fat body (Up–Down (yellow)) (Fig. 1D). The opposing regulation of shared DEGs suggests a model in which reduced insulin signaling bifurcates into tissue-specific downstream programs. We hypothesized that a subset of these genes functions as regulatory “decision nodes,” allowing the same upstream insulin cue to be interpreted differently depending on tissue context.

To test this hypothesis and highlight key regulation patterns, we generated heatmaps showing the top 30 genes in each regulatory category (Supplementary Fig. 1A-D). Genes downregulated in both OSNs and Fb (Down–Down) included core metabolic and stress-related genes such as Gapdh2 (glycolysis) [2], glutathione S-transferases (GstE7, GstD1, GstD9) [16], and structural or immune genes like Spds, and Tsf-1 [17–19], suggesting that reducing insulin signaling leads to a shared suppression of glycolytic and detoxification functions (Supplementary Fig. 1A). Among genes upregulated in both tissues (Up–Up), we found Sik2 (energy metabolism) [20], neprilysin family genes (Nep4, Nep7, Nepl12) [21], bw (involved in pigmentation and neural function) [22], Rgk2 (G-protein signaling), and several cytochrome P450s (Cyp6a22, Cyp309a2) [23], suggesting that reducing insulin signaling increases metabolic and signaling responses in both tissues (Supplementary Fig. 1D). Genes with opposing regulation between tissues revealed additional insights. For instance, Dcr-2 and chico were downregulated in OSNs but upregulated in Fb (Down–Up), possibly reflecting different roles in post-transcriptional regulation and insulin sensitivity [7, 24, 25]. This group also included immune genes (BomS2, BomBc3) [26, 27], cytoskeletal elements (alphaTub85E) [28], and translation-related genes (eIF3l, eIF3b) [29] (Supplementary Fig. 1B). Finally, some genes were upregulated in OSNs but downregulated in Fb (Up–Down). These included innate immune genes (PGRP-SB2, PGRP-SD), Yp3 (yolk protein), and neural/developmental genes (Jon66Cii, Nt5a, eya), indicating neuron-specific activation of immune and signaling pathways, while these same pathways were suppressed in the fat body [30–33] (Supplementary Fig. 1C). These observations support our hypothesis, which provides a mechanistic framework that can be directly tested through tissue-specific genetic manipulation of these shared but oppositely regulated genes.

Pathway-level analysis reveals both common and tissue-specific responses to reduced insulin signaling

Next, we conducted pathway enrichment analysis on significantly up- and downregulated genes from OSN and Fb tissue upon InR knockdown in these tissues [34–36]. We identified both shared and distinct pathway-level changes between the two tissues (Fig. 2).

Fig. 2.

Convergent and divergent pathway enrichment under reduced insulin signaling in OSNs and fat body. Top 20 significantly enriched KEGG and Reactome pathways (adjusted p < 0.05) from up- and downregulated genes in OSNs and fat body. Triangle orientation indicates direction of regulation; bubble size reflects significance (−log₁₀ adjusted p); fill color denotes functional category

A prominent shared feature of reducing insulin signaling in both OSNs and Fb was the downregulation of pathways related to protein synthesis and mRNA surveillance. Specifically, genes involved in cap-dependent translation initiation, ternary complex formation, ribosomal scanning, SRP-dependent co-translational targeting, and nonsense-mediated mRNA decay (both canonical and enhanced) were significantly repressed.

Beyond these shared effects, we noted distinct tissue-specific responses to reduced insulin signaling: 1) In OSNs, we observed significant enrichment of pathways related to neuroactive ligand–receptor interactions, G protein-coupled receptor (GPCR) signaling, calcium signaling, and processes associated with ciliary assembly and synaptic transmission. These results point to enhanced neuronal signaling activity, potentially suggesting a functional remodeling of sensory circuits under low insulin conditions [8, 10, 14, 37]. 2) In OSNs, genes associated with NCAM1-mediated cell adhesion, complement signaling, extracellular matrix remodeling, and organelle biogenesis were upregulated, while these structural and immune-related pathways were not significantly affected in the fat body tissue under low insulin conditions. 3) OSNs showed a notable reduction in genes involved in protein homeostasis and neuroprotective mechanisms, including pathways related to ubiquitination, post-translational modification, and protein turnover. Despite the upregulation of synaptic genes, this decrease in proteostasis-related pathways may suggest a reduced capacity for maintaining neuronal integrity under prolonged reduction in insulin signaling [38]. 4) The fat body tissue exhibited widespread downregulation of metabolic processes, including the tricarboxylic acid (TCA) cycle, oxidative phosphorylation, gluconeogenesis, glucose metabolism, lipid biosynthesis, and amino acid catabolism. These transcriptional changes suggest a broad suppression of both energy production and metabolic flux under low insulin conditions. [12]. Since enrichment was performed separately on upregulated and downregulated gene sets in each tissue, pathways shown in Fig. 2 represent coherent directional changes rather than mixtures of up- and downregulated genes, highlighting selective and modular regulation of pathway components under reduced insulin signaling.

Differential impact of reduced insulin signaling on core insulin signaling components across tissues

We examined the expression of 29 core genes in the insulin-like signaling cascade, as annotated in FlyBase (FlyBase GO: FBgg0000904) [39]. Among these, 18 genes were significantly differentially expressed (adjusted p < 0.05) in at least one tissue. Five genes were differentially expressed in both OSNs and Fb, ten were uniquely regulated in OSNs, and three were uniquely regulated in the Fb (Fig. 3A). This pattern indicates a partially overlapping but distinct deployment of core insulin signaling components across tissues.

Fig. 3.

Insulin signaling exhibits shared architecture but tissue-specific deployment across OSNs and fat body. (A) Venn diagram of differentially expressed insulin signalling genes (adjusted p < 0.05): OSN-specific (pink), fat body–specific (blue), shared (purple). (B) Scatterplot of log₂ Fold changes for shared insulin genes across tissues. Point size reflects mean |log₂ Fold change|; color indicates regulation pattern as in Fig. 1d(C) Circos plot of enriched gene Ontology terms (adjusted p < 0.05): biological process (red/orange), molecular function (teal), cellular component (blue/purple). Sector size denotes the number of terms; ribbon width, the number of genes; grey ribbons, shared terms. (D) Schematic of insulin–PI3K–Akt pathway transcriptional changes. Node color indicates tissue-specific significance: red (OSN), blue (fat body), and purple (both), based on adjusted p < 0.05. Triangle glyphs represent the direction of regulation - upward for upregulation, downward for downregulation

Next, we assessed whether the five insulin pathway genes shared between OSNs and Fb exhibit coordinated regulation by comparing their log₂ fold changes across tissues (Fig. 3B). Among these, Sik2 (red circle) was upregulated in both tissues, consistent with its role in energy stress [20, 40]. However, S6k, Pdk1, Chico, and Akt (purple circles) showed opposing regulation—down in OSNs, up in Fb. These findings highlight tissue-specific modulation of core insulin signaling components, potentially reflecting neuronal plasticity in OSNs versus metabolic scaling in the fat body [24, 41, 42].

Next, we performed Gene Ontology (GO) enrichment analysis on all significantly regulated genes (adjusted p < 0.05) in each tissue to investigate how transcriptional differences among shared insulin pathway genes translate into broader functional outcomes [43, 44]. We used a grouped chord diagram to visualize enrichment across the Biological Process (BP), Molecular Function (MF), and Cellular Component (CC) ontologies (Fig. 3C). OSNs exhibited broad and diverse enrichment across all three GO domains: Unique BP terms included vesicle trafficking, cytoskeleton organization, protein translation, and synaptic signaling. MF enrichments highlighted ATP binding and transcriptional regulation. CC-specific terms were enriched for mitochondrial and cytoskeletal components. In contrast, in the fat body, enrichment was more narrowly focused, dominated by metabolic process terms within the BP domain. No unique terms were identified in the MF or CC categories, and all remaining enriched terms overlapped with OSNs.

Finally, to integrate the gene-level expression changes (Fig. 3B) with functional enrichment patterns (Fig. 3C), we constructed a curated schematic of the insulin–PI3K–AKT pathway. We incorporated tissue-specific expression directionality and statistical significance based on FlyBase and Reactome annotations [36] (Fig. 3D). This diagram maps upstream insulin signaling components and downstream FOXO- and mTOR-linked targets across OSNs and Fb. While several insulin-like peptides (Ilp2, Ilp3, Ilp5, and Ilp7) showed no significant differential expression, Ilp6, a stress-responsive insulin-like peptide primarily expressed in the fat body, was strongly upregulated, consistent with its proposed role in systemic feedback during nutrient stress [45]. In OSNs, Foxo was downregulated alongside suppression of several FOXO-regulated pathways identified in the GO analysis. Key signaling intermediates such as Chico and Pdk1 were also downregulated in OSNs but upregulated in Fb, suggesting divergent upstream regulation. Akt showed mild upregulation in OSNs, indicating potential decoupling of downstream branches. mTOR-associated components exhibited clear tissue-specific patterns: S6k, Myc, and CycD were downregulated in OSNs but maintained or upregulated in Fb, reflecting selective repression of anabolic growth in neurons. Additionally, sgg (GSK3β), a node linking metabolism and neurodegeneration, was significantly downregulated in OSNs, possibly reflecting a stress-adaptive response. Together, this pathway-level integration reveals that insulin signaling is not uniformly downregulated but rather restructured in a tissue-specific manner: OSNs emphasize FOXO-dependent rewiring and suppress TOR outputs, while the fat body retains mTOR activity and activates feedback mechanisms to maintain systemic metabolic control [46, 47].

Additionally, the divergent regulation of core insulin signaling components suggests potential systemic feedback effects. Fat-body upregulation of upstream signaling intermediates and Ilp6 may help stabilize organismal metabolic homeostasis, while suppression of TOR- and FOXO-linked outputs in OSNs may reduce neuronal anabolic demand and indirectly influence whole-animal energy allocation during nutrient stress. Together, they suggest a testable hypothesis that tissue-specific restructuring of insulin signaling not only serves local functional needs but also contributes to cross-tissue coordination during nutrient stress.

Differential impact of reduced insulin signaling on five gene groups across tissues

We profiled the expression of five key gene groups—Transcription Factors, Kinases, Phosphatases, Ion Channels, and GPCRs—to assess how insulin signaling affects cellular programs in OSN and Fb tissues.

Transcription Factors: Out of 628 curated TFs, 345 were differentially expressed in OSNs, compared to 20 in Fb, with only 14 shared (Supplementary Fig. 2A). A scatterplot of the shared TFs revealed tissue-specific regulatory trajectories (Supplementary Fig. 2B). OSNs showed greater TF family diversity, including Forkhead, Sp1/KLF, Runt, and NR-type factors, while Fb expression favored NR5A, Dorsal-class, and immune-related TFs (Supplementary Fig. 2C).

Kinases: Among 189 annotated kinases, 143 were differentially expressed in OSNs and 48 in Fb, with 28 shared (Supplementary Fig. 3A). A scatterplot of the shared kinases revealed tissue-specific regulatory trajectories, with several canonical insulin effectors (e.g., Akt, S6k, Pdk1) downregulated in OSNs but upregulated in Fb (Supplementary Fig. 3B). OSNs preferentially expressed MAPK, CK1, and GSK3-like families, while Fb expressed energy-related kinases like glycerol and adenylyl transferases (Supplementary Fig. 3C).

Phosphatases: Of 180 phosphatases analyzed, 67 were altered in OSNs and only 9 in Fb, with 12 shared (Supplementary Fig. 4A). A scatterplot of the shared phosphatases revealed tissue-specific regulatory trajectories (Supplementary Fig. 4B). OSNs predominantly expressed phosphatases involved in mitochondrial import, RNA turnover, and cytoskeletal control, while Fb favored RNA exonucleases and phosphatases regulating extracellular signaling (Supplementary Fig. 4C).

Ion Channels: Out of 262 ion channels, 133 were differentially expressed in OSNs versus 8 in Fb, with 6 shared (Supplementary Fig. 5A). A scatterplot of the shared ion channels revealed tissue-specific regulatory trajectories, with most channels upregulated in both tissues (Supplementary Fig. 5B). OSNs uniquely expressed voltage-gated, TRP, Pickpocket, and ligand-gated channels, while Fb retained only minimal ionic components (Supplementary Fig. 5C).

GPCRs: Among 117 GPCRs, 79 were differentially expressed in OSNs and 16 in Fb, with 10 shared (Supplementary Fig. 6A). A scatterplot of the shared GPCRs revealed tissue-specific regulatory trajectories (Supplementary Fig. 6B). OSNs showed enrichment of neuromodulatory GPCRs (e.g., muscarinic, dopamine, octopamine receptors), while Fb expressed endocrine-related Class B GPCRs (Supplementary Fig. 6C).

Functional associations among genes oppositely regulated in OSNs and fat body

To investigate the functional associations among genes oppositely regulated in OSN and Fb tissues under low insulin signaling, we constructed a protein–protein interaction (PPI) network [48]. Out of 754 shared DEGs, 407 (~54%) were identified as oppositely regulated between OSNs and Fb. Among these, 345 gene products (≈85%) formed a connected PPI network (Fig. 4), indicative of extensive functional association.

Fig. 4.

Protein–protein interaction network of oppositely regulated genes in OSNs and fat bodies under insulin signaling. Network of 345 genes oppositely regulated in OSNs and the fat body. Nodes represent proteins; edges, interactions. Node color indicates functional category: red, insulin signaling; green, ion channels; orange, kinases; magenta, phosphatases; blue, transcription factors; yellow, GPCRs. Node shape shows OSN regulation (upward triangle, upregulated; downward triangle, downregulated); size reflects degree centrality. Top 10 hub genes in bold

Network topology analysis revealed 10 hub genes with the highest degree of connectivity: RpS13, eIF4A, Rad23, RpL4, RpL40, pix, eIF3b, Gp93, RpL27, and Akt1 (Supplementary Table 1). All were consistently downregulated in neurons and upregulated in Fb tissue. Many of these hubs encode core components of the translational machinery, including multiple ribosomal proteins and translation initiation factors. Additional hubs included proteins involved in ubiquitin-mediated proteolysis (Rad23), endoplasmic reticulum chaperoning (Gp93), and insulin signaling (Akt1).

Functional classification of the 345 interconnected proteins revealed distinct patterns of expression across key gene categories. Core insulin signaling components (chico, Akt, InR, Pdk1, S6k) were uniformly downregulated in neurons. Similar expression trends were observed for ion channels (Stim, Tmem63), multiple kinases (babo, CG16898, CkIIalpha, Egfr, Hipk, Ire1, MAPk-Ak2, Pdk, put, Pvr, wnd), and several phosphatases (CG11050, CG6707, CG9449, NT5E-2, Ptp69D). In contrast, CG5577 and eya were the only phosphatases found to be upregulated in neurons. Most transcription factors within the network (cnc, CrebA, dbr, Mondo, Rel, REPTOR, Smox, SREBP) were downregulated in neurons, while pnr was uniquely upregulated. Additionally, the GPCR, fz2, exhibited decreased expression in neurons (Fig. 4).

Discussion

Our comparative bioinformatics analysis of tissue-specific datasets revealed that reducing insulin signaling triggers both common and tissue-specific transcriptional responses in Drosophila olfactory sensory neurons (OSNs) and fat body (Fb) tissue. Upon reducing insulin signaling specifically in either OSNs or fat bodies, both tissues exhibited downregulation of pathways related to glycolytic and detoxification functions, as well as protein synthesis and mRNA surveillance (Figs. 1, 2). However, reducing insulin signaling also led to broader responses that diverged substantially: OSNs showed enhanced expression of genes involved in neuronal signaling, synaptic function, and structural remodeling, alongside suppression of proteostasis mechanisms, suggesting neuronal adaptation, and is consistent with potential vulnerability of neurons observed under low insulin conditions (Fig. 2). In contrast, the fat body exhibited a pronounced downregulation of metabolic pathways, indicative of a systemic shift toward hypometabolism (Fig. 2). Furthermore, our data highlight the distinct modulation of core insulin signaling components across tissues, with OSNs emphasizing FOXO-dependent rewiring and suppression of mTOR outputs, whereas the fat body maintains mTOR activity and activates compensatory feedback mechanisms (Fig. 3). The divergent regulation of key gene families—including transcription factors, kinases, phosphatases, ion channels, and GPCRs—further underscores the tissue-specific strategies employed in response to insulin disruption (Supplementary Figures 2–6). Finally, network analysis of oppositely regulated genes revealed functional associations centered on translational control and insulin signaling. The apparent discrepancy between global translation pathway suppression (Fig. 2) and oppositely regulated translational hubs (Fig. 4) reflects regulation at different levels of organization. While this could suggest a coordinated yet contrasting remodeling of cellular programs between the nervous and metabolic systems, this would need to be tested with further experiments.

Our results strongly support our working hypothesis that reduced insulin signaling engages distinct regulatory mechanisms in different tissues. First, the tissue-specific regulation of core insulin pathway components illustrates how a shared hormonal signal can elicit divergent molecular responses to meet the functional demands of different tissues (Fig. 3) [20, 24, 40–42, 46, 47]. Second, our data specific to OSNs suggest that reduced insulin signaling prompts neural circuit remodeling, potentially as a compensatory response to maintain sensory function under nutrient stress (Figs. 1 and 2). Third, our data specific to Fb indicate a broad repression of metabolic genes in response to reduced insulin signaling, consistent with Fb’s role in coordinating systemic energy balance (Figs. 1, 2). Fourth, Fb had a more restricted transcriptional response compared to OSNs (Figs. 1C, 2, and 3C). This suggests a limited, energy-conserving transcriptional response to insulin suppression in Fb compared to OSNs [12, 49–51]. Engaging distinct mechanisms in OSNs and Fb reflects the diverse physiological roles of these tissues [9, 12, 14, 52]. Together, these data support the idea that insulin signaling is not a uniform modulator of gene expression, but rather a context-dependent regulator fine-tuned to the unique requirements of different tissues.

The opposing regulation of 407 shared DEGs suggests that reduced insulin signaling triggers a bifurcation into tissue-specific transcriptional strategies. For example, the insulin-receptor substrate chico is strongly downregulated in OSNs but upregulated in fat body, consistent with a model in which neurons suppress nutrient-responsive signaling to promote sensory plasticity, whereas fat body compensates to maintain metabolic responsiveness [3]. Similarly, opposing patterns in translation-related genes (e.g., eIF3l, eIF3b) imply that each tissue deploys tailored translational control mechanisms under nutrient stress [53]. Together, these patterns suggest that divergently regulated genes act as regulatory “decision nodes,” reflecting biologically meaningful tissue-specific responses to reduced insulin signaling.

Our results align with and extend previous studies demonstrating tissue-specific roles of insulin signaling in Drosophila and other organisms. In neurons, insulin-dependent modulation has been previously linked to synaptic plasticity, neuroprotection, and sensory function [8, 14]. Our findings that reduced insulin signaling enhances the expression of genes involved in synaptic signaling and GPCR-mediated pathways in OSNs align with these prior findings. The tissue-specific regulation of mTOR and FOXO targets observed in our study mirrors prior reports that insulin pathway components can differentially engage downstream effectors to balance growth, metabolism, and stress responses [2, 42]. Our findings regarding the transcriptional landscape in the Fb also align with prior work, which has demonstrated that insulin signaling in the Fb is crucial for systemic metabolic homeostasis, with reductions in pathway activity resulting in suppressed lipid and carbohydrate metabolism [12, 45]. Similarly, the upregulation of Ilp6 that we detected in the Fb supports its established role as a feedback signal during nutrient stress [45]. By combining gene-level, pathway-level, and network-level analyses, our study further builds upon these prior observations, offering new insights into how insulin signaling orchestrates tissue-specific gene regulatory programs.

The tissue-specific transcriptional responses we uncovered suggest that insulin signaling serves distinct functional roles across tissues—enhancing sensory readiness in OSNs while supporting metabolic function in the fat body. In OSNs, reduced insulin signaling appears to promote neural plasticity through upregulation of synaptic and signaling genes, potentially sustaining sensory processing during nutrient stress. However, the downregulation of proteostasis and neuroprotective pathways could predispose neurons to damage over time, linking insulin dysregulation to cognitive decline and neurodegenerative vulnerability. In the fat body, broad suppression of metabolic and biosynthetic pathways reflects a shift toward hypometabolism, likely an energy-conserving adaptation, while feedback activation of stress-responsive insulin-like peptides such as Ilp6 may help maintain systemic homeostasis.

The markedly larger number of DEGs in OSNs relative to fat body likely reflects intrinsic differences in cellular complexity and transcriptional responsiveness between these tissues: neurons routinely exhibit broad transcriptomic rewiring due to their integrative sensory role, whereas the fat body mounts more focused metabolic responses [1, 2, 15, 54, 55]. Importantly, despite this scale difference, both tissues converge on core insulin-signaling principles—including mTOR suppression, translation downregulation, and insulin-peptide feedback—indicating that the numerical disparity reflects tissue-specific transcriptional architecture rather than fundamentally different modes of insulin-pathway engagement.

Our findings underscore the importance of tissue context in determining the outcomes of insulin signaling dysregulation and its relevance to aging, disease susceptibility, and environmental adaptation. Insulin and insulin-like signaling pathways are conserved regulators of lifespan across species, with tissue-specific modulation implicated in both protective and deleterious outcomes [2]. While some of the changes we observe may help tissues adapt to reduced insulin signaling, long-term or severe reduction could also have harmful effects, such as impaired metabolic function in the fat body or gradual loss of neuronal integrity leading to neurodegenerative conditions. Our results, therefore, likely capture both helpful short-term adaptations and possible longer-term pathological risks, depending on the extent and duration of insulin signaling disruption. The neuronal suppression of FOXO- and mTOR-linked pathways we identified may contribute to neuroprotective adaptations that extend lifespan under conditions of nutrient stress, as suggested by studies linking reduced insulin signaling to delayed neurodegeneration [56, 57]. In metabolic tissues like the Fb, the shift toward hypometabolism and feedback activation of stress-responsive insulin-like peptides (Ilp6) may reflect an adaptive strategy to cope with environmental challenges such as food scarcity, but could also parallel pathological states like cachexia or metabolic syndrome when dysregulated [58, 59]. Thus, strategically targeting tissue-specific insulin mechanisms may have more favorable outcomes while treating specific diseases or conditions.

While our study provides examples of tissue-specific transcriptional responses to reduced insulin signaling, several limitations and open questions remain. Because our analysis is based solely on RNA-level changes, future work should directly test the causal roles of the most strongly and oppositely regulated insulin-pathway components—such as Akt, chico, and Pdk1—using tissue-specific knock-in/knock-out or rescue approaches to determine how these genes differentially govern neuronal plasticity versus metabolic scaling. Targeted proteomic and metabolomic profiling in OSNs and fat body would further clarify whether divergent transcriptional programs translate into similarly divergent signaling outputs and physiological states. Expanding these mechanistic tests to additional tissues (e.g., muscle, gut, reproductive organs) will help define whether the bifurcation we observe represents a broader organizational principle of insulin signaling. Finally, longitudinal experiments examining how these tissue-specific programs change across aging or environmental stress will be essential for understanding when reduced insulin signaling drives adaptive remodeling versus pathological decline.

A key limitation of this study is that the OSN and fat-body transcriptomes were generated in separate experiments using different tissue-enrichment strategies. Although we re-processed all raw reads through a unified computational pipeline and the datasets are matched in developmental stage, genetic manipulation, and sequencing platform, not all transcriptional differences can be attributed solely to tissue identity. These comparisons should therefore be viewed as hypothesis-generating rather than fully controlled cross-tissue experiments. Future work using co-isogenic lines, standardized husbandry, and parallel sample collection will be needed to confirm our findings.

We also acknowledge that tissue-specific perturbations—particularly in neurons—can produce systemic, cell-non-autonomous effects. However, since this study analyzed tissue-specific RNA datasets after tissue-specific knockdown of insulin signaling, analyzing the impact of reduced insulin signaling in one tissue on gene expression in other larval tissues is beyond the scope of this study.

Conclusions

In summary, our findings demonstrate that insulin signaling drives distinct, tissue-specific transcriptional programs that reflect the unique functional demands of the tissue. The results presented here for OSNs and FB tissues advance our understanding of how insulin signaling coordinates organismal adaptation to metabolic stress and lay the groundwork for future studies of its role in aging and disease.

Methods

Comparative transcriptome datasets

Comparative transcriptomic analysis was performed on two independent RNA-seq datasets: (i) insulin receptor (InR) knockdown in Olfactory Sensory Neurons (OSNs) of the Drosophila melanogaster third instar larvae, (ii) InR knockdown of a Drosophila melanogaster fat body dataset. The OSN dataset (Orco-Gal4 > UAS-InR-RNAi) was generated in-house and submitted to NCBI under BioProject PRJNA1187561 (reviewer access link available upon request). The fat body (Fb) dataset (r4-GAL4 > UAS-InRi) was previously published [13] (GEO: GSE97447; BioProject: PRJNA381810). Both datasets consisted of three biological replicates per condition (control and InR knockdown), totaling 50 single-end Illumina RNA-seq libraries.

Preprocessing and quality control of transcriptome datasets

For both data sets, raw FASTQ files were processed using a unified analysis pipeline by the Nevada Bioinformatics Center (RRID:SCR_017802). Adapter trimming and quality filtering were performed using fastp (v0.20.0) [60]. Read quality was evaluated using FastQC (v0.11.9) and summarized with MultiQC (v1.12) [61]. Reads were aligned to the D. melanogaster reference genome (FlyBase r6.60, FB2024_05) using STAR (v2.7) [62]. Gene-level quantification was performed using featureCounts (v2.0.0) [63].

To ensure the suitability of both datasets for downstream differential expression and functional pathway analysis, low-abundance genes (fewer than 10 counts across all samples) and two outlier samples from the OSN dataset with poor complexity were excluded. This resulted in 15,396 genes for the OSN dataset and 7980 for the fat body dataset. Substantially more expressed genes were retained in OSNs (15,396) than in fat body tissue (7,980), reflecting both biological complexity and technical factors. Such differences are not uncommon in RNA-seq experiments comparing tissues with inherently distinct transcriptional architectures and sequencing efficiencies [64, 65]. OSNs likely exhibit greater transcriptional heterogeneity due to the involvement of diverse neuronal subtypes, a broad array of molecular targets, including ion channels, receptors, and signaling molecules, as well as various functions [66]. In contrast, the fat body tissue is more homogeneous and metabolically specialized [12, 15].

Differential expression analysis

Each dataset was analysed independently using DESeq2’s (v1.44.1) internal size factor estimation [67]. Differentially expressed genes (DEGs) were identified using an adjusted p-value < 0.05 (Benjamini–Hochberg correction) [68]. For more stringent comparisons, an additional fold-change threshold of |log₂FC| ≥ 1 was applied. Full differential expression tables and DESeq2 results are available on [http://GitHub].

Gene overlap and pairwise expression analysis and visualizations

DEGs shared between the OSN and fat body datasets (padj < 0.05) were classified into four regulatory (OSN–Fb) categories: Up–Up, Down–Down, Up–Down, and Down–Up. Gene-wise log₂ fold changes were plotted in two-dimensional space and color-coded by regulatory pattern. The top five genes per quadrant were labeled using ggrepel [69]. Visualizations were generated using ggplot2 [70], scales, and base R. This analysis served as the basis for subsequent analyses involving pattern classification, Venn diagrams, and quadrant-specific heatmaps. For each regulatory category (Up–Up, Down–Down, Up–Down, Down–Up), the top 30 genes were selected based on a combined significance score calculated as –log₁₀ (padj_OSNs + padj_Fatbody), prioritizing genes that were statistically significant in both tissues regardless of fold change magnitude. These genes were then visualized using log₂ fold changes and –log₁₀(padj) values in clustered heatmaps. Heatmaps for each regulation pattern (Up–Up, Down–Down, Up–Down, Down–Up) were generated using pheatmap (v1.0.12), with rows clustered using Euclidean distance. All code and output files are available on [http://GitHub].

Functional pathway enrichment and visualization

First, for each dataset, genes were separated into upregulated and downregulated subsets. Next, gene symbols were mapped to Entrez IDs using bitr() from org.Dm.eg.db. Finally, functional pathway enrichment was carried out using: enrichKEGG() (KEGG API) [35, 71], enrichPathway() (ReactomePA [72]), and enricher() (MSigDB C2 KEGG gene sets) [73] with hypergeometric testing and Benjamini–Hochberg adjustment [68]. Redundant terms were filtered using stringdist (Jaro-Winkler similarity > 0.8) [74]. All enrichment tables and scripts are available on [http://GitHub].

The top 20 significantly enriched pathways (padj < 0.05) from KEGG and Reactome were visualized using ggplot2 (v3.5.1) as triangular bubble plots. Pathways were grouped into biological themes (e.g., “Mitochondrial Function,” “Protein Synthesis”), trimmed to ≤ 40 characters, and styled with triangle orientation (indicating up- or down-regulation), fill color (indicating category), and bubble size (indicating –log₁₀(padj) values). Code and outputs are available on [http://GitHub].

Gene Ontology (GO) enrichment and visualization

GO enrichment was performed using topGO (v2.56.0) [75–77] with the elim algorithm and Fisher’s exact test. Backgrounds included all expressed genes (≥10 counts) per tissue. p-values were adjusted using the Benjamini–Hochberg method [68]. The significant GO terms (padj < 0.05) were manually grouped into high-level supercategories using stringr-based pattern matching.

Visualizations were created using circlize (v0.4.15) [78, 79], with ribbon width representing term count and color-coded sectors for ontology class (Biological Process = red/orange, Mitochondrial Function = teal, Cellular Components = blue/purple). Scripts and term mappings are available on [http://GitHub].

Analysis of core insulin signaling components

Canonical insulin signaling pathway genes (FlyBase GO: FBgg0000904) were identified and compared across OSNs and fat body. Differentially expressed insulin-related genes were classified using the same regulation pattern and plotting framework described in the gene overlap and pairwise expression analysis pipeline. Scatter plots were scaled by average |log₂FC| and color-coded by directionality. Gene overlaps were visualized using VennDiagram [80], and Scatterplots for pairwise analysis were completed using the same code base. Figures and outputs are available on [http://GitHub].

Pathway mapping of insulin signaling

Canonical insulin signaling pathway genes were visualized using a Reactome-style schematic in DiagrammeR (v1.0.10) [81]. Nodes were styled by tissue-specific significance (fill color: Red-OSNs, Blue-FB, Purple-OSNs + Fb) and regulation pattern (with triangular glyphs indicating up/downregulation).

Analysis of key regulatory gene families

Curated gene families from FlyBase were analyzed: (i) Transcription factors (FBgg0000745), (ii) Kinases (FBgg0001142), (iii) Phosphatases (FBgg0000268), (iv) Ion channels (FBgg0000582), and (v) GPCRs (FBgg0000172).

DEGs were classified as OSN-specific, fat body-specific, or shared. Regulation patterns and Venn overlaps were analyzed using the same R-based framework described in the section on gene overlaps. Shared genes were visualized via bubble plots (average |log₂FC|, colored by pattern). Top five families were highlighted in circlize-based chord diagrams [79], with segments representing families and tissues. All visualizations and annotation files are available on [http://GitHub].

Protein–protein interaction (PPI) network analysis

PPI networks were constructed using Cytoscape (v3.10.3) [48] with stringApp (v2.2.0) [82]. Genes showing inverse regulation patterns (Up–Down or Down–Up) were identified using the same classification pipeline described above and queried against STRING v11.5 [83] using a confidence score ≥ 0.4. Interactions were filtered for experimental and curated evidence. Network topology analysis was conducted using the built-in NetworkAnalyzer tool within Cytoscape. This analysis identified hub genes based on degree. The top 10 genes ranked by degree were identified as key hub genes. The Functional annotation of the hub genes was obtained using the UniProt database [84]. Networks were visualized using a yFiles Layout, and nodes were annotated by FlyBase gene families. Files and network data are available on [http://GitHub].

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

RJ undertook the majority of the bioinformatic analysis in consultation with CH and JP, who also contributed to the analysis strategy. RK contributed significantly to the bioinformatic analysis, particularly the functional association analysis data presented in Fig. 4 and Supplementary Table 1. Both RJ and RK assisted in preparing the figures. DM conceived the study, consulted on the analysis strategy, and wrote the manuscript with input from RJ and RK. All authors read and approved the final manuscript.

Funding

This project was supported by grants from the National Institute of General Medical Sciences (GM103440) of the NIH and from the NSF under award number 2341202.

Data availability

OSN RNA-seq dataset: NCBI BioProject http://PRJNA1187561. Fat body dataset: GEO accession http://GSE97447. All preprocessing outputs, quality control reports (FastQC, MultiQC), and DESeq2 differential expression analysis were completed by the Nevada Bioinformatic Center (NBC) (RRID:SCR_017802). Codes related to these steps are available from the NBC upon request. All the downstream analysis codes, including ones for gene classification, enrichment results, visualizations, and network analyses, were performed in-house and are fully shared on our [http://GitHub] page.

Declarations

Ethics approval

Not applicable.

Competing interests

The authors declare no competing interests.