Adenosine-Specific Transcriptional Programs in Murine Connective Tissue-Type Mast Cells

Abstract

Mast cells are tissue-resident immune cells that are critical for the pathogenesis of allergic and inflammatory disorders. Their physiological functions include host defense against parasites and, more recently, food quality control through antigen avoidance. The purine nucleoside adenosine (ADO), like other mast cell activators, such as antigens or Mrgprb2 agonists, increases intracellular Ca²⁺ concentration; however, it fails to induce degranulation of preformed mediators when applied to mast cells alone, and there is limited knowledge about whether ADO evokes the de novo synthesis and release of inflammatory mediators in tissue mast cells. An unbiased genome-wide analysis of gene expression triggered by various mast cell activators should enable the identification of the gene program specifically activated by ADO in mast cells and thereby reveal new components of the associated inflammatory responses. Here, we performed bulk RNA sequencing on primary murine peritoneal mast cells (PMCs) representing connective tissue mast cells. By comparing responses evoked by ADO stimulation with those of the Mrgprb2 agonist compound 48/80 and antigens activating FcεRI receptors, we identified 393 genes uniquely regulated by ADO, including genes encoding the de novo synthesized mediators transforming growth factor α and interleukin 7. Transcription factor activity inference, protein classification, functional enrichment analysis, protein interaction network analysis, and topology analysis revealed a distinct ADO-specific transcriptional gene program involved in phosphoinositide signaling, vesicle trafficking, glycolysis, mitochondrial activity, and cell cycle arrest. The functional relevance of the identified de novo synthesized mediators for ADO-evoked inflammatory reactions can be evaluated in future studies.

1. Introduction

Mast cells are ancient, evolutionarily conserved, resident cells in many tissues. They are located at the host’s environmental interfaces, such as the skin, respiratory tract, and gastrointestinal lining, and act as critical sentinels of the immune system. , While traditionally viewed for their detrimental roles in allergic disorders and asthma, , as well as in itch and food allergy, mast cells have increasingly been recognized for their beneficial functions in wound healing, host defense, and antigen avoidance. −

A morphologic feature of mast cells is their abundance of electron-dense secretory granules, which contain large amounts of preformed compounds, including biogenic amines (histamine and serotonin), , specific preformed cytokines (for example, tumor necrosis factor and vascular endothelial growth factor), , serglycin proteoglycans, various lysosomal enzymes, and many mast cell-specific proteases. , In mice, mast cells are classified into two major subtypes: connective tissue mast cells (CTMCs) and mucosal mast cells (MMCs). Peritoneal mast cells, the focus of our study, correspond to CTMC subtype. ,− CTMCs predominantly express two distinct chymases: the β-chymase, mouse mast cell protease 4 (mMCP-4), and the α-chymase, mMCP-5. In addition, they express tryptases mMCP-6 and mMCP-7, as well as carboxypeptidase A3 (CPA3). −

Mast cell activation follows a triphasic cascade. The immediate response, occurring within seconds to minutes, involves the degranulation and release of the above-mentioned preformed compounds into the extracellular space. The rapid intermediate-phase synthesis of lipid mediators follows this. To this end, enzymes process membrane phospholipids to generate arachidonic acid derivatives, such as leukotrienes and prostaglandins. Finally, the late-phase response is initiated as transcription factors, notably NFAT and NF-κB, drive the de novo synthesis of a wide array of cytokines and chemokines. ,

Extracellular adenosine is an important signaling molecule in the extracellular environment, acting as a neuromodulator and regulator of immune/cardiovascular functions. Canonically, extracellular ADO is produced from ATP through the enzymatic activities of CD39, which dephosphorylates ATP and ADP to produce AMP, followed by subsequent conversion to ADO by CD73. ATP can be released from immune cells via pannexin-1 (PANX1) hemichannels, where it activates P2 × 4 and P2 × 7 channels in an autocrine manner, resulting in an influx of Ca²⁺. However, ATP levels can increase in the extracellular space due to the release of ATP from dying or damaged cells. Thus, cells expressing both CD39 and CD73 possess the enzymatic capacity to generate extracellular ADO, as demonstrated in T cells, B cells, and macrophages.

Adenosine (ADO) signals to cells in an autocrine or paracrine manner. Mast cells show either anti-inflammatory or proinflammatory responses upon ADO stimulation, depending on the receptor engaged. There are four distinct G-protein-coupled receptor subtypes for ADO: A1, A2a, A2b, and A3. Of these, A2a and A2b receptors are coupled to G alpha (s) protein. They activate adenylyl cyclase (AC) and increase the level of cAMP production. In contrast, A1 and A3 receptors inhibit AC and are coupled to G alpha (i), leading to a reduction in cAMP production. Moreover, A2b and A3 can couple to G alpha (q) protein, activating phospholipase C (PLC). PLC catalyzes the hydrolysis of phosphatidylinositol 4,5-bisphosphate (PIP2) into diacylglycerol (DAG) and inositol 1,4,5-trisphosphate (IP3), with IP3 initiating Ca²⁺ release from the endoplasmic reticulum (ER). As with other Ca²⁺-mobilizing pathways, ER depletion triggers calcium entry. In a mast cell/basophil cell line, ADO receptor agonist NECA was shown to activate a class of Ca²⁺ entry channels named “Calcium Release Activated channels” (CRAC). Activation of A2a is predominantly reported to be anti-inflammatory, whereas signaling through A1, A2b, and A3 can be either pro- or anti-inflammatory. ,

ADO acts synergistically with canonical stimuli; when combined with an antigen challenge, it enhances FcεRI-mediated degranulation. However, ADO alone is insufficient to trigger a significant degranulation. ADO has been shown to potentiate mediator release from mast cells upon antigen stimulation. − In the human HMC-1 mast cell line, IL-8 release can be evoked by ADO-receptor stimulation; however, data from different research groups are controversial. −

The classical immunologic pathway for mast cell activation involves the binding of an antigen to IgE antibodies, which are already bound to the mast cell’s high-affinity receptor for IgE (FcεRI). This receptor activation induces downstream signaling cascades mediated by PLC. The subsequent depletion of intracellular Ca²⁺ activates Ca²⁺ influx through plasma membrane channels that partially depend on Orai1 expression, leading to a robust increase in cytosolic Ca²⁺ that drives degranulation and the de novo synthesis of inflammatory mediators. For example, antigen-evoked release of TNF-alpha and IL-6 was partially reduced in Orai1-deficient primary mast cells.

In addition, mast cells can degranulate via activation of the Mas-related G protein-coupled receptor b2 (Mrgprb2), the ortholog of human MRGPRX2. , Activation of Mrgprb2 initiates a signaling cascade involving activation of the PLC-IP3 axis, followed by an intracellular Ca²⁺ rise and mast cell degranulation, along with the de novo synthesis of inflammatory mediators.

The pathological significance of ADO signaling is evident in several human inflammatory disorders. Notably, inhaled ADO provokes bronchoconstriction in patients with asthma or chronic obstructive pulmonary disease but not in healthy individuals. Consistent with its role in cutaneous inflammation, plasma ADO levels have also been found to be significantly elevated in patients with chronic urticaria compared with healthy controls. These elevations persisted over a 1-month follow-up period. Moreover, animal studies have demonstrated that elevating ADO levels in mice can activate features of chronic diseases, such as pulmonary inflammation and airway remodeling. −

Despite its pathophysiological relevance, the ADO-mediated transcriptional program in primary mast cells remains largely unexplored. A search of PubMed Central using keywords related to adenosine, transcriptome, and mast cell identified only one relevant study, which performed transcriptomic profiling of the human mast cell line HMC-1 following ADO receptor stimulation. However, a comprehensive analysis of the ADO-specific transcriptional program in primary peritoneal mast cells (PMCs) is still lacking.

Here, we used bulk RNA sequencing to profile transcriptional programs in PMCs in response to three Ca²⁺-mobilizing agonists: ADO, compound 48/80 (C48/80), and antigen. As an ADO-specific response, we identified 393 genes that were differentially expressed exclusively under ADO treatment as compared with the other two stimuli. Our computational functional analysesincluding transcription factor activity, protein classification, functional enrichment, interaction network, and topology analysisshow involvement in phosphoinositide signaling, vesicle trafficking, glycolysis, mitochondrial function, and cell-cycle arrest. The approach confirmed known mast cell pathways and revealed ADO-specific induction of de novo mediators such as Tgfa and Il7, suggesting new regulatory components in ADO-driven inflammatory mast cell responses.

2. Results

2.1. ADO, C48/80, and DNP Induce Ca²⁺ Release Followed by Ca²⁺ Influx

To demonstrate that the PMC mast cell model functionally expresses the receptors and downstream signaling molecules necessary for Ca²⁺-dependent activation by ADO, C48/80, and antigens, we used Fura-2 microfluorometry to monitor changes in the intracellular Ca²⁺ concentration ([Ca²⁺]_i) in PMCs in response to any of the three agonists. To separate calcium release from intracellular stores and calcium entry across the plasma membrane, we applied the so-called “re-addition” protocol. In the absence of extracellular Ca²⁺, PMCs responded to all three agonists with a transient elevation of [Ca²⁺]_i resulting from calcium release from intracellular stores (Figure A and C). After [Ca²⁺]_i returned to the baseline level, extracellular calcium was “re-added” to its physiological concentration of 2 mM. It evoked a fast and prominent [Ca²⁺]_i rise due to the agonist-evoked Ca²⁺ entry. All three stimuli: adenosine (10 μM), compound 48/80 (50 μg/mL), and antigen (dinitrophenyl (DNP)-human serum albumin 100 ng/mL) elicited a reaction with a similar pattern of Ca²⁺ release and Ca²⁺ entry response (Figure A and C). These results demonstrated that all three agonists are able to evoke Ca²⁺-dependent activation pathways in PMCs.

1.

PMCs, Ca²⁺ release, followed by subsequent Ca²⁺ entry, is induced by ADO, C48/80, and DNP. Fura-2-based calcium imaging of PMCs following treatment with (A) adenosine (ADO, 10 μM), (B) compound 48/80 (C48/80, 50 μg/mL), and (C) dinitrophenyl-human serum albumin conjugate (DNP, 100 ng/mL). Changes in [Ca²⁺]_i over time are presented as the fluorescence ratio F340/F380 mean values of 3 independent preparations. The horizontal bars above the mean traces indicate the time of agonist applications, as well as the time of Ca²⁺ removal from the extracellular space (dotted bars) and the time of Ca²⁺ readdition (solid bars). Error bars represent the standard error of the mean.

2.2. Agonists’ Impact on Calcium Ion Channel Expression

All three tested agonists evoked calcium signals in the PMCs. We then analyzed the expression profile and agonist-evoked transcriptional changes in genes encoding calcium channel constituents and regulators across four stimulation conditions: ADO for 2 h, compound 48/80 (C48/80) for 6 h, and antigen (DNP) for 2 and 6 h. A complete summary of the expression levels under all conditions is provided in Table S1. In unstimulated PMCs, we found abundant expression (defined as DESeq2 normalized count >100) of the following genes encoding channels, subunits, and modulators: Cacnb4, Cacng7, Ryr3, Itpr1, Itpr2, Itpr3, Tpcn1, Tpcn2, Tmem63a, Tmem63b, Trpv2, Trpm2, Trpm4, Trpm7, Mcoln1, Pkd2, Orai1, Orai2, Orai3, Stim1, Stim2, P2rx1, P2rx4, P2rx7, Piezo1, Grin2c, Grin2d, and Calhm2. Some of these calcium channels, such as inositol 1,4,5-trisphosphate-gated calcium channel (ITPR1, ITPR2, ITPR3), Transmembrane protein TMEM63A, Transient receptor potential channels TRPV2, Stromal interaction molecule 1 (STIM1), and P2X purinoceptor (P2RX1, P2RX4, P2RX7), were also identified at the protein level in a proteome data set from unstimulated primary mouse peritoneal mast cells. Furthermore, ITPR1, ITPR2, Two-pore segment channel (TPCN1), TMEM63A, TRPV2, STIM1, P2RX1, and P2RX4 were identified in the proteome data set of both human skin and fat mast cells.

Among the 11 calcium channel genes differentially regulated by ADO, two were upregulated while nine were downregulated (Table S1). Notably, the NMDA receptor Grin2d showed the highest ADO-specific log₂ fold change (log₂FC = −1), representing the most strongly induced transcript. In contrast, the transient receptor potential channel Trpm4 was upregulated in both ADO (2h) and DNP (2h) treatments. A critical molecular regulator of Store-Operated Calcium Entry Stim1 was consistently downregulated in both ADO (2h) and DNP (2h) conditions. Tmem63b (also known as OCaR2) encoding for a mechanosensitive cation channel in laminar bodies of AT1 and AT2 cells , displayed upregulation across all agonist treatments, whereas P2X receptor genes P2rx4 and P2rx7 were uniformly downregulated across all conditions.

2.3. Transcriptional Expression of PMC Proteases

Given that mast cell proteases are among the most abundantly expressed transcripts, reaching or surpassing the levels of classical housekeeping genes, we examined the expression profiles of major murine mast cell proteases under CONTROL conditions (Figure S1). Mcpt1 (mMCP-1) transcripts were undetectable, and Mcpt2 (mMCP-2) expression remained minimal (Figure S1). In contrast, Cma1 (mMCP-5), Mcpt4 (mMCP-4), Tpsb2 (mMCP-6), Tpsab1 (mMCP-7), and Cpa3 (CPA3) were highly expressed (Figure S1). These results are consistent with previous findings about proteases expressed in CTMCs. , Among these, Cpa3 and Tpsb2 displayed the highest transcript levels (77 039 ± 11 727 and 82 027 ± 5376, respectively), followed by Cma1 (34 382 ± 1897), Tpsab1 (11 699 ± 2514), and Mcpt4 (7191 ± 502). These quantitative data demonstrate that Cpa3 and Tpsb2 dominate the protease transcriptome in PMCs under basal conditions. In addition, Ctsg (cathepsin G) and Prss34 (mast cell protease 11) were abundantly expressed, aligning with protein-level evidence reported in mouse connective tissue mast cells.

2.4. ADO Induces a Distinct Transcriptional Response in Mast Cells

To assess global transcriptional differences among treatment groups, a principal component analysis was performed (Figure S2). The x- and y-axes represent the variance explained by principal components (PC) 1 and 2, respectively. When all samples were analyzed together (Figure S2A), distinct clustering was observed between the CONTROL and treated groups, with PC1 accounting for 31.44% of the total variance. ADO-, C48/80-, and DNP-treated samples formed separate clusters, indicating treatment-specific transcriptional responses. Treated groups were restricted to ADO (2h) and C48/80 (6h) (Figure S2B), further demonstrating clear separation from CONTROL samples. Similarly, DNP-treated samples (Figure S2C) were segregated from CONTROL along PC1, with distinctions between 2-h and 6-h treatments along PC2, suggesting time-dependent transcriptional changes.

To evaluate the transcriptional program induced by ADO, we performed differential expression analysis comparing ADO (2h) to CONTROL and identified 821 upregulated (Figure A) and 630 downregulated genes (Figure B). A notable portion of this response was unique to ADO, as no other stimuli modulated these 223 upregulated and 170 downregulated genes. We visualized the global changes of these ADO-specific genes in a heatmap across all five experimental conditions (seven biological replicates each) (Figure C). The ADO (2h) stimulation induced a visually distinct gene expression pattern compared to CONTROL, which was not seen after C48/80 or DNP stimulation.

2.

Common and distinct transcriptional responses to mast cell agonists. Venn diagrams showing the overlap of significantly (A) upregulated and (B) downregulated genes across four conditions (ADO (2h, red), C48/80 (6h, yellow), DNP (2h, light blue), and DNP (6h, dark blue)) relative to CONTROL (p _adj < 0.01). (C) Heatmap visualizing the DESeq2-normalized expression of ADO-specific genes across all five conditions: CONTROL, ADO (2h), C48/80 (6h), DNP (2h), and DNP (6h). Each condition included seven biological replicates, except for the DNP (6h) treatment, which comprised six replicates due to the unavailability of replicate 2.2.

We next assessed the overlap with other activation pathways. The ADO (2h) transcriptional profile most closely resembled the 2 h antigen stimulation (DNP (2h)) response, sharing 548 commonly upregulated (Figure A) and 405 commonly downregulated genes (Figure B). These data demonstrate that ADO drives a unique transcriptional profile while also sharing a significant number of DEGs with the canonical antigen-mediated activation pathway.

Overlap with the later DNP (6h) stimulation (255 upregulated, 217 downregulated genes) and the C48/80 (6h) stimulation (196 upregulated, 62 downregulated genes) was also observed (Figure A, B), although to a lesser extent. A core set of 102 genes was upregulated across all four conditions, likely representing a general mast cell activation signature (Figure A). Conversely, only 30 genes were commonly downregulated (Figure B).

To better visualize genes exhibiting both large-magnitude and statistically significant changes in expression, we generated Volcano plots comparing each treatment condition (ADO (2h), C48/80 (6h), DNP (2h), and DNP (6h)) to CONTROL (Figure S3). These analyses further confirmed distinct transcriptional responses of PMCs to each stimulus.

The differential expression profile of ADO-specific genes is visualized in a Volcano plot (Figure A). Detailed statistics for all ADO-specific protein-coding genes, whose biotype was annotated as “protein-coding” in Ensembl Release 115, are provided in Table S2. The most significantly upregulated ADO-specific genes were involved in cell adhesion, such as the scaffolding protein Mpp7 and the immunoglobulin superfamily member Igsf5, as well as those related to lipid metabolism, including the oxysterol-binding protein family Osbpl6 and the prostaglandin transmembrane transporter Slco3a1. A strong induction was also observed for signaling molecules, such as cAMP-responsive element modulator Crem and the small GTPases Rap2b and Rab4a. Other highly induced genes included the apoptosis inhibitor Niban1, the pre-mRNA splicing factor Isy1, and the calcium-binding protein Hpcal1. Osbpl6, Mpp7, Crem, and Isy1 were the four most induced genes by ADO with high mean expression across all conditions (DESeq2 normalized counts > 100) and the largest absolute log₂ fold changes. They all demonstrated a robust and specific upregulation after 2 h of ADO treatment (Figure B). In contrast, a smaller cohort of genes was significantly downregulated, most prominently the growth differentiation factor Gdf11 and the glutamate receptor subunit Grin2d (Figure B).

3.

ADO-specific transcriptional program in mast cells. (A) Volcano plot of DEGs in mast cells following a 2-h ADO treatment versus CONTROL. Genes meeting thresholds for both significance (p _adj < 0.01) and fold-change (|log₂ FC| > 1) were colored red, while genes meeting only the significance threshold were colored blue. The most significant DEGs (p _adj < 1 × 10^–5) were labeled. (B) Normalized expression plots for the top four most upregulated and abundantly expressed (DESeq2 normalized counts > 100) ADO-specific DEGs across five conditions: CONTROL (CTRL), ADO (2h), C48/80 (6h), DNP (2h), and DNP (6h). Each condition included seven biological replicates, except for the DNP (6h) treatment, which comprised six replicates due to the unavailability of replicate 2.2. Statistical significance was assessed using DESeq2, with significance levels defined as follows: ns, not significant; *p _adj < 0.01; **p _adj < 0.005; ***p _adj < 0.001. (C) Normalized expression plots for ADO-specific DEGs that are also tissue-resident mast cell-specific; all other conditions are as described in (B).

We next compared our ADO-specific gene signature against a 128 mast cell signature gene set defined by the Immunological Genome Project Consortium. This reference signature was established based on genes exhibiting at least a 2-fold higher transcript expression in all mast cell populations compared to other analyzed immunocytes. We found an overlap of four upregulated genes: RAS oncogene family member Rab27b, latexin Lxn, MAS-related GPR family member Mrgprx2, neutral cholesterol ester hydrolase Nceh1, and two downregulated genes: NCK-associated protein Nckap1 and endothelin receptor Ednra. In particular, the expression of the top 4 most significant onesRab27b, Lxn, Nckap1, and Ednrawas shown in a bar graph (Figure C).

2.5. Integrated Functional and Regulatory Enrichment Analysis of ADO-Specific Genes

To estimate the ADO-evoked activation of transcription factors (TFs), transcription factor activity inference was performed. This analysis revealed significant modulation of two TFs: Ctnnb1 and E2f1, whose target gene expression was visualized in Figure A and B, respectively.

4.

Integrated functional and regulatory enrichment analysis of ADO-specific genes. (A, B) Volcano plots displaying the inferred activity of TFs based on their ADO-specific target genes, as determined by ULM (univariate linear model) analysis. Labeled targets are ADO-specific genes. (C) A bar graph showing the top 10 annotated protein classes in upregulated and downregulated ADO-specific genes according to the PANTHER classification system. (D) Bar graph showing significantly enriched pathways from GO Molecular Function, KEGG, and Reactome databases that satisfy p-value < 3 × 10^–03. The gene count within each term is visualized by the color density.

The TF Ctnnb1 encodes β-catenin, which is a central component of the Wnt signaling pathway. Although Ctnnb1 activity was not significantly altered across all target genes (activity score = 1.63, p = 0.1), its activity was significantly suppressed across the ADO-specific gene set (activity score = −2.07, p = 0.04) by ADO treatment. This indicates that the ADO engages a dedicated pathway to suppress the transcriptional output of the canonical Wnt signaling pathway. E2f1 is a transcription factor that controls the cell cycle. Its activity was strongly and globally inhibited across both the full target set (activity score = −3.97, p = 7.25 × 10^–5) and the ADO-specific subset (activity score = −2.55, p = 0.01), indicating that E2f1 target genes were broadly downregulated upon ADO treatment.

Gene annotation via the PANTHER (protein analysis through evolutionary relationships) classification system identified protein classes for 137 of 184 upregulated and 85 of 136 downregulated ADO-specific genes. The classes with at least 5 annotated proteins were listed in a bar graph (Figure C). The most enriched protein class was the metabolite interconversion enzyme (PC00262), which included a broad set of genes essential for core metabolic processes. Key examples include enzymes involved in glycolysis (Aldoa, Pfkp), fatty acid metabolism (Elovl5, Cpt1a), and mitochondrial energy production (Impdh1, Sdhb, Ldha) (Table S3).

The second major class, protein-binding activity modulators (PC00095), suggests widespread activation of intracellular signaling networks. This is highlighted by the upregulation of numerous small GTPases (Arf2, Rab27b, Rhoq, Rap2b, Rab11fip5, and Rab4a) and their regulators, such as the guanyl-nucleotide exchange factors Trio and Cyth1, and GTPase-activating protein Arhgap25 and Sipa1l1 (Table S3). The prominence of GTPase-related signaling also drew our attention to the unique upregulation of GTPases Gvin1 and Gvin2, which encode GTPase, very large interferon-inducible 1, and GTPase, very large interferon-inducible 2, following ADO treatment (Table S3).

The third class, transporters (PC00227), shows an increase in the number of genes involved in the movement of molecules across cellular membranes. This includes components of the ATP synthase complex (Atp5f1d, Atp5mk), various ion channels (Cbarp, Kcnn4, Vdac3), and multiple solute carrier organic anion transporter family members like Slco3a1, Slc20a2, and Slc16a3 (Table S3). In the enrichment of cytoskeletal protein (PC00085), we observed an upregulation of genes involved in both actin and microtubule networks, such as Myo5a, Tpm4, and Arpc3 (Table S3). Enrichment was also observed for the transmembrane signal receptor class (PC00197), which included G-protein-coupled receptors (Grm5, Lpar2, Mrgprx2), pattern recognition receptor Tlr4, Frizzled family members Fzd6 and Fzd7, and cytokine receptor subunits Csf2rb and Csf2rb2 (Table S3). The scaffold/adaptor protein class (PC00226) was also represented, including the highly upregulated Mpp7, as well as signaling regulators such as Ywhaz and Arrb1 (Table S3).

In contrast to the upregulation of metabolic and structural genes, a strong trend of downregulation was observed in two key regulatory classes: Gene-Specific Transcriptional Regulators (PC00264) and Protein-Modifying Enzymes (PC00260) (Figure C). The first class contains many transcription factors of gene expression programs, including a large family of C2H2 zinc-finger proteins (Glis2, Zeb1, and Ikzf2) as well as key developmental factors such as Gata1, Gata2, and Hoxb4 (Table S4). This widespread downregulation suggests a major shift or shutdown of established transcriptional programs. The second class contains genes that regulate protein function through post-translational modifications, as highlighted by the decreased expression of numerous nonreceptor serine/threonine protein kinases (Ksr1, Taok3, Prkcd) and components of the ubiquitin system, including multiple E3 ubiquitin-protein ligases (Pias2, Itch, Trim58) (Table S4). This points to significant alteration in cellular signaling pathways and protein degradation.

To further understand the molecular functions and signaling pathways involved, we performed functional enrichment analysis of genes specifically upregulated or downregulated by ADO treatment using databases from Gene Ontology (GO) Molecular Function terms, , KEGG, − and Reactome. The analysis revealed a diverse set of functions with ADO-promoting pathways involved in cell signaling while suppressing those involved in nuclear regulation and gene expression (Figure D).

Among the upregulated genes, the most significant enrichment was found for pathways associated with membrane signaling and lipid metabolism. The top enriched terms included lipid binding (GO: 0008289), Synthesis of PIPs at the plasma membrane (R-MMU-1660499), the Phospholipase D signaling pathway (mmu04072), and PI Metabolism (R-MMU-1483255) (Figure D). Furthermore, multiple G-protein signaling pathways, such as G alpha (s) and G alpha (q) signaling events, were significantly enriched, indicating a broad activation of signal transduction cascades originating at the cell membrane. In contrast, the most prominent downregulated categories included DNA binding (GO: 0003677), transcription regulator activity (GO: 0140110), and protein-modifying enzymes (PC00260) (Figure D), reflecting a broad suppression of genes involved in transcriptional regulation, signaling, and post-translational modification in response to ADO stimulation.

2.6. Protein–Protein Interaction Network Revealing ADO-Evoked Changes in Key Hubs of Metabolism and Signaling

To further investigate the functional relationships among the ADO-specific genes, we constructed a protein–protein interaction (PPI) network. This analysis identified distinct, highly interconnected modules corresponding to core metabolic processes and phosphatidylinositol signaling, highlighting key hub genes that likely orchestrate these cellular responses (Figure ).

5.

Protein–protein interaction network analysis of ADO-regulated genes and identification of hub genes. The network illustrates the predicted physical and functional interactions among genes whose expression is specifically regulated by ADO. The color of each node (gene) corresponds to its log₂ FC, with red indicating upregulation and blue indicating downregulation. Seven highly connected hub genes, outlined in black, are central to the network’s function. These hub genes are grouped into three key functional modules indicated by their respective node background colors: Glycolysis (Gray), Mitochondrial Activity (Purple), and Phosphatidylinositol-Related Activity (Green).

Four of the top seven hub genes identified by the MNC algorithm were central components of the metabolic process network. Notably, lactate dehydrogenase A (Ldha) and aldolase A (Aldoa) are key enzymes in glycolysis. In contrast, iron–sulfur subunit B of the succinate dehydrogenase complex (Sdhb) and the F1 subunit delta of ATP synthase (Atp5f1d) are integral to mitochondrial function. A second central module was organized around Phosphatidylinositol-Related signaling. This network connected numerous phospholipases, kinases, and their associated proteins, indicating a comprehensive remodeling of this critical second messenger pathway. Three hub genes were central to this module: Phospholipase C gamma 1 (Plcg1), Phospholipase C beta 3 (Plcb3), and Phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit beta (Pik3cb) (Figure ). Unlike the metabolic module, this signaling network comprised a mix of both upregulated and downregulated genes, suggesting a complex, fine-tuned regulation of phosphoinositide signaling. In summary, the network analysis pinpoints critical hub genes that connect ADO stimulation to two primary cellular outcomes: a coordinated upregulation of central energy metabolism and a complex remodeling of the phosphatidylinositol signaling cascade.

2.7. Topology Analysis Revealing a Causal Structure of Key ADO-Modulated Pathways

To elucidate the causal regulatory architecture of pathways significantly affected by ADO, we performed topology analysis on Reactome pathways using SEMgraph. We identified six significantly modulated regulatory networks, each involving at least three ADO-specific genes, highlighting key activation events in metabolism, endocytosis, and nuclear processes (Figure ). These networks meet the criteria of having a standardized root-mean-square residual (srmr) < 0.1 and a deviance-to-degrees-of-freedom ratio (dev/df) <3. In addition, pNodeAct <0.05 indicates a significant activation of the combined effect of all nodes, whereas pNodeIhn <0.05 indicates a significant inhibition.

6.

Topological analysis of Reactome pathways seeded with ADO-specific genes. Six significant regulatory networks were identified from the Reactome database using SEMgraph: (A) clathrin-mediated endocytosis; (B) phosphatidylinositol metabolism; (C) the TCA cycle and respiratory electron transport; (D) glycolysis; (E) SUMOylation; and (F) nuclear envelope breakdown. Networks shown were selected for containing at least three ADO-specific genes and meeting model fit criteria (srmr < 0.1; deviance/df < 3). The combined effect of all nodes was considered significantly activated if pNodeAct < 0.05 and/or inhibited if pNodeIhn < 0.05.

Among these, clathrin-mediated endocytosis exhibited the best model fit (dev/df = 2.05, srmr = 0.08) (Figure A). This network involved a mixed expression profile, with Tgfa and Arrb1 being significantly upregulated, while Fcho1 was repressed. The phosphatidylinositol metabolism pathway was also significantly perturbed (pNodeAct = 3.43e ^–14, pNodeInh = 9.80e ^–10), implicating key upstream kinases Pik3cb, Pi4kb, and Pip5k1c that regulate downstream effectors such as Plcg1, Adgpk, and Rab4a (Figure B). Two central metabolic pathways, the citric acid (TCA) cycle and respiratory electron transport (pNodeAct = 9.21e ^–10) and glycolysis (pNodeAct = 3.24e ^–09), were significantly activated, driven by the upregulation of enzymes like Sdhb, Ldha, Pfkp, and Aldoa (Figure C, D). Conversely, the SUMOylation pathway exhibited strong inhibition (pNodeInh = 1.58e ^–08) (Figure E). Finally, the nuclear envelope breakdown pathway showed bidirectional regulation with Emd and Lmnb1 being activated and Nup153 being repressed (Figure F). In summary, this topological analysis provides a mechanistic framework for understanding how ADO orchestrates mast cell responses through the activation of metabolic and signaling pathways and the modulation of nuclear and post-translational regulatory circuits.

2.8. ADO’s Impact on the Expression of Genes Involved in the Release of De Novo Synthesized Mediators

To identify ADO’s effect on de novo synthesized mediators in mast cells, we curated a list of genes encoding lipid mediators, cytokines, chemokines, growth factors, and their receptors (see Table S5). We then checked the expression patterns of genes that were either uniquely responsive to ADO or significantly modulated by ADO treatment (p _adj < 0.01; Figure S4). Notably, Pla2g4a and Dagla, key enzymes that mediate arachidonic acid release from membrane phospholipids and the synthesis of 2-arachidonoylglycerol, respectively, were potently upregulated following ADO stimulation. Additionally, the growth factor Tgfa and the cytokine Il7 were upregulated in response to ADO. Among ADO-specific immune receptors, we found upregulation of the pattern recognition receptor Tlr4, the Frizzled family members Fzd6 and Fzd7, the cytokine receptor subunits Csf2rb and Csf2rb2, and downregulation of the endothelin-1 receptor Ednra.

3. Discussion

ADO is a mediator implicated in a variety of inflammatory processes and also plays an essential role in mast cell activation and its fine-tuning. Since ADO alone cannot induce massive mast cell degranulation and typically enhances the release of mediators evoked by immunological stimuli, studying ADO-stimulated transcriptional responses could provide new insights into the intracellular signaling pathways triggered by this agonist. The number of reports on this is very limited and includes only studies using immortalized mast cell lines.

In most of the studied mast cell models, ADO stimulation evokes an elevation of intracellular Ca²⁺ concentration. Intracellular Ca²⁺ is a ubiquitous and essential trigger of specific functions in eukaryotic. For example, elevation of [Ca²⁺]_i in mast cells activates not only exocytosis but also the transcription factor NFAT, which regulates gene expression. , Mast cell degranulation is indispensable for [Ca²⁺]_i elevation. In PMCs, the stimulation of FcεRI, Mrgprb2, and ADO receptors evoked Ca²⁺ release followed by Ca²⁺ entry in a similar pattern for all three stimuli. Thus, in our mast cell model used in this study, ADO triggered functional responses that were principally similar to those evoked by other mast cell activators.

3.1. ADO’s Effects on Calcium Channel Expression

Transcriptomic analysis provides a complementary approach to understanding how PMCs regulate channel abundance and their regulators, thereby determining calcium homeostasis during agonist stimulation over time at the transcriptional level. Among Ca²⁺ channel-related genes, only Trpm4 and Tmem63b were upregulated upon ADO stimulation. Our previous work identified TRPM4 as a critical limiting factor for antigen-evoked calcium rise in mouse PMCs, in which loss of TRPM4 leads to exaggerated calcium influx and enhanced degranulation. Tmem63b was identified to be upregulated by 7-fold after LPS treatment in bone marrow-derived mast cells, suggesting that it may participate in mast cell activation. In our study, Tmem63b expression increased modestly (1.4-fold, 1.6-fold, 1.5-fold, and 1.9-fold) in response to ADO (2h), C48/80 (6h), DNP (2h), and DNP (6h), respectively. Conversely, the downregulation of Grin2d, Itpr1, Stim1, P2rx4, and P2rx7 may indicate the engagement of a transcriptional negative-feedback mechanism to prevent excessive calcium entry and maintain calcium homeostasis in PMCs following ADO stimulation.

3.2. ADO-Specific Transcriptional Responses

Our analysis reveals that ADO induces a specific transcriptional program in mast cells. Namely, while a majority (65%) of ADO-induced genes overlap with the canonical antigen-triggered response, a significant portion (27%) is unique to ADO stimulation. The major shared signature with antigen is comprehensible since ADO is known to augment the canonical FcεRI-induced degranulation. In particular, we observed significant upregulation of Hdc and Tpsab1 following both ADO (2h) and DNP (2h) stimulation. Hdc encodes histidine decarboxylase, the rate-limiting enzyme responsible for histamine synthesis, while Tpsab1 encodes tryptase mMCP-7, a major protease stored in mast cell granules. On the other hand, the unique ADO-induced molecular signature is of particular interest to us. It could provide insight into the unique ADO signaling pathway that contributes to the release of inflammatory mediators without massive degranulation. For example, we found Hdac7 downregulation specifically in ADO. According to the CollecTRI database, Hdac7 negatively regulates Hdc expression, suggesting an ADO-specific pathway regulating histamine synthesis.

3.3. Effects of ADO on the Expression of Molecules Determining GPCR Signaling

ADO exerts its effects primarily by activating ADO receptors in the plasma membrane. Upon stimulation, these receptors signal their respective G proteins, initiating downstream signaling cascades. In line with this, pathway enrichment analysis of ADO-responsive genes demonstrated significant enrichment for both G alpha (s) and G q signaling events.

Activation of G alpha (s) protein increases AC activity and cAMP production. The significant upregulation (∼9-fold) of Crem (cAMP-responsive element modulator) gene in our data suggests the activation of a canonical GPCR-cAMP signaling axis. This is consistent with the previous findings that the human gene CREM is significantly upregulated upon ADO receptor activation in the human mast cell line HMC-1, and our study further demonstrates that the transcriptional response of Crem is unique to ADO. Concurrently, activation of G alpha (q) protein leads to PLC activation. Our PPI analysis identified Plcb3 (Phospholipase C β3) and Plcg1 (Phospholipase C, gamma 1) as hub genes in the ADO-specific response network, and both were downregulated. This highlights their regulatory role and suggests a transcriptional negative feedback loop after activation.

3.4. Effects of ADO on the PI3K/Akt Signaling Axis

Another hub gene, Pik3cb (Phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit beta), was upregulated in response to ADO. A previous study in RBL-2H3 cells indicated that ADO activates Gi-coupled A3 receptors, leading to protection against apoptosis via a pathway involving the Gi_βγ subunits, PI3Kβ, and protein kinase B (Akt). Consistent with this, our data show that G gamma subunits Gng4 and Pik3cb, and their docking protein Gab2, were uniquely upregulated in ADO, supporting an enhancement of the PI3K/Akt signaling axis and suggesting a pathway involved in promoting mast cell survival and metabolic activity.

3.5. Effects of ADO on the Expression of Molecules Determining Vesicle Trafficking

Our protein classification analysis revealed ADO-specific upregulation of genes associated with vesicle trafficking, including small GTPases, cytoskeletal-associated proteins, and scaffold/adaptor molecules. Among these, Rab27b plays a crucial role in mast cell degranulation, particularly by regulating the transition of vesicle transport from microtubule- to actin-based motility. Myo5a, encoding the motor protein MYO5A, is part of the RAB27a–Mlph–MYO5A complex, which regulates distinct steps in the BMMC degranulation pathway. Likewise, Rab4a has been identified as a key regulator of mast cell exocytosis through vesicle trafficking pathways.

In addition to exocytic machinery, enrichment of clathrin-mediated endocytosis pathways was observed, suggesting active internalization and recycling of receptors or membrane components during ongoing vesicle turnover and signaling. Arrb1, encoding β-arrestin 1, functions as a molecular scaffold linking G-protein-coupled receptors to clathrin-mediated endocytosis. Notably, β-arrestin 1 mediates agonist-dependent internalization and desensitization of the MRGPRX2 receptor in human mast cells. Moreover, the growth factor, Tgfa, was uniquely upregulated by ADO treatment. The membrane-bound Tgfa precursor or Tgfa ligand–receptor complexes can undergo clathrin-dependent internalization. Collectively, these findings highlight that ADO stimulation reprograms the vesicular trafficking machinery in PMCs, which may contribute to the potentiating effect of ADO on antigen-evoked degranulation.

3.6. Effects of ADO on the Expression of Inflammatory Mediators

We observed an increased expression of a range of lipid-mediator-producing enzymes, cytokines, chemokines, and immune receptors in response to ADO. Notably, the cytokine Il7 was significantly upregulated in response to ADO. In contrast, microarray analysis using HMC-1 cells reported a 2-fold downregulation of IL7 following Cl-IB-MECA treatment. We also observed a unique upregulation of immune receptors, including Csf2rb, Csf2rb2, and Tlr4, possibly reflecting a feed-forward mechanism that sensitizes the cell to the newly produced cytokines.

3.7. ADO-Dependent Transcriptional Effects in Metabolism and Cell Cycle

Glycolysis and mitochondria are critical for mast cell activation since the inhibition of glycolysis and ATP production attenuated IL-33-mediated and lipopolysaccharide-induced mast cell function. , Through PPI and topology analysis, we also identified activated core metabolic pathways, including glycolysis and mitochondrial activity.

In addition, our analysis revealed an ADO-specific suppression of signaling pathways associated with cell proliferation. Our TF activity analysis indicated a reduced activity of Ctnnb1 (β-catenin) and E2f1. Dysregulated β-catenin signaling has been shown to promote the expansion of bone marrow-derived connective tissue-type mast cells, systemic inflammation, and colon cancer. Furthermore, the Wnt/β-catenin pathway is a well-established driver of cell proliferation. , E2F1, in turn, is a key regulator of G1/S transition and DNA synthesis, driving the expression of genes required for S-phase entry. Notably, E2F1 is a novel TF regulating Ctnnb1 expression. The suppression of both TFs suggests a shutdown of the proliferation-promoting Wnt/β-catenin-E2F axis and indicates cell cycle arrest under ADO stimulation.

We also noted a significant topological pathway of nuclear envelope dynamics involving the upregulation of Lmnb1 (Lamin B1) and Emd (Emerin) and downregulation of Nup153. Upregulation of Lmnb1 and Emd suggests reinforcement of the nuclear lamina, potentially increasing its rigidity and resistance to remodeling, − while the downregulation of Nup153, a key component of the nuclear pore complex required for pore disassembly during mitosis, may impair nuclear envelope dynamics and thereby limit cell cycle progression. Together, these changes suggest reduced nuclear plasticity and cell-cycle arrest specific to the ADO treatment.

3.8. Limitations of Our Study

While transcriptomic analysis provides a comprehensive framework for hypothesis generation, it has limitations. First, the mRNA abundance does not always correlate with downstream protein expression or enzyme activity. It will be essential to validate these findings at the protein level (e.g., proteomic approach) and by measuring functional outputs (e.g., inflammatory mediator release, cAMP, lactate production, oxygen consumption rate, ATP levels, degranulation, cell proliferation, etc.), which should be a matter for our future studies. Second, because calcium signals exhibit spatial and temporal dynamics, our study could not distinguish transcriptional responses driven by calcium release from those triggered by calcium influx. This limitation could be addressed in future work by using pharmacological blockers or genetic models to delineate the specific contribution of defined calcium signals to the observed transcriptional changes. Finally, calcium-independent signaling pathways that influence gene expression warrant further investigation.

4. Conclusions

Our study reveals that ADO (compared with activators of FcεRI and Mrgprb2 receptors) alone can elicit a distinct mast cell activation program characterized by transcriptional remodeling of cellular/intracellular signaling, metabolic, and vesicular pathways. This response may promote the synthesis and release of specific inflammatory mediators, as well as the synthesis of components of the exocytosis machinery. Future studies should validate these ADO-driven mechanisms at the protein level, assess their associated functional responses, and further investigate the transcriptional programs triggered by ADO-specific, defined calcium signaling.

5. Materials and Methods

5.1. Mice

All animal experiments were approved by the Regional Council Karlsruhe and were performed according to their ethical guidelines (approval no. T-64/18). Mice were bred and maintained at the central animal facility of the University of Heidelberg under specific pathogen-free conditions. They were provided with drinking water and food ad libitum. Mice were killed by a lethal dose of CO₂. The mice used for isolation of PMC mast cells in this study were control mice with the genotype Orai1^flox/flox; Orai2^flox/flox , of the C57Bl6/N genetic background (backcrossed with C57Bl/6N mouse strain obtained from Charles River, USA at least 8 generations). Adult 9- to 12-week-old male mice were used for the experimental procedures.

5.2. Peritoneal Mast Cells: Primary Culture and Agonist Stimulation

PMCs were isolated and cultured as previously described. Briefly, the cells were obtained by peritoneal lavage with RPMI medium, centrifuged, and subsequently resuspended in culture medium (4 mL per mouse). The RPMI-based culture medium additionally contained 20% fetal calf serum (FCS), 1% penicillin-streptomycin, 10 ng/mL IL-3, and 30 ng/mL Stem Cell Factor. For one preparation, the cells isolated from 2 to 3 mice were pooled. The isolated cell suspension was cultured at 37 °C and 5% CO₂. Two days after the isolation, the culture medium was changed, and all nonadherent cells were removed by aspiration. On day 9, the cells were split and cultivated further at a concentration of 1 × 10⁶ of cells/mL. The PMCs were used for the experiments 12–16 days after isolation. The purity of PMCs, estimated according to the double-positive immunoreactivity to FcεRI and c-Kit antigens, as previously reported was at the level of ∼98.5%. For the antigen stimulation experiments, the cells were treated overnight with 300 ng/mL of anti-Dinitrophenyl IgE antibodies. The PMCs for RNA sequencing were stimulated in the RPMI-based culture medium at a concentration of 2 × 10⁵ cells/mL at 37 °C and 5% CO₂.

5.3. Microfluorimetric Intracellular Free Ca²⁺ Concentration Measurements

PMCs were loaded with Fura-2 by incubating them in a physiological salt solution containing 2.5 μM Fura-2 AM and 0.1% Pluronic F-127 for 30 min at room temperature on coverslips coated with concanavalin A (0.1 mg/mL) for cell immobilization. The intracellular free Ca²⁺ concentration was measured on the stage of an AxioObserver-A3 inverted microscope (Zeiss, Germany) equipped with a 40× (1.3 NA) immersion oil objective (Zeiss, Germany) in a perfused 0.5 mL chamber in Physiological Salt Solution at room temperature. The Physiological Salt Solution (PSS) contained (in mM): NaCl 135, KCl 6, CaCl₂ 2, MgCl₂ 1.2, glucose 12, and HEPES 10; pH 7.4 (NaOH). Nominally Ca²⁺-free PSS was made by excluding the CaCl₂ salt from the composition. At the beginning of each experiment, the cells were washed thoroughly with PSS. The fluorescence signal was obtained by alternately exciting the Fura-2 with light of 340 and 380 nm wavelengths (50 ms exposure time) using a pE-800fura (CoolLED, United Kingdom). Emitted fluorescent signal was filtered at >510 nm and detected by a charged-coupled device camera AxioCam MRm (Zeiss, Germany). The fluorescent ratio F ₃₄₀/F ₃₈₀ was measured with an acquisition rate of 5 s per cycle. Both the camera and the light source were controlled by the Zen 3.2 software (Zeiss, Germany), allowing for the recording of fluorescent signal intensities in particular, cell-attributed Regions of Interest (ROIs).

5.4. Information about Samples Undergoing RNA Sequencing

A total of seven independent PMC cell preparations were generated, each from 2 or 3 mice. Each PMC preparation was stimulated under four experimental conditions: ADO (10 μM) for 2 h (ADO (2h)), C48/80 (50 μg/mL) for 6 h (C48/80 (6h)), DNP (100 ng/mL) for 2 h (DNP (2h)), and DNP (100 ng/mL) for 6 h (DNP (6h)).

The time points were selected according to the results of a pilot transcriptome experiment in which one PMC preparation was stimulated with ADO (10 μM), C48/80 (50 μg/mL), and DNP (100 ng/mL) for three time periods, i.e., 2, 6, and 18 h (Figure S5). The agonist concentrations were chosen to evoke sustained Ca²⁺ transients in the cells. Among the 100 top-changed genes (for each agonist stimulus), we selected 36 genes that were changed in at least 2 agonist treatments and plotted the changes in their mean values over time. As shown in Figure S5, the maximal responses were observed after 2 h of ADO treatment and after 6 h of treatment with DNP and C48/80, respectively. The results of this pilot study served as the basis for the experimental design in this manuscript.

From each PMC preparation, two technical replicates with vehicle control treatment (CONTROL) were further processed. RNA isolation and subsequent sequencing were performed in two batches. The corresponding samples were termed 1.1, 1.2, and 1.3 (Batch 1), and 2.1, 2.2, 2.3, and 2.4 (Batch 2). For the second biological replicate from batch 2 (2.2), the data for the DNP 6-h-treated condition were unavailable for this analysis, so six independent biological replicates were analyzed for this condition.

For RNA isolation, each sample (0.2–2 × 10⁵ PMCs) was lysed in 400 μL of RLT Buffer supplemented with 4 μL β-mercaptoethanol. Isolation of mRNA was performed using RNA Micro KIT (Qiagen 56304); at the end, mRNA was eluted in 14 μL buffer, and 10 μL RNA of each sample was utilized for further transcript library synthesis. Transcript libraries were prepared using “Ovation SoLo RNA-Seq Library Preparation Kit” (NuGEN), and 25 ng of the labeled cDNA per sample was utilized for further sequencing. For the deep sequencing, a “NextSeq 2000” (Illumina) sequencing platform was used (EMBL Heidelberg Genomics Core Facility). The sequencing reads included three types: 75 bp (single-end) for sample 1.1, 61 bp (paired-end) for samples 1.2 and 1.3, and 100 bp (paired-end) for samples 2.1–2.4. For each sample, at least 20 million reads were obtained. The ENSEMBL mouse genome database was used for transcript identification.

5.5. Bioinformatic Analysis of RNA-Seq Data

5.5.1. Preprocessing and Quantification

Raw RNA-Seq reads were processed using the nf-core/rnaseq pipeline (version 3.14.0). The pipeline was executed using Nextflow with the Singularity profile to ensure reproducibility via containerized environments. Read alignment was performed against the Mus musculus GRCm39 reference genome (GRCm39.primary_assembly.genome.fa) using STAR (version 2.7.11a). Gene-level quantification was carried out using RSEM (version 1.3.1) with annotations from GENCODE release M35 (gencode.vM35.primary_assembly.annotation.gtf).

5.5.2. Differential Expression Analysis

Differential gene expression analysis was conducted using the DESeq2 package (version 1.42.0) in R. Prior to analysis, gene counts were filtered to exclude genes with low expression; specifically, only genes exhibiting a raw count of at least 10 in a minimum of 7 samples were retained. This threshold ensures that analyzed genes demonstrate substantive expression in at least one experimental condition, considering the seven biological replicates per condition.

Technical replicates for CONTROL samples were merged by summing their counts using the collapseReplicates function in DESeq2 to increase the sequencing depth for those biological samples. The experimental design formula, ∼ batch + treatment, was employed to model and account for potential batch effects arising from batch 1 and batch 2 data, thereby enhancing the sensitivity for detecting treatment-specific expression changes.

The filtered count matrix input to DESeq2 contained 16 177 genes and 34 samples (after merging technical replicates and accounting for the missing DNP (6h) sample). Differentially Expressed Genes (DEGs) between conditions were identified using the Wald test. The resulting p-values were adjusted for multiple comparisons by using the Benjamini–Hochberg (BH) procedure. A gene was considered significantly differentially expressed if its BH-adjusted p-value (p _adj) was less than 0.01.

5.5.3. Identification and Visualization of ADO-Specific Genes

To identify genes specifically modulated by ADO treatment, DEGs from four comparisons (ADO (2h) vs CONTROL, C48/80 (6h) vs CONTROL, DNP (2h) vs CONTROL, and DNP (6h) vs CONTROL) were compared using Venn diagrams. Genes significantly up- or downregulated exclusively in the ADO (2h) vs CONTROL comparison, but not in the other three comparisons, were classified as ADO-specific DEGs. Expression data were batch-adjusted using ComBat-seq and normalized using size factors from DESeq2 prior to visualization in heatmaps and bar graphs.

5.5.4. Transcription Factor Activity Inference

Transcription factor (TF) activity inference was conducted using the decoupleR package (version 2.9.1). The input comprised stat, p _adj, and log₂ fold change values derived from DESeq2 differential expression analysis. TF–target associations were sourced from the CollecTRI database, a comprehensive collection of TF–target relationships aggregated from 12 distinct resources. TF activities were estimated using the default method in decoupleR: the univariate linear model (ULM). TFs were considered significantly activated or repressed if they exhibited an absolute activity score (|score|) > 2 and a p-value < 0.05. To identify TFs specifically associated with the ADO response, TF activity inference was performed on ADO-specific genes, and the resulting significantly regulated TFs were considered ADO-specific TFs.

5.5.5. Protein Class Classification and Functional Enrichment Analysis

Protein class classification was conducted using PANTHER classification system (version 19.0). The input consisted of 393 ADO-specific DEGs (annotated with Ensembl IDs).

Pathway over-representation analysis was conducted using clusterProfiler and ReactomePA. The input for GO term enrichment consisted of Ensembl ID-annotated ADO-specific DEGs tested against a background of 16 177 Ensembl-annotated genes. For KEGG and Reactome enrichment, which need Entrez gene IDs, the input comprised 360 ADO-specific DEGs (Entrez-annotated) tested against a background of 13 280 Entrez-annotated genes. This reduction in gene numbers for Entrez-based analyses was due to some Ensembl IDs lacking direct Entrez ID mapping. All gene annotations were derived using the org.Mm.eg.db R package. The hypergeometric test was used to calculate significance, and analyses were restricted to terms or pathways containing 10–500 genes.

5.5.6. PPI Network Analysis

Interaction data were sourced from the STRING database (v12.0; minimum confidence score >0.4). The resulting network was then imported into Cytoscape, where the top seven hub proteins were identified using the Maximal Neighborhood Component (MNC) algorithm of the cytoHubba plugin (v0.1).

5.5.7. Topology Analysis

The R package SEMgraph was used for topological analysis of RNA-Seq data. The analysis is predicted on three key inputs: a gene interaction network, the processed gene expression data set, and the sample group design (in this instance, ADO (2h) versus CONTROL).

Molecular networks were constructed using pathway information from the Reactome databases using the Graphite R package. Pathways from this database were merged to create a comprehensive network. To prepare the network for directional analysis, bidirectional edges were removed, and only nodes (vertices) with at least one remaining edge were retained. This resulted in initial graphs for Reactome with 5575 edges and 152 052 nodes.

For this analysis, gene expression data (previously filtered for low read counts and with technical replicates collapsed) were further processed. Batch effects were adjusted using the ComBat-seq function. Gene identifiers were converted to Entrez IDs. Subsequently, expression values were rank-based inverse normal transformed using the huge.npn function from the R package huge to mitigate normality constraints. The resulting expression matrix for topology analysis comprised 13 025 genes for the CONTROL and ADO (2h)-treated conditions (n = 7 biological replicates per condition). Edges in the merged network were weighted based on gene expression data. Pairwise Pearson correlations between connected genes were calculated across samples, transformed using Fisher’s r-to-z transformation, and corresponding z-scores (z-sign) and p-values were derived to represent edge weights and significance.

A Steiner Tree (ST) algorithm, specifically a fast Kou’s algorithm implementation, was employed to extract relevant subnetworks from the weighted graph using the z-sign values as edge weights. , The ST approach identifies a minimal subnetwork connecting a predefined set of “seed” genes. In our case, the seed genes for each pathway were those that were part of the pathway and specific to the ADO treatment.

Extracted subnetworks were evaluated using Structural Equation Modeling (SEM) via the SEMrun function in the SEMgraph R package to assess both perturbation effects and overall model fit. Perturbation effects, representing the local fit, were considered significant when the p-value was less than 0.05. The combined effect of all nodes was considered significantly activated if pNodeAct < 0.05 and/or inhibited if pNodeIhn < 0.005. Global model fit was assessed using the standardized root-mean-square residual (srmr) and the deviance-to-degrees-of-freedom ratio (dev/df). A srmr value below 0.1 and a dev/df ratio less than 3 were considered indicative of an acceptable global fit. It is noteworthy that global and local fit indices are not necessarily dependent; thus, even when the global fit is suboptimal, locally significant relationships may still be valid and informative.

The resulting SEM graphs were visualized using the igraph R package. Nodes were color-coded to represent regulatory direction and significance: red for significant activation, blue for significant repression, and white for nonsignificant expression. Edges were colored in black if the connection between nodes was significant, with the arrow and tee sign indicating positive and negative regulation.

Glossary

Abbreviations

ADO

adenosine

BH

Benjamini–Hochberg

C48/80

compound 48/80

DAG

diacylglycerol

DEGs

Differentially Expressed Genes

dev/df

deviance-to-degrees-of-freedom ratio

DNP

2,4-dinitrophenyl human serum albumin

ER

endoplasmic reticulum

PLC

phospholipase C

PMCs

peritoneal mast cells

PIP2

phosphatidylinositol 4,5-bisphosphate

IgE

immunoglobulin E

IP3

inositol 1,4,5-trisphosphate

SEM

structural equation modeling

SOCE

store-operated calcium entry

srmr

standardized root-mean-square residual

ST

Steiner tree

TF

transcription factor

ULM

univariate linear model

The original data of the work will be uploaded to GEO upon acceptance.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsptsci.5c00741.

• Supplementary figures include Mast cell protease expression, Principal component analysis for all conditions, Volcano plots comparing all four treated groups to CONTROL, Expression profiling of inflammatory mediators regulated by ADO, and Time course of the mean fold-change (normalized to untreated group (0h)) of mRNA expression level of 36 genes of interest; Supplementary tables provide DESeq2 normalized expression of calcium ion channels in CONTROL, Differential expression statistics of all ADO-specific protein-coding genes, Summary of the top six most abundant protein classes among upregulated ADO-specific genes, Summary of the top two most abundant protein classes among downregulated ADO-specific genes, and Summary of genes involved in inflammatory mediator biosynthesis and immune receptors in mast cells (PDF)

∇.

Q.L. and V.T. share first authorship. Conceptualization: M.F. Data Curation: Q.L., V.T., M.B., and N.L. Formal Analysis: Q.L., V.T., M.B., C.R. Funding acquisition: A.B., A.K., M.F. Investigation: Q.L., V.T., M.B. Methodology: Q.L., V.T., A.B., C.R., M.B., N.L. Supervision: V.T., A.B., A.K., M.F. Visualization: Q.L., V.T., M.F. Writing, original draft: Q.L., V.T. Writing, review, and editing: Q.L., V.T., A.B., A.K., M.F.

This research was funded by the German Research Foundation (DFG) through the Collaborative Research Centre CRC1328 (FKZ 335447717), project A21 to M.F., BE 6934/3-1 to A.B., the DZHK (German Center for Cardiovascular Research), and the BMBF (German Ministry of Education and Research).

The authors declare no competing financial interest.

Published as part of ACS Pharmacology & Translational Science special issue “Purinergic Signaling”.