Systems biology-based drug repurposing for neuroinflammation treatment in activated human microglia Authors: Martina Cirinciani, Lorenzo Germelli, Eleonora Da Pozzo, Corrado Priami, Claudia Martini, Paolo Milazzo Published: 2026 ABSTRACT AbstractNeuroinflammation is a physiological response triggered by alterations in tissue homeostasis within the Central Nervous System (CNS); depending on the magnitude and chronicity of inflammation, it is considered a common state in the pathophysiology of several neurodegenerative and psychiatric diseases. Neuroinflammation is a very complex and context-dependent condition, mediated by the activity of several pathways and molecules, thus the search of valuable targets and therapeutic strategies is a priority. This study proposes a systems biology approach to create a data network about genes, drugs, and related targets in the context of human neuroinflammation to identify new potential repurposed drug candidates. Each candidate drug was associated with a score that considered both the topological properties of the network and the biological functions of the proteins. The computational pipeline identified Fostamatinib as a potential repurposed candidate in neuroinflammation. To confirm the computational results, R406, the active metabolite of Fostamatinib inhibiting the Syk pathway, was assessed in two different human microglial in vitro models to verify its potential beneficial effects. Results evidenced the efficacy of R406 in counteracting the pro-inflammatory response in both models.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-026-56079-3.Subject terms: Computational biology and bioinformatics, Diseases, Drug discovery, Neuroscience Introduction Neuroinflammation is a very complex and multifactorial phenomenon affecting the Central Nervous System (CNS) following several kinds of insults1. The neuroinflammatory process can evolve into an excessive response, causing a chronic inflammatory condition that itself damages the NS2. Specifically, activated microglia cells are the pivotal pathological hallmark of the neuroinflammation occurring in several pathological states, and their activation leads to the release of proinflammatory cytokines and chemokines3. This is a common event in many pathological conditions, also quite different from each other, such as brain injuries4,5, depressive disorder6, and neurodegenerative diseases7–9 where the neuroinflammatory response has different degrees of intensity and involves various pathways and mediators, but the common signature is the activated microglia. Finding efficacious treatment strategies for neuroinflammation is extremely important because it is both a cause and a consequence in the manifestation of these disorders. Drug repurposing, the process of identifying new therapeutic applications for already approved drugs, has become an effective strategy in drug development10. Several in silico methods have been developed to conduct drug repurposing studies11, ranging from machine learning12,13 to network-based methods14–16. Among the multiple approaches available, we chose a computational pipeline previously developed by us in10, already positively employed in a drug repurposing study focused on metabolic syndrome. Being a network-based method, it does not require a training process or large curated datasets to effectively represent biological processes and complex mechanisms. Moreover, differently from other network-based methods17 that consider only the graph’s topological properties for identifying repurposing candidates, our computational pipeline integrates such information with functional data, better including the biological aspect of the analysed system10. This computational pipeline was herein adopted to repurpose drugs with potential unidentified anti-neuroinflammatory activity, then experimentally validated with in vitro models of activated microglia. The pipeline includes the collection of disease-related genes and drug data, their network representation, and processing, following a workflow characterized by high flexibility and versatility to novel applications. The pipeline and the related code developed in10 were employed here for a neuroinflammatory-based study. A set of dysregulated genes linked to the neuroinflammatory processes in different pathological conditions was identified starting from a search of transcriptomics literature and filtered based on a network- and drug-target information- analysis. Then, already approved drugs with potential anti-neuroinflammatory activity were identified by calculating a proximity score that integrates the network’s topological properties and the biological similarities of the players involved. The prodrug Fostamatinib had one of the highest proximity scores and was selected from the final list of candidates. Its active metabolite R406, an inhibitor of spleen tyrosine kinase (Syk) acting on Fc receptor signaling, has already been investigated for pathologies characterized by neuroinflammation in murine model18–21; its effects on inflammation of human CNS cells are not well-known, except for one paper, at the best of our knowledge, in which phagocytosis in an Alzheimer disease (AD) model was investigated22. Anyway, Syk signalling has been successfully modulated in human therapy against various peripheral diseases, such as autoimmune diseases, leukaemia, and lymphoma23. Thus, considering the importance of microglia in regulating human neuroinflammatory processes by activating its anti-inflammatory role, the effect of R406 and its ability to counteract inflammation in two different activated human microglial models were herein studied. Results Genes associated with neuroinflammation The workflow applied for drug repurposing in the context of neuroinflammation is shown in Fig. 1. The selection of genes was conducted on GeneCards24 by searching for the term “neuroinflammation” and identified 1,373 genes (Supplementary Table S5). This list was combined with an additional 777 genes, upon duplicates removal, derived from text mining analysis on PubMed, which considered transcriptomic studies related to neuroinflammation (Supplementary Table S5). Overall, 2,150 genes were identified. Schematic illustration of the computational workflow. Schematic workflow describing the steps and results of the analysis. The computational pipeline developed in10 is here applied to data related to the neuroinflammatory response to investigate potential repurposing drugs for neuroinflammation treatment. Genes related to neuroinflammation were retrieved from GeneCards and literature search. The gene set was filtered with an enrichment analysis and then used to build a protein-protein interaction (PPI) network. The integration of the PPI with a regulatory network identified the tissue-specific background network. Then, drug data were retrieved from DrugBank and LINCS. Through network analysis, drug modules and disease modules were identified and used to compute a proximity score based on the topological properties of the network and a functional similarity measurement. Filtering for a score > 1.63 (i.e., the median value. Differently from the original paper by Misselbeck et al., we considered the median value as a filtering threshold to enlarge the drug candidates, as better explained in Materials and Methods and Results), and considering only pharmacologically active drug-target interactions, a list of 90 drugs was identified. In the final steps, drugs were excluded if already investigated for their therapeutic use in diseases associated with neuroinflammation and if associated with side effects on the CNS. (Adapted from Misselbeck et al.10). Tissue-specific network Given the heterogeneity of the data sources, a gene-set enrichment analysis was performed using the Database for Annotation, Visualization and Integrated Discovery (DAVID, https://david.ncifcrf.gov)25, which identified 673 pathways involved in neuroinflammation and having a p-value < 0.01 (Supplementary Table S1). This filter identified 1,472 genes belonging to at least one significant pathway. Using HIPPIE26], these genes created the tissue-specific protein-protein interaction (PPI) network characterized by 7,678 nodes and 30,533 edges. The PPI network was built, setting on HIPPIE the highest predefined confidence level available (i.e., 0.72) for only representing interactions with high confidence scores. The PPI network was merged with a regulatory network characterized by 1,308 nodes and 4,676 edges, which produced the background network defined by 8,535 nodes and 35,209 edges (Supplementary File S2). On this network, 7 significant neuroinflammation modules were identified by using the community detection function developed by Misselbeck et al.10. Among them, two largest modules having 369 nodes and 977 edges, and 739 nodes and 950 edges, respectively) can be distinguished from the other five smaller hubs ranging from 10 to 66 nodes. Drug repurposing 4,467 approved drugs were selected from DrugBank (https://go.drugbank.com)27 to identify drug-repurposing candidates for our filtered genes. The drug proximity score based on drug modules, including up-and down-regulated genes, was calculated for 277 drugs that had targets mappable on our network of genes with appropriate semantic similarity of expression profile retrieved from the Library of Integrated Cellular Signatures (LINCS) (https://clue.io)28 (Supplementary Table S6). Despite being a dataset largely retrieved from perturbation studies on cancer cell lines, it has been demonstrated how drug signatures are often conserved across diverse cell types and settings28,29. Indeed, LINCS has been frequently employed in several drug repurposing projects even when not focused on tumors studies10,30. Filtering drugs based on a score included in the most significant 5% of the reference distribution resulted in a list in which several candidates were highly represented, such as Fostamatinib, despite their several interactions with secondary targets. Thus, to extend the drug list, the filtering step was conducted based on a score > 1.63 (corresponding to the median value). 152 drugs with 110 different targets were identified, among which 90 candidates were known for being pharmacologically active drug-target interactions (Supplementary Table S6). Through OpenTargets (https://www.opentargets.org)31, 77 drugs were additionally excluded for being already investigated for their therapeutic use in diseases associated with neuroinflammation. Last, the drugs with side effects on the CNS were excluded, leading to the final list of 20 drug candidates with 17 different targets (Table 1). Table 1List of identified drugs repurposing candidates with associated targets and their scores. The proximity score combines a network-based distance and a functional similarity measurement between proteins in the drug modules and the disease modules. The network-based distance between the modules considered topological information of the network, while the functional similarity measurement included biological knowledge about the proteins. The score ranges in [0,2]: the closer the score is to 2, the better the potential drug repurposing candidate is.NameTargetScoreBaricitinibJAK11755Hydrocortisone phosphateANXA11751Flufenamic acidAR1745MidostaurinPDGFRA1742DienogestAR1726DacomitinibEGFR1720Fluocinolone acetonideANXA11705OxandroloneAR1700BalsalazidePPARG1693BaricitinibJAK21687RucaparibPARP11678Hydrocortisone valerateANXA11678PimecrolimusMTOR1677FostamatinibSYK1665Testosterone propionateAR1663CisaprideHTR2A1660TriclosanPPARG1659AlitretinoinRARB1659TazaroteneRXRB1656AdenosineADORA11644AdapaleneRARG1644AlitretinoinRXRB1639AdapaleneRXRG1633AlitretinoinRARA1631TazaroteneRARA1630 List of identified drugs repurposing candidates with associated targets and their scores. The proximity score combines a network-based distance and a functional similarity measurement between proteins in the drug modules and the disease modules. The network-based distance between the modules considered topological information of the network, while the functional similarity measurement included biological knowledge about the proteins. The score ranges in [0,2]: the closer the score is to 2, the better the potential drug repurposing candidate is. Among them, considering the important and peculiar role of the Syk receptor in inflammatory processes (as already described in the introduction), the Syk inhibitor Fostamatinib was selected for subsequent experiments. Fostamatinib is an FDA-approved drug for chronic immune thrombocytopenia purpura32, investigated for the treatment of rheumatoid arthritis33, renal diseases34, different types of cancer35,36, and severe COVID-1937. To the best of our knowledge, Fostamatinib and its active metabolites have not yet been studied in human in vitro models of microglia-derived neuroinflammation. In vitro validation of Fostamatinib The human microglia C20 and HMC3 cell lines treated with the metabolically active compound R406 were used to assess the possible use of Fostamatinib in neuroinflammation. Dexamethasone (DEX) was used as anti-inflammatory drug reference38. First, using the MTS assay, both cell lines were treated with different drug concentrations for 24 h to determine the non-toxic concentration range of R406. The IC50 of R406 on C20 cells resulted to be 0.56 × 10− 6 M (Fig. 2A), indeed the compound did not significantly affect the viability below the concentration of 100 nM. Regarding HMC3 cells, the IC50 was 2.62 × 10− 6 M (Fig. 2A), and the cell viability was not significantly affected below the concentration of 50 nM. Both cell lines showed similar IC50 in the low micromolar concentration, and the different responses could be due to their various stages of development, as C20 cells are adult microglia and HMC3 are embryonic microglia. R406 effects on cell viability. (a) R406’s effects on C20 and HMC3 cells viability. The IC50 of R406 on C20 cells was 0.56 × 10− 6 M. No significant effects on viability below the concentration of 100 nM were observed. The IC50 of R406 on HMC3 cells was 2.62 × 10− 6 M. No significant effects on viability below the concentration of 50 nM were evidenced. (b) R406’s effects on C20 and HMC3 cells, concomitantly treated with inflammatory stimuli (i.e. IL-1β 20 ng/mL for C20 cells, and IL-1β + IFN-γ both at 10 ng/mL for HMC3 cells). *p < 0.05, **p < 0.01, ***p < 0.001 and ****p < 0.0001 vs. CTRL, # p < 0.05 and ## p < 0.01 vs. INFL+DMSO. Results are reported as mean ± standard error of the mean (SEM) from three independent experiments. From these results, we decided to use R406 at 10 nM for the subsequent assays, as this concentration of the compound is not toxic at all in both microglial models. Thus, the cell viability was also analysed during inflammatory conditions, evidencing that R406 can exert significant protection against neuroinflammation in both cell lines (Fig. 2B). Indeed, the viability of inflamed cells was significantly increased (p < 0.01) following the co-treatment with R406. As a positive control, DEX was used at 1 µM, according to the literature39,40. Then, C20 and HMC3 cells were specifically inflamed39,41 and treated with R406 10 nM or DEX 1 µM to assess the ability of R406 to modulate the neuroinflammatory response; the expression of selected genes related to pro- and anti-inflammatory mediators was evaluated (Figs. 3 and 4). In particular, we assessed the mRNA of interleukins 6 and 10 (IL-6 and IL-10), the best-known markers of pro- and anti-inflammation state, respectively. Furthermore, tumour necrosis factor-alpha (TNF-α), NLR family pyrin domain containing 3 (NLRP3), Transforming Growth Factor- β (TGF-β), and the nuclear factor kappa-light-chain-enhancer (NF‑κB) genes’ expression was also analysed. TNF-α is a cytokine exploiting various cellular effects, and it is considered one of the major regulators of inflammation, as it triggers other inflammatory molecules42. NLRP3 is an inflammasome component that triggers immune responses43. TGF-β is a multifunctional regulator factor that can stimulate or inhibit proliferation, depending on the context of its action44. Finally, NF‑κB is among the main pathways activated by IL-1β-induced inflammation39. R406 can counteract neuroinflammation in human microglial C20. (a–f) C20 cells were treated with 10 nM of R406 concomitantly with IL-1β 20 ng/mL. In particular, the figure shows the evaluation of mRNA levels of IL-6 (a), TNF-α (b), NLRP3 (c), IL-10 (d), TGF-β (e), and NF‑κB (f). *p < 0.05, **p < 0.01, and ***p < 0.001. All mRNA transcript levels were expressed as fold of change versus the control (CTRL). Histograms report the mean relative fold change ± standard error of the mean (SEM) from three independent experiments. R406 can counteract neuroinflammation in human microglial HMC3 cellular model. (a–f) HMC3 cells were treated with 10 nM of the R406 concomitantly with IL-1β + IFN-γ at 10 ng/mL. In particular, the figure shows the evaluation of mRNA levels of IL-6 (a), TNF-α (b), NLRP3 (c), IL-10 (d), TGF-β (e), and NF‑κB (f). *p < 0.05, **p < 0.01, and ***p < 0.001. All mRNA transcript levels were expressed as relative fold of change versus the control (CTRL). Histograms report the mean relative fold change ± standard error of the mean (SEM) from three independent experiments. Herein, both the human inflammatory cell models showed a significant increase in the expression of pro-inflammatory markers upon challenge (p < 0.05) (Figs. 3 and 4). As reported in Fig. 3, R406 was able to reduce the expression levels of inflammation markers in C20 cells, namely IL-6, TNF-α, NLRP3, and NF‑κB. At the same time, the drug treatment was associated with an increase in the expression of the anti-inflammatory cytokine IL-10. As expected, a similar trend was also observed after DEX treatment, characterized by a decrease in IL-6, TNF-α, and NF‑κB production, as well as an increase in IL-10 expression levels. Interestingly, even though not significantly when compared to the inflammatory stimulus, R406 showed a clearer trend in reducing TNF-α and NLRP3 than DEX; on the other hand, DEX is more efficient to increase the production of IL-10. Regarding TGF-β, the homeostatic cytokine that is activated during reparative processes, its production is increased upon inflammatory stimulus, while DEX and R406 treatments were both associated with a reduction to control levels. In particular, DEX had a higher efficacy in reducing TGF-β expression than R406. R406 showed to counteract the inflammation also in HMC3 cells, as reported in Fig. 4. In particular, R406 was associated with a decrease in pro-inflammatory mediators’ expression, namely IL-6, TNF-α, NLRP3, and NF‑κB, and with an increase in IL-10 expression levels. The positive control, the DEX treatment, showed similar effects. Similar to C20 cells, despite not having significant effects if compared with the inflamed condition, the R406 treatment showed a higher efficacy in reducing TNF-α than DEX, and DEX was associated with a larger increase of IL-10 transcript than R406. Moreover, R406 had a higher ability to reduce IL-6 than DEX, and DEX treatment was associated with a more significant NLRP3 decreasing. An opposite trend was observed for TGF-β mRNA levels during inflammation with respect to C20 cells; a significant decrease in TGF-β was found in HMC3 and DEX was able to restore TGF-β at levels similar levels to control. To further investigate whether R406 modulates downstream inflammatory signaling, we quantified NF‑κB activation by measuring p65 nuclear translocation in C20 and HMC3 microglial cells following cytokine stimulation. We also investigated this aspect at concentrations higher than 10 nM, while remaining within a range that, based on previous results, did not appear to be significantly cytotoxic. Notably, in C20 cells exposed to IL‑1β, R406 induced a clear dose‑dependent reduction in p65 nuclear accumulation, with significant inhibition already detectable at 10 nM and further decreases at 50 and 100 nM. These results indicate that also partial Syk inhibition is sufficient to attenuate NF‑κB activation in this cellular context. In HMC3 cells stimulated with the IL‑1β/IFN‑γ cocktail, R406 reduced p65 nuclear translocation at 10 and 50 nM; however, the highest concentration (100 nM) failed to inhibit NF‑κB and instead resulted in a modest increase in nuclear p65 signal compared to CTRL. This divergent response suggests that, under stronger inflammatory stimulation, HMC3 cells may engage compensatory pathways that sustain NF‑κB activation when Syk inhibition becomes more pronounced. Overall, these findings confirm that R406 effectively modulates the Syk–NF‑κB axis at non‑toxic concentrations, while also revealing cell‑type‑ and stimulus‑dependent differences in pathway sensitivity. Syk inhibition by R406 prevents inflammation-induced p65 (NF‑κB) nuclear translocation in human microglia. (a) Representative confocal images of p65 (red) and DAPI (blue) staining of C20 and HMC3 microglial cells following treatments. (b, c) Mean of fluorescence intensity of p65 nuclear signal. Bars represent the mean ± SEM of 30 nuclei counted in three independent experiments. Statistical analysis was performed by One-way ANOVA following Tukey’s post’s test: *p < 0.05, ***p < 0.001, **** p < 0.0001. Discussion The neuroinflammatory response is a multifactorial process, common to different diseases, involving various cell types and mediators, and systems biology is one of the most used approaches in drug repurposing studies regarding neuroinflammation11. Network-based methods frequently employ topological network properties for ranking the drug candidates45 such as the closeness centrality value, like in several algorithms of NeDRexApp17. Here we adapt the computational pipeline in10, an already validated effective procedure for drug repurposing, by which a potential therapeutic candidate was identified and subsequently experimentally tested10. Indeed, disease and drug data investigation by enrichment analysis, network representation, and a final score calculation based on topological graph properties, as well as gene ontology knowledge and drug perturbation data, make the workflow versatile and adaptable to other applications. Research on GeneCards and literature was conducted to identify genes associated with neuroinflammation. The gene set retrieved from GeneCards, which collects genomic, proteomic, and genetic data, was increased with genes derived from the literature with the search terms “neuroinflammation” and “transcriptomics”. This step allowed us to produce an important gene set integration with several genes significantly associated with the neuroinflammatory condition. To mitigate the bias that could affect initial gene collection toward most characterized genes, and reduce the risk of only considering biased proteins and interactions, the following steps of enrichment analysis, integration with tissue-specific PPI and regulatory interactions, and the multi-parametric scoring method were fundamental. The tissue-specific network built by merging a PPI with a regulatory network allowed us to consider both protein interactions and their gene expression regulation by the activity of transcription factors. This aspect was particularly important in considering how protein expression is regulated, which is crucial to better understanding a complex pathological condition10. On this network, disease-related modules were identified by employing the community detection function proposed in10. Among the 7 modules, two can be distinguished for their size, namely 369 and 739 nodes. In particular, the module composed of 369 nodes and 1080 edges had the highest number of connections and included the Syk target (Table 2). As expected, the modules prevalently included genes associated with immune response, neuroinflammatory processes, signal transduction, and nervous system development (Supplementary File S3). By integrating all the filtering and enriching steps, we focused on common inflammatory mechanisms shared among various diseases characterized by chronic neuroinflammation and altered microglial homeostasis, such as neurodegenerative conditions like Alzheimer’s Disease, Parkinson’s Disease, Amyotrophic Lateral Sclerosis, or Schizophrenia. As already observed in metabolic syndrome10, an essential aspect of the applied computational analysis was defining the proximity score assigned to each drug. The score calculation relies on the topological properties of the network and the functional similarity of genes, allowing also considering biological knowledge about proteins. Moreover, it effectively assigned a high score to drugs previously investigated or already employed as a treatment for diseases involving neuroinflammation, including DEX, which was here selected as a reference drug in the in vitro validation; notably, such drugs, already employed against neuroinflammation, were filtered out during the last step of the analyses to produce the final drug ranking. Such efficiency of the scoring method further confirmed the approach’s suitability for drug repurposing studies focused on pathologies different from metabolic syndrome. According to the obtained ranking Fostamatinib was identified for the subsequent in vitro tests because of its Syk inhibitory activity. Syk, although expressed in various cell types, including epithelial cells, fibroblasts, and neuronal and vascular endothelial cells, plays a critical role in signal transduction in immune cells46 Syk contains two src homology 2 (SH2) domains that bind to tyrosine residues on the immunoreceptor tyrosine-based activating motif (ITAM), which is located in the cytoplasmic regions of several immune receptors such as Fc receptors, TCRs, BCRs, and natural killer cell receptors23. The tyrosine kinase activity leads to the subsequent activation of downstream signalling molecules, resulting in the induction of expression of proinflammatory genes, such as TNF-α, COX-2, and iNOS, and the production of inflammatory mediators, such as NO and PGE2 in macrophages47. The inhibition of Syk suppresses downstream signal transduction48, it has been positively tested in rat neuron-glia cocultures against neuronal loss LPS-induced49 and on neuroinflammation murine models by observing a reduction in cognitive dysfunctions50,51. Table 2Disease modules. Disease Modules identified by employing the community detection function reported in10. For each module are indicated the number of nodes, of edges, and the most represented biological processes and molecular functions retrieved from DAVID.NodesEdgesBiological meaning1017Inflammatory response, Chemotaxis, Cytokine1218Complement pathway, Immunity, Protease2331Cell adhesion, Hydrolase739999Transport, Cell adhesion, Angiogenesis, Stress response, Endocytosis, Host-virus interaction, Cell cycle, Protease, Integrin, Growth factor, RNA-binding66127Transport, Ion channel, Receptor1622Complement pathway, Immunity, Protease3691080Cell adhesion, Growth regulation, Angiogenesis, Phagocytosis, Protein transport, Apoptosis, Immunity, Neurogenesis, Inflammatory response, Kinase, Receptor, Phosphatases Disease modules. Disease Modules identified by employing the community detection function reported in10. For each module are indicated the number of nodes, of edges, and the most represented biological processes and molecular functions retrieved from DAVID. Fostamatinib is an FDA-approved drug for chronic immune thrombocytopenia purpura (ITP)32, investigated for the treatment of other diseases such as rheumatoid arthritis33 different types of cancer35,36, and more recently for severe COVID-1937,52. EMA approved Fostamatinib for ITP treatment in patients with an insufficient response to other therapies. The efficacy and safety of the prodrug has been demonstrated in two Phase III, randomised, doubleblind, placebo-controlled studies in adult patients with previously treated persistent or chronic ITP53. As reported in the EMA report of Tavlesse, Fostamatinib is highly bound to plasma proteins (98.3% in human plasma); following oral administration, Fostamatinib is rapidly converted to its active metabolite R406 that has a mean absolute bioavailability of 55% and the median Tmax of approximately 1.5 h53. Data regarding the drug’s ADME, following oral dosing of 14C-Fostamatinib, suggested a wide distribution to all tissues with the exception of the CNS53. Indeed, Fostamatinib (but not R406) was determined to be a P-gp inhibitor in vitro, and P-gp acts as an efflux pump at the blood–brain barrier54. Thus, different formulations to administer directly R406 required further investigation. For instance, interestingly, some reversible distribution of drug to red blood cells was seen53. Lastly, Fostamatinib has been recently identified as an interesting drug candidate in multiple networks in-silico studies and drug repurposing investigations focusing on peripheral inflammation17,54. In line with these studies, we decided to focus our analyses on Fostamatinib’s active metabolite, R406, which recently resulted to be effective in the most common in vitro model of ischemic stroke (oxygen-glucose deprivation (OGD) model) and chronic migraine in murine microglia18,19. In particular, Xu and colleagues highlighted that the pre-treatment of R406 was able to reduce the levels of IL-1β and IL-18 in primary murine microglia injured with ischemic stroke18, while Fan et al., to study Metformin effects in regulating neuroinflammation through the TREM2-SYK pathway, confirmed the R406 ability to reduce NLRP3 and IL-1β levels19. Moreover, R406 was employed as a Syk inhibitor for investigating specific neuroinflammatory mechanisms in various disease conditions, such as neuromyelitis optica spectrum disorder (NOSD) or in mice with diabetic retinopathy20,21. Zhang et al. reported how the detrimental effects of CD22 blockade in a NOSD mouse model were diminished by R406 treatment, helping in understanding CD22’s role in microglial activity21. Liu et al. employed R406 to deepen the role of microglial Syk in diabetic retinopathy progression, observing how R406 reduced retinal inflammation and inhibited microglia activation caused by high glucose treatment20. However, all these studies were conducted on murine models of diseases presenting neuroinflammation, having different characteristics than human neuroinflammatory response41. To the best of our knowledge, there is a single article22 using R406 on human iPSC-derived microglia but in a model of AD-transformed cells and not on cells treated with a proper inflammatory stimulus. McQuade et al. demonstrated that the R406 pre-treatment partially counteracted phagocytosis in AD-like cellular models, but the amount of inflammation markers was not assessed. Considering the importance of phagocytosis in AD, R406 effects could be explained by the modulatory effect on TGF-β levels44,55, which are highly involved in phagocytosis, and by the very high drug concentration (5 µM) employed, five hundred times greater than the dose we used, and that showed cell toxicity44,55. Such discrepancy in R406 toxicity levels could be related with the differences between iPSC-derived microglia and immortalized cell lines in gene expression profiles and proliferation rate56,57, that could make C20 and HMC3 cells more sensitive to R406 concentrations and the Syk pathway a highly responsive therapeutic target. Thus, being the previous studies about R406 effects on neuroinflammation mainly conducted in murine models18–21, we decided to investigate the Syk inhibitor’s activity on human cell lines of microglia, able to mount a typical neuroinflammatory response to further confirm the R406 potential in counteracting the human processes involved in neuroinflammation41. Herein, the two microglial cell lines utilized were the C20 and HMC3 cells; C20 are immortalized microglia derived from cortical tissue of adults, and HMC3 cells are human embryonic microglia (embryos aged between 8 and 12 weeks). Despite iPSC-derived microglia has a more similar gene expression profile with primary microglia then immortalized cell lines, both C20 and HMC3 have been recently confirmed for being two effective models to study inflammatory cascades and regulatory mechanisms56. Indeed, both cell lines are commonly used in literature to assess the molecular processes at the basis of human neuroinflammation39,58–60. IL-1β-stimulated C20 cell line is a widely used model of activated microglia39,60. HMC3 cells, treated with IL-1β and IFN-γ as inflammatory stimuli both at a concentration of 10 ng/mL41, are a model of activated microglia, characterized by a significant increase of pro-inflammatory cytokines expression. Our results, obtained on human microglia inflammatory models, support the potential of Fostamatinib active metabolite in regulating the neuroinflammatory response. Indeed, after an initial evaluation to exclude the concentration range of R406 cell toxicity, microglial cells were inflamed and treated with 10 nM R406 to assess its effects on modulating inflammatory mediator production. Such a dose, being below the previously reported Ki of 30 nM48, allowed us to operate at a non-toxic concentration able to partially inhibit the Syk cascade, well known for involving several mediators and confirmed by our network analysis for being a central regulatory hub. 10 nM R406 was used in comparison with 1 µM DEX, as a well-known anti-inflammatory drug, with the highest potency among all glucocorticoids (a potency 25 times higher than hydrocortisone)38. Interestingly, the NF‑κB activation data further support the effectiveness of R406 even at the lowest concentration tested, while also indicating that a broader dose range may exert additional modulatory effects. In particular, the 50 nM dose produced a marked reduction in p65 nuclear translocation, highlighting that higher, but still non‑toxic, concentrations can enhance the anti‑inflammatory response. However, the dose dimension requires careful consideration in microglial systems. As already suggested by the viability assays, R406 displays a narrow window between effective pathway modulation and the onset of cellular stress, especially in HMC3 cells. This is reflected in the divergent behaviour observed between the two microglial models. While C20 cells exhibited a clear dose‑dependent decrease in NF‑κB activation, the highest concentration (100 nM) failed to reduce p65 nuclear translocation in HMC3 cells compared with cytokine‑stimulated controls. This discrepancy may arise from the greater sensitivity of HMC3 cells to R406‑induced cytotoxicity, pointing to a potential cell‑type‑specific response. Additionally, the presence of IFN‑γ used in the inflammatory cocktail for HMC3 may activate other compensatory signaling pathways, such as JAK/STAT‑dependent mechanisms, that sustain NF‑κB activity when Syk inhibition becomes more pronounced61. Together, these findings underscore that while R406 can attenuate NF‑κB activation in human microglia, its efficacy and directionality depend on both the cellular background and the inflammatory context. This highlights the importance of dose selection when targeting central signaling hubs such as Syk in neuroinflammatory settings. Overall, in both in vitro models, despite not significant effects were observed compared to the inflammatory stimulus, R406 was able (i) to reduce the expression levels of the pro-inflammatory factors IL-6, TNF-α, and NF‑κB and (ii) to promote a mild anti-inflammatory response, herein represented by increased levels of IL-10 transcript. Interestingly, the comparative evaluation of R406 and DEX effects showed an analogue trend of activity in the modulation of gene expression in both cell models, with R406 showing a clearer trend in counteracting the pro-inflammatory response and DEX in triggering the anti-inflammatory one. Differences between the two drugs are simply related to the entity of the effects. For instance, DEX was associated with a higher increase in IL-10 levels in both cell lines with respect to R406. A larger reduction of IL-6 in C20 cells was evidenced for DEX with respect to R406, whereas in HMC3 cell lines, R406 was more effective in reducing IL-6 mRNA than DEX. Notably, R406 was associated with a higher reduction of TNF-α in both models, with respect to DEX treatment. The marked effect of R406 on TNF-α expression, compared with DEX, might suggest a particular efficacy in counteracting the triggering of the inflammatory response, being this cytokine one of the first involved in the initiation of inflammatory processes. Consequently, the reduced upregulation of IL-10 and of TGF-β in HMC3 following treatment with R406 could be due to a pro-inflammatory response that is not particularly intensive enough to promote the anti-inflammatory phenotypic shift, as it can be observed in cells treated with only the inflammatory stimulus. However, some differences in the response can be highlighted between the two models, and that could be due to both the different cellular backgrounds (C20 vs. HMC3) and the different types of inflammatory stimulus (IL-1β vs. IL-1β/ IFNγ). Results suggested that C20 cells spontaneously respond to inflammation by increasing the expression of TGF-β, a cytokine involved in restoring the microglial homeostatic signature. In this case, for DEX and R406 treatments, no changes in TGF-β expression were observed. Another substantial difference was observed in the expression of NLRP3. Indeed, while in C20, R406 showed a higher reduction of NLRP3 expression compared to DEX, the same effect was not found for HMC3. However, the discrepancy could be explained by the presence of IFN-γ in the inflammatory mix used for HMC3, as it is reported that IFN-γ is able to induce the expression and activation of NLRP362, probably via Syk-independent mechanisms. This, together with a limited anti-inflammatory response induced by R406, may not be sufficient to counteract the increased expression of NLRP3 in HMC3 significantly. These results together strengthen the observations previously made on different neuropathology models, confirming the effective beneficial role of R406 in partially modulating the human microglial inflammatory response. R406 and Fostamatinib were already tested in pathologies showing an inflammatory component, however, some important insights arise from the comparison between our data and those present in currently available literature. Indeed, Fostamatinib has already been identified as an interesting repurposing candidate in human peripheral tissue inflammation by in silico analyses17, but the mechanisms involved in the central inflammatory response can be widely different from the peripheral ones. For example, central and peripheral inflammatory processes mainly differ in the cell types involved in the response, such as microglia and lymphocytes, respectively. Moreover, R406 effects on counteracting neuroinflammation were mostly investigated on murine models that can present significant differences with human-specific neuroinflammation; mice and rats are more resistant than humans to inflammatory stimuli and have different response signaling. For instance, some immunological similarities between rats and humans are not present in mice, and although many murine immune markers and mechanisms have been introduced as the gold standard in human immunology, some of these emerged to be not completely transferable to the human immune system63. Fostamatinib, the methylene phosphate prodrug, was synthesized to address the low aqueous solubility of R40664, and it was suggested as a repurposed candidate from this computational pipeline; however, some aspect regarding pharmacokinetics must be considered before continuing further studies in neuroinflammation. As previously reported in this discussion, to overcome the low distribution in the CNS, various alternative drug formulations or administration methods beyond oral could be explored. For instances, potential strategies to bypass the blood–brain barrier include the intranasal or a direct and localized administration, as well as novel formulations, such as lipid nanocarriers, that could be designed to cross the BBB and enhance R406 stability, optimizing therapeutic levels in the brain. Another critical issue is that Fostamatinib has multiple targets, and its administration requires further attention for secondary effects and potential interactions with other drugs. Nevertheless, Syk inhibition remains a promising strategy, and Fostamatinib a drug potentially effective to investigate, also considering these results about R406. Overall, these data confirm the effectiveness of a systems biology approach in drug repurposing studies and, more specifically, the easy adaptability of the computational pipeline developed in10. The workflow was successfully fit to analyze the neuroinflammatory condition and identify potential drug repurposing candidates. Indeed, Fostamatinib active metabolite R406 was confirmed as a valid repurposed drug with a partially modulatory effects on the neuroinflammatory cascade in the two in vitro models of human microglia employed. Despite the necessity of additional experiments to further support Fostamatinib/R406 translation, such as its validation in more physiologically complex systems better mimicking human neuroinflammatory mechanisms, and additional investigation of its pharmacokinetics and novel drug formulations, these results on human microglial models confirm Syk inhibition as a promising strategy to be further explored in neuroinflammation, suggesting a possible adjuvant role of R406 in counteracting the neuroinflammatory response. Methods Identification of genes associated with neuroinflammation The list of genes associated with neuroinflammation was created from GeneCards and PubMed. A direct search of the term “neuroinflammation” was used with GeneCards. Literature research of gene names co-occurring with “neuroinflammation” and “transcriptomics” was conducted by a text mining analysis on PubMed abstracts using PubTator Central65. The two gene lists were integrated by removing 81 duplicates genes to ensure a non-redundant input for network construction66. Genes located on sex chromosomes were individuated and excluded using the R package biomaRt67. Pathway analysis and tissue-specific network Given the heterogeneity of the data sources, a gene-set enrichment analysis was performed using the Functional Annotation tools on DAVID (accessed in August 2023) (v. 2021 update), focusing on pathways. Pathways related to neuroinflammation with nominal p-value < 0.01 were retained and used as an intermediate filtering step to define the gene set for downstream network construction. Because this step was used for feature prioritization rather than for stand-alone pathway-level inference, the enrichment results are interpreted cautiously. Only genes included in those pathways were considered for the following steps. The PPI network was created using HIPPIE starting from the gene set previously identified. A threshold value of 0.73 was chosen to consider only the most reliable interactions. A regulatory network for the brain, which represents the interaction between genes and transcription factors, was obtained from the online resource “regulatory circuits”10 and filtered as previously described10. Cytoscape68 was used for opening and analyzing both tissue-specific PPI and regulatory networks: duplicate edges and self-loops were eliminated, and the two networks were merged forming the tissue-specific background network characterized by 8,535 nodes and 35,209 edges. Network modules associated with neuroinflammation were detected using the walktrap algorithm in combination with an overrepresentation test as previously described10. These modules were used to calculate the proximity score to be associated with drug candidates. Drug repurposing To identify new therapeutic uses for drugs already available on the market, we selected from DrugBank (version 5.1.10, release date: 2023-01-04, database downloaded in August 2023) approved drugs having as a target one of the genes of our set. Data about the gene expression profile of drug-perturbated cells were obtained from the Library of Integrated Cellular Signatures (LINCS) (https://clue.io). LINCS is a database that provides gene expression signatures from a variety of cell types treated with a specific perturbagen. The top 100 up-and down-regulated genes were retrieved for each drug with a high technical and functional quality signature parameter (is_hiq = 1) and with the highest signature strength parameter (distil_ss). Following10, these up-and down-regulated genes were additionally filtered for having a medium/high semantic similarity with the target (ThrSimScore = 0.5). The target and the filtered up-and down-regulated genes were connected to generate the network-based drug modules. The drug modules and the disease modules previously generated were used for calculating a score for each drug. The proximity score calculated using the functions described in10, integrates a network-based distance and a functional similarity measurement, considering both the topological properties of the network and the biological aspect of the genes: a drug has a high score if its module is close to the neuroinflammation modules and if the genes in them have a similar biological function. In particular, the nodes distance within the network referred to the shortest path connecting drug module’s proteins with the closest protein of a disease module. On the other hand, the functional similarity is a measurement referring to the genes’ biological function according to Gene Ontology annotations. Only drugs with a score greater than the median value were retained. The remaining drugs with pharmacologically active drug-target interactions were filtered following research conducted on the OpenTargets platform (v. 23.06): drugs were excluded for being already used in therapy for the treatment of diseases characterized by neuroinflammation, or for being already studied in the past, also considering their participation in clinical trials. To exclude drugs with adverse effects on the CNS, a final filtering procedure was conducted using DrugCentral (https://drugcentral.org) (Database Update Sept 2022)69. Cell culture C20 cell line The human microglial C20 cell line was developed through SV40/hTERT immortalization of primary cultures of human microglia derived from cortical tissue of adult subjects, kindly provided by Dr Álvarez-Carbonell (Case Western Reserve University, Cleveland, OH, USA)70. It was cultured in DMEM-F12 medium supplemented with 10% fetal bovine serum (FBS), 100 U/mL penicillin, 100 mg/mL streptomycin, and 600 µg/mL neomycin (Sigma-Aldrich S.r.l., Milan, Italy). Cells were maintained in an atmosphere of 5% CO2 at 37 °C. HMC3 cell line The Human Microglial Clone 3 (HMC3) cell line was developed through the SV40-dependent immortalization of microglial cells from human spinal cord and cortical tissues, obtained from embryos aged between 8 and 12 weeks. It was purchased from ATCC, and cultured in EMEM medium supplemented with 10% fetal bovine serum (FBS), 100 U/mL penicillin, 100 mg/mL streptomycin, and 600 µg/mL neomycin (Sigma-Aldrich S.r.l., Milan, Italy). Cells were maintained in an atmosphere of 5% CO2 at 37 °C. All materials for cell culture were purchased by Corning, New York, USA. Cell treatments and evaluation of cell viability C20 and HMC3 cells were seeded in 96-well microplates respectively at a density of 10,000 cells/well and 7,000 cells/well, and maintained in their specific, complete culture media. The day after the seed, cells were treated with increasing concentrations (10, 50, 100, 500 nM and 1, 10, 50, and 100 µM) of Syk inhibitor R406 (Selleckchem, Houston, TX, USA) for 24 h. Lyophilized R406 was reconstituted in DMSO and diluted in complete cell media or starvation media according to the experimental design for in vitro treatments. Cell viability was examined by MTS colorimetric assay according to the manufacturer’s instructions (Promega, Milano, Italy) by taking the reported IC50 41nM of R406 into account. The absorbance was read at 490 nm using the EnSight multimode plate reader (PerkinElmer, Waltham, MA, USA), equipped with Kaleido Data Acquisition and Analysis Software. To evaluate the cell viability in the presence of both the Syk inhibitor and inflammatory stimulus, the same procedure has been applied to challenged cells concomitantly treated with a selected concentration of R406 (10 nM). The inflammatory stimulus used was IL-1β 20 ng/mL for C20 cells39 and IL-1β + IFNγ, each at the concentrations of 10 ng/mL, for HMC3 cells, as reported in previous literature41. Immunocytochemistry The day before the experiment, C20 and HMC3 cells were seeded in a six-well plate with a density of 5 × 104 cells/well and maintained in complete cell media. Then, cells were concomitantly treated with the inflammatory stimulus and three different concentrations of R406 (10, 50, and 100 nM) in serum-free medium for 24 h. DEX at 1 µM concentration was used as a positive control. Cells were fixed with PFA 4% for 10 min, washed with phosphate buffered solution (PB), permeabilized for 10 min using PB-BSA 2.5% + 0,1 Triton X-100 (Bio-Rad, United States), incubated for 1 h at room temperature in blocking solution, and then incubated overnight at 4°C with primary antibody anti-p65 (Santa-Cruz, sc-8008, mouse, dilution 1:100). The day after, cells were washed with PB and incubated with anti-mouse IgG conjugated with an Alexa Fluor 594 (Invitrogen, A-21207, dilution 1:100) dye for 2 h at room temperature. Coverslips were then mounted with Vectashield containing DAPI and images were acquired by confocal microscopy (Nikon Eclipse Ti-A1 MP) using a 60x/1.4NA oil objective. All images were analysed by ImageJ program by quantifying the mean of fluorescence intensity in the nuclear region identified in the DAPI field. mRNA analysis in real-time PCR (qPCR) The day before the experiment, C20 and HMC3 cells were seeded in a six-well plate with a density of 2 × 105 cells/well and maintained in complete cell media. After treatments, total RNA was extracted using the RNeasy Mini Kit (Qiagen, Hilden, Germany), and 1 µg RNA was converted into cDNA through iScript cDNA Synthesis Kit (Bio-Rad, Hercules, CA, USA) using random hexamers (Promega). RT-PCR reactions were performed with 50 ng of cDNA in a final volume of 20 µl using SsoAdvanced Universal SYBR Green Supermix (Bio-Rad). Previously validated primers39 were used to assess NLRP3, TNF-α, IL-6, IL-10, TGF-β, and NF‑κB. β-actin was used as a housekeeping gene (F: 5’-GCACTCTTCCAGCCTTCCTTCC-3’; R: 5’-GAGCCGCCGATCCACACG-3’) (Supplementary Table S4). Reactions were performed in triplicate, and expression data was calculated as the relative expression index of 2-ΔΔCt. Statistical analysis Statistical analyses and figures were produced by the Graph-Pad Prism program (GraphPad Software Inc., San Diego, CA, USA). The significance of the cell viability in the presence of inflammatory stimulus was determined by an unpaired t-test. The significance of the differences in mRNA expression was determined by non-parametric Anova (Friedman test) followed by multiple comparisons (Mann-Whitney test). Statistical significance was considered P-value ≤ 0.05. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 2 Supplementary Material 3 Supplementary Material 4 Supplementary Material 5 Supplementary Material 6 Supplementary Material 7 Author contributions Conceptualization: M. Cirinciani, P. Milazzo, C. Priami, E. Da Pozzo. Methodology and data analysis: M. Cirinciani, L. Germelli. Writing and editing: M. Cirinciani, L. Germelli, P. Milazzo, C. Priami, C. Martini, E. Da Pozzo. Funding This work is partially funded by (i) PNRR ECS00000017 “THE - Tuscany Health Ecosystem” - Spoke 6 “Precision medicine & personalized healthcare”, funded by the European Commission under the NextGenerationEU programme; (ii) Ministry of University and Research (MUR) as part of the PON 2014–2021 “Research and Innovation” resources - Green - DM MUR 1062/2021 (iii) “Hub multidisciplinare e interregionale di ricerca e sperimentazione clinica per il contrasto alle pandemie e all’antibioticoresistenza (PAN-HUB)” funded by the Italian Ministry of Health (POS 2014–2020, project ID: T4-AN-07, CUP: I53C22001300001). Data availability The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. These authors contributed equally: Martina Cirinciani and Lorenzo Germelli. References Associated Data Supplementary Materials Supplementary Material 1 Supplementary Material 2 Supplementary Material 3 Supplementary Material 4 Supplementary Material 5 Supplementary Material 6 Supplementary Material 7 Data Availability Statement The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request.