Defining blood-induced microglia functions in neurodegeneration through multiomic profiling

Blood protein extravasation through a disrupted blood–brain barrier and innate immune activation are hallmarks of neurological diseases and emerging therapeutic targets. However, how blood proteins polarize innate immune cells remains largely unknown. Here, we established an unbiased blood-innate immunity multiomic and genetic loss-of-function pipeline to define the transcriptome and global phosphoproteome of blood-induced innate immune polarization and its role in microglia neurotoxicity. Blood induced widespread microglial transcriptional changes, including changes involving oxidative stress and neurodegenerative genes. Comparative functional multiomics showed that blood proteins induce distinct receptor-mediated transcriptional programs in microglia and macrophages, such as redox, type I interferon and lymphocyte recruitment. Deletion of the blood coagulation factor fibrinogen largely reversed blood-induced microglia neurodegenerative signatures. Genetic elimination of the fibrinogen-binding motif to CD11b in Alzheimer’s disease mice reduced microglial lipid metabolism and neurodegenerative signatures that were shared with autoimmune-driven neuroinflammation in multiple sclerosis mice. Our data provide an interactive resource for investigation of the immunology of blood proteins that could support therapeutic targeting of microglia activation by immune and vascular signals.

Blood protein extravasation through a disrupted blood-brain barrier and innate immune activation are hallmarks of neurological diseases and emerging therapeutic targets. However, how blood proteins polarize innate immune cells remains largely unknown. Here, we established an unbiased blood-innate immunity multiomic and genetic loss-of-function pipeline to define the transcriptome and g lo bal p ho sp ho pr oteome of blood-induced innate immune polarization and its role in microglia neurotoxicity. Blood induced widespread microglial transcriptional changes, including changes involving oxidative stress and neurodegenerative genes. Comparative functional multiomics showed that blood proteins induce distinct receptor-mediated transcriptional programs in microglia and macrophages, such as redox, type I interferon and lymphocyte recruitment. Deletion of the blood coagulation factor fibrinogen largely reversed blood-induced microglia neurodegenerative signatures. Genetic elimination of the fibrinogen-binding motif to CD11b in Alzheimer's disease mice reduced microglial lipid metabolism a nd n eu ro de ge ne rative signatures that were shared with autoimmune-driven n eu roinflammation in multiple sclerosis mice. Our data provide an interactive resource for investigation of the immunology of blood proteins that could support therapeutic targeting of microglia activation by immune and vascular signals.
Vascular and immune signals are potent activators of the innate immune response in a wide range of autoimmune, inflammatory and infectious diseases in the brain and the periphery [1][2][3][4] . Innate immune cells integrate environmental signals to rapidly activate target genes and perform specialized cellular functions 5 . Pathogenic activation of microglia contributes to oxidative stress, inflammation and neurodegeneration in both Alzheimer's disease (AD) and multiple sclerosis (MS) 6 . Blood-brain barrier (BBB) disruption is an early pathological feature  Tables 2 and 3). Indeed, the number of genes in the blood-induced microglia gene network was reduced by 97%, with significant downregulation of 52% of the genes, when plasma was derived from Fga −/− mice (Fig. 1f, Extended Data Fig. 3 Table 3), suggesting that fibrinogen is a key protein in the blood that induces microglia activation. Through unbiased KEGG pathway analysis of DEGs between Fga −/− and WT plasma-stimulated microglia, we identified the 12 top pathways induced by fibrinogen, including ROS (for example, Hmox1, Cox7a2, Slc25a5), COVID-19 (for example, Ccl12, Rps8, Rpl35) and AD (for example, Atp5e, Psmd2, Tubb5) (Fig. 1e,g). Similarly, microglia gene expression was reduced in response to plasma from Fgg γ390-396A mice, in which fibrinogen had been mutated to lack the CD11b-CD18 binding motif but retained normal clotting function 19,23 compared with WT plasma (Fig. 1c and Supplementary Table 2). Whereas the effect of C3 −/− or Alb −/− plasma on microglia was largely similar to that of WT plasma (five and three DEGs, respectively), Fga −/− and Fgg γ390-396A plasma induced major gene expression changes in microglia (348 and 331 DEGs, respectively) (Fig. 1c, Extended Data Fig. 2 and Supplementary Table 2). These results are consistent with reduced demyelination in the corpus callosum induced by Fga −/− or Fgg γ390-396A plasma compared with WT plasma or fibrinogen administration 21,24 . Collectively, these results suggest that there is specificity among blood proteins in the induction of microglia transcriptional changes, indicating that fibrinogen signaling is a critical regulator in the blood for the induction of oxidative stress and disease-induced signatures in microglia following BBB leakage.
We report a blood-induced microglia gene network and show that blood proteins elicit distinct receptor-mediated transcriptional changes and signaling programs in innate immune cells. We provide a transcriptomic and phosphoproteomic atlas of fibrin-, iC3b-and lipopolysaccharide (LPS)-selective activation of innate immunity and reveal ligand-selective pathways with differential functions in MS and AD mice. We identify fibrin-CD11b signaling as causal for neurotoxic microglial programming in disease. Moreover, our study provides a resource for the investigation of the immunology of blood proteins in inflammatory, autoimmune and neurodegenerative diseases.

Blood-induced microglial transcriptomic profiling
To discover the molecular programs controlling microglial and macrophage polarization by blood proteins, we developed an unbiased blood-innate immunity multiomic and genetic loss-of-function pipeline consisting of deep sequencing of blood-induced transcriptomes, functional single-cell and oxidative stress transcriptomics, global phosphoproteomics and integration with innate immune signatures from AD and MS models (Extended Data Fig. 1). To determine the blood-induced transcriptome in microglia, we stereotactically delivered wild-type (WT) plasma to the brain, followed by RNA sequencing (RNA-seq) analysis of sorted microglial cells ( Fig. 1a and Supplementary Table 1). By unbiased analyses of differentially expressed genes (DEGs), gene ontology (GO) networks and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways, we showed that WT plasma induced widespread microglial transcriptional changes, including changes involving genes related to oxidative stress (for example, Hmox1, Romo1, Gpx1), disease-associated microglia (DAM) (for example, Cst7, Spp1) and the cell cycle (for example, Top2a, Cdkn2d), as well as reactive oxygen species (ROS), oxidative phosphorylation, neurodegeneration, AD, glutathione metabolism and COVID-19 pathways (Fig. 1b-e and Supplementary Table 2). These changes were largely absent following stimulation with plasma derived from fibrinogen-deficient Fga −/− mice but were relatively preserved following stimulation with complement 3-deficient (C3 −/− ) or albumin-deficient (Alb −/− ) plasma   Table 5). By contrast, GO analysis of the BMDM-complement-cluster showed enrichment of pathways for host defense response, myeloid cell differentiation, innate immune and lymphocyte activation and of related genes (for example, S100a8, S100a9, Cxcr4, Glul and Maf) (Fig. 3d,e and Extended Data Fig. 4j). Antiviral gene signatures were identified as shared GO terms in both the fibrin and iC3b clusters ( Fig. 3e and Extended Data Fig. 4l). As expected for LPS-primed macrophages 5 , LPS clusters were enriched in gene pathways for MAPK activity, inflammatory cytokine and chemokine production, and chemotaxis (for example, Il1b, Il6 and Tnf) (Fig. 3d, Extended Data Fig. 4j and Supplementary Table 5). Pseudotime analysis showed a two-path transcriptional bifurcation from unstimulated to iC3b to the fibrin-induced state (path 1) or unstimulated to fibrin to the LPS-induced state (path 2) (Extended Data Fig. 5a,b and Supplementary Table 6), suggesting that the CR3 and TLR4 ligands induce distinct activation pathways. All cell clusters were enriched for monocyte-derived macrophage markers (for example, Lyz2, Cd14 and Ctsd) and had low expression of monocyte-derived dendritic cell markers (for example, Itgax) (Extended Data Fig. 5c), suggesting that undifferentiated or contaminating cells were not a major driver of clustering. Together, these results show that fibrin polarization primarily promotes prooxidant, lipid metabolism and IFN-I responses, whereas complement iC3b and LPS induce host defense and classical inflammatory states, respectively.

Phosphoproteomics reveal distinct fibrin and iC3b signaling
Although macrophage signal transduction pathways have been extensively studied for Toll-like receptor ligands 25 , the downstream signaling cascades for CR3 ligand activation have not been characterized. We performed unbiased quantitative phosphoproteomics 26 using mass spectrometry to globally characterize the dynamics of protein phosphorylation in response to complement iC3b or fibrin stimulation in RAW 264.7 macrophages. Hierarchical clustering analysis of detected phosphorylation sites revealed distinct signaling profiles for fibrin and iC3b (Extended Data Fig. 6a,b and Supplementary Tables 7 and 8). Fibrin initially induced a greater increase in global phosphorylation relative to iC3b, as evidenced by a significant increase in the detection and abundance of differentially expressed phosphosites (DEPs) 1 h after treatment (Extended Data Fig. 6a-c and Supplementary Table 9). iC3b stimulation induced a greater number of detected phosphosites at 3 h after stimulation (Extended Data Fig. 6a-c and Supplementary Table 9), suggesting differential phosphorylation kinetics.
The regulation of phosphorylation signaling cascades is largely mediated by protein kinases. To predict kinase-substrate relationships, we bioinformatically calculated kinase activities from our phosphoproteomics data 26 . We identified significant activation of MAPK1 (also known as extracellular signal-regulated kinase 2, ERK2) and serum-glucocorticoid kinase 1 (SGK1) at 1 h after fibrin but not iC3B treatment (Fig. 4d and Supplementary Table 11). At 3 h after stimulation, we observed significant activation of casein kinase II (CSNK2A1), S6 kinase (RPS5KA1), protein kinase C and D kinases, and pyruvate dehydrogenase kinases (PDK1-4) in iC3b-treated samples relative to fibrin-treated samples ( Fig. 4d and Supplementary Table 11). Together, these results show that fibrin and iC3b induce differential phosphorylation events and kinase activities, suggesting that blood proteins induce distinct signal transduction pathways in innate immune cells.
Next, we validated the top proteins phosphorylated by fibrin in BMDMs, primary microglia and brain tissue from AD mice. Paxillin binds to the cytoplasmic domain of β2 subunits of integrins including CD11b-CD18, and its phosphorylation initiates focal adhesion complex formation upon integrin engagement with ECM 27 . Fibrin induced dynamic phosphorylation of paxillin at residue 83 (p-PXN) in BMDMs and primary microglia (Fig. 4e,f), consistent with fibrin activation of focal adhesions in platelets 28 . MEK2 phosphorylates ERK1/2, resulting in increased cellular proliferation and migration, oxidative stress and inflammation. We tested the effects of fibrin on MEK2 phosphorylation and proinflammatory gene activation. Fibrin induced robust phosphorylation of MEK2 at residue 394 (p-MEK2) in BMDMs and microglia (Fig. 4e,g). Specific MEK2 inhibitor trametinib blocked p-MEK2 and reduced expression of fibrin-induced gene Il1b in fibrin-treated BMDMs (Extended Data Fig. 6e,f), suggesting that MEK2 activation mediates fibrin-induced proinflammatory gene expression. Similarly, treatment with the therapeutic monoclonal 5B8 antibody, Resource https://doi.org/10.1038/s41590-023-01522-0 which targets the fibrin-binding site to the CD11b I-domain without affecting fibrin polymerization 29 , blocked p-MEK2 in fibrin-treated BMDMs (Fig. 4h), suggesting that fibrin-induced phosphorylation is receptor mediated. The cytosolic NADPH oxidase subunit NCF2 translocates to the plasma membrane upon phosphorylation by ERK1/2 and phosphatidylinositol-3-kinase, leading to NADPH oxidase activation   and ROS production 30 . Fibrin induced phosphorylation of p-NCF2 in BMDMs and primary microglia (Fig. 4e,g), consistent with fibrin activation of NADPH oxidase and ROS generation 20,29,31 . NADPH oxidase activation has been identified in progressive MS 32 and has been implicated in neurodegeneration and cognitive impairment in AD mice 10,20 .
To test whether NCF2 was phosphorylated in disease models in vivo, we assessed NCF2 expression and activation in the brains of 5XFAD mice, a model of AD. p-NCF2 and total NCF2 were higher in 5XFAD than in nontransgenic (NTG) control mice at 12 months of age (Fig. 4i). Together, these results identify fibrin as a CD11b-CD18 ligand coupling integrin signaling with NADPH oxidase activation (Extended Data Fig. 7). They also reveal the phosphoproteome of fibrin and iC3b and demonstrate the specificity of blood proteins in controlling and integrating innate immune signaling pathways in disease.

Fibrin drives neurotoxic innate immune programs in MS mice
Oxidative injury is associated with neuronal loss and myelin damage and has been proposed as a key contributor to disease pathogenesis in MS and AD 6,33,34 . As both transcriptomic and phosphoproteomic analyses identified oxidative stress as a key fibrin-induced pathway, we performed an unbiased overlay of the fibrin, iC3b and LPS gene signatures (this study) with oxidative stress central nervous system (CNS) innate immune signatures that have been defined in a model of MS 31 . Single-cell RNA-seq of oxidative stress producing cells (Tox-seq) in an experimental autoimmune encephalomyelitis (EAE) model previously identified distinct cell subsets polarized toward oxidative stress (MgV and MpI clusters), whereas others were enriched in antigen-presenting and phagocytic genes (MgIII and MpIII clusters) 31 . We found that stimulation by fibrin or LPS but not iC3b recapitulated the core oxidative stress signature (for example, Ncf2, Cybb, Sod2 and Irg1) expressed by ROS + microglia and macrophages in EAE ( Fig. 5a Table 12). The microglial-shared iC3bfibrin gene signature (for example, C1qc, C1qa, Fcrls, Clec7a, Apoe, Sepp1) was enriched in the EAE microglia antigen-presenting MgIII cluster 31 (Fig. 5a,b and Supplementary Table 12). The BMDM iC3b gene signature was enriched in the EAE macrophage clusters MpV and MpVI ( Fig. 5a,b), which were identified as phagocytic subsets based on their gene expression programs 31 . All treatments significantly downregulated the homeostatic microglia gene signature (for example, Cx3cr1, Trem2, Bin1 and Cst3) 31,35 ( Fig. 5b and Extended Data Fig. 8b). The fibrin transcriptomic signature is consistent with reduced oxidative stress, demyelination, axonal damage and protection from paralysis in Fgg γ390-396A mice and WT mice treated with fibrin-targeting antibody 5B8 in EAE 19,24,29 . These data suggest that fibrin and complement iC3b are potent signals in neuroinflammatory lesions that recapitulate the polarization of function-specific innate immune responses, with fibrin serving as a potent inducer of oxidative stress gene programs in microglia and peripheral macrophages.

Fibrin drives neurotoxic microglia programs in AD mice
We next used Tox-seq to analyze the transcriptomes of CD11b + cells labeled for ROS production (as assessed by 2′,7′-dichlorofluorescein diacetate (DCFDA)) via scRNA-seq from brains of NTG control or 5XFAD mice (Extended Data Fig. 8c-f). Using unbiased clustering analysis superimposed with functional ROS characterization, we identified four transcriptionally distinct CD11b + clusters containing ROS + and ROS − cell populations as visualized by UMAP (Fig. 6a,b and Supplementary Table 13). Most ROS + microglia (50% of cells) from 5XFAD mice were found in microglia cluster 4 (cluster Mg4) (Fig. 6b), which had enrichment of genes involved in neurodegenerative microglial cell activation and iron transport (for example, Apoe, Tyrobp and Fth1) (Fig. 6c,d, Extended Data Fig. 8g,h and Supplementary Table 13). By contrast, genes known to negatively regulate ROS production, superoxide metabolism and maintenance of the microglial homeostatic signature were increased in ROS − microglia from 5XFAD mice (for example, Nrros, Clk1 and Zeb2, respectively) ( Fig. 6c,d). Differential gene expression analysis showed few changes in ROS + compared with ROS − microglia from NTG mice (Extended Data Fig. 8i). Functional subcluster analysis of both 5XFAD microglia clusters revealed that subcluster 0 had the highest single-cell expression of ROS + microglia ( Fig. 6e, Tox-seq cluster overlay). Fibrin-induced genes were enriched in subcluster 0 ( Fig. 6e, fibrin signature overlay). ROS + microglia were significantly enriched for fibrin-induced and iC3b-fibrin-induced genes but not LPS-induced genes ( Fig. 6f and Supplementary Table 12).

Fibrin microglia signatures shared between AD and MS mice
We next compared the Tox-seq transcriptomic profiles of microglia between the 5XFAD and EAE models. The oxidative stress core signature identified in ROS + microglia from EAE mice was also present in 5XFAD mice ( Fig. 8a and Supplementary Table 16). Although 67 DEGs were shared between EAE and 5XFAD ROS + microglia (for example, Apoe), the majority of genes were specific for either 5XFAD (132 DEGs) or EAE (170 DEGs), such as Igf1 and Il1b, respectively (Fig. 8b,c). These results are in line with human microglial transcriptomics identifying partial overlap between MS and AD 37 . Pathway analysis of the microglial oxidative stress genes shared among EAE and 5XFAD models identified enrichment in pathways related to blood coagulation and hemostasis (for example, Plaur, Slc16a3, Eno1), antigen presentation (for example, H2-Ab1, H2-K1, Cd74), neutrophil degranulation (for example, B2m, Cstb, Bst2) and the tyrosine kinase binding protein Tyrobp network (Fig. 8d, Extended Data Fig. 10a and Supplementary  Table 16). We next overlaid the shared AD and EAE oxidative stress signature with the blood-induced microglia profiles (Fig. 8e). The WT plasma signature overlapped with the shared oxidative stress microglia signature, indicating that the dataset alignment identified blood-induced microglia genes both in MS and AD mice. Fga −/− plasma largely reduced oxidative stress and disease-associated transcripts to control levels ( Fig. 8e and Extended Data Fig. 10b,c), suggesting that fibrinogen is a key ligand in the blood that activates neurodegenerative microglia responses. Taken together, these data suggest a pathogenic role for fibrin-induced microglia polarization in neurodegeneration in both MS and AD, demonstrating shared and unique drivers of innate immune-driven neurotoxicity.

Discussion
We report the first unbiased transcriptome and phosphoproteome of blood-induced polarization of innate immunity, revealing the selectivity and causal role of blood proteins in mediating neurotoxic microglial functions. Traditionally, blood leaks were considered to be secondary to inflammation, with largely interchangeable functions once they extravasated into the brain 7,38 . Through in vivo genetic loss-of-function studies combined with unbiased comparative transcriptomics analysis, our study shows the specificity of blood proteins in differentially activating receptor-mediated immune responses required for pathogenic microglial gene programs in AD and MS models. Blood-induced prooxidant programming of neurotoxic microglia occurs along common molecular pathways across neurodegeneration and CNS autoimmunity. Fibrin-CD11b signaling was necessary for neurotoxic microglia programs in AD mice consisting of neurodegenerative, oxidative stress and lipid metabolism that were shared with an MS model. Given that immune, vascular and blood signals are key players in aging, neurological and peripheral diseases 2  biomarkers and for discovery of drugs that selectively target pathogenic innate immunity in aging and inflammatory, autoimmune and neurodegenerative diseases.
Our study provides a mechanistic link between cerebral vascular pathology and neurodegeneration by identifying fibrin-CD11b signaling as an apical inducer of neurotoxic pathways in innate immune cells.    Our model proposes that fibrin binding to CD11b-CD18 induces an outside-in integrin signaling cascade to initiate focal adhesion complex formation, activate MAPK and transactivate NADPH oxidase to induce proinflammatory, oxidative stress and IFN-I signaling responses. Fibrin also induces phosphorylation of SMARCA5, RANBP3 and NUP98, suggesting regulation of nuclear import, chromatin remodeling and transcription 40 . Formation of multiprotein adhesion complexes that link the ECM to the cytoskeleton, MAPK signaling, oxidative stress and gene expression identify fibrin as a mechanoregulator of CD11b-CD18 integrin effector functions in innate immune cells 27,41,42 . Indeed, increased fibrin deposition, paxillin, MAPK and sustained MEK2 activation, and NADPH oxidase activity promote oxidative injury and mediate neurodegeneration and synaptic dysfunction in both MS and AD 6,20,[43][44][45][46][47] . In addition to NADPH oxidase, we identified mitochondrial VDAC1 and ApoE as induced by fibrin, suggesting that fibrin-induced oxidative stress may be mediated by multiple pathways. We also identified fibrin as a potent activator of STATs, ISGs and IFN-I response, which induces neuronal dysfunction in AD 48 and T cell effector functions in MS and autoimmune diseases 49 . Fibrin could thus be a driver of IFN-I response in disease. Together, these results identify fibrin as a key signal required for pathogenic polarization of immune cells at sites of vascular damage. We discovered unique and shared transcriptomic and phosphoproteomic signatures induced by fibrin and iC3b, indicating that ligand-biased CR3 signaling may underlie the pleiotropic functions of CR3 in innate immune cell polarization [50][51][52] . Our results suggest that limiting fibrin-induced innate immune responses may suppress oxidative injury and neurodegeneration, whereas suppressing complement may preferentially reduce phagocytosis and antiviral immune responses. iC3b-specific signatures could also be relevant to complement diseases such as C3 glomerulopathy 1 . Ligation of CR3 by fibrin and iC3b may lead to differential conformational changes in its ectodomain, leading to ligand-biased outside-in signaling 41 . The stoichiometry and spatial distribution of fibrin and iC3b may also contribute to biased ligand signaling and transactivation of other receptors. Fibrin deposits and  NTG 5XFAD 5XFAD: Oxidative stress Resource https://doi.org/10.1038/s41590-023-01522-0 complement activation may affect immunopathogenesis of intracerebral hematoma in conditions such as stroke and traumatic brain injury, in addition to neurodegenerative diseases 53 . Future studies will be necessary to determine how the cross-talk of the fibrinogen and complement pathways orchestrate oxidative injury and phagocytic signaling cascades, respectively. As CR3 mediates both protective   and damaging immune functions, strategies to target ligand-selective activation pathways may have therapeutic benefits 8 . Fibrin-CD11b signaling is required for pathogenic innate immune activation in the brain and periphery 19,20,23,29,[54][55][56] . Fgg γ390-396A mice and those treated with the fibrin-targeting 5B8 antibody show protection against neurodegeneration and cognitive impairment in AD models and against paralysis and axonal damage in EAE [19][20][21]24 . Protection from neurodegeneration upon inhibition of fibrin-CD11b signaling may be due to selective suppression of neurotoxic pathways in microglia, such as ApoE, IFN-I and oxidative stress pathways identified in this study 32,34,49,57 . Thus, fibrin-targeting immunotherapy could be a therapeutic strategy in AD and MS without adverse anticoagulant effects or global suppression of innate immunity.
The resource we provide here should be considered with some caveats. As in vitro microglia cultures do not fully recapitulate in vivo homeostatic signatures 58 , we also validated the ligand-selective signatures in vivo. Antibody-depleted plasma for fibrin, C3 or other blood proteins could complement the genetic depletion used in this study. Our phosphoproteomic analysis was performed in RAW 264.7 cells owing to their transcriptomic similarity to BMDMs, their use in phosphoproteomic studies 25,59 and because of the technical demand for high cell numbers. We validated fibrin-induced phosphosites identified in RAW 264.7 cells in BMDMs and primary microglia and in 5XFAD mice, suggesting that the phosphorylation events also occur in vivo. Future studies in primary cells with additional concentrations and time points could be used to assess differential signaling pathways. For our comparative transcriptomic analysis, we selected three ligands-fibrin, iC3b and LPS-in macrophages and microglia owing to their broad and diverse immune roles in vascular, inflammatory and infectious disease. We selected fibrin and iC3b to study ligand-biased signaling of CD11b-CD18. Future studies could use our platform to compare additional complement and coagulation activators in other cell types. Another potential limitation was that protein aggregates could induce signaling independent of specific ligand-receptor interactions. However, we found specific receptor-mediated transcriptomic programs for blood proteins in vivo and suppression of the neurodegenerative gene signature in 5XFAD:Fgg γ390-396A mice, indicating the distinct transcriptional programs of blood proteins are not due to aggregate protein formation. Although the results of transcriptomic analysis in 5XFAD:Fgg γ390-396A mice support the fibrin-induced microglia response being CD11b dependent, they do not exclude direct or indirect effects of fibrin on other cellular targets in the brain 3 . The Trem2 pathway was induced in microglia in 5XFAD mice but not in the healthy brain by the blood, potentially owing to the difference in pathology as well as the age of the mice and the time points in the study. We performed Tox-seq analysis in 5XFAD mice, an AD model dependent on immune and vascular mechanisms 20,60 . Future studies could use Tox-seq to characterize neurotoxic innate immune responses in other neurodegeneration models with different etiologies to discover additional pathways related to disease progression.
In summary, we demonstrated that blood-induced polarization of innate immunity is causal for the induction of neurotoxic microglial programming in disease. By establishing a blood-innate immunity multiomic and genetic loss-of-function pipeline, we defined fibrin as a unique blood protein required for microglial polarization to oxidative stress and neurodegenerative phenotypes in MS and AD mice. Our study uncovers principles of distinct transcriptomic and phosphoproteomic events induced by immune and vascular signals and their contributions to immune diversity in autoimmune and neurodegenerative disease. Furthermore, we lay the groundwork for future experiments to define the spatiotemporal regulation of blood-induced innate immune cell polarization, which may enable discovery of selective therapeutic strategies in inflammatory, neurological and infectious diseases.

Online content
Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41590-023-01522-0. Violin plots of EAE oxidative stress signature enrichment in 5XFAD and NTG mice (right); n = 1,579 cells from brains of three 5XFAD and three NTG mice. Violin plots depict minimum, maximum and median expression, with points showing single-cell expression levels. Box plots show the first to third quartiles (25-75% box bounds) with median values indicated and upper and lower whiskers extending to 1.5× interquartile range. b, Venn diagram of oxidative stress genes in CD11b + ROS + microglia from 5XFAD or EAE. c, UMAP plots of shared and unique oxidative stress microglia genes in 5XFAD and EAE. Gene expression overlays for Apoe, Igf1 and Il1b are shown. Gene expression is depicted as log-normalized scaled expression. The red outline demarcates microglial cells in a ROS-enriched cluster. Representative genes shared between or specific to 5XFAD and EAE Toxseq are shown. d, Metascape analysis of top significant gene pathways shared in ROS + microglia from 5XFAD and EAE mice. e, Dot plot of gene expression from blood microglia profiles overlaid with the 5XFAD and EAE shared oxidative stress gene signature in microglia as shown in b.

Bulk RNA-seq of microglia
Samples from the plasma injection experiment were processed for RNA-seq using an Ovation RNA-seq System V2 low input kit (NuGEN) as previously described 31 . Libraries were equimolar pooled and sequenced on a NovaSeq 6000 S4 (Illumnia) with 200 paired-end reads to a depth of >50 million reads per library. Paired-end fastq files were processed using the Nextflow RNA-seq pipeline 67 Table 3). GO pathways were determined using functional enrichment analysis in String 69 with default parameters visualized in Cytoscape.

Cell culture
Primary microglia were prepared from neonatal rats at postnatal day 5 or from C57BL/6J mice at postnatal days 2-3 as previously described 21,31 . Viability was assessed using trypan blue. Microglia cultures from one litter were used for each independent experiment. BMDMs were prepared from male and female 12-20-week-old mice as previously described 21,31 and used for experiments after 6 or 7 days of differentiation. For BMDM scRNA-seq, two male or two female mice were pooled per experiment. Individual animals were used for biological replicates unless otherwise stated. RAW 264.7 macrophages were obtained from ATCC and cultured in Dulbecco's modified Eagle medium (DMEM) supplemented with 10% fetal bovine serum (FBS).
Control wells were treated with the same buffer void of fibrin or iC3b. For LPS stimulation, BMDMs or microglia were seeded into 24-well tissue-culture-treated plates at 5.0 × 10 5 cells ml −1 , allowed to adhere overnight, then treated with 100 ng ml −1 LPS (Sigma-Aldrich, O55:B5) for the durations indicated in figure legends. Following stimulation, adherent cells were lifted with accutase (StemCell Technologies) and viable cells were counted by trypan blue, then resuspended in RPMI with 5% FBS and used for scRNA-seq. For scRNA-seq library preparation, BMDMs or primary microglia were processed with a Chromium Single Cell 3′ v.2 kit according to the manufacturer's guidelines (10x Genomics). Libraries were balanced to achieve a minimum of 75,000 reads per cell and run on two lanes of a NovaSeq 6000 (Illumnia) with 150 paired-end reads. Samples were demultiplexed, and fastq files were used to align reads to the mm10 reference assembly (downloaded 2019) and aggregated using the Cell Ranger count and aggr packages (10x Genomics).

5XFAD Tox-seq
Samples were prepared for Tox-seq analysis as previously described 31 with the following modifications. Male and female 12-month-old 5XFAD and NTG mice were perfused with 4 °C DPBS, and cortical and hippocampal regions were dissected. Tissues were incubated with lysis buffer without ActD for 30 min at 37 °C. Using a FACSAria II, live Sytox blue − CD11b + ROS − and live Sytox blue − CD11b + ROS + cell populations were sorted into tubes containing FACS buffer. Sorted cells were resuspended in 4 °C DPBS supplemented with 2% FBS and immediately processed for scRNA-seq.
For scRNA-seq library preparation, live Sytox blue − CD11b + ROS − and live Sytox blue − CD11b + ROS + sorted cell populations were processed using the Chromium Single Cell 3′ v.2 kit following the manufacturer's instructions. Balanced library pools were sequenced across three lanes of a HiSeq 4000 system (Illumnia) with a targeted sequencing depth of 100,000 reads per cell. Reads were mapped to the mm10 genome (downloaded 2019), and samples were combined and sequence-depth normalized using Cell Ranger count v.3.0.2 and aggr packages, respectively. https://doi.org/10.1038/s41590-023-01522-0

Brain CD11b scRNA-seq
Brains from male and female 6-month-old 5XFAD, 5XFAD:Fgg γ390-396A , NTG and Fgg γ390-396A mice were processed for FACS as described for 5XFAD Tox-seq analysis. Live Sytox blue -CD11b + cells from cortical/ hippocampal tissues were sorted into tubes containing DPBS supplemented with 5% FBS at 4 °C and then resuspended in 4 °C DPBS supplemented with 2% FBS at 333 cells μl −1 and processed for scRNA-seq with the Chromium Single Cell 3′ v.3 kit following manufacturer's instructions. Balanced library pools were run across three lanes of Hiseq4000, reads mapped to the mm10 genome and samples combined and sequence-depth normalized using the Cell Ranger count v.3.0 and aggr packages, respectively.

scRNA-seq data analysis
The R toolkit Seurat 70 was used for QC, clustering analysis and differential gene expression analysis of scRNA-seq data in R v.4.0.2 unless otherwise stated. For scRNA-seq data visualizations, dittoseq package v.1 was used to produce UMAPs, dot plots and violin plots 71 .
For microglia scRNA-seq analysis (Fig. 2), the QC parameters were: nFeature_RNA > 1000; nFeature_RNA < 5500; and < 5% and 20% mitochondrial and ribosomal genes, respectively. nCount_RNA in the 93rd percentile (nCount_RNA < 26,206) was used for downstream analysis. Data were normalized and scaled, and a percentage of mitochondrial and cell cycle genes were regressed out using Seurat SCTransform. Jackstraw was performed with num.replicate of 100. The FindNeighbors and FindClusters functions in Seurat were used with the first eight significant principal components (PCs) and a resolution of 0.4, respectively. A total of 16,186 microglial cells passed QC, with an average of 3,469 genes per cell and 20,228 genes. Consistent with the literature 72 , canonical microglial markers were expressed at varying levels in the identified clusters (Extended Data Fig. 3e). Cluster DEGs were determined by FindAllMarkers with default parameters. Genes that met the log 2 FC threshold of >0.25 with adjusted P < 0.05 (Benjamini-Hochberg correction) were used for downstream analysis.
For BMDM scRNA-seq analysis (Fig. 3), two independent experiments were integrated and corrected for batch effects as previously described 70 . The batch-corrected dataset QC parameters were: nFea-ture_RNA > 200; nFeature_RNA < 5000; and < 5% and 25% mitochondrial and ribosomal genes, respectively. The 17,625 QC-passed BMDMs were used with the Seurat integration workflow using default parameters. Jackstraw was performed with num.replicate of 100. The RunUMAP, FindNeighbors and FindClusters functions were used with the first 20 significant PCs and a resolution of 0.5. DEGs were determined by FindAllMarkers with default parameters. Genes that met the log 2 FC threshold of 0.25 with adjusted P < 0.05 (Benjamini-Hochberg correction) were used for downstream analysis. Pseudotime trajectories were performed on the UMAP embeddings and Seurat clusters using  Table 6). To generate the heat map, a pseudocount of 1 was added to the raw scRNA-seq counts for the top 50 DEGs, the rows were log 2 row-normalized and K-means clustering was performed on the rows.
For 5XFAD Tox-seq analysis (Fig. 6), the QC parameters were 200-5,000 nFeature_RNA, <7,500 nCount_RNA, and <5% and 20% mitochondrial and ribosomal genes, respectively. Data were normalized and scaled, and a percentage of mitochondrial genes were regressed out using Seurat SCTransform. Jackstraw was performed with num. replicate of 100. FindNeighbors and FindClusters Seurat functions were used with the first 30 significant PCs and a resolution of 0.6, respectively. To remove variation in sex-linked genes, the dataset was integrated using the Harmony algorithm 75 with runHarmony: group. by.vars = sex and assay.use = SCT. Clustering analysis was performed using Harmony dims = 15 and resolution = 0.4. In accordance with previous literature 76 , all four CD11b + cell clusters had high expression of core microglial genes (Extended Data Fig. 9e and Supplementary  Table 13). DEGs for each cluster were determined by FindAllMarkers with default parameters using MAST statistical test. Genes that met the log 2 FC threshold of 0.25 with adjusted P < 0.05 (Benjamini-Hochberg correction) were used for downstream analyses.

scRNA-seq signature enrichment
Average expression levels for a given gene list were computed across single-cell transcriptomes using the AddModuleScore function in Seurat with default parameters. The modular scores of a gene list were visualized in UMAP or violin plot. The list of genes used is provided in Supplementary Table 12.

Functional enrichment and network analysis of scRNA-seq data
Functional enrichment analysis of DEGs was performed in Metascape using default parameters 77 , and significant GO terms were identified by false discovery rate (FDR) P < 0.05 unless otherwise stated. Gene network analyses were performed with GSEA with C5.bp.v7.1symbols.gmt using default settings. GO terms with P < 0.10 were used for enrichment map visualization in Cytoscape v.3.7.2 and unbiasedly clustered using the AutoAnnotate v.1.3.2 plugin with default settings. For the microglial dataset, cluster gene signatures were determined using ClusterProfiler 78 and the gseGO function with the following parameters: ont = BP, nPerm = 10000, minGSSize = 3, maxGSSize = 800, pvalueCutoff = 0.1, OrgDB = org.Mm.eg.db, pAdjustMethod = BH.

Phosphoproteomics sample preparation
RAW 264.7 macrophages (10 × 10 6 cells, 10 mg of protein per sample) were prepared for global phosphorylation protein sample digestion for mass spectrometry analysis as previously described 79 . Macrophages were plated on fibrin-coated (final concentration, 12.5 μg ml −1 ) or iC3b-coated (final concentration, 10 μg ml −1 ) plates for 1 or 3 h. Fibrin concentration was based on our previous studies 21,29,31 . We selected a comparable concentration for iC3b based on previous studies using similar concentrations for macrophage effector responses and CD11b binding 80 . Under these conditions, in our previous studies, we observed phosphorylation at longer time points in primary Schwann cells 81 . RAW cells were used owing to the high protein concentration needed for phosphoproteomic analysis, which could not be feasibly obtained from primary BMDM cultures using instrumentation at the time of the study.

Mass spectrometry analysis
Samples were analyzed on a Thermo Scientific Orbitrap Fusion mass spectrometry system equipped with an Easy nLC 1200 uHPLC system interfaced with the mass spectrometer via a Nanoflex II nanoelectrospray source as previously described 79 . https://doi.org/10.1038/s41590-023-01522-0

Mass spectrometry data processing and statistical analysis
Quantitative analysis was performed in R v.4.1.3. Initial QC analyses, including interrun clusterings, correlations, PCA, and peptide and protein counts and intensities were completed with the R package artMS v.1.12.0. Two sample outliers in intensities and peptide detections were discarded before quantitative analysis: fibrin 1 h (PRIDE sample ID FU20180420-23) and iC3b 1 h (PRIDE sample ID FU20180420-05). Statistical analyses of phosphorylation changes between stimulated and control runs were carried out using peptide ion fragment intensity data output from MaxQuant with preprocessing using artMS. Quantifications of phosphorylation based on peptide ions were performed with artMS::doSiteConversion and artMS::artmsQuantification with default settings using artMS 82 . All peptides containing the same set of phosphorylated sites were grouped and quantified together into phosphorylation site groups, and equal median normalization was performed across runs to control for differences in sample preparation. Statistical tests in MSstats compared phosphopeptide intensities between stimulated and control conditions for each time point. We compared each stimulation condition with its time-matched control and compared stimulations with each other (that is, fibrin versus iC3b). We used defaults for MSstats for adjusted P values (Student's t test and Benjamini-Hochberg correction), even in cases of n = 2 biological replicates. We quantified between 2,000 and 6,000 phosphorylated peptides per sample, mapping to 300-3,000 different proteins per sample.

Kinase activity analysis
FC values from MSstats were reduced to a single FC per site by choosing the FC with the lowest P value (noninfinite log 2 -transformed FC values) and used for kinase activity and enrichment analysis. Mus musculus phosphorylation sites were converted to their Homo sapiens orthologous sites. Orthologous pairs of gene identifiers between M. musculus and H. sapiens were downloaded from Ensembl using BioMart. Ensembl gene identifiers were mapped to UniProt identifiers, and orthologous pairs of sequences were aligned using the Needleman-Wunsch global alignment algorithm implemented using the Biostrings v.2.62.0 function pairwiseAlignment with default parameters in R. The resulting alignments were used to convert the sequence positions of phosphorylations in M. musculus to positions in H. sapiens protein sequences, if possible. Kinase activities were estimated using known kinase-substrate relationships 83 and inferred as a z score calculated using the mean log 2 FC of phosphorylated substrates for each kinase in terms of standard error (z = (M − u)/s.e.), comparing FCs in phosphosite measurements of the known substrates against the overall distribution of FCs across the sample. P values were calculated using two-tailed z test 84 . We collected substrate annotations for 400 kinases with available data. Kinases with two or more measured substrates were considered to be predicted kinases (Supplementary Table 11).

Network reconstruction and enrichment analysis of phosphoproteomics data
Proteins with changes in phosphorylation state were selected based on an FDR threshold of 0.05. Protein phosphorylation site pairs significant for at least one time point were maintained. After filtering, iC3b resulted in 44 phosphoproteins, and fibrin resulted in 68 phosphoproteins. The STRING database was queried using Cytoscape. Proteins with STRING interaction scores higher than 0.4 were connected by edges with widths and opacities reflecting the score level. Phosphorylation state changes were visualized using Omics Visualizer 85 as two outer ring circles representing phosphorylation at 1 h and 3 h. To enhance the signal, we included up to ten additional nodes identified by the STRING database as functionally related to our phosphoproteins using stringApp 86 . Final results were filtered based on an FDR threshold of 0.05, and redundant results were removed using a redundancy cutoff of 0.5. Two significant GO terms were selected and visualized as node fill colors. STRING-provided proteins and unconnected proteins were removed for visualization.

Fibrin phosphorylation cell assays
BMDMs were cultured for 18 h in RPMI-1640 supplemented with 1% FBS (RPMI 1% FBS). Cells were plated on fibrin-coated dishes for 15-90 min in RPMI 1% FBS and then processed for either immunocytochemistry (ICC) or immunoblotting. Unstimulated BMDMs served as controls. Primary rat microglia were used on day 4 in vitro and plated on fibrin-coated dishes for 15-75 min in DMEM supplemented with 2% FBS. Unstimulated microglia served as controls.

Pharmacologic inhibition assays
BMDMs were cultured for 18 h in RPMI-1640 supplemented with 1% serum. For MEK inhibition, cells were preincubated with 20 nM trametinib (S2673, Selleckchem) for 2 h and then plated on fibrin-coated plates for 90 min for ICC or 6 h for quantitative PCR. Cells unstimulated in RPMI 1% FBS for 90 min or 6 h were used as time point zero controls. Dimethyl sulfoxide was used as a vehicle control. Fibrin-CD11b blockade using 5B8 monoclonal antibody was performed as previously described 29 . In brief, 5B8 or IgG2b isotype control antibodies were preincubated (each 50 μg ml −1 ) in fibrin-coated plates for 90 min at 37 °C before cell plating. Cells were incubated on fibrin for 90 min and processed for ICC. Cells incubated with 5B8 or IgG2b in the absence of fibrin served as controls.

Quantitative real-time PCR
Quantitative PCR and data analysis were performed as previously described 31 . The following primer sequences were used. Gapdh: forward, caaggccgagaatgggaag; and reverse, ggcctcaccccatttgatgt. Il1b: forward, agttgacggaccccaaaag; and reverse, agctggatgctctcatcagg.

Statistical analyses for nonsequencing data
Data are presented as mean ± s.e.m. with overlaid scatter plot. Data distribution was assumed to be normal, but this was not formally tested. Two-tailed unpaired t test or Mann-Whitney test, one-way analysis of variance (ANOVA) and two-way ANOVA tests were performed with GraphPad Prism v.9. No statistical method was used to predetermine sample sizes, but sample sizes were similar to those used in our previously published studies 29,31 . All mice survived until the end of the study, and all of the data were analyzed. Mice were randomized and blindly coded for group assignment and data collection for immunohistochemistry and ICC experiments. For in vivo stereotactic plasma injections, mice were randomized and blindly coded for group assignment and data collection. For all scRNA-seq experiments, mice were randomized by sex and genotype before sample preparation. All injections, histological analyses and quantification were done in a blinded fashion. Quantification of immunohistochemistry data was performed independently by two blinded observers.

Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability
The scRNA-seq and bulk RNA-seq datasets are deposited in the Genome Expression Omnibus under SuperSeries accession number GSE229376. Searchable web resources from this study of the microglia and BMDM ligand-activation scRNA-seq data are available at https://toxseq. shinyapps.io/ligand_activation/, and the single-cell 5XFAD Tox-seq data are available at https://toxseq.shinyapps.io/5xfad_toxseq/. The EAE Tox-seq data are available at https://toxseq.shinyapps.io/scrnaseq_ viewer/. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD021230. The phosphoproteomic interaction networks have been made available through NDEx at https://doi.  Fig. 4 | Quality control for primary microglia scRNA-seq and BMDM scRNA-seq datasets. a, b, Violin plots of number genes (a) and unique molecular identifiers (UMI, b) per cell post-normalization and shown for each biologically independent sample from scRNA-seq of primary microglia stimulated with fibrin, iC3b, LPS or left unstimulated. Data are from two biologically independent samples of fibrin, two biologically independent samples of iC3b, two biologically independent samples of LPS, and one biologically independent sample of unstimulated primary microglia. c, Elbow plot of top PC used to select for clustering analysis. d, Distribution of cells in each biologically independent sample across each seurat clusters. e, Dot plot of selected microglial gene markers across each cluster from scRNA-seq dataset of primary microglia as shown in Fig. 3a. Average gene expression and cell population expression is depicted as log expression and percent, respectively. f, g, Violin plots of number genes (f) and UMI (g) per cell post-normalization and shown for each biologically independent sample from scRNA-seq of primary BMDMs stimulated with fibrin, iC3b, LPS or left unstimulated. Data are from three biologically independent samples of fibrin, one biologically independent sample of iC3b, two biologically independent samples of LPS and four biologically independent samples of unstimulated BMDMs. h, Elbow plot of top PC used to select for clustering analysis. i, Distribution of cells in each biologically independent sample across each seurat clusters. j, Gene-set enrichment plots of top GO terms for a given cluster (adjusted P value < 0.05 with BH correction). Violin plots depict minimum, maximum, and median expression, with points showing single-cell expression levels (a, b, f, g). Box plots show the 1st to 3rd quartiles (25-75% box bounds) with median values indicated and upper and lower whiskers extending to 1.5*inter-quartile range (a, b, f, g). Fibrin binding to CD11b-CD18 leads to the conversion of the integrin to the high-affinity extended-open (active) conformation that induces signal transduction in macrophages. This outside-in signaling is propagated by the formation of focal adhesions through recruitment and phosphorylation of scaffold proteins and signaling kinases such as paxillin and focal adhesion kinase (FAK) resulting in phosphorylation of PI3K and cytoskeleton organization. In parallel, the MAPK cascade MEK2 and ERK1/2 components are phosphorylated leading to 1) transactivation of NADPH oxidase complex (NOX2) and mitochondria responses to induce ROS release and oxidative stress, 2) phosphorylation of SMARCA5, NUP98, and the ERK1/2 nuclear transporter RANBP3 to regulate nuclear import, 3) phosphorylation of IRF2B2 regulating IFN signaling and 4) transcriptional activation of fibrin-induced genes involved in inflammatory, oxidative stress, and IFN-I responses. Phosphorylation (P); fibrininduced proteins identified in this study, red filled shapes; fibrin-induced genes identified in this study are shown in box. Created with Biorender.com. https://doi.org/10.1038/s41590-023-01522-0 Extended Data Fig. 8 | Ligand-induced profile overlays with EAE innate immune cell signatures, and quality control data related to Fig. 6. a, Dot plot of selected gene markers across scRNA-seq datasets of primary microglia or BMDMs unstimulated or stimulated with fibrin, iC3b, or LPS. Gene expression is depicted as scaled log-normalized expression. b, Violin plot of primary microglia overlaid with microglial homeostatic gene signature from healthy mice as previously identified 31 . Treatments in x-axis are rank ordered from highest to lowest expression. P < 0.0001 by one-way ANOVA with Tukey's multiple comparison test. c, Flow cytometry plots of live CD11b + ROSand live CD11b + ROS + cells from brains of 12-month-old 5XFAD and NTG mice. Cell population (%) shown inside plot. Data representative of two independent experiments. d, Quantification of total live CD11b + cells and CD11b + ROS + cells from brains of 12m 5XFAD and NTG mice. Data from n = 3 mice per genotype shown as mean ± s.e.m. ROS production assessed via DCFDA (c,d). P < 0.05 as determined by two-tailed, unpaired t-test with Welch's correction. e, f, Violin plots of number genes (e) and UMI (f) per cell post-normalization, shown for each biologically independent sample for 5XFAD and NTG Tox-seq analysis. Data from n = 3 mice per condition. Box plots show the 1st to 3rd quartiles (25-75% box bounds) with median values indicated and upper and lower whiskers extending to 1.5*interquartile range. g, UMAP plots as shown in Fig. 6a overlaid with microglial gene marker expression. Expression depicted as log-fold change expression. h, Heat map of top DEGs per single cell cluster from 5XFAD Tox-seq dataset. Gene expression depicted as scaled z-score. i, Volcano plot of DEGs in microglia between CD11b + ROS + compared to CD11b + ROScells in NTG mice. Dots depict average log 2 FC and -log 10 adjusted P values (log 2 > 0.25, adjusted P value < 0.05, MAST test with BH correction). Violin plots depict minimum, maximum, and median expression, with points showing single-cell expression levels (b, e, f).

Resource
Corresponding author(s): Katerina Akassoglou, PhD Last updated by author(s): Apr 18, 2023 Reporting Summary Nature Portfolio wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency in reporting. For further information on Nature Portfolio policies, see our Editorial Policies and the Editorial Policy Checklist.

Statistics
For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.

n/a Confirmed
The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly The statistical test(s) used AND whether they are one-or two-sided Only common tests should be described solely by name; describe more complex techniques in the Methods section.
A description of all covariates tested A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient) AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals) For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted

Software and code
Policy information about availability of computer code Data collection Single cell transcriptomes were prepared using 10X Genomics Chromium Controller and sequenced using Illumnia Novaseq6000 and

March 2021
were found to be significant at least at one time point were maintained. After filtering, iC3b resulted in 44 phosphoproteins, and fibrin resulted in 68 phosphoproteins. To investigate the functional relatedness of proteins, STRING database was queried using the network analysis tool Cytoscape. Proteins with STRING interaction scores higher than 0.4 were connected by edges with widths and opacities reflecting the score level. Phosphorylation state changes were visualized using Omics Visualizer as two outer ring circles, with each layer representing phosphorylation at 1 h and 3 h. In order to enhance the signal for enrichment analysis, we also included up to ten additional nodes (proteins) identified by the STRING database as functionally related to our phosphoproteins using the stringApp. Final results are filtered based on an FDR threshold of 0.05 and redundant results were removed using a redundancy cutoff of 0.5. Two significant Gene Ontology (GO) Biological Process terms were selected and visualized as node fill colors. STRING-provided proteins and unconnected proteins were removed for visualization.
For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors and reviewers. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Portfolio guidelines for submitting code & software for further information.

Data
Policy information about availability of data All manuscripts must include a data availability statement. This statement should provide the following information, where applicable: -Accession codes, unique identifiers, or web links for publicly available datasets -A description of any restrictions on data availability -For clinical datasets or third party data, please ensure that the statement adheres to our policy The scRNA-seq and bulk RNA-seq datasets are deposited in the Genome Expression Omnibus under the SuperSeries accession number GSE229376. A searchable web resource of the microglia and BMDM ligand-activation scRNA-seq data are available at https://toxseq.shinyapps.io/ligand_activation/ and the single-cell 5XFAD Tox-seq data at https://toxseq.shinyapps.io/5xfad_toxseq/. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD021230. The protein interaction networks have been made available through NDEx at https:// doi:10.18119/N9CK5Z and https://doi:10.18119/N9H89M.

Field-specific reporting
Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.

Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see nature.com/documents/nr-reporting-summary-flat.pdf

Life sciences study design
All studies must disclose on these points even when the disclosure is negative.

Sample size
No statistical methods were used to pre-determine sample sizes but sample sizes are similar to our previous publications (Ryu et  Data exclusions From bulk RNA-seq experiment, three samples were removed that did not pass RNA and cDNA library quality control testing. One sample was removed due to large deviation on PCA and poor sequence alignment. For phosphoproteomic data, two sample outliers in intensities and peptide detections were discarded prior to quantitative analysis: Fibrin 1 h (PRIDE sample ID FU20180420-23) and one iC3b 1 h (PRIDE sample ID FU20180420-05) samples. No samples or animals were excluded from any other analyses.

Replication
The number of experimental repeats is detailed at the bottom of each legend for each figure. All attempts at replication following the protocols described in the methods were successful.
Randomization Mice were randomized and blindly coded for group assignment and data collection for IHC and ICC experiments. For in vivo stereotactic plasma injections, mice were randomized and blindly coded for group assignment and data collection. For all scRNA-seq experiments, mice were randomized by sex and genotype prior to sample preparation.

Blinding
Stereotactic surgery was performed blinded to the experimental groups. For IHC experiments, image acquisition and quantification was performed by observers blinded to experimental conditions. Images were quantified independently by two blinded observers.
Reporting for specific materials, systems and methods We require information from authors about some types of materials, experimental systems and methods used in many studies. Here, indicate whether each material, system or method listed is relevant to your study. If you are not sure if a list item applies to your research, read the appropriate section before selecting a response. Methods n/a Involved in the study ChIP-seq

Animals and other organisms
Policy information about studies involving animals; ARRIVE guidelines recommended for reporting animal research

Laboratory animals
Male and Female C57BL/6J and B6.Cg-Tg(APPSwFlLon,PSEN1*M146L*L286V) 6799Vas/Mmjax (5XFAD) mice were purchased from the Jackson Laboratory. Fga-/-and Fgg390-396A mice were obtained from Dr. Jay Degen (University of Cincinnati, OH, USA). 5XFAD mice were crossed with Fgg390-396A mice to generate 5XFAD:Fgg390-396A mice. Male and female mice were used in this study. Sprague-Dawley female rats with litters were purchased from Charles River. Animals were housed under IACUC guidelines in a temperature and humidity-controlled facility with 12 h light-12 h dark cycle and ad libitum feeding.

Wild animals
The study did not involve wild animals Field-collected samples The study did not involve samples collected from the field Ethics oversight All animal protocols were approved by the Committee of Animal Research at the University of California, San Francisco, and were in accordance with the National Institutes of Health guidelines.
Note that full information on the approval of the study protocol must also be provided in the manuscript.