TRAM promotes Toll-like-receptor-free myddosome signal transduction

We established a flow-cytometry-based assay to separate myddosome-containing cells from cells lacking myddosomes (referred to as bystanders). This assay is based on the concept that the large number of μm-sized myddosomes within individual cells would change the side scatter (SSC) properties, when assessed using flow cytometry³². To test this idea, we used immortalized bone-marrow-derived macrophages (iBMDMs) that contain a biallelic knock-in of an AGF (Apex2-mEGFP-Flag) tag at the endogenous locus encoding the myddosome component MyD88 (ref. ¹). MyD88–AGF iBMDMs treated with bacterial lipopolysaccharide (LPS), a TLR4 ligand, exhibited a shift from homogenous SSC^low properties to the majority of cells being SSC^high (Fig. 1a). Imaging of cells after sorting confirmed phenotypic separation, with SSC^high cells (referred to as responders) containing clusters of MyD88–AGF (that is, myddosomes). SSC^low bystander cells, representing a minority population, did not contain myddosomes (Fig. 1a). We performed RNA-sequencing (RNA-seq) analysis to compare responder and bystander cells after sorting and additionally analysed bulk (non-sorted) populations of untreated and LPS-treated cells (Fig. 1a–c, Supplementary Fig. 1 and Supplementary Data 2). In total, 2,380 genes were differentially expressed across all samples (DEGs; adjusted P (P_adj) < 0.01, |log₂[fold change (FC)]| > 1; Supplementary Fig. 1a). Principal component analysis (PCA), clustering and individual comparisons revealed that SSC^high responders resembled the bulk LPS-treated cells, whereas SSC^low bystanders displayed a distinct transcriptome (Fig. 1b and Supplementary Fig. 1). Relative to bystanders and untreated cells, responder cells selectively induced primary response genes (PRGs) and secondary response genes (SRGs)^25,26,27 (Fig. 1c). Although bystander cells were defective for the induction of TLR-response genes, these cells were not dormant. Like the responders, bystanders exhibited an interferon (IFN)-stimulated gene (ISG) signature, suggesting that these cells have responded to cytokines and other factors (such as IFNs) produced by responder cells (Fig. 1c). Consistent with this idea, flow cytometry analysis of LPS-treated cells (co-treated with brefeldin A to prevent cytokine secretion) demonstrated that responder cells, but not bystanders, produced IL-6 and TNF (Fig. 1d). Staining for the plasma-membrane-localized ISG product tetherin confirmed that brefeldin A treatment prevented IFN activities in all cells (Fig. 1d). In the absence of brefeldin A, both populations equally expressed tetherin (Fig. 1d). Thus, myddosome-containing cells represent the primary source of the LPS-induced transcriptional program.

To identify regulators of myddosome assembly, we used this sorting-based approach to conduct a genome-wide CRISPR screen. To increase the fidelity of SSC^high cells to report on TLR signalling, we introduced a fluorescent NF-κB reporter into MyD88–AGF iBMDMs and stably expressed Cas9 (Extended Data Fig. 1a,b). The NF-κB reporter was validated by electroporation of single-guide RNA (sgRNAs) targeting Myd88, Tlr2 or Tlr4, followed by TLR stimulation³³ (Extended Data Fig. 1c). Cells from all genotypes maintained the NF-κB reporter response to TNF, but selectively lost responsiveness to LPS or the TLR2 ligand bacterial lipoteichoic acid (LTA) in a receptor-specific manner (Extended Data Fig. 1c). We introduced the Julianna sgRNA library³⁴ into these cells, stimulated with LPS or LTA, and sorted SSC^lowNF-κB^low cells to enrich mutants defective in myddosome assembly and NF-κB activation (Extended Data Fig. 1d). Sequencing analysis of sgRNAs showed minimal guide dropout and high complexity were retained (Extended Data Fig. 1e) and essential genes were depleted in untreated controls versus the plasmid input (Extended Data Fig. 1f), together confirming the screen and library quality (Extended Data Fig. 1e,f). Robust rank aggregation analysis (using MAGeCK)³⁵ identified positive regulators of myddosome assembly and NF-κB signalling (Fig. 1e), which we confirmed at the sgRNA and gene level (Extended Data Fig. 1g and Supplementary Data 3). Genes encoding canonical TLR signalling components, chaperones and membrane sorting factors were enriched in the SSC^lowNF-κB^low cell populations³⁶ (Fig. 1e). Rela was the top hit, validating the NF-κB reporter (Fig. 1e). Notably, Ticam2 (encoding TRAM) ranked among the top hits for LPS and LTA treatment. This finding was unexpected, as TRAM is a regulator of the MyD88-independent TLR signalling pathways. TRAM links TLR4 within endosomes to TRIF to mediate IFN expression after TLR4 endocytosis^20,37,38,39. TRAM also mediates the weak IFN expression induced by TLR2 (refs. ^40,41,42). The gene encoding TRIF (Ticam1) was not enriched in non-responder cells treated with LPS or LTA (Fig. 1e). On the basis of these findings, we re-examined the RNA-seq data described above (Fig. 1c). Among the 248 genes of which the expression was lower in bystanders than responders and unchanged by LPS treatment, Ticam2 was the only gene overlapping with the hits from the CRISPR screen (Fig. 1c,e). Reduced Ticam2 expression in bystander cells (Supplementary Fig. 1c) was confirmed at the protein level (Fig. 1f). These datasets, along with reports of TRAM–MyD88 interactions in HEK293 cells^41,43, prompted deeper investigations.

Using CRISPR–Cas9 knockout (KO) in MyD88–AGF iBMDMs, most of the signalling factors examined did not impair myddosome formation and were probably identified as screen hits due to defective NF-κB activation (Extended Data Fig. 2a). By contrast, cells lacking TLRs or the myddosome assembly regulator TIRAP^44,45,46,47 were unable to assemble myddosomes (Fig. 1g). In the case of TRAM, a unique phenotype was observed. TRAM-deficient cells lacked cytosolic myddosomes but retained MyD88 complexes at the plasma membrane (Fig. 1g). We quantified this phenotype using total internal reflection (TIRF) microscopy after PAMP stimulation (Fig. 1h and Extended Data Fig. 2b). Proto-myddosomes formed rapidly in all cells except those lacking TIRAP (Fig. 1h). In wild-type (WT) and ∆Ticam1 cells, proto-myddosomes mostly disappeared from the plasma membrane within 1 h (Fig. 1h). In ∆Ticam2 cells, MyD88 complexes persisted at the membrane (Fig. 1h). TIRF resolution revealed that the retained proto-myddosomes in ∆Ticam2 cells did not change in size over 2 h (Fig. 1h). Live-cell imaging (Supplementary Video 1) and quantification of MyD88 cluster formation confirmed similar response frequencies across genotypes and that ∆Ticam2 cells formed MyD88 clusters that failed to depart from the membrane (Fig. 1i). These findings suggest that TRAM controls myddosome assembly.

To dissect MyD88 activities biochemically, we performed co-immunoprecipitations of MyD88 from MyD88–AGF iBMDMs at 5 min intervals during the first hour after PAMP stimulation (Extended Data Fig. 3a). We found that MyD88 interacted rapidly with TIRAP and the seeding receptor (TLR4 after LPS stimulation). These interactions diminished over time (Extended Data Fig. 3a). After dissociation from TIRAP, MyD88 engaged core myddosome components (IRAK2 and IRAK4 (IRAK2/4), TRAF6), coinciding with signal transduction (phosphorylated p38α (p-p38α); Extended Data Fig. 3a). This kinetic analysis therefore provided biochemical evidence for the concept of myddosome maturation: initial TLR–TIRAP binding (that is, proto-myddosome formation), followed by receptor release and recruitment of signalling enzymes (that is, myddosome maturation; Fig. 2a).

We observed MyD88–TRAM complexes after PAMP stimulation, which occurred during the proto-myddosome to myddosome transition (Fig. 2a and Extended Data Fig. 3a). MyD88–TIRAP–TLR4 complexes were assembled and disassembled at comparable rates in WT and ∆Ticam1 cells (Fig. 2a). By contrast, ∆Ticam2 cells maintained MyD88–TIRAP–TLR4 complexes over time, which ultimately recruited core myddosome components (Fig. 2a and Extended Data Fig. 3a). ∆Irak4, ∆Irak2 and ∆Traf6 cells and WT cells treated with zimlovisertib (IRAK4 kinase inhibitor) or Dynasore (TLR4 endocytosis inhibitor) did not alter MyD88 interactions with the receptor, TIRAP or TRAM (Extended Data Fig. 3b). TLR2 and TLR4 cell surface levels and LPS-induced TLR4 endocytosis were unaffected in cells of all genotypes (Extended Data Fig. 3c).

Extending co-immunoprecipitation times after PAMP stimulation revealed that the unusual interactions between MyD88–TIRAP–TLR4 and downstream signalling enzymes (occurring uniquely in ∆Ticam2 cells) persisted for up to 6 h (Fig. 2b). By contrast, WT cells transitioned within minutes from the MyD88–TIRAP–TLR4 complex (proto-myddosome) to the MyD88–IRAK–TRAF6 complex (myddosome). The MyD88 complexes in ∆Ticam2 cells were unstable and subsequently disappeared. By contrast, MyD88 interactions with IRAK2/4 and TRAF6 persisted for over 12 h in WT cells, with TIRAP and TLR4 interactions being evident for less than 1 h (Fig. 2b).

By 30–60 min after PAMP stimulation, WT myddosomes matured by incorporating IKKs, TBK1 and the terminal effectors NF-κB p65 and p38α (Extended Data Fig. 3d). Although MyD88 also formed complexes with these factors in ∆Ticam2 cells, interactions were lost prematurely and MyD88 complexes were targeted by the autophagy regulator p62 earlier (Extended Data Fig. 3d). These data indicate that TRAM associates with MyD88 with kinetics that differ from other proteins, and that TRAM directs myddosome maturation by releasing proto-myddosomes from receptors.

We assessed the role of TRAM in TLR signalling by immunoblotting for NF-κB (p-IKKα/β, p-p65, IκBα degradation), MAPKs (p-ERK1/2, p-JNK, p-p38α) and TBK1 (p-TBK1) activities for 12 h after PAMP stimulation (Supplementary Fig. 2). Densitometric analysis and normalization to untreated cells enabled calculation of activity scores and total pathway output (Extended Data Fig. 4a). Compared with WT cells, ∆Myd88 and ∆Tirap cells were defective in all of the examined TLR-induced pathways. ∆Ticam1 cells responded normally to LTA but displayed reduced NF-κB activity in response to LPS, while MAPK and TBK1 activation remained unaffected (Extended Data Fig. 4a). Notably, ∆Ticam2 cells initiated all TLR-induced activities, but these activities terminated prematurely (Extended Data Fig. 4a). Early signal termination in ∆Ticam2 cells was most evident for NF-κB and TBK1, which remained active for 12 h in WT cells. We further analysed NF-κB dynamics using live-cell imaging of p65–mScarlet3 reporters in the presence of soluble TNFRII to block autocrine TNF effects (Fig. 2c and Extended Data Fig. 4b). The nuclear:cytoplasmic p65 ratios demonstrated sustained NF-κB activation in LPS-treated WT cells, as reported previously^21,48,49, whereas LTA stimulated oscillatory activity (Fig. 2c and Extended Data Fig. 4b). ∆Myd88 and ∆Tirap cells were largely defective for NF-κB activation (Fig. 2c and Extended Data Fig. 4b). Consistent with previous work²³, LPS-treated ∆Ticam1 cells exhibited an oscillatory pattern that resembled LTA stimulation. ∆Ticam2 cells displayed a unique phenotype. These cells had a near-normal first NF-κB activity peak but failed to sustain signalling, indicating premature signal attenuation (Fig. 2c and Extended Data Fig. 4b). Quantitative PCR with reverse transcription (RT–qPCR) analysis of pre-mRNAs and mRNAs encoding the PRG Tnf and the SRG Il6 showed comparable expression patterns in WT and ∆Ticam1 cells and no induction in ∆Myd88 and ∆Tirap cells (Fig. 2d). The oscillatory Tnf pre-mRNA bursts detected in WT cells were largely absent after the initial peak in ∆Ticam2 cells. Il6 pre-mRNA was similarly reduced, leading to lower mature Il6 mRNA levels (Fig. 2d) and diminished cytokine secretion (Fig. 2e). This effect was largely TRIF independent (Fig. 2e). These findings were validated in primary BMDMs from WT and Ticam2^−/− mice (Fig. 3a–d and Supplementary Fig. 3a), in which we additionally assessed secretion of the SRG product IL-12p40 (Fig. 3d).

We performed RNA-seq analysis of LTA- or LPS-stimulated WT and Ticam2^−/− BMDMs after 2 h and 6 h of PAMP treatment and analysed PRGs (at 2 h), SRGs and ISGs (at 6 h) (Fig. 3e,f). We identified 4,963 DEGs (P_adj < 0.05, |log2[FC]| > 1; Supplementary Fig. 3b). PCA revealed that WT and Ticam2^−/− macrophages similarly diverged from the baseline at 2 h after PAMP treatment. However, in contrast to WT cells, Ticam2^−/− cells returned to near the baseline by 6 h (Fig. 3e). Gene subset analysis confirmed that PRG activation was comparable across genotypes and that ISGs were induced by LPS only in WT cells. SRGs were not activated in Ticam2^−/− macrophages irrespective of stimulus (Fig. 3f). The inability of TLRs to release proto-myddosomes in ∆Ticam2 cells therefore results in diminished myddosome stability and inflammatory gene expression, particularly involving SRGs.

We analysed cytokine secretion in WT and TRAM-deficient macrophages stimulated with a panel of PAMPs (Supplementary Fig. 4). In immortalized and primary BMDMs, TRAM was required for the secretion of TNF, IL-6 and IL-12p40 induced by several TLR2 ligands (LTA, P2C, P3C) and LPS. By contrast, TRAM-deficient cells were not defective for IFNβ or IP-10 secretion induced by TLR3 (p(I:C)), STING (DMXAA) or RIG-I (Sendai virus (SeV); Supplementary Fig. 4). As expected^20,37,38,39, TRAM-deficient cells were defective for LPS-induced IFN responses (Supplementary Fig. 4). To assess the role of TRAM in other MyD88-dependent pathways, we stimulated WT and TRAM-deficient macrophages with the TLR7 and TLR8 agonist R848, TLR9 agonist CpG or the IL-1-family cytokine IL-33 (Supplementary Fig. 4). Overall, the severity of the cytokine production defects associated with TRAM deficiency was highest for plasma-membrane-localized TLRs, with IL-6 and IL-12p40 being the most sensitive to TRAM deficiency across all TLRs (Supplementary Fig. 4). TRAM was not required for IL-33 activities.

To extend these findings to infectious contexts, we examined macrophage responses to bacteria by infecting BMDMs with Escherichia coli and Staphylococcus aureus (Fig. 3g,h). Primary Ticam2^−/− macrophages displayed increased survival upon infection, as reported previously⁵⁰. IL-6 production was reduced in Ticam2^−/− BMDMs after E. coli or S. aureus infection, whereas TNF production was minimally affected (Fig. 3g). Moreover, after E. coli infection, WT macrophages exhibited cytosolic myddosomes, whereas MyD88 in TRAM-deficient cells was predominantly associated with the cell membrane (Fig. 3h). Escherichia coli induced an IFNβ response that was reduced in Ticam2^−/− cells, consistent with TRAM’s role in TRIF signalling^20,37,38,39 (Fig. 3g). By contrast, IFNβ production induced by S. aureus was largely preserved in the absence of TRAM (Fig. 3g), consistent with a role for STING⁵¹.

In iBMDMs, a panel of five bacterial species (E. coli, Salmonella enterica serovar Typhimurium (S. Typhimurium), Listeria monocytogenes, S. aureus and Lactococcus lactis) demonstrated that TRAM was required for IL-6 but not TNF production across all infections (Extended Data Fig. 5). IFNβ responses were pathogen-dependent and did not show a uniform requirement for TRAM (Extended Data Fig. 5). Importantly, LTA or LPS elicited comparable responses to those observed across a range of bacterial multiplicities of infection (MOIs). The PAMP concentrations used in this study are therefore within a range that applies to infectious scenarios (Extended Data Fig. 5).

To elucidate how TRAM releases proto-myddosomes from the TLR–TIRAP complex, we identified elements in TRAM that mediate these activities. We reconstituted ∆Ticam2 MyD88–AGF iBMDMs with TRAM WT or the following mutants: G2A (myristoylation-defective and cytosolic)³⁹, 3X (phosphoinositide-binding defective and restricted to endosomes)³⁸, E180A (TRAF6-binding defective)⁵², C114H (TIR mutant)⁵³ and D88A/E89A⁵⁴ (membrane-binding defective; Extended Data Fig. 6a). Each mutant displayed their expected localization pattern (Extended Data Fig. 6b). After PAMP stimulation, only TRAM WT and E180A supported proto-myddosome release into the cytosol (Fig. 4a,b). Co-immunoprecipitation showed that the release-defective mutants neither disrupted the MyD88–TIRAP complex nor formed a complex with MyD88 (Fig. 4c and Supplementary Fig. 5). Accordingly, only TRAM WT and E180A fully supported TLR-induced TNF and IL-6 secretion (Fig. 4d). We compared these phenotypes to the function of TRAM in TRIF-mediated IFN induction from endosomes. Only TRAM WT, 3X, E180A and, partially, D88A/E89A restored IFNβ production after LPS stimulation (Fig. 4d). LTA did not induce IFNβ production (Fig. 4d). MyD88–TRAM interactions probably occurred at the plasma membrane, as TRAM was not enriched in cytosolic myddosomes (Fig. 2b) and showed no changes in its localization during the signalling process (Extended Data Fig. 6c). In contrast to these plasma-membrane-based myddosome activities, the endosomal TRAM 3X mutant³⁸ maximally promoted IFNβ secretion but not myddosome activities or cytokine secretion (Fig. 4a–d). These data suggest that TRAM functions from the plasma membrane to mediate myddosome maturation and from endosomes to activate TRIF.

Plasma membrane localization of TRAM requires its N-terminal myristoylation motif³⁹. To examine alternative localization mechanisms, we replaced the N terminus with a dual acylation motif from Fyn (Fyn–TRAM), Lyn (Lyn–TRAM) or the TIRAP plasma-membrane-binding motif (TIRAP^PBM–TRAM; Extended Data Fig. 6d). While Fyn–TRAM localized predominantly to the plasma membrane, Lyn–TRAM targeted endo-membranes (Extended Data Fig. 6e). Only TIRAP^PBM–TRAM restored myddosome release after LTA or LPS stimulation (Fig. 4e and Extended Data Fig. 6f). Consistent with the primary requirement of TRAM for SRG expression, we found that all chimeras restored TNF secretion to TRAM-deficient cells, but only TIRAP^PBM–TRAM promoted secretion of IL-6 (Fig. 4f). As these chimeras restrict TRAM to the plasma membrane, no production of IFNβ was detected after LPS stimulation (Fig. 4f). TRAM reconstitutions of ∆Tirap/Ticam2 double-KO macrophages failed to restore cytokine secretion after PAMP stimulation, but restored IFNβ secretion after LPS stimulation, providing further evidence that TRAM–TRIF signalling operates independently of the MyD88 pathway (Extended Data Fig. 6g).

The finding that TIRAP^PBM–TRAM was the only chimera to maximally support TLR signalling suggested that pools of TRAM and TIRAP within the same membrane subdomain control myddosome maturation. The TIRAP^PBM mediates interactions principally with phosphatidylinositol 4,5-bisphosphate (PI(4,5)P₂)^46,47. We used 3D microscopy to quantify the distribution of TIRAP and TRAM variants at the plasma membrane and within PI(4,5)P₂-rich microdomains (using the PI(4,5)P₂-specific pleckstrin homology (PH) domain from PLCδ (PH-PLCδ) as a marker; Fig. 4g). Intensity analysis showed modest enrichment of TRAM WT in PI(4,5)P₂-rich domains, which we could not observe for other TRAM variants, whereas TIRAP^PBM–TRAM displayed strong enrichment similar to TIRAP and PH-PLCδ (Fig. 4g). We conclude that a pool of TRAM within PI(4,5)P₂-rich plasma membrane microdomains is responsible for promoting proto-myddosome release and maturation into myddosomes.

As TIRAP and TRAM localized to PI(4,5)P₂-rich microdomains, we hypothesized that their ratio may affect myddosome activities. To test this idea, we implemented the CRISPR synergistic activation mediator⁵⁵ system, a three-component gene activation (CRISPRa) tool to boost transcription of endogenous genes coding for TIRAP or TRAM in macrophages (Extended Data Fig. 7a). Testing six sgRNAs per gene yielded 2-fold to around 60-fold mRNA induction levels, which was mirrored at the protein level (Extended Data Fig. 7b). After PAMP stimulation, increased TIRAP expression phenocopied ∆Ticam2 cells, retaining MyD88 at the plasma membrane (Extended Data Fig. 7c,d). Enhancing TRAM expression had no effect on MyD88 localization (Extended Data Fig. 7c,d). Correspondingly, after PAMP stimulation, IL-6 secretion decreased with rising TIRAP levels, whereas TNF secretion was unaffected (Extended Data Fig. 7e). Simultaneous activation of both genes using sgRNAs in pairwise combinations showed that a high TIRAP:TRAM ratio abolished IL-6 secretion (Extended Data Fig. 7f), while TNF remained unchanged. Crucially, co-upregulation of TRAM alongside TIRAP restored IL-6 secretion to WT levels, even when both exceeded 50-fold expression (Extended Data Fig. 7g). Thus, the TIRAP:TRAM ratio, rather than protein abundance, is the determinant of function.

To further skew the TIRAP:TRAM ratio, we ectopically overexpressed TIRAP. As the NCBI (NP_001171316.1) and UniProt (Q99JYI) sequence annotations differ by eight N-terminal residues, we performed mass spectrometry (MS) analysis of TIRAP immunoprecipitated from iBMDMs (Supplementary Fig. 6 and Supplementary Data 4), confirming an acetylated alanine as the N terminus and no additional residues (Supplementary Fig. 6). We therefore expressed the shorter Q99JYI version as full-length (TIRAP(FL)) or a version lacking the TIR domain (TIRAP(∆TIR)), achieving over 670-fold overexpression (Extended Data Fig. 8a). While both versions localized to the plasma membrane (Extended Data Fig. 8b), only TIRAP(FL) sequestered proto-myddosomes at the membrane (Extended Data Fig. 8c,d). TIR-mediated interactions therefore govern proto-myddosome assembly and release. At this extreme overexpression level, low basal cytokine secretion occurred, precluding further analysis. These data indicate that, within PI(4,5)P₂-rich plasma membrane microdomains, a balanced TIRAP–TRAM ratio controls MyD88 activities.

We performed tandem kinetic immunoprecipitations to determine whether MyD88 interacts with TIRAP and TRAM separately or within a ternary complex. MyD88 first associated with TIRAP (10 min), then both proteins (20 min), then exclusively with TRAM (40 min; Extended Data Fig. 8e), engaging around 60–70% of the cellular TIRAP and TRAM pools. Subsequent immunoprecipitations for TRAM from the unbound fraction showed no TIRAP–TRAM interaction outside proto-myddosomes. Further fractionation of the elution revealed a ternary MyD88–TIRAP–TRAM complex at 20 min (Extended Data Fig. 8e), which we confirmed to localize to PI(4,5)P₂-rich microdomains by microscopy (Extended Data Fig. 8f). These data suggest that TRAM targets the initial TIRAP–MyD88 containing proto-myddosome complex.

To determine whether release of MyD88 from the plasma membrane is sufficient to sever TIRAP from MyD88 and permit myddosome maturation, we depleted PI(4,5)P₂ with ionomycin⁵⁶ (Extended Data Fig. 8g,h). LPS was used to arrest proto-myddosomes at the plasma membrane of ∆Ticam2 iBMDMs, followed by ionomycin treatment. Live-cell imaging confirmed PI(4,5)P₂ depletion and cytosolic relocalization of MyD88 (Supplementary Video 2 and Extended Data Fig. 8h). Immunoprecipitation after ionomycin treatment demonstrated that MyD88 rapidly lost interaction with TLR4. However, in contrast to the scenario in WT cells, in which MyD88 dissociation is mediated by TRAM, ionomycin did not disrupt interactions with TIRAP (Extended Data Fig. 8h). Moreover, ionomycin-released MyD88 complexes did not mature into larger assemblies (Supplementary Video 2 and Extended Data Fig. 8h). Thus, receptor-free proto-myddosomes cannot mature into myddosomes while TIRAP remains bound.

These results suggest that TIRAP seeds MyD88 assembly but also restricts further growth. To test this idea, we reconstituted this TIR-domain-dependent switch mechanism in vitro. We purified recombinant TIR domains of MyD88 (MyD88^TIR), TIRAP (TIRAP^TIR) and TRAM (TRAM^TIR) (Supplementary Fig. 7) and measured supramolecular assembly into turbidity-causing particles (TCPs; Extended Data Fig. 9). We found that MyD88^TIR self-assembled in a concentration-dependent manner, as reported previously³² (Extended Data Fig. 9a). TIRAP^TIR did likewise, albeit to a lesser extent, whereas TRAM^TIR did not (Extended Data Fig. 9a). Mixing TIRAP^TIR and MyD88^TIR reduced turbidity by over 80%, even at sub-stoichiometric ratios (10% molar concentration of TIRAP^TIR), suggesting inhibition of MyD88^TIR self-assembly (Extended Data Fig. 9b). TRAM^TIR alone had no effect on MyD88^TIR (Extended Data Fig. 9b). Cross-linking and SDS–PAGE confirmed that TIRAP^TIR prevented MyD88^TIR TCP formation whereas TRAM^TIR did not (Extended Data Fig. 9c). We repeated these assays with a TIRAP(P125H) mutant that is non-functional in cells⁴⁵. Contrary to the WT TIR domain, TIRAP(P125H) was unable to inhibit MyD88^TIR oligomerization (Extended Data Fig. 9a,b), providing symmetry between the turbidity assays and activities in cells.

To determine whether TRAM can release MyD88 self-assembly from TIRAP inhibition, we combined all three TIR domains. Under these conditions, the TRAM^TIR reversed TIRAP^TIR-mediated inhibition (Extended Data Fig. 9d). The TIR domains are therefore sufficient to recapitulate the activities of these proteins in cells, with TRAM functioning to release MyD88 self-assembly from TIRAP restriction (Extended Data Fig. 9e).

Our analysis of ∆Ticam2 cells revealed that TRAM is required for (1) releasing proto-myddosomes from the receptor complex, and for (2) myddosome assembly and durable signalling. To determine the importance of the latter activity (durable signalling), a method was necessary to disrupt myddosomes after they have formed. Such a method was offered by methylpiperidinopyrazole (MPP) and its derivative TSI-13-57, inhibitors of MyD88^TIR self-assembly⁵⁷. We confirmed that pretreatment of BMDMs with MPP or TSI-13-57 was non-toxic and reduced cytokine secretion after PAMP stimulation, without affecting IFN-dependent IP-10 secretion induced by LPS (Extended Data Fig. 10a,b). LTA-induced IP-10 production was reduced by MPP or TSI-13-57, suggesting that TLR2 induces IP-10 via MyD88 (Extended Data Fig. 10a).

We determined whether MPP could disassemble pre-formed myddosomes, phenocopying the premature disassembly in ∆Ticam2 cells. We stimulated MyD88–AGF iBMDMs with LPS for 1 h and followed myddosomes by live imaging during MPP treatment at this time (Supplementary Video 3). MPP reduced the fraction of cells containing myddosomes and the number of myddosomes per cell (myddosome half-life, t_1/2, of around 30 min), with smaller complexes disappearing first and larger myddosomes persisting longer, leading to an increase of the average size (Extended Data Fig. 10c). Fluorescence recovery after photobleaching (FRAP) assays demonstrated rapid MyD88 dynamics during signal transduction, but 5 min of MPP treatment abolished the recovery of bleached myddosomes (Extended Data Fig. 10d). MPP treatment of LPS-stimulated cells disrupted MyD88 interactions with IRAK2/4, TRAF6 and signalling enzymes within 1 h (Extended Data Fig. 10e), whereas DMSO-treated WT cells maintained MyD88 interactions with signalling enzymes for over 12 h (Extended Data Fig. 10e,f). Loss of myddosomes by MPP did not appear to be autophagy mediated, as MyD88 was not targeted by p62 after MPP treatment (Extended Data Fig. 10e).

Immunoblotting (Supplementary Fig. 8) and densitometry analysis revealed that adding MPP 1 h after PAMP stimulation caused premature termination of NF-κB, MAPK and TBK1 activity (Extended Data Fig. 10g), as well as Tnf and Il6 transcription (Extended Data Fig. 10h) and cytokine secretion (Extended Data Fig. 10i), resembling the phenotype of TRAM-deficient cells. Additional analysis showed TNF secretion was affected only when MPP was added either before or within 2 h of TLR stimulation, whereas MPP suppressed IL-6 secretion even when added 4 h after TLR stimulation (Extended Data Fig. 10j), consistent with delayed SRG activation. Long-lived myddosomes, which are naturally induced by TRAM, are therefore an essential aspect of the TLR pathway.

To determine whether long-lived myddosomes are required for inflammatory activities in vivo, we focused on the MPP derivative TSI-13-57. TSI-13-57 is a more specific MyD88 inhibitor than the parent chemical, is non-toxic and enters the bloodstream after intraperitoneal (i.p.) administration⁵⁷. We performed i.p. injections of LPS or LTA into mice, followed by TSI-13-57 (Fig. 5a). MyD88 immunoprecipitations from splenocytes isolated from mice revealed PAMP-induced MyD88 interactions with IRAKs, TRAF6, IKKα/β, TBK1 and NF-κB p65 (Fig. 5b), therefore establishing that myddosomes form in vivo.

Injection of mice with TSI-13-57 that were pre-injected with PAMPs disrupted myddosomes without affecting MyD88 abundance (Fig. 5b). In PAMP-treated mice, IL-6 and IL-12p40 serum levels were sustained or increased over 4.5 h. TSI-13-57 addition 30 min after PAMP treatment abolished IL-6 and IL-12p40 in the serum at 4.5 h (Fig. 5c). In contrast to these cytokines, LPS-induced ISG expression in splenocytes was unaffected by TSI-13-57, highlighting its specificity for the MyD88 pathway (Fig. 5d). RT–qPCR analysis of splenocyte RNA confirmed reduced Il6 and Il12b transcripts but no changes in the expression of the TRIF-dependent genes Ifnb, Stat1 or Cxcl10 (Fig. 5e). These results demonstrate the importance of long-lasting myddosomes for sustained cytokine production invivo.

Material availability

All unique and newly generated materials (cell lines, plasmids) are available from the corresponding authors on request.

Cell culture, primary BMDM generation and treatments

iBMDM lines and HEK293T cells (CRL-11268, ATCC) were cultured in DMEM supplemented with glutamine, sodium pyruvate, penicillin–streptomycin and 10% FBS (cDMEM), at 37 °C under 5% CO₂. An overview of all cell lines is provided in Supplementary Table 1.

BMDMs were differentiated from bone marrow from tibiae and femurs of C57BL/6J, WT or Ticam2^−/− mice (aged 10 weeks, female, provided by L. Li). In brief, bones were surgically removed and cleaned of surrounding tissue, washed once in 70% ethanol, once in PBS, 2.5 mM EDTA and penicillin–streptomycin and stored in PBS, 2.5 mM EDTA and penicillin–streptomycin. The bones were cut open and placed into a sterile 0.6 ml tube with holes at the base. The 0.6 ml tube with bones was placed into a sterile 2 ml tube and the nested tubes were centrifuged at 10,000g for 1 min. Obtained bone marrow pellets were resuspended in PBS + 2.5 mM EDTA and centrifuged at 400g for 3 min followed by RBC lysis using ACK buffer for 1 min at room temperature. Lysis was stopped by adding 10 vol of PBS + 2.5 mM EDTA and cells were pelleted at 400g for 3 min. A total of 4 × 10⁶ cells was seeded per 10 cm suspension culture dish and differentiated in 10 ml cDMEM supplemented with 30% L929 supernatants containing M-CSF for 6 days. Cells were fed with 5 ml additional differentiation medium after 3 days. BMDMs were detached using PBS + 2.5 mM EDTA by incubating for 10 min at room temperature and reseeded for experiments at the required density and numbers in cDMEM supplemented with 10% L929 supernatants. Experiments were performed the next day.

Cells were treated with 10 ng ml⁻¹ LPS O55:B5 (ALX-581-013, Enzo) or 100 ng ml⁻¹ LTA-SA (tlrl-pslta, Invivogen) unless otherwise indicated. Other treatments were performed with 100 ng ml⁻¹ Pam2CSK4 (P2C; tlrl-pm2s, Invivogen), 100 ng ml⁻¹ Pam3CSK4 (P3C; vac-pms, Invivogen) 100 μg ml⁻¹ poly(I:C) (tlrl-pic, Invivogen), 1 µg ml⁻¹ Fla-ST (tlrl-epstfla, Invivogen), 100 ng ml⁻¹ R848 (vac-r848, Invivogen), 5 µM CpG-DNA (ODN1826; tlrl-1826, Invivogen), 50 ng ml⁻¹ mIL-33 (3626-ML, R&D Systems), 50 μg ml⁻¹ DMXAA (tlrl-dmx, Invivogen) or 20 ng ml⁻¹ mIFNγ (315-05, Peprotech). RIG-I–MAVS pathway activation used Sendai virus (SeV, Cantell strain) supplied as allantoic fluid (10100774, Charles River Laboratory). Cells were treated with 20 µM Zimlovisertib/PF-06650833 (HY-19836, MedChemExpress) to inhibit IRAK4 kinase activity. Cells were stimulated with 50 ng ml⁻¹ recombinant mouse TNF (315-01A, Thermo Fisher Scientific) to activate NF-κB activity or kept in the presence of 200 ng ml⁻¹ mouse sTNFRII (426-R2-050, R&D Systems) to block autocrine TNF signalling. To deplete PI(4,5)P₂, cells were treated with 10 µM ionomycin (I24222, Invitrogen). The MyD88 inhibitors MPP (13863, Cayman Chemical) and TSI-13-57 (BD01861091, BLD Pharm) were used at 5 µM for in vitro experiments. Secretion of cytokines was blocked by treatment with 5 µg ml⁻¹ brefeldin A (420601, BioLegend) and TLR endocytosis was blocked by treatment with 10 µM hydroxy Dynasore (Dyngo-4a; HY-13863, MedChemExpress).

Bacteria and virus infections

All bacterial strains used in this study are summarized in Supplementary Table 1. Staphylococcus aureus (strain SA113), S. Typhimurium (strain SL1344) and E. coli K12 (strain MG1655) were cultured on Luria Broth (LB; 12-780-052, Invitrogen) agar (1.5% w/v; 214010, BD) at 37 °C. Listeria monocytogenes (strain 10403S) was cultured on brain–heart infusion (BHI; 237500, BD) agar at 37 °C. Lactococcus lactis (strain NCTC 6681) was cultured on BHI agar at 37 °C under 5% CO₂.

To prepare bacteria for infection, S. aureus, S. Typhimurium and E. coli colonies were inoculated into liquid LB and grown overnight at 37 °C. Listeria monocytogenes colonies were inoculated into liquid BHI and grown overnight at 30 °C without shaking. Lactococcus lactis colonies were inoculated into liquid BHI supplemented with 0.042% (w/v) NaHCO₃ and grown overnight at 37 °C. Stationary phase S. aureus and E. coli in liquid cultures were then subjected to two back dilutions in liquid LB to enrich for highly viable, logarithmic phase cultures for infections. Salmonella enterica serovar Typhimurium was back diluted twice into LB supplemented with 300 mM NaCl and grown in a closed container. Lactococcus lactis was back diluted once into BHI supplemented with 0.042% (w/v) NaHCO₃ and L. monocytogenes overnight cultures were used right away.

At the time of infection, bacteria were centrifuged at 10,000g for 3 min, supernatants discarded, pellets washed with PBS twice and resuspended in cDMEM without antibiotics. Bacterial suspensions were serially diluted and plated to confirm the MOI. To initiate synchronized infections, the target cells (0.5 × 10⁶ cells per well of a 24-well plate) were placed onto ice for 5 min, the culture supernatants removed, exposed to 1 ml bacteria at the indicated MOI and centrifuged at 750g for 10 min. Infected cells were incubated for 1 h at 37 °C under 5% CO₂ to allow for bacterial uptake, after which the cells were washed with warm PBS and the medium was replaced with cDMEM supplemented with 100 μg ml⁻¹ gentamicin (15750060, Gibco) to remove and eliminate remaining extracellular bacteria. Infections were allowed to proceed at 37 °C with 5% CO₂ for the remaining time.

For SeV (Cantell strain) infection, SeV was diluted in DMEM supplemented with 1% FBS for an MOI of 0.00001. Cells were exposed to SeV for 1 h at 37 °C with 5% CO₂. After 1 h, cells were washed twice with PBS, the medium was replaced with 1 ml cDMEM and infection incubated at 37 °C with 5% CO₂ for the remaining time.

Lentiviral transductions

Plasmid transfections were performed by using a 10× mix in Opti-MEM I reduced serum medium (31985062, Thermo Fisher Scientific) containing DNA plasmids and Lipofectamine 2000 transfection reagent (11668019, Thermo Fisher Scientific) at a 1:2 ratio according to the manufacturer’s instructions. For lentiviral packaging, HEK293T cells were transfected with equimolar ratios of the transfer plasmid, pMD2.G (Addgene, 12259) and psPAX2 (Addgene, 12260, both gifts from D. Trono). The medium was replaced after 16 h with cDMEM containing 5 mM sodium butyrate (B5887, Sigma-Aldrich) and the cells were left to produce lentiviral particles for over 24 h. The virus-containing supernatant was filtered through a 0.43-μm syringe filter and supplemented with 5 μM polybrene (TR-1003, Sigma-Aldrich). The target cells were resuspended in virus-containing medium and spinfected for 1 h at 1,250g and 30 °C. After centrifugation, 1 ml cDMEM was added and the cells left to rest.

Myddosome flow cytometry sorting and RNA-seq analysis

Historically, most flow cytometers use a 488 nm (blue) laser as standard configuration for SSC detection because the 488 nm wavelength provides robust sensitivity to intracellular refractive index heterogeneity and μm-scale structures. We therefore assessed the formation of cytosolic myddosomes by flow cytometry using SSC. After stimulation, endogenously GFP-tagged MyD88 in MyD88–AGF iBMDMs forms abundant μm-scale cytoplasmic complexes that increase intracellular heterogeneity and granularity. The accumulation of these protein-dense assemblies results in measurable changes in SSC.

SSC shift was used to separate myddosome-containing cells from non-responding bystander cells. To do so, PAMP-stimulated cells were collected, washed in PBS, resuspended in PBS + 1% FCS to a concentration of 4 × 10⁶ cells per ml and passed through a 35 µm nylon mesh strainer. Subsequently, cells were sorted using a BD FACSMelody Cell Sorter using BD FACSChorus software (BD Biosciences) by gating on singlets and SSC^high versus SSC^low cells.

To assess separation of subpopulations, we sorted cells as described to obtain 1 × 10⁶ cells for each bin and collected the cells into 5 ml flow cytometry tubes containing 0.5 ml cDMEM, pelleted the cells by centrifugation at 400g, washed them once with cDMEM and subsequently plated onto six-well glass-bottom imaging plates (P06-1.5H-N, Cellvis). Cells were centrifuged at 400g for 5 min to speed up adhering and immediately imaged to test for presence/absence of cytosolic myddosomes. Alternatively, collected cells were lysed and analysed by immunoblotting.

To analyse myddosome-induced transcriptional changes by RNA-seq, eight 10-cm tissue-culture-treated dishes were seeded each with 2.0 × 10⁷ MyD88–AGF iBMDMs and the cells stimulated with 10 ng ml⁻¹ LPS. After 6 h, cells were washed with PBS and prepared for sorting as described above. All of the subsequent steps were carried out at 4 °C. First, LPS-stimulated cells were sorted into responder and bystander populations based on SSC^high and SSC^low bins. Subsequently, LPS-treated cells and untreated controls were passed through the flow cytometer without gating to obtain pooled samples for comparison. Then, 1 × 10⁶ cells for each sample/replicate/bin were collected into 5 ml flow cytometry tubes containing 0.5 ml TRIzol LS (10296028, Thermo Fisher Scientific). Total RNA was extracted according to the manufacturer’s instructions. RNA pellets were resuspended in 100 μl 1× TurboDNase buffer and potential gDNA contaminants were removed by adding 5 μl TurboDNase (AM2238, Thermo Fisher Scientific) and incubating at 37 °C for 45 min. RNA was repurified by phenol–chloroform extraction and ethanol precipitation. In brief, RNA solutions were transferred to 1.5 ml low bind tubes, and 100 μl double-distilled H₂O and 200 μl (1 vol) of phenol:chloroform:isoamyl alcohol (25:24:1, v/v; 15593031, Thermo Fisher Scientific) were added. The samples were thoroughly mixed and centrifuged at 12,000g for 5 min. The aqueous phase (about 250 μl) was transferred to a new tube, and 1 vol of chloroform was added. The samples were thoroughly mixed and centrifuged again, and chloroform extraction repeated two more times. Finally, the RNA was precipitated by adding 1 ml ice-cold 100% ethanol and 1 μl glycol blue (5 μg; AM9515, Thermo Fisher Scientific) and centrifugating at 21,000g for 1 h at 4 °C. The obtained RNA pellets were washed twice with 80% ethanol, dried for 10 min at room temperature and resuspended in 100 μl double-distilled H₂O. The RNA quantity and quality were assessed by spectrophotometry (NanoDrop One, Thermo Fisher Scientific) and confirmed by fragment analysis showing RNA-integrity numbers of >7.1 for all of the samples (performed by Genewiz). RNA-seq and initial data processing was performed by Genewiz and further data analysis was performed using the Galaxy Server⁶⁰. Adapter trimming was conducted using Cutadapt (Galaxy v.3.7), applying a minimum read length of 20 bp and Phred quality score of >20. Reads were filtered using fastp (Galaxy v.1.1.0) with poly(G) trimming. Trimmed paired-end reads were aligned to the mouse reference genome (GRCm38, mm10) using HISAT2 (Galaxy v.2.2.2). Alignment statistics were evaluated with MultiQC, showing mapping rates exceeding 90%. Gene-level expression quantification was performed using featureCounts (Galaxy v.2.1.1). Additional alignment quality control was performed using RSeQC, including gene body coverage analysis (Galaxy v.5.0.3), and read distribution across genomic features (Galaxy v.5.0.3). Duplicate reads were evaluated using Picard MarkDuplicates (Galaxy v.3.1.1.0), and chromosomal mapping statistics were generated using SAMtools idxstats (Galaxy v.2.0.8), together demonstrating low duplication levels, even gene body coverage, predominant exon mapping and high-quality paired-end alignments. Differential expression analysis was conducted using DESeq2 (Galaxy v.2.11.40.8). The samples were grouped by condition with four replicates per group. Low-count genes were prefiltered before normalization. PCA and MA plots were generated to assess sample relationships and differential expression patterns. Genes with P_adj < 0.01 and |log₂[FC]| > 1 were considered to be differentially expressed. Per-gene z scores were calculated, and heat maps were generated using Morpheus (https://software.broadinstitute.org). Hierarchical clustering was performed using 1 − Pearson correlation as the distance metric and average linkage as the clustering method. RNA-seq data are available from NIH Gene Expression Omnibus (GEO) under identifier GSE321692.

Genome-wide CRISPR KO screen for myddosome assembly regulators

To express Cas9 in the parental MyD88–AGF iBMDMs, the sgRNA expression cassette was removed from pLentiCRISPR-v2 (52961, Addgene; a gift from F. Zhang)⁶¹ by digesting with KpnI-HF (R3142, NEB) and EcoRI-HF (R3101, NEB) and religating with annealed repair oligonucleotides that restore the restriction sites using Quick Ligation Kit (M2200, NEB). Next, the resulting plasmid was digested with BamHI-HF (R3136, NEB) and RsrII (R0501, NEB) to remove the PuroR gene and a bsr open reading frame (ORF) conferring blasticidin resistance amplified from pLenti-dCAS-VP64_Blast (61425, Addgene; a gift from F. Zhang)⁵⁵ inserted using Gibson assembly to obtain pLenti-CRISPR-Blast-no-sgRNA. MyD88–AGF WT iBMDM were transduced with lentiviral particles and selected with 15 μg ml⁻¹ blasticidin S (A1113903, Thermo Fisher Scientific). Successful transduction was validated by immunoblotting.

To generate an NF-κB activity reporter plasmid, a DNA sequence containing a synthetic poly(A) signal, a transcriptional pause site (to reduce background expression) and a minimal promoter was synthesized by gBlock (Integrated DNA Technologies, IDT) and a pentameric NF-κB response element (5×NF-κB-RE) was added to the promoter through annealed oligonucleotides by performing overlap extension PCR using Q5 DNA polymerase (M0491, NEB). An mScarlet-3 ORF was synthesized by gBlock and the two fragments were inserted into EcoRI-HF- and BamHI-HF-digested pLenti-P2A-PuroR⁶² by Gibson assembly (E2611, NEB). The resulting pLenti-mScarlet3-NF-κB-reporter plasmid was confirmed by whole-plasmid sequencing (Plasmidsaurus) and transduced into MyD88–AGF WT + Cas9 iBMDMs with lentiviral particles. Transduced cells were stimulated with 50 ng ml⁻¹ recombinant TNF for 6 h, sorted for uniform NF-κB-induced mScarlet3 expression and rested for >1 week, in the presence of 15 μg ml⁻¹ blasticidin S to maintain Cas9 expression until fluorescence had faded.

To validate the functionality of the NF-κB reporter and Cas9, cells were electroporated with recombinant sgRNAs against Tlr2, Tlr4, Myd88 or non-targeting control (IDT) using the Neon Transfection System (Thermo Fisher Scientific). In brief, 1.2 × 10⁶ cells were resuspended in 120 μl buffer R, 2 μl 100 µM sgRNA was added and the cells electroporated using 100 μl electroporation tips, 1,400 V, 2 pulses, 10 ms each. Cells were immediately placed into wells of a six-well plate containing 3 ml cDMEM. The medium was replaced after 4 h with fresh cDMEM + 15 μg ml⁻¹ blasticidin S to remove dead cells and maintain Cas9 expression. Then, 48 h later, the cells were collected and the KO efficiency was assessed by flow cytometry (TLR2/4) or immunoblot analysis (MyD88). Moreover, 0.2 × 10⁶ cells were seeded per well of a black-wall 96-well plate (12 wells for each genotype) and left to adhere for >4 h. Cells were then treated in triplicates with LPS, LTA and TNF in 200 μl Fluobrite DMEM (A1896701, Gibco) supplemented with 20 mM HEPES or left unstimulated. Plates were sealed and induction of mScarlet-3 fluorescence measured over 6 h on a Tecan Spark 10 M Multi-Mode Plate Reader heated to 37 °C (operated using Tecan SparkControl software). Fluorescence was recorded every 10 min and the resulting kinetic analysed with GraphPad/PRISM. At the end point, cells were collected, pooled from the replicate wells and single-cell fluorescence induction was analysed by flow cytometry.

To execute the screen, we used the recently published Julianna library³⁴. Plasmids of the pooled Library were obtained from Addgene (247028, a gift from J. Doench and D. Root) and propagated in ElectroMAX Stbl4 Competent Cells (11635018, Thermo Fisher Scientific) according to the depositor’s instructions. After DNA preparation by Maxiprep (740414.50, Macherey-Nagel), the plasmids were further purified to remove any contaminants. In brief, around 200 μg plasmid DNA (pDNA) was placed into a 1.5 ml tube and 1 vol 2× protein digestion buffer (final concentration 10 mM Tris pH 8.0, 1 mM EDTA and 0.5% SDS) was added. pDNA solution was supplemented with proteinase K to a final concentration of 0.1 mg ml⁻¹ (25530049, Thermo Fisher Scientific) and incubated at 37 °C for 30 min. Subsequently, pDNA was purified by phenol–chloroform extraction and ethanol precipitation (see above) and resuspended to a concentration of 1 μg μl⁻¹ in double-distilled H₂O. Library virus production followed the same protocol as described above using eight 10-cm dishes with HEK293T cells. Virus was produced in 7 ml cDMEM + 5 mM butyrate per dish for 72 h. Virus-containing supernatants were collected, centrifuged for 3 min at 400g to remove cellular debris, filtered through a 0.43 µm syringe filter, aliquoted and stocks were stored at −80 °C until use.

Viral tittering before library creation optimized the transduction conditions. Increasing amounts of Julianna library virus stocks (0–300 μl) were added to 3 × 10⁶ cells in cDMEM with 5 μg ml⁻¹ polybrene in a constant final volume of 2 ml per well of a 12-well plate. Transduction was performed by centrifugation at 1,250g for 1 h at 30 °C. After spinfection, 2 ml of cDMEM was added. The next day, 0.1 × 10⁶ cells from each condition were replated into two wells of a six-well plate each and half the wells supplemented with 4 μg ml⁻¹ puromycin. Test selection was carried out over 5 days. At the end of the selection, viable cells were counted and transduction efficiency calculated as the ratio of viable cells from the conditions with or without puromycin. The target was around 30% transduction efficiency to ensure a maximum of 0.5–1 infections per cell and that the final enrichment analysis can correlate guide abundance to a single perturbation⁶³.

To guarantee statistical robustness during enrichment analysis, the downstream screening assays should maintain >500 cells per sgRNA. For our library creation, a total of 4.8 ml of virus stock was used to transduce 1.5 × 10⁸ cells. Cells were transduced as described above and, 24 h after transduction, the cells were reseeded onto six 15 cm tissue culture dishes. After 1 day, selection was started using 4 μg ml⁻¹ puromycin and 15 μg ml⁻¹ blasticidin S. The selection medium was replaced every 2 days and cells expanded as necessary resulting in 15 confluent 15-cm tissue culture dishes with around 1 × 10⁹ cells after 1 week, when untransduced controls had died—marking the successful creation of iBMDM MyD88–AGF + 5×NFkB-RE-mScarlet3 + Cas9 +Julianna library cells (Julianna cells).

From these, 0.9 × 10⁹ cells were seeded on a total of eighteen 15-cm tissue culture dishes and left to rest overnight. Cells on six dishes each were either treated with 10 ng ml⁻¹ LPS with 100 ng ml⁻¹ LTA or left untreated. After 6 h of PAMP stimulation, cells were collected by detaching with 8 ml PBS + 2.5 mM EDTA per dish and all cells from one condition were pooled into 50 ml tubes. Cells were pelleted by centrifugation at 400g for 5 min, washed with PBS and fixed by resuspending the pellets in 100 ml of PBS + 0.04% BSA + 4% PFA and left to fix at room temperature for 15 min while mixing continuously with end-over-end rotation. Fixed cells were pelleted by centrifugation at 400g for 5 min, resuspended in 20 ml PBS + 0.04% BSA (cell density of around 1.5 × 10⁷ cells per ml) passed through a 70 µm nylon mesh strainer, aliquoted into 5 ml flow cytometry tubes and stored at 4 °C. Cells were sorted on a BD FACSMelody Cell Sorter as described above by gating on singlets and NF-κB^lowSSC^low cells (around 15% lowest for each). We obtained around 7.5–10 × 10⁶ cells per condition and these were collected into PBS + 1 mM EDTA + 30% FBS. All 3 × 10⁸ cells from the untreated condition were saved as input control.

Genomic DNA was extracted by resuspending the cells at a density of 1.0 × 10⁷ in nucleus lysis buffer (A7941, Promega) and adding proteinase K to a final concentration of 0.1 mg ml⁻¹. The samples were thoroughly mixed by vortexing and incubated 55 °C overnight to reverse PFA cross-linking of chromatin. The next morning RNase A (A7973, Promega) was added to a final concentration of 20 µg ml⁻¹, the samples mixed and incubated at 37 °C for 30 min. The samples were chilled on ice, one-third vol protein precipitation solution (A7951, Promega) was added and the samples were mixed by vortexing until they turned visibly cloudy. Proteins were precipitated by incubating on ice for 5 min followed by centrifugation at 4 °C and 16,000g for 5 min. The cleared supernatants were transferred into 1.5 ml tubes, and DNA was precipitated by adding 1 vol isopropanol. Tubes were inverted 50 times to mix and incubated at −20 °C for 15 min. Genomic DNA was pelleted by centrifugation at 4 °C and 21,000g for 20 min, washed twice with 70% ethanol, air dried and resuspended in double-distilled H₂O to a concentration of 1 μg μl⁻¹.

Next, the locus encoding the sgRNA was amplified by PCR using the following primers Fwd: 5′-GGACTATCATATGCTTACCGTAA-3′; rev: 5′-GTTGATAACGGACTAGCCTTAT-3′ and Q5 DNA polymerase. As each diploid nucleus contains around 6 pg of gDNA and 500× sgRNA coverage needs to be maintained, the PCRs used 210 μg of gDNA from the control sample and all gDNA from the LPS/LTA-treated and enrichment samples (about 70 μg each). For comparison, the locus was also amplified from 80 ng input pDNA. Each PCR reaction used 2.5 µg DNA per 50 μl reaction resulting in linear amplification over 28 cycles (annealing temperature, 58 °C; and extension time, 30 s). The resulting amplicons were gel purified and extracted using the crush and soak method. In brief, amplicon bands were excised and crushed into bits by passing through a 20 G needle. Crushed gel bits were soaked in 3 vol of soak buffer (300 mM sodium acetate + 1 mM EDTA pH 8.0) at 4 °C overnight with end-over-end rotation. The next day, gel suspensions were filtered through a 40 µm nylon mesh strainer and 3 vol of ice-cold 100% ethanol was added. The samples were inverted 50 times, incubated at −80 °C for 10 min and centrifuged at 7,000g for 2 h at 4 °C. DNA pellets were washed twice with 70% ethanol, air-dried and resuspended in 100 μl double-distilled H₂O. To ensure sufficient DNA purity, the samples were further purified using phenol–chloroform re-extractions as described above. The final samples (library input control, untreated control, LPS-treated and enriched, LTA-treated and enriched) were sent to Genewiz to perform library preparation and high-throughput sequencing.

All further data analysis was performed using Galaxy Server. Adapter trimming was performed using Cutadapt (Galaxy v.5.2) with the 5′ adapter sequence 5′-TTGTGGAAAGGACGAAACACCG-3′, discarding untrimmed reads and filtering for a minimum read length of 20 bp. This step selectively retained reads containing the Julianna plasmid-derived sequence upstream of the sgRNA cassette and ensures that reads began directly with the 20 bp sgRNA protospacer. Quality assessment of trimmed reads was performed using FastQC (Galaxy v.0.74), incorporating a custom adapter reference containing the sgRNA scaffold sequence immediately 3′ of the protospacer (5′-GTTTAAGAGCTA-3′). The sgRNA scaffold adapter was detected in over 90% of reads beginning at position 21, confirming correct library structure and trimming.

sgRNA abundance quantification was performed using MAGeCK count (Galaxy v.0.5.8.4) with trimmed and filtered fastq files and the corresponding sgRNA library annotation to generate a count summary and normalized counts. The sequencing depth was sufficient across all samples, with the minimum sgRNA coverage exceeding 350×. Mapping efficiency to the sgRNA library exceeded 87%, zero-count sgRNAs represented <0.1% of the library and Gini indices below 0.05 indicated that a high library complexity was maintained through all steps. sgRNA- and gene-level enrichment analyses were performed using the MAGeCK test (Galaxy v.0.5.8.1). Analyses were conducted for LPS versus untreated, LTA versus untreated as well as plasmid input versus untreated samples, producing ranked gene and sgRNA enrichment summaries. Guide dropout for essential genes was compared against datasets from ref. ⁶⁴. Gene lists were annotated to highlight non-targeting controls and known TLR signalling factors. Enrichment score distributions were visualized using GraphPad/Prism after log-transformation of positive selection scores.

RNA-seq analysis of primary WT and Ticam2 ^−/− BMDMs

RNA was extracted from 0.5 × 10⁶ cells differentiated and treated as described above using PureLink RNA Mini Kit (12183025, Thermo Fisher Scientific). During purification, on-column DNase treatment using PureLink DNase (12185010, Thermo Fisher Scientific) was performed to remove contaminating gDNA according to the manufacturer’s instructions. BMDMs were derived from three WT and three Ticam2^−/− mice representing biological replicates per genotype/treatment condition. The RNA quantity and quality was assessed by spectrophotometry (NanoDrop One, Thermo Fisher Scientific).

RNA-seq and initial data processing were performed by Plasmidsaurus. In brief, raw reads were filtered using fastp v.0.24.0 with poly(X) tail trimming, 3′ quality-based tail trimming, Phred quality score > 15, and a minimum length > 50 bp. Filtered reads were aligned to the reference genome (GRCm38, mm10) using STAR aligner v.2.7.11 with non-canonical splice junction removal, followed by coordinate sorting using samtools v.1.22.1. PCR and optical duplicates were removed using UMI-based deduplication with UMIcollapse v.1.1.0. Gene-level expression quantification was performed using featureCounts (subread package v.2.1.1) with strand-specific counting, multi-mapping read fractional assignment, exons and three prime UTR as the feature identifiers, and grouped by gene_id. Differential expression analysis was done with edgeR v.4.0.16 using filtering for low-expressed genes. Sample-sample correlations and PCA were calculated on normalized counts (trimmed mean of M-values, TMM) using Pearson correlation.

For creating heat maps, DEGs were extracted from the dataset with a threshold of P_adj < 0.05 and |log₂[FC]| > 1. Subset analysis of PRGs, SRGs and ISGs was performed using comparisons against gene lists compiled from refs. ^25,26,27,59 and Hallmark IFNα/IFNγ response datasets (MSigDB)⁶⁵. Per-gene z scores were calculated, and heat maps were generated using Morpheus (Broad Institute). Hierarchical clustering was performed using 1 − Pearson correlation as the distance metric and average linkage as the clustering method. RNA-seq data are available from NIH GEO under identifier GSE321707.

Generation of KO cells using CRISPR–Cas9

Guide RNA (gRNA) sequences targeting the genes of interest were designed using CRISPick⁶³. Complementary DNA oligonucleotides were annealed and cloned into BsmBI-v2-digested (R0739, NEB) pLentiCRISPR-v2 and transduced into target cells using lentiviral particles. After selection with 5 µg ml⁻¹ puromycin, the sgRNA-efficiency was assessed by Sanger sequencing of the gDNA surrounding the sgRNA targeting site followed by CRISPR ICE analysis (EditCo Bio). Knockout pools were subcloned by serial dilution into 96-well plates. The obtained clones were screened for absence of the target proteins by immunoblotting. For creating ∆Ticam1 cells, KO was reconfirmed by ICE analysis with scores over 95% for each clone, as no antibody is available. For each cell line, over five clones with confirmed absence of target proteins were pooled before undergoing a final screen using CRISPR ICE analysis and immunoblotting. All crRNA sequences and CRISPR ICE genotyping primers are provided in Supplementary Tables 2, 3.

Generation of p65–mScarlet3-expressing cells

To generate an mScarlet3-p65 expression plasmid, the mScarlet3 ORF was amplified from pLenti-mScarlet3-NF-κB-reporter plasmid and the p65 ORF from RelA-cFlag-pcDNA3 (20012, Addgene; a gift from S. Smale)⁶⁶ using Q5 DNA polymerase and cloned into pLenti-P2A-Zeo¹ using Gibson assembly. Successful cloning was confirmed by whole-plasmid sequencing. The obtained pLenti-mScarlet3-p65-P2A-Zeo plasmid was transduced into target cells with lentiviral particles and the cells selected with 200 µg ml⁻¹ Zeocin (R25001, Thermo Fisher Scientific). Surviving cells were sorted for uniform mScarlet3-p65 expression and successful transduction validated by immunoblotting. To validate functionality of the reporter, cells were treated with recombinant TNF for 1 h and live imaged by confocal microscopy, ensuring a homogenous and comparable response between cell lines.

Generation of TRAM-expressing cells

To generate TRAM WT and mutant expression plasmids, the TRAM ORF within pLenti-mTRAM-HA-P2A-Puro was first rendered CRISPR-resistant using site-directed mutagenesis and the resistance cassette was swapped for a neo (neomycin resistance) gene to confer mammalian resistance to G418 (Geneticin). This resulted in pLenti-mTRAM-HA-P2A-Neo and successful cloning was confirmed by whole-plasmid sequencing. Next, the TRAM ORF was mutated using site-directed mutagenesis to generate the G2A, 3X, E180A, C114H and D88A/E89A mutants. To generate expression plasmids for chimeric TRAM versions, the TRAM ORF was removed from pLenti-mTRAM-HA-P2A-Neo by restriction digest with XhoI (R0146, NEB) and XbaI (R0145, NEB). The TRAM ORF was amplified from the original plasmid using Q5 DNA polymerase but removing its N-terminal plasma membrane-binding domain (amino acids 1–39) and adding the sequence encoding the first 15 amino acids of either Fyn or Lyn by primer extension. Moreover, the sequence encoding the TIRAP^PBM (amino acids 1–102) was amplified from pcDNA3-mTIRAP (J.C.K. Laboratory). Finally, the empty backbone and amplicons were assembled using Gibson assembly, resulting in pLenti-P2A-Neo plasmids containing the ORFs of Fyn–TRAM, Lyn–TRAM or TIRAP^PBM–TRAM and successful cloning was confirmed by whole-plasmid sequencing. The obtained plasmids were transduced into MyD88–AGF ∆Ticam2 or ∆Tirap/Ticam2 iBMDMs with lentiviral particles and the cells selected with 1.5 mg ml⁻¹ G418 (ant-gn-5, Invivogen). Successful transduction was validated by immunoblotting.

Generation of TIRAP-overexpressing cell lines

To generate TIRAP WT and ∆TIR mutant expression plasmids, the TIRAP ORF was amplified from pcDNA3-mTIRAP (J.C.K. laboratory) with primers adding a C-terminal HA-tag and either amplifying the entire ORF or truncating the last 156 amino acids (∆TIR) using Q5 DNA polymerase and inserted into BamHI-HF and XhoI-digested pLenti-P2A-PuroR using Gibson assembly. This resulted in pLenti-mTIRAP-HA-P2A-Puro or pLenti-mTIRAP(∆TIR)-HA-P2A-Puro, which were confirmed by whole-plasmid sequencing. The obtained plasmids were transduced into iBMDM MyD88–AGF WT target cells with lentiviral particles and the cells selected with 5 µg ml⁻¹ puromycin. Successful transduction was validated by immunoblotting and overexpression was assessed after RNA extraction and RT–qPCR.

Generation of PH-PLCδ–mScarlet3-expressing PI(4,5)P₂-reporter cell lines

To generate a plasmid expressing a marker for PI(4,5)P₂ localization in live-cells, the ORF of the PIP₂-binding domain of human PLCδ1 (PH-PLCδ1; amino acids 1–170) was synthesized by gBlock (Integrated DNA Technologies, IDT), the mScarlet3 ORF was amplified from pLenti-mScarlet3-NF-κB-reporter plasmid using Q5 DNA polymerase and the fragments were inserted into BamHI-HF-digested pLenti-P2A-PuroR using Gibson assembly. Successful cloning was confirmed by whole-plasmid sequencing. The obtained pLenti-PH-PLCδ1-mScarlet3-P2A-Puro plasmid was transduced into all target cells (Supplementary Table 1) with lentiviral particles and the cells selected with 5 µg ml⁻¹ puromycin (R25001, Thermo Fisher Scientific). Surviving cells were sorted for homogenous mScarlet3 fluorescence across cell lines and successful transduction validated by fluorescence imaging.

Implementing CRISPR-SAM for activating Tirap and Tram genes in mouse macrophages

We implemented the CRISPR-SAM system⁵⁵ in mouse macrophages. This three-component gene activation (CRISPRa) tool can boost transcription of endogenous genes. To do so, MyD88–AGF iBMDMs were transduced with pLenti-dCAS-VP64_Blast using lentiviral particles and selected with 15 μg ml⁻¹ blasticidin S. Successful transduction was validated by immunoblotting.

Next, complementary DNA oligonucleotides encoding the crRNA sequences of the activation gRNAs (Supplementary Table 2) were annealed and cloned into BsmBI-v2-digested pXPR_502 (96923, Addgene; a gift from J. Doench and D. Root)⁶⁷ encoding the modified gRNA scaffold as well as the MS2-P65-HSF1 activator complex. Obtained plasmids were transduced into MyD88–AGF +dCas9 iBMDMs using lentiviral particles and selection performed using 15 μg ml⁻¹ blasticidin and 5 μg ml⁻¹ puromycin. However, transductions of iBMDMs with pXPR_502 were unproductive. Note that our standard lentivirus production procedure yielded 1,000× lower titres for pXPR_502 when compared side by side with a control (pLentiCRISPR-v2; determined using Lenti-X GoStix Plus; 631281, Takara). Furthermore, the hPGK promoter driving the activator expression appeared to be non-functional (inactivated) in iBMDMs as we were unable to detect any transcripts by qPCR.

We decided to abandon pXPR_502 and instead transfer the cassettes encoding the modified sgRNA and the MS2-P65-HSF1 activator complex into the pLentiCRISPR-v2 backbone, which produced superior titres. First, the Cas9 ORF in pLentiCRISPR-v2 was replaced with the activator complex encoding ORF from pXPR_502. For this, pLentiCRISPR-v2 was digested with AgeI-HF (R3552, NEB) and BamHI-HF, the activator ORF amplified from pXPR_502 using Q5 polymerase and Gibson assembly performed to obtain an intermediate plasmid now expressing the MS2-P65-HSF1 activator complex from an EF1α promoter. Second, the hU6 and sgRNA cassette in pLentiCRISPR-v2 was replaced with the modified sgRNA expression cassette from pXPR_502. For this, the intermediate plasmid was digested with PacI (R0547, NEB) and EcoRI-HF, the hU6/modified sgRNA cassette from pXPR_502 amplified using Q5 polymerase and the final pLenti-CRISPRa-empty plasmid was assembled by Gibson assembly. Successful cloning was confirmed by whole-plasmid sequencing. Next, the annealed oligos encoding the activation sgRNAs for Tirap or Ticam2 were inserted and the resulting plasmids or EV transduced into MyD88–AGF + dCas9 iBMDMs either individually or in binary combinations using lentiviral particles and cells selected using 15 μg ml⁻¹ blasticidin and 5 μg ml⁻¹ puromycin. This resulted in successful activation of Tirap and/or Ticam2 genes as detected by RT–qPCR and immunoblotting.

Flow cytometry and cell sorting

For analysis of TLR surface localization, 2 × 10⁶ iBMDMs were seeded per well of a six-well plate and treated as described above. Cells were collected by scraping in ice-cold PBS, washed twice with PBS, passed through a 35 µm nylon mesh strainer (352235, Corning) and resuspended in PBS containing mouse TruStain FcX PLUS (anti-mouse CD16/32; 156604, BioLegend) and primary antibodies against TLRs or isotype control (Supplementary Table 4). Cells were stained by incubation at 4 °C for 30 min in the dark, washed twice with ice-cold PBS and analysed on a LSRFortessa flow cytometer using FACSDiva software (BD Biosciences). The proportion of surface localized TLRs was determined by calculating the MFI using FlowJo v.10.8.2 and comparing with untreated cells and TLR-KO controls after gating for single cells. Analysis of NF-κB reporter activity also used live cells that were stained with LIVE/DEAD Fixable Violet Dead Cell Stain Kit (L34963, Thermo Fisher Scientific) according to the manufacturer’s instructions. The induction of the fluorescent NF-κB reporter was plotted as histograms of MFI using FlowJo v.10.8.2 after gating for live cells and single cells.

For sorting cells based on fluorescent protein expression, cells were collected, washed in PBS, resuspended in PBS + 1% FCS to a concentration of 4 × 10⁶ cells per ml and passed through a 35 µm nylon mesh strainer. Subsequently, cells were sorted into 500 μl complete medium on a BD FACSMelody Cell Sorter using BD FACSChorus software (BD Biosciences) by gating on singlets and fluorescent-protein-positive cells. Immediately after sorting, cells were seeded into six-well plates with complete medium.

SDS–PAGE and immunoblotting

For immunoblotting, 1 × 10⁶ cells were seeded per well of a 24-well plate and stimulated as described above. Cells were washed with ice-cold PBS and lysed for 15 min on ice in 250 μl RIPA buffer (25 mM Tris-HCl pH 7.4, 150 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS) supplemented with protease inhibitors (Protease Inhibitor Cocktail set III, EDTA free; 539134, Millipore) and 1 mM activated sodium orthovanadate (567540, Sigma-Aldrich), 5 mM sodium fluoride (201154, Sigma-Aldrich) and 3 mM β-glycerophosphate (G9422, Sigma-Aldrich) phosphatase inhibitors. The lysates were cleared by centrifugation at full speed for 10 min at 4 °C in a table-top centrifuge. Then, 15–20 μg of total protein per sample was mixed with 5× loading dye (125 mM Tris-HCl pH 6.8, 10% SDS, 50% glycerol, 0.06% bromophenol blue) supplemented with 5% β-mercaptoethanol, denatured at 95 °C for 10 min and separated by SDS–PAGE. For immunoblots against phosphorylated proteins, cells were washed with PBS and then lysed in 250 μl 1× loading dye supplemented with 5% β-mercaptoethanol. The cell lysates were homogenized by several passages through a 29 G needle, denatured at 95 °C for 10 min and separated by SDS–PAGE using home-made running buffer (25 mM Tris-HCl pH 8.3, 192 mM glycine, 0.1% (w/v) SDS) and 10%, 12% or 15% polyacrylamide gels for around 90 min at 140 V.

After SDS–PAGE, proteins were transferred onto methanol-activated PVDF membranes (IPVH00010, Millipore) using wet-transfer at 400 mA for 1 h in Towbin buffer (25 mM Tris-HCl pH 8.3, 192 mM glycine, 20% (v/v) methanol). Membranes were blocked with 5% BSA in TBS-T (0.05% Tween-20) for >30 min at room temperature. Incubation with primary antibodies (Supplementary Table 5) was performed at 4 °C overnight (typical dilution 1:1,000). Blots were developed by washing with TBS-T, probing with 1:10,000 diluted HRP-conjugated secondary antibodies (Supplementary Table 6), washing again and imaging on a ChemiDoc MP imaging system (Bio-Rad) with Image Lab 6.1 (Bio-Rad) software using SuperSignal West Pico PLUS (34580, Thermo Fisher Scientific) or SuperSignal West Femto Maximum Sensitivity Chemiluminescent Substrate (34096, Thermo Fisher Scientific). For reprobing membranes with different primary antibodies, membranes were stripped using Restore PLUS Western Blot Stripping Buffer (46430, Thermo Fisher Scientific). Densitometry analysis used quantification of protein band intensities with ImageJ/Fiji⁶⁸ and normalization to the loading control and/or holo-enzyme when assessing phosphorylated protein species. To determine the activity of TLR-responsive signalling pathways, obtained values were used to calculate the FC versus untreated cells and the total signalling pathway activity was determined by measuring the area under the curve as implemented in GraphPad/Prism.

Immunoprecipitations

Immunoprecipitation of endogenous MyD88 was performed as previously described⁶⁹ using 0.5 μg of anti-MyD88 antibody (Supplementary Table 5) and 15 μl (bed volume) of protein G Sepharose beads (for endogenous MyD88; P3296, Millipore) or 15 μl (bed volume) of anti-Flag(M2) agarose beads (for MyD88–AGF; A2220, Millipore). Immunoprecipitation of HA-tagged TRAM species used 15 μl (bed volume) of anti-HA agarose beads (26181, Thermo Fisher Scientific). Immunoprecipitation was performed by incubation at 4 °C for 4 h on a tube rotator and proteins were eluted using either 200 ng ml⁻¹ 3×Flag peptide (F4799, Sigma-Aldrich) in lysis buffer, by acidification with 50 μl of 0.2 M glycine pH 2.0 followed by an immediate wash with 50 μl of lysis buffer or by resuspending in 1× loading dye and heating at 95 °C for 15 min. The input, unbound and elution fractions were kept during immunoprecipitations and analysed by immunoblotting.

Analysis of gene expression by RT–qPCR

RNA was extracted from 0.5 × 10⁶ cells using the PureLink RNA Mini Kit (12183025, Thermo Fisher Scientific). During purification, on-column DNase treatment using PureLink DNase (12185010, Thermo Fisher Scientific) was performed to remove contaminating gDNA according to the manufacturer’s instructions. The purified RNA was reverse transcribed using high-capacity cDNA synthesis kit (4368813, Thermo Fisher Scientific). qPCR used iTaq Universal SYBR Green Supermix (1725121, Bio-Rad), 20–40 ng cDNA in a 10 μl reaction and primers at 1 μM final concentration (the sequences are provided in Supplementary Table 7). Data were recorded on a CFX384 Real-Time system (Bio-Rad) and analysed using CFX Manager v.3.1 (Bio-Rad). Primer specificity was ensured by designing primers to span exon–exon junctions, whenever possible. Recorded C_t values were normalized to the recorded C_t of the mouse Hprt1 housekeeping gene, and data were plotted as ∆C_t (relative expression) or as the FC compared with the untreated controls.

Cytokine ELISAs and Cell Titer Glow assays

ELISAs were performed to measure TNF, IL-6, IL12-p40, IFNβ or IP-10 secretion. Cell culture supernatants were cleared of cell debris by centrifugation at 400g for 3 min. Cleared supernatants were diluted between 1:2 and 1:100 with ELISA diluent to be within the assay range. The concentrations of cytokines were measured according to the manufacturer’s protocols using uncoated mouse IL-6 (88-7064-88), uncoated mouse IL-12p40 (88-7120-88), uncoated mouse TNF (88-7324-88, all from Thermo Fisher Scientific), LumiKine Xpress mIFN-β (luex-mifnbv3, Invivogen) or mouse IP-10 (DY466, R&D Systems) ELISA kits.

Cell viability was quantified using the CellTiter-Glo Luminescent Cell Viability Assay (G7571; Promega) according to the manufacturer’s instructions. The background signal from medium-only wells was subtracted, and viability was expressed relative to the indicated control. ELISA absorbance and luminescent readings were measured with a Tecan Spark 10 M Multi-Mode Plate Reader.

Fixed immunofluorescence imaging

For imaging, 0.25 × 10⁶ cells were seeded onto coverslips in 24-well plates. Cells were fixed with 4% methanol-free formaldehyde (50-980-487, EMS) added directly to the cell culture medium followed by incubation for 15 min at room temperature. Fixed cells were washed with PBS and kept at 4 °C overnight. Fixed specimens were permeabilized with PermQuench buffer (0.2% (w/v) BSA and 0.02% (w/v) saponin in PBS) for 30 min at room temperature and stained with primary antibodies (Supplementary Table 5) for 1 h at room temperature. After three washes with PBS, cells were incubated with dye-conjugated secondary antibodies (Supplementary Table 6) and 1 μg ml⁻¹ DAPI (D1306, Thermo Fisher Scientific) diluted in PermQuench buffer for 1 h at room temperature. The coverslips were extensively washed with PBS, once with double-distilled H₂O and mounted using 5 μl ProLong Glass Antifade Mountant (P36982, Thermo Fisher Scientific).

Imaging of prepared specimens was performed using a LSM 880 point-scanning confocal microscope with fast Airyscan controlled with Zeiss Zen Blue software (Zeiss Instruments). Imaging used a Plan Apochromat ×40/1.3 NA or a Plan Apochromat ×63/1.4 NA oil-immersion objective (Zeiss Instruments). In a typical experiment, 9–25 fields of view were imaged using multiposition acquisition and the following settings: bidirectional scanning, 2–4× frame averaging, 16 bit bit-depth, ROI 2,000 × 2,000 px and optical zoom 1 resulting in a pixel size of 65 nm px⁻¹ and pinhole: 89.3, resulting in 0.5 µm sections. Acquired raw microscopy data were processed using ImageJ/Fiji to produce representative images and analysed for quantification of myddosome formation and localization.

Live-cell confocal microscopy

Live-cell imaging was performed using a LSM 880 point scanning confocal microscope (Zeiss Instruments) with an environment control chamber. Imaging used a Plan Apochromat ×63/1.4 oil immersion objective (Zeiss Instruments) and the following settings: bidirectional scanning, 2× line averaging, 16 bit bit-depth, ROI 2,000 × 2,000 px and optical zoom 1 resulting in a pixel size of 65 nm px⁻¹, pinhole: 89.3, resulting in 0.5 µm sections and sampling every 60 s with minimal laser power to minimize acquisition bleaching. In a typical experiment, 10–20 fields of view were imaged simultaneously, and z focus was controlled every five acquisitions using the Definite Focus 2 system (Zeiss Instruments).

To prepare samples for live-cell imaging, 2 × 10⁶ MyD88–AGF iBMDMs were seeded onto 35 mm dishes with an integrated coverslip (P35G-1.5-14-C, Mattek). If more than one cell line was imaged, 0.2 × 10⁶ cells were seeded per well of an 8 high-well µ-Slide (80806, ibidi) instead. Then, 1 h before imaging, the culture medium was replaced with 2 ml prewarmed FluoroBrite DMEM supplemented with 20 mM HEPES. Simultaneously, the environmental control chamber on the confocal microscope was set to pre-heat to 37 °C and left to equilibrate. To image myddosome formation, the prepared cells were placed on the microscope and left to equilibrate for around 15 min. Myddosome formation was induced by PAMP-treating cells using a microfluidics setup followed by live imaging for up to 12 h. Similarly, MPP was added during live imaging to disrupt myddosomes. Raw imaging data were further processed as described below to quantify myddosome formation and localization.

Quantification of myddosome plasticity was performed and analysed using FRAP as described previously^1,70. Myddosome formation was induced by LPS treatment for 1 h with subsequent addition of 10 µM MPP or DMSO control. The experiment was performed 5 min after MPP addition.

TIRF microscopy

TIRF microscopy was performed using a Nikon Ti inverted microscope equipped with an Andor Zyla 4.2 Plus sCMOS camera, an environment control chamber, and Nikon TIRF Lun-f Illumination for 488 nm excitation of MyD88–AGF. TIRF microscopy was performed using an Apo TIRF ×100/1.49 DIC N2 oil-immersion objective (Nikon) and the following settings: no pixel binning, 12-bit & Gain 4 bit depth, 150 ms exposure, 1% laser power, TIRF angles between 1955–2010, ROI 1,024 × 1,024 px resulting in a pixel size of 65 nm px⁻¹ and sampling every 15 s. Raw imaging data were further processed as described below to quantify proto-myddosome formation.

Imaging analysis

Myddosome detection and quantification was performed as described previously¹. Other object detection used Cellpose-SAM as implemented in Cellpose (v.4.0.8)⁷¹ to detect cells, nuclei were detected using the ImageJ/Fiji plugin for StarDist^72,73. Proto-myddosome density from TIRF imaging was calculated by dividing the number of detected myddosomes by the cell-covered area. The average myddosome position over time was extracted by determining the centroid of each myddosome label within single cells. Membrane distance was calculated as minimal Euclidian distance to the plasma membrane, that is, the outline of the cell label containing the respective myddosome. Distances of less than 2 µm were considered as membrane proximal/associated.

Label images of nuclei and cells were used for tracking of objects using the ImageJ/Fiji plugin Trackmate⁷⁴ and the simple LAP tracker. Tracking was performed using the following empirically determined parameters: max. linking distance, 8 µm; max. gap closing and max. distance, both 10 µm; and max. frame gap, 2. Detected nuclei and cell labels were used to analyse NF-κB activation in PAMP-treated iBMDMs. Peak analysis as implemented in GraphPad/PRISM was used to quantify the NF-κB activity peaks and oscillations.

MS analysis of endogenous mouse TIRAP

Immunoprecipitation of TIRAP from iBMDM WT was performed as described above but starting with 1.00 × 10⁸ cells using 2 μg of primary antibody against TIRAP (Supplementary Table 5) and 40 μl (bed volume) of protein G Sepharose beads. After immunoprecipitation, beads were washed three times with lysis buffer and twice with PBS and retained proteins eluted by boiling in 50 μl 1× loading dye for 15 min. The entire eluate was run on a 12% SDS–PAGE gel and the gel stained with Coomassie Instant Blue (ab119211, Abcam). The band corresponding to the expected size of mouse TIRAP was excised and submitted for MS analysis by the HMS Taplin Biological Mass Spectrometry Facility.

In brief, excised gel bands were cut into approximately 1 mm³ pieces. Gel pieces were then subjected to a modified in-gel trypsin digestion procedure⁷⁵. Gel pieces were washed and dehydrated with acetonitrile for 10 min followed by removal of acetonitrile. The pieces were then completely dried in a speed-vac. Rehydration of the gel pieces was with 50 mM ammonium bicarbonate solution containing 12.5 ng μl⁻¹ modified sequencing-grade trypsin (Promega) at 4 °C. After 45 min, the excess trypsin solution was removed and replaced with 50 mM ammonium bicarbonate solution to just cover the gel pieces. The samples were then placed in a 37 °C room overnight. Peptides were later extracted by removing the ammonium bicarbonate solution, followed by one wash with a solution containing 50% acetonitrile and 1% formic acid. The extracts were then dried in a speed-vac (for around 1 h) and stored at 4 °C until analysis.

On the day of analysis, the samples were reconstituted in 5–10 µl of HPLC solvent A (2.5% acetonitrile, 0.1% formic acid). A nano-scale reverse-phase HPLC capillary column was created by packing 2.6 µm C18 spherical silica beads into a fused silica capillary (100 µm inner diameter × approximately 30 cm length) with a flame-drawn tip⁷⁶. After equilibrating the column each sample was loaded through a Thermo EASY-LC (Thermo Fisher Scientific). A gradient was formed and peptides were eluted with increasing concentrations of solvent B (90% acetonitrile, 0.1% formic acid).

As peptides eluted, they were subjected to electrospray ionization and then entered into an Orbitrap Fusion Lumos mass spectrometer (Thermo Fisher Scientific). Peptides were detected, isolated and fragmented to produce a tandem mass spectrum of specific fragment ions for each peptide. Peptide sequences (and therefore protein identity) were determined by matching protein databases with the acquired fragmentation pattern by the software program, Sequest (Thermo Fisher Scientific)⁷⁷. All databases include a reversed version of all sequences, and the data were filtered to a 1–2% peptide FDR.

Recombinant protein production and purification

Mouse MyD88^TIR (residues 129–296), mouse TIRAP^TIR (residues 74–249), and mouse TRAM^TIR (63–232) were subcloned into a pET28a expression vector using Gibson Assembly. The TIRAP^TIR(P125H) mutant was generated using site-directed mutagenesis with Q5 DNA polymerase. All constructs were designed with an N-terminal His₆ tag, MBP, and a TEV recognition sequence, and verified by whole-plasmid sequencing. Clones were transformed into E. coli BL21 (DE3) cells, grown at 37 °C in LB broth (Miller) to an optical density at 600 nm (OD₆₀₀) of 0.4–0.6, and induced with 1 mM IPTG (I56000, Research Product International) for 4 h at 30 °C. Bacterial cells were collected by centrifugation at 5,000g for 5 min, flash-frozen in liquid nitrogen and pellets stored at −80 °C until purification. Bacterial cells were lysed by sonication in buffer containing 50 mM HEPES pH 8.0, 150 mM NaCl, 10 mM β-mercaptoethanol and 1 mM PMSF. The soluble supernatant fraction was extracted by ultracentrifugation at 50,000g for 1 h, filtered through a glass fibre filter and loaded over a Ni-NTA affinity column (cOmplete His-Tag Purification Resin; 5893682001, Roche) pre-equilibrated with lysis buffer. The column was washed in 10 column volumes (CV) of wash buffer A (50 mM HEPES pH 8.0, 300 mM NaCl, 10 mM β-mercaptoethanol), followed by 20 CV of wash buffer B (50 mM HEPES pH 8.0, 500 mM NaCl, 10 mM β-mercaptoethanol, 5 mM imidazole), and eluted in 5 CV of elution buffer (50 mM HEPES pH 8.0, 150 mM NaCl, 10 mM β-mercaptoethanol, 150 mM imidazole). The resulting eluate was supplemented with 1% (w/w) TEV protease and dialysed overnight at 4 °C in 10 kDa cut-off SnakeSkin Dialysis Tubing (68100, Thermo Fisher Scientific) in 1 l buffer (50 mM HEPES pH 8.0, 150 mM NaCl, 10 mM β-mercaptoethanol). The nickel resin column was re-equilibrated in 20 CV dialysis buffer and passed with the dialysate three times. The TEV-cleaved TIR domain containing flow-through was further concentrated using Amicon Ultra Centrifugal Filters (3/10 kDa MWCO; UFC900324, Millipore Sigma) and loop-loaded over a single Superdex 200 Increase 10/300 GL (Cytiva) column in 500 µl volumes. The fractions corresponding to the protein of interest were verified by SDS–PAGE and Coomassie Instant Blue staining, pooled and concentrated to about 4 mg ml⁻¹, flash-frozen in liquid nitrogen and aliquots stored at −80 °C until further use. Protein was stored in a final buffer composition of 20 mM HEPES pH 7.0, 50 mM NaCl and 5 mM DTT.

Turbidity assays

Protein aliquots were thawed on ice, centrifuged at 15,000g for 5 min to remove precipitate and the protein concentration of the supernatant was measured. To assess TIR domain assembly formation, 0–200 µM concentration ranges of MyD88^TIR, TRAM^TIR and TIRAP^TIR, alone, and in combination, were investigated. The assay was performed in a UV-STAR 96-well half area plate (Greiner Bio-One) using a TECAN plate reader, shaken continuously at 180 rpm for 5 h at 37 °C and OD₃₅₀ measurements recorded at 5 min intervals. Reactions were performed in triplicate conditions in 30 µl volumes in assay buffer (20 mM HEPES pH 7.0, 50 mM NaCl, and 5 mM DTT), and their average used as one repeat. Assay data shown contain data from at least two independent purifications.

BS³ cross-linking and immunoblot analysis

MyD88^TIR, TRAM^TIR and TIRAP^TIR were mixed in 30 µl duplicate volumes in assay buffer, with one maintained at 4 °C and the other incubated at 37 °C for 5 h to mimic turbidity assay start and end point conditions, respectively. Bis(sulfosuccinimidyl)suberate (BS³) cross-linker (21586, Thermo Fisher Scientific) was reconstituted to 100 mM in turbidity assay buffer at room temperature, immediately before use. Each reaction received 1.5 µl reconstituted BS³ (final concentration, 5 mM) and was incubated at room temperature for 30 min for complete cross-linking. This was followed by a 15 min quench reaction at room temperature, by adding Tris HCl pH 7.5 to a final concentration of 50 mM. Cross-linked samples were analysed by SDS–PAGE and immunoblotting as described above. For complete transfer of the larger protein assemblies and TCPs, protein transfer onto PVDF membranes was performed at 45 V and 4 °C overnight.

In vivo myddosome inhibition experiments

For all in vivo studies, 12-week old female C57BL/6N mice purchased from Envigo were used. Mice were allowed at least a 72 h acclimatization period on arrival at Boston Children’s Hospital (BCH) and were maintained under specific-pathogen-free conditions with continuous access to food and water. All animals were maintained under a 12 h–12 h light–dark cycle in a temperature- and humidity-controlled vivarium, with lights on from 07:00 to 19:00, and the mice were examined before the initiation of the studies to ensure that they were healthy and acclimated to the laboratory environment. All of the animal studies were approved by the BCH Institutional Animal Care and Use Committee (IACUC) and conducted under the supervision of the department of Animal Resources at Children’s Hospital (ARCH). Mice were i.p. injected with 1 mg per kg LTA or 1 mg per kg LPS in 200 μl PBS or with PBS only. After 30 min mice were i.p. injected with 150 mg per kg TSI-13-57 formulated in 200 μl 5% N-methyl-2-pyrrolidone (NMP; 328634, Millipore-Sigma), 5% Solutol HS 15 (HY-Y1893, MedChem Express), 90% normal saline or vehicle control. Then, 90 min after the PAMP injection, the mice were euthanized with CO₂ and spleens surgically removed. Spleens were kept in wash buffer (PBS + 1 mM EDTA + 2% FCS) on ice and mechanically dissociated by gentle homogenization through a 70 µm nylon mesh strainer using a syringe plunger. While on ice, the splenocyte suspension was washed with 20 ml wash buffer and cells were centrifuged at 500g for 5 min and 4 °C. Red blood cells were lysed by resuspending in 1 ml ACK buffer, incubating for 1 min at room temperature and quenching with 10 vol wash buffer. The cell suspension was filtered through a 40 µm strainer and the cells washed twice with wash buffer using centrifugation at 500g for 5 min at 4 °C. Splenocytes were resuspended in 1 ml PBS + phosphatase inhibitors and counted. Immunoprecipitation of MyD88 and associated proteins was performed as described above using lysates from 5.0 × 10⁷ splenocytes.

To analyse serum cytokines and splenic IFN responses, mice were injected as described above. Then, 90 min after the PAMP injection, mice were bled retro-orbitally. After 4.5 h, a submandibular bleed was performed and mice were immediately euthanized with CO₂, after which spleens were surgically removed. Blood was collected in EDTA-coated containers (BD Microtainer K₂EDTA, 365974) and centrifuged at 2,000g for 10 min at 4 °C. The obtained serum was sampled for TNF, IL-6 and IL-12p40 by ELISA. Spleens were processed as described above and splenocyte lysates were assessed for an IFN response by immunoblotting and RT–qPCR.

Data handling and statistics

Data were plotted using Prism v.10.6.1 (GraphPad) and are presented as mean values of multiple experiments, with error bars showing the s.d. or s.e.m. unless otherwise noted. Statistical significance of results was determined using nonparametric one-way ANOVA, unpaired t-tests or two-way ANOVA as indicated in the figure legends. Benjamini, Krieger and Yekutieli false-discovery-rate-based (Q = 5%) correction for multiple comparisons as implemented in Prism was used when making more than three comparisons. Uncropped immunoblot images and source data are provided in Supplementary Data 1 and the Source Data.

Reporting summary

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

1. Fisch, D. et al. Molecular definition of the endogenous Toll-like receptor signalling pathways. Nature 631, 635–644 (2024).

2. Motshwene, P. G. et al. An oligomeric signaling platform formed by the Toll-like receptor signal transducers MyD88 and IRAK-4. J. Biol. Chem. 284, 25404–25411 (2009).

   Article  CAS  PubMed  PubMed Central  Google Scholar 

3. Lin, S.-C., Lo, Y.-C. & Wu, H. Helical assembly in the MyD88–IRAK4–IRAK2 complex in TLR/IL-1R signalling. Nature 465, 885–890 (2010).

   Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

4. Medzhitov, R. et al. MyD88 is an adaptor protein in the hToll/IL-1 receptor family signaling pathways. Mol. Cell 2, 253–258 (1998).

   Article  CAS  PubMed  Google Scholar 

5. Latty, S. L. et al. Activation of Toll-like receptors nucleates assembly of the MyDDosome signaling hub. eLife 7, e31377 (2018).

   Article  PubMed  PubMed Central  Google Scholar 

6. Deliz-Aguirre, R. et al. MyD88 oligomer size functions as a physical threshold to trigger IL1R myddosome signaling. J. Cell Biol. 220, e202012071 (2021).

   Article  CAS  PubMed  PubMed Central  Google Scholar 

7. Cao, F., Deliz-Aguirre, R., Gerpott, F. H., Ziska, E. & Taylor, M. J. Myddosome clustering in IL-1 receptor signaling regulates the formation of an NF-kB activating signalosome. EMBO Rep. 24, e57233 (2023).

   Article  CAS  PubMed  PubMed Central  Google Scholar 

8. Poltorak, A. et al. Defective LPS signaling in C3H/HeJ and C57BL/10ScCr mice: mutations in Tlr4 gene. Science 282, 2085–2088 (1998).

   Article  ADS  CAS  PubMed  Google Scholar 

9. Kawagoe, T. et al. Sequential control of Toll-like receptor–dependent responses by IRAK1 and IRAK2. Nat. Immunol. 9, 684–691 (2008).

   Article  CAS  PubMed  Google Scholar 

10. Medzhitov, R. Approaching the asymptote: 20 years later. Immunity 30, 766–775 (2009).

    Article  CAS  PubMed  Google Scholar 

11. Kagan, J. C., Magupalli, V. G. & Wu, H. SMOCs: supramolecular organizing centres that control innate immunity. Nat. Rev. Immunol. 14, 821 (2014).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

12. Horng, T., Barton, G. M., Flavell, R. A. & Medzhitov, R. The adaptor molecule TIRAP provides signalling specificity for Toll-like receptors. Nature 420, 329–333 (2002).

    Article  ADS  CAS  PubMed  Google Scholar 

13. Yamamoto, M. et al. Essential role for TIRAP in activation of the signalling cascade shared by TLR2 and TLR4. Nature 420, 324–329 (2002).

    Article  ADS  CAS  PubMed  Google Scholar 

14. Kawai, T., Adachi, O., Ogawa, T., Takeda, K. & Akira, S. Unresponsiveness of MyD88-deficient mice to endotoxin. Immunity 11, 115–122 (1999).

    Article  CAS  PubMed  Google Scholar 

15. Strickson, S. et al. Roles of the TRAF6 and Pellino E3 ligases in MyD88 and RANKL signaling. Proc. Natl Acad. Sci. USA 114, E3481–E3489 (2017).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

16. Cohen, P. & Strickson, S. The role of hybrid ubiquitin chains in the MyD88 and other innate immune signalling pathways. Cell Death Differ. 24, 1153–1159 (2017).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

17. Arthur, J. S. C. & Ley, S. C. Mitogen-activated protein kinases in innate immunity. Nat. Rev. Immunol. 13, 679–692 (2013).

    Article  CAS  PubMed  Google Scholar 

18. Fitzgerald, K. A. et al. IKKε and TBK1 are essential components of the IRF3 signaling pathway. Nat. Immunol. 4, 491–496 (2003).

    Article  CAS  PubMed  Google Scholar 

19. Tan, P., He, L. & Zhou, Y. Engineering supramolecular organizing centers for optogenetic control of innate immune responses. Adv. Biol. 5, e2000147 (2021).

    Article  Google Scholar 

20. Yamamoto, M. et al. TRAM is specifically involved in the Toll-like receptor 4–mediated MyD88-independent signaling pathway. Nat. Immunol. 4, 1144–1150 (2003).

    Article  CAS  PubMed  Google Scholar 

21. Lee, T. K. et al. A noisy paracrine signal determines the cellular NF-κB response to lipopolysaccharide. Sci. Signal. 2, ra65 (2009).

    Article  PubMed  PubMed Central  Google Scholar 

22. Kawai, T. & Akira, S. Signaling to NF-κB by Toll-like receptors. Trends Mol. Med. 13, 460–469 (2007).

    Article  CAS  PubMed  Google Scholar 

23. Covert, M. W., Leung, T. H., Gaston, J. E. & Baltimore, D. Achieving stability of lipopolysaccharide-induced NF-κB activation. Science 309, 1854–1857 (2005).

    Article  ADS  CAS  PubMed  Google Scholar 

24. Yamamoto, M. et al. Role of adaptor TRIF in the MyD88-independent Toll-like receptor signaling pathway. Science 301, 640–643 (2003).

    Article  ADS  CAS  PubMed  Google Scholar 

25. Hargreaves, D. C., Horng, T. & Medzhitov, R. Control of inducible gene expression by signal-dependent transcriptional elongation. Cell 138, 129–145 (2009).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

26. Ramirez-Carrozzi, V. R. et al. A unifying model for the selective regulation of inducible transcription by CpG islands and nucleosome remodeling. Cell 138, 114–128 (2009).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

27. Bhatt, D. M. et al. Transcript dynamics of proinflammatory genes revealed by sequence analysis of subcellular RNA fractions. Cell 150, 279–290 (2012).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

28. Janeway, C. A. Approaching the asymptote? Evolution and revolution in immunology. Cold Spring Harb. Symp. Quant. Biol. 54, 1–13 (1989).

    Article  CAS  PubMed  Google Scholar 

29. Ma, M., Jiang, W. & Zhou, R. DAMPs and DAMP-sensing receptors in inflammation and diseases. Immunity 57, 752–771 (2024).

    Article  CAS  PubMed  Google Scholar 

30. Fitzgerald, K. A. & Kagan, J. C. Toll-like receptors and the control of immunity. Cell 180, 1044–1066 (2020).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

31. Kawai, T., Ikegawa, M., Ori, D. & Akira, S. Decoding Toll-like receptors: recent insights and perspectives in innate immunity. Immunity 57, 649–673 (2024).

    Article  CAS  PubMed  Google Scholar 

32. Ve, T. et al. Structural basis of TIR-domain-assembly formation in MAL- and MyD88-dependent TLR4 signaling. Nat. Struct. Mol. Biol. 24, 743–751 (2017).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

33. Takeuchi, O. et al. Differential roles of TLR2 and TLR4 in recognition of Gram-negative and Gram-positive bacterial cell wall components. Immunity 11, 443–451 (1999).

    Article  CAS  PubMed  Google Scholar 

34. Drepanos, L. M. et al. Balancing off-target and on-target considerations for optimized CRISPR-Cas9 knockout library design. Cell Genom. 6, 101190 (2026).

35. Li, W. et al. MAGeCK enables robust identification of essential genes from genome-scale CRISPR/Cas9 knockout screens. Genome Biol. 15, 554 (2014).

    Article  PubMed  PubMed Central  Google Scholar 

36. Parnas, O. et al. A genome-wide CRISPR screen in primary immune cells to dissect regulatory networks. Cell 162, 675–686 (2015).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

37. Oshiumi, H. et al. TIR-containing adapter molecule (TICAM)-2, a bridging adapter recruiting to Toll-like receptor 4 TICAM-1 that induces interferon-β *. J. Biol. Chem. 278, 49751–49762 (2003).

    Article  CAS  PubMed  Google Scholar 

38. Kagan, J. C. et al. TRAM couples endocytosis of Toll-like receptor 4 to the induction of interferon-β. Nat. Immunol. 9, 361–368 (2008).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

39. Rowe, D. C. et al. The myristoylation of TRIF-related adaptor molecule is essential for Toll-like receptor 4 signal transduction. Proc. Natl Acad. Sci. USA 103, 6299–6304 (2006).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

40. Sacre, S. M. et al. Selective use of TRAM in lipopolysaccharide (LPS) and lipoteichoic acid (LTA) induced NF-κB activation and cytokine production in primary human cells: TRAM is an adaptor for LPS and LTA signaling. J. Immunol. 178, 2148–2154 (2007).

    Article  CAS  PubMed  Google Scholar 

41. Stack, J. et al. TRAM is required for tlr2 endosomal signaling to type I IFN induction. J. Immunol. 193, 6090–6102 (2014).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

42. Nilsen, N. J. et al. A role for the adaptor proteins TRAM and TRIF in Toll-like receptor 2 signaling *. J. Biol. Chem. 290, 3209–3222 (2015).

    Article  CAS  PubMed  Google Scholar 

43. Ohnishi, H. et al. TRAM is involved in IL-18 signaling and functions as a sorting adaptor for MyD88. PLoS ONE 7, e38423 (2012).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

44. Fitzgerald, K. A. et al. Mal (MyD88-adapter-like) is required for Toll-like receptor-4 signal transduction. Nature 413, 78–83 (2001).

    Article  ADS  CAS  PubMed  Google Scholar 

45. Horng, T., Barton, G. M. & Medzhitov, R. TIRAP: an adapter molecule in the Toll signaling pathway. Nat. Immunol. 2, 835–841 (2001).

    Article  CAS  PubMed  Google Scholar 

46. Kagan, J. C. & Medzhitov, R. Phosphoinositide-mediated adaptor recruitment controls Toll-like receptor signaling. Cell 125, 943–955 (2006).

    Article  CAS  PubMed  Google Scholar 

47. Bonham, K. S. et al. A promiscuous lipid-binding protein diversifies the subcellular sites of toll-like receptor signal transduction. Cell 156, 705–716 (2014).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

48. Sakai, J. et al. Lipopolysaccharide-induced NF-κB nuclear translocation is primarily dependent on MyD88, but TNFα expression requires TRIF and MyD88. Sci Rep. 7, 1428 (2017).

    Article  ADS  PubMed  PubMed Central  Google Scholar 

49. Zambrano, S., De Toma, I., Piffer, A., Bianchi, M. E. & Agresti, A. NF-κB oscillations translate into functionally related patterns of gene expression. eLife 5, e09100 (2016).

    Article  PubMed  PubMed Central  Google Scholar 

50. Cook, P. et al. Salmonella-induced SipB-independent cell death requires Toll-like receptor-4 signalling via the adapter proteins Tram and Trif. Immunology 122, 222–229 (2007).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

51. Ragland, S. A. et al. Purine and pyrimidine-based bacterial cyclic dinucleotides egress the phagosome and activate the innate immune sensor STING. Immunity 59, 2086–2105 (2026).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

52. Verstak, B. et al. The TLR signaling adaptor TRAM interacts with TRAF6 to mediate activation of the inflammatory response by TLR4. J. Leukoc. Biol. 96, 427–436 (2014).

    Article  PubMed  PubMed Central  Google Scholar 

53. Enokizono, Y. et al. Structures and interface mapping of the TIR domain-containing adaptor molecules involved in interferon signaling. Proc. Natl Acad. Sci. USA 110, 19908–19913 (2013).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

54. Funami, K. et al. Identification of a regulatory acidic motif as the determinant of membrane localization of TICAM-2. J. Immunol. 195, 4456–4465 (2015).

    Article  CAS  PubMed  Google Scholar 

55. Konermann, S. et al. Genome-scale transcriptional activation by an engineered CRISPR-Cas9 complex. Nature 517, 583–588 (2015).

    Article  ADS  CAS  PubMed  Google Scholar 

56. Várnai, P. & Balla, T. Visualization of phosphoinositides that bind pleckstrin homology domains: calcium- and agonist-induced dynamic changes and relationship to Myo-[³H]inositol-labeled phosphoinositide pools. J. Cell Biol. 143, 501–510 (1998).

    Article  PubMed  PubMed Central  Google Scholar 

57. Ippagunta, S. K. et al. Identification of Toll-like receptor signaling inhibitors based on selective activation of hierarchically acting signaling proteins. Sci. Signal. 11, eaaq1077 (2018).

    Article  PubMed  PubMed Central  Google Scholar 

58. Kim, A. H., Krummenacher, B., Yeung, J., Koes, D. R. & Lee, R. E. C. NEMO recruitment at single cytokine-receptor complexes shows quantized dynamics independent of ligand affinity. Cell Rep. 44, 116637 (2025).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

59. Sen, S. et al. Gene regulatory strategies that decode the duration of NFκB dynamics contribute to LPS-versus TNF-specific gene expression. Cell Syst. 10, 169–182 (2020).

60. Blankenberg, D. et al. Dissemination of scientific software with Galaxy ToolShed. Genome Biol. 15, 403 (2014).

    Article  PubMed  PubMed Central  Google Scholar 

61. Sanjana, N. E., Shalem, O. & Zhang, F. Improved vectors and genome-wide libraries for CRISPR screening. Nat. Methods 11, 783–784 (2014).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

62. Fisch, D. et al. Human GBP1 differentially targets Salmonella and Toxoplasma to license recognition of microbial ligands and caspase-mediated death. Cell Rep. 32, 108008 (2020).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

63. Doench, J. G. et al. Optimized sgRNA design to maximize activity and minimize off-target effects of CRISPR-Cas9. Nat. Biotechnol. 34, 184–191 (2016).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

64. Hart, T., Brown, K. R., Sircoulomb, F., Rottapel, R. & Moffat, J. Measuring error rates in genomic perturbation screens: gold standards for human functional genomics. Mol. Syst. Biol. 10, MSB145216 (2014).

    Article  Google Scholar 

65. Castanza, A. S. et al. Extending support for mouse data in the Molecular Signatures Database (MSigDB). Nat. Methods 20, 1619–1620 (2023).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

66. Sanjabi, S. et al. A c-Rel subdomain responsible for enhanced DNA-binding affinity and selective gene activation. Genes Dev. 19, 2138–2151 (2005).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

67. Sanson, K. R. et al. Optimized libraries for CRISPR-Cas9 genetic screens with multiple modalities. Nat. Commun. 9, 5416 (2018).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

68. Schindelin, J. et al. Fiji: An open-source platform for biological-image analysis. Nat. Methods 9, 676–682 (2012).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

69. Tan, Y. & Kagan, J. C. Biochemical isolation of the myddosome from murine macrophages. Methods Mol. Biol. 1714, 79–95 (2018).

70. Kang, M., Andreani, M. & Kenworthy, A. K. Validation of normalizations, scaling, and photofading corrections for FRAP data analysis. PLoS ONE 10, e0127966 (2015).

    Article  PubMed  PubMed Central  Google Scholar 

71. Pachitariu, M., Rariden, M. & Stringer, C. Cellpose-SAM: superhuman generalization for cellular segmentation. Preprint at bioRxiv https://doi.org/10.1101/2025.04.28.651001 (2025).

72. Schmidt, U., Weigert, M., Broaddus, C. & Myers, G. Cell detection with star-convex polygons. Lect. Notes Comput. Sci. 11071, 265–273 (2018).

73. Tinevez, J. Y. et al. TrackMate: an open and extensible platform for single-particle tracking. Methods 115, 80–90 (2017).

    Article  CAS  PubMed  Google Scholar 

74. Ershov, D. et al. TrackMate 7: integrating state-of-the-art segmentation algorithms into tracking pipelines. Nat. Methods 19, 829–832 (2022).

    Article  CAS  PubMed  Google Scholar 

75. Shevchenko, A., Wilm, M., Vorm, O. & Mann, M. Mass spectrometric sequencing of proteins from silver-stained polyacrylamide gels. Anal. Chem. 68, 850–858 (1996).

    Article  ADS  CAS  PubMed  Google Scholar 

76. Peng, J. & Gygi, S. P. Proteomics: the move to mixtures. J. Mass Spectrom. 36, 1083–1091 (2001).

    Article  ADS  CAS  PubMed  Google Scholar 

77. Eng, J. K., McCormack, A. L. & Yates, J. R. An approach to correlate tandem mass spectral data of peptides with amino acid sequences in a protein database. J. Am. Soc. Mass. Spectrom. 5, 976–989 (1994).

    Article  ADS  CAS  PubMed  Google Scholar 

Download references

Authors and Affiliations

1. Division of Gastroenterology, Boston Children’s Hospital and Harvard Medical School, Boston, MA, USA

   Daniel Fisch, Vaani Ohri, Lucia J. Wesemann, Joon H. Choi, Katherine E. Lindblad & Jonathan C. Kagan

2. Department of Medical Oncology, Dana-Farber Cancer Institute, Harvard Medical School, Boston, MA, USA

   Eleni Anastasakou

3. Broad Institute of MIT and Harvard, Cambridge, MA, USA

   Eleni Anastasakou

4. Ludwig Center, Harvard Medical School, Boston, MA, USA

   Eleni Anastasakou

Contributions

D.F. and J.C.K. conceived the study. D.F. performed experiments with contributions from L.J.W., J.H.C. and K.E.L. assisted with mouse experiments. V.O. performed production of recombinant proteins and in vitro assays. E.A. assisted with designing and implementing CRISPRa experiments. D.F. and J.C.K. acquired and contributed funding and experimental oversight, analysed the data and wrote the manuscript, with input from all of the authors. All of the authors discussed the results and commented on the manuscript.

Corresponding authors

Correspondence to Daniel Fisch or Jonathan C. Kagan.

Competing interests

J.C.K. receives compensation from and holds equity in Corner Therapeutics, Larkspur Biosciences, MindImmune Therapeutics and Neumora Therapeutics. None of these relationships impacted this study. The other authors declare no competing interests.

Peer review information

Nature thanks Dominic De Nardo, Nick Gay and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available.

Reporting Summary (download PDF )

Peer Review File (download PDF )

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions