N4BP1 negatively regulates NF-κB by binding and inhibiting NEMO oligomerization

Many immune responses depend upon activation of NF-κB, an important transcription factor in the elicitation of a cytokine response. Here we show that N4BP1 inhibits TLR-dependent activation of NF-κB by interacting with the NF-κB signaling essential modulator (NEMO, also known as IκB kinase γ) to attenuate NEMO–NEMO dimerization or oligomerization. The UBA-like (ubiquitin associated-like) and CUE-like (ubiquitin conjugation to ER degradation-like) domains in N4BP1 mediate interaction with the NEMO COZI domain. Both in vitro and in mice, N4bp1 deficiency specifically enhances TRIF-independent (TLR2, TLR7, or TLR9-mediated) but not TRIF-dependent (TLR3 or TLR4-mediated) NF-κB activation, leading to increased production of proinflammatory cytokines. In response to TLR4 or TLR3 activation, TRIF causes activation of caspase-8, which cleaves N4BP1 distal to residues D424 and D490 and abolishes its inhibitory effect. N4bp1−/− mice also have diminished numbers of T cells in the peripheral blood. Our work identifies N4BP1 as an inhibitory checkpoint protein that must be overcome to activate NF-κB, and a TRIF-initiated caspase-8-dependent mechanism by which this is accomplished.

The inducible transcription factor NF-κB plays a key role in the development and function of the immune system by activating or repressing hundreds of genes that carry out the cellular response to specific stimuli1. Canonical activation of NF-κB requires its release from the inhibitory IκB proteins that normally retain NF-κB in the cytoplasm2. This occurs when the IκB kinase (IKK) complex phosphorylates IκB proteins resulting in their subsequent K48-ubiquitination and degradation by the proteasome. In this role, the IKK complex serves as the gatekeeper for NF-κB activation and as such is subject to numerous modifications that modulate its activity. Composed of three subunits, the kinases IKKα and IKKβ and the regulatory subunit NEMO, the IKK complex is activated by combinations of phosphorylation (on IKK), ubiquitination (on NEMO), and sumoylation (on NEMO) events, and negative regulators target these modifications to either remove or prevent them3,4. Persistent NF-κB activation may lead to pathologic consequences such as chronic inflammation or autoimmunity5.

Although receptor-proximal signaling pathways differ, inflammatory stimuli such as tumor necrosis factor (TNF), interleukin (IL)-1, and ligands for the Toll-like receptors (TLRs) all activate NF-κB, impinging on the IKK complex to do so. TLR activation results in the recruitment of up to four adapter proteins, MyD88 (myeloid differentiation 88), MAL (MyD88-adapter like), TRIF (Toll-interleukin 1 receptor (TIR) domain-containing adapter inducing IFNβ), and TRAM (TRIF related adapter molecule), which assemble distinct signaling complexes depending on the combination of adapters utilized. All TLRs except TLR3 depend upon MyD88 for full signaling activity whereas TRIF is utilized only by TLR3 and TLR41,6,7. Signaling from TLRs leads to IKK complex activation; NF-κB then enters the nucleus to activate the transcription of many proinflammatory genes including those encoding TNF and IL-6.

In this work, we carried out a forward genetic screen to discover proteins that mediate TLR signaling, testing peritoneal macrophages from N-ethyl-N-nitrosourea (ENU)-mutagenized mice for altered TNF responses to stimulation with TLR ligands. We identify several mutations in the neural precursor cell expressed, developmentally downregulated 4 binding protein 1 gene (N4bp1) that cause elevated TNF production by macrophages stimulated with ligands for TLRs other than TLR3 and TLR4. Mechanistic studies show how N4BP1, initially identified as a NEDD4 binding protein, inhibits NF-κB activation downstream from all TLRs except TLR3 and TLR4. We also show how TLRs that recruit TRIF overcome the inhibitory effect of N4BP1 on NF-κB.

Multiple N4bp1 mutations enhance TRIF-independent TLR signaling

Peritoneal macrophages from third-generation (G3) descendants of ENU-mutagenized mice were assayed for TNF secretion after stimulation with TLR ligands. Three distinct mutations of N4bp1 (stash, winter, and acorn; Fig. 1a) in three ancestrally unrelated pedigrees were associated with increased production of TNF in response to R848, a TLR7 ligand (Fig. 1b, cand Supplementary Fig. 1a–c). The acorn (ac) allele contained the nonsense mutation Q135* and was a predicted null allele. We detected no N4BP1 protein in N4bp1ac/ac macrophages, consistent with nonsense-mediated decay of the transcript (Fig. 1d).

Further analysis of responses to TLR stimulation demonstrated dose-dependent increased production of TNF by N4bp1ac/ac macrophages stimulated with Pam3CSK4 (TLR2-TLR1 ligand, Fig. 1e) or R848 (Fig. 1f). However, N4bp1ac/ac macrophages produced similar amounts of TNF as N4bp1+/+ macrophages when stimulated with poly(I:C) (TLR3 ligand, Fig. 1g) or lipopolysaccharide (LPS; TLR4 ligand, Fig. 1h). Peritoneal macrophages (Fig. 1i, j), bone marrow-derived macrophages (BMMs; Supplementary Fig. 1e), and bone marrow-derived dendritic cells (BMDCs; Supplementary Fig. 1f) from mice homozygous for a CRISPR-Cas9-targeted null allele of N4bp1 (Fig. 1m) recapitulated the acorn phenotype, responding to Pam3CSK4 and R848 by producing excessive amounts of TNF and interleukin (IL)-6. Time course analysis of the cytokine response in N4bp1−/− peritoneal macrophages supported elevated production rather than a merely accelerated response (Supplementary Fig. 1h). In contrast, the TNF and IL-6 responses to poly(I:C) and LPS were normal (Fig. 1k, land Supplementary Fig. 1e, f). IFNα production induced by dsDNA stimulation was comparable between wild-type and N4bp1−/− BMMs and BMDCs (Supplementary Fig. 1g). These findings indicate that TLR2-TLR1 and TLR7 signaling, but not TLR3 and TLR4 signaling, were enhanced in N4bp1−/− macrophages and dendritic cells. Since TLR3 and TLR4 depend on the TRIF adapter protein to propagate signaling, whereas all other TLRs signal independently of TRIF, we concluded that N4BP1 inhibits TRIF-independent signaling.

Among tissues examined, the 893-amino acid N4BP1 protein was highly expressed in lymphoid organs, brain, and ovary (Supplementary Fig. 2a). Cellular fractionation of peritoneal macrophages stimulated with R848 for up to 4 h showed that N4BP1 was predominantly localized in the cytosol before and throughout the period of stimulation (Supplementary Fig. 2b). We also detected reduced numbers of CD4+ and CD8+ T cells in the peripheral blood of N4bp1−/− mice (Supplementary Fig. 2c). The numbers of peripheral blood monocytes, neutrophils, and B cells were normal (Supplementary Fig. 2c). N4bp1−/−mice appeared grossly normal and showed no internal anatomical abnormalities. The mice were not overtly suffering from autoimmunity since the concentration of anti-dsDNA antibodies in the serum was comparable with that in wild-type mice (Supplementary Fig. 2d).

The amino acid sequence of N4BP1 is 81% identical in humans and mice. We knocked out endogenous N4BP1 via CRISPR-Cas9 gene targeting in the human monocytic cell line THP1 and observed more TNF secretion in response to stimulation with Pam3CSK4 and R848, but not LPS, relative to that of N4BP1-sufficient THP1 cells (Supplementary Fig. 1i–k). These data indicate that the function of N4BP1 in inhibiting TRIF-independent TLR signaling is conserved between humans and mice.

N4BP1 inhibits NF-κB activation and NF-κB-dependent gene expression

We examined MAPK and NF-κB signaling in N4bp1−/− peritoneal macrophages. We found that phosphorylated NF-κB p65 was markedly increased in N4bp1−/− macrophages compared to N4bp1+/+ macrophages before or after Pam3CSK4 stimulation (Fig. 2a). Phosphorylated IKKα/β was also increased in N4bp1−/− macrophages after Pam3CSK4 stimulation (Fig. 2a) and this correlated with increased IKKβ association with NEMO (Fig. 2b), an interaction required for NF-κB activation3. Indeed the IKKα/β kinase activity was significantly elevated in N4bp1−/− macrophages (Fig. 2c). By contrast, we observed no appreciable difference between N4bp1−/− and wild-type macrophages in their levels of phosphorylated p38 and c-Jun in response to Pam3CSK4 (Fig. 2a). Moreover, chromatin immunoprecipitation and quantitative PCR (ChIP-qPCR) indicated that greater amounts of IL-6 promoter DNA was associated with NF-κB p65 and acetyl-histone H4 in N4bp1−/− macrophages than in wild-type macrophages stimulated with Pam3CSK4 (Fig. 2d). These findings demonstrate that NF-κB activation and recruitment to the IL-6 promoter was increased in N4bp1−/− macrophages in response to Pam3CSK4.

The increased association between acetyl-histone H4 and the IL-6 promoter in N4bp1−/− cells suggested that Il6 was more transcriptionally active in N4bp1−/− cells than in N4bp1+/+ cells. To investigate transcript expression of Il6, Tnf, and other genes, we performed RNA sequencing of N4bp1+/+ and N4bp1−/− peritoneal macrophages before and after R848 or LPS stimulation. We found that the expression of NF-κB-dependent genes, including Tnf and Il6, was higher in N4bp1−/− macrophages than in N4bp1+/+ macrophages after R848 but not LPS stimulation (Supplementary Fig. 3a, b and Supplementary Data 1). These observations were confirmed by RT-qPCR of Il1b, Il6, Il10, Il12 p40, Tnf, and Ccl5 (Fig. 2e). Moreover, we found that NF-κB-dependent genes were upregulated in unstimulated N4bp1−/− macrophages relative to their expression levels in N4bp1+/+ macrophages (Supplementary Fig. 3c and Supplementary Data 1).

To analyze the gating effect of N4BP1 on various promoter elements, we used CRISPR-Cas9 gene targeting to generate N4BP1 knockout HEK293T cells and transfected them with luciferase reporter constructs driven by (1) an IL-6 promoter, (2) an IL-6 promoter with the NF-κB binding site deleted, (3) an IL-6 promoter with AP-1 and CEBPb binding sites deleted8, (4) a pure NF-κB binding site, or (5) a pure AP-1 binding site. In the absence of stimulation, wild-type IL-6 promoter-dependent luciferase activity was elevated in N4BP1−/− cells compared to that in N4BP1+/+ cells; this activity was abolished by deletion of the NF-κB binding site, but not deletion of the AP-1 and CEBPb binding sites, from the IL-6 promoter (Fig. 2f). Similarly, a consensus NF-κB binding site, but not an AP-1 binding site, drove increased reporter activity in N4BP1−/− cells relative to N4BP1+/+ cells (Fig. 2f). Furthermore, the level of the NF-κB inhibitor IκBα was decreased in N4BP1−/− HEK293T cells compared to that in N4BP1+/+ cells, both at steady state (t = 0) and at various times after imposition of translational blockade with cycloheximide (Fig. 2g). Together, these data demonstrate that N4BP1 negatively regulates NF-κB-dependent gene expression in macrophages both before and in response to TLR stimulation.

N4BP1 associates with NEMO to inhibit NF-κB

To identify the functional target(s) of N4BP1, we co-transfected different NF-κB pathway components, each together with N4BP1, into HEK293T cells. By immunoprecipitation (IP), NEMO was identified as a binding partner of N4BP1 (Fig. 3a). The interaction between endogenous N4BP1 and NEMO was also detected in peritoneal macrophages and was not influenced by cell stimulation with Pam3CSK4 (Fig. 3b).

N4BP1 was reported to have a K Homology (KH) domain, putatively capable of RNA binding, near the N-terminus, and a N4BP1, YacP-like nuclease (NYN) domain near the C-terminus9. To test whether N4BP1 directly degraded NF-κB-dependent gene mRNA, we treated wild-type and N4BP1−/− HEK293T cells with IL-1β together with actinomycin D, an inhibitor of transcription. Increased IκBα and IL-8 mRNA levels were observed in N4BP1−/− cells, however the decay rates were comparable in wild-type and N4BP1−/− cells (Supplementary Fig. 4a, b). Furthermore, mice with N4BP1 NYN domain mutation Y774H did not affect R848 induced TNF production (Supplementary Fig. 1d). The results suggest that N4BP1 does not function as a nuclease to negatively regulate NF-κB-dependent gene expression.

By searching COLIS10 and HHpred (https://toolkit.tuebingen.mpg.de/#/tools/hhpred), we further identified two putative ubiquitin binding domains in N4BP1: a ubiquitin-associated (UBA)-like domain in the middle region of N4BP1 (a.a. 341-400) and a ubiquitin conjugation to ER degradation (CUE)-like domain at the C-terminus (a.a. 843-896; Fig. 3c). To map the region responsible for N4BP1-NEMO association, we generated constructs encoding HA-tagged N4BP1 deletion fragments (Fig. 3c), expressed them in HEK293T N4BP1−/− cells, and examined the ability of the encoded proteins to bind to NEMO. The KH domain, UBA-like domain, and CUE-like domain were necessary for interaction with NEMO since deletion of any one of them reduced the association between N4BP1 and NEMO (Fig. 3c). The NYN domain in N4BP1 was dispensable for NEMO binding (Fig. 3c). In NEMO, the coil zipper (COZI) domain was necessary and sufficient for binding to N4BP1 (Fig. 3d).

We tested the effect of deletion of the UBA-like domain, CUE-like domain, or the KH domain of N4BP1 on NF-κB-dependent gene expression in N4BP1−/− HEK293T cells. NF-κB driven luciferase reporter activity was increased in N4BP1−/− HEK293T cells compared to that in N4BP1+/+ cells at steady state and after IL-1β stimulation (Fig. 3e). Transfection of N4BP1−/−cells with either full-length N4BP1 or N4BP1 with NYN domain deletion reduced the luciferase activity to levels similar to those in wild-type HEK293T cells after IL-1β stimulation. N4BP1 with KH domain deletion or with one or both ubiquitin-binding domains deleted failed to normalize NF-κB-dependent luciferase activity when expressed in N4BP1−/−HEK293T cells (Fig. 3e). These findings indicate that the KH domain and ubiquitin-binding domains of N4BP1 are necessary for its association with NEMO and the inhibition of NF-κB.

N4BP1 inhibits NEMO oligomerization

We next explored the mechanism by which N4BP1 affected the function of NEMO. The UBA-like domain and CUE-like domain are potential sites for ubiquitin binding. Indeed, when incubated in vitro, purified recombinant HA-N4BP1 immunoprecipitated together with K63-linked and linear tetra-ubiquitin chains, but not with K48-linked tetra-ubiquitin chains (Fig. 4a). However, N4BP1 lacking the UBA-like and CUE-like domains (ΔUBAL&CUEL) failed to bind to linear and K63-linked ubiquitin chains (Fig. 4b and Supplementary Fig. 5a). NEMO ubiquitination and NEMO binding to linear ubiquitin chains are both required for NF-κB activation11,12. We therefore hypothesized that ubiquitin chains might regulate the interaction between NEMO and N4BP1. We confirmed the direct interaction of purified recombinant Flag-NEMO and HA-N4BP1 proteins by immunoprecipitation (Fig. 4c and Supplementary Fig. 5b). The addition of purified recombinant polyubiquitin chains enhanced in a dose-dependent manner the interaction between NEMO and N4BP1 (Fig. 4cand Supplementary Fig. 5b). These data indicate that linear and K63-linked ubiquitin chains, which bind to both N4BP1 and NEMO, promote N4BP1-NEMO interaction.

Previous reports have shown that NEMO homo-oligomerization is required for recruitment of IKKα and IKKβ and subsequent NF-κB activation13,14. In addition, such oligomerization may be mediated by a C-terminal region of NEMO that overlaps with the binding site for N4BP113,15,16. Thus, we tested whether NEMO-N4BP1 interaction could inhibit NEMO homo-oligomerization. Purified recombinant V5-tagged NEMO immunoprecipitated with Flag-tagged NEMO and this interaction was enhanced in the presence of linear polyubiquitin chains (Fig. 4d). The addition of HA-tagged N4BP1 dose-dependently reduced the amount of V5-NEMO that immunoprecipitated with Flag-NEMO (Fig. 4d). Similar results were observed when K63-linked polyubiquitin chains were substituted for linear polyubiquitin chains in the same experiment (Supplementary Fig. 5c). We also observed more NEMO dimers and oligomers in N4bp1−/− peritoneal macrophages (Fig. 4e). These findings suggest that N4BP1 blocks NEMO oligomerization by binding to the C-terminal oligomerization domain of NEMO, thereby precluding IKKα/β recruitment and NF-κB activation.

TRIF-dependent N4BP1 cleavage by caspase-8

TLR proximal signaling pathways differ depending in part on the adapters that serve them, but all converge downstream from TRAF6 to signal activation of MAPKs and NF-κB17. To understand how N4BP1 deficiency specifically enhanced NF-κB signaling in response to Pam3CSK4 and R848, but had no effect on NF-κB activation in response to poly(I:C) or LPS, we first examined N4BP1 expression levels in peritoneal macrophages after LPS, Pam3CSK4, R848, and poly(I:C) stimulation. We found that N4BP1 expression decreased by 30 min after LPS treatment, was undetectable at 60 and 120 min after LPS treatment and recovered to barely detectable levels by 180 min after stimulation (Fig. 5a). The LPS-induced downregulation of N4BP1 was not observed in TRIF-deficient (Triflps2/lps2) macrophages but remained intact in IRAK4-deficient macrophages homozygous for the otiose allele, which fail to propagate signals from MyD88 (Fig. 5a). Poly(I:C) stimulation of wild-type macrophages induced N4BP1 downregulation with a similar time course and level of reduction as that induced by LPS (Fig. 5b). In contrast, N4BP1 failed to be downregulated in wild-type macrophages in response to Pam3CSK4 and R848 (Fig. 5b). To summarize, LPS stimulation of wild-type or IRAK4-deficient macrophages, or poly(I:C) stimulation of wild-type macrophages, led to downregulation of N4BP1 expression. Downregulation of N4BP1 was not observed in TRIF-deficient macrophages treated with LPS or in wild-type macrophages treated with Pam3CSK4 or R848. These data indicate that stimulation of TRIF-dependent TLR signaling pathways results in N4BP1 downregulation.

We hypothesized that TRIF might direct the downregulation and inhibition of N4BP1. To investigate which pathway(s) might function downstream from TRIF to downregulate N4BP1, we tested whether inhibitors of the proteasome, autophagy, cysteine proteases, or protein translation could block LPS-induced N4BP1 downregulation, but none did so (Supplementary Fig. 6). Neither did the necrosis inhibitor Necrostatin-1 block LPS-induced N4BP1 downregulation (Fig. 5c). However, the pan-caspase inhibitor Z-VAD-FMK successfully blocked N4BP1 downregulation such that N4BP1 was detectable by immunoblot of wild-type macrophage lysates at all time points up to 180 min after LPS stimulation (Fig. 5c). We tested more specific caspase inhibitors and found that Z-IETD-FMK, a caspase-8 inhibitor, abrogated N4BP1 downregulation in response to LPS (Fig. 5d). Furthermore, we confirmed that caspase-8 knockdown inhibited LPS-induced N4BP1 downregulation (Fig. 5d). These data suggest that LPS-induced TRIF-dependent signaling activates caspase-8 to cleave N4BP1, resulting in reduced levels of full-length N4BP1. This is consistent with a previous report that TRIF but not MyD88 activates caspase-818.

To further probe the functional relationship between TRIF and N4BP1, we tested the effect of N4BP1 deficiency on LPS-induced TNF production by Triflps2/lps2 macrophages. We observed increased TNF production by Triflps2/lps2N4bp1−/− cells compared to Triflps2/lps2 cells, consistent with the hypothesis that TRIF signaling inhibits N4BP1 (Fig. 5e). As expected, wild-type and Triflps2/lps2 macrophages produced similar amounts of TNF in response to Pam3CSK4 or R848 stimulation, and N4BP1 deficiency significantly increased those amounts (Supplementary Fig. 7a, b). Since TLR3 signaling is fully dependent on TRIF, no TNF production was detected in either Triflps2/lps2N4bp1−/− or Triflps2/lps2 macrophages after poly(I:C) stimulation (Supplementary Fig. 7c). These data support the conclusion that TRIF-dependent signaling leads to inhibition of N4BP1.

TNF also induces caspase-8 activation19, and we indeed observed N4BP1 downregulation after we treated peritoneal macrophages with TNF but not with IL-1β (Supplementary Fig. 7d, e). In line with these findings, IL-1β but not TNF induced more Il6 and Tnf mRNA expression in N4bp1−/− macrophages than in wild-type macrophages (Supplementary Fig. 7f, g).

Caspase-8 preferentially cleaves protein and peptide substrates immediately after D residues in ‘XEXD’ and ‘LETD’ motifs, respectively, where ‘X’ represents any residue20. By scanning the human N4BP1 protein sequence, we identified two putative caspase-8 cleavage sites (LTTD424 and PETD490), which were also present in the mouse protein sequence. To investigate whether N4BP1 is cleaved at one or both sites in a TRIF-dependent manner, we expressed HA-tagged N4BP1 with or without Flag-tagged TRIF in HEK293T cells and immunoblotted the cell lysates for HA-N4BP1. In addition to the full-length protein, two N4BP1 fragments were detected in lysates of cells expressing HA-N4BP1 and Flag-TRIF, and these fragments corresponded in size to the 424-aa and 490-aa products expected to result from caspase-8 cleavage (Fig. 5f, lanes 1–6). A D424A mutation abrogated the 424-aa cleavage product while preserving the 490-aa product (Fig. 5f, lanes 7–8). An analogous outcome resulted from a D490A mutation (Fig. 5f, lanes 9–10). Mutation of both D424 and D490 to alanine blocked detectable TRIF-dependent N4BP1 cleavage (Fig. 5f, lanes 11–12). We also observed that HA-N4BP1 cleavage products immunoprecipitated reduced amounts of Flag-NEMO compared to full-length HA-N4BP1 (Fig. 5g), and could not reduce NF-κB driven luciferase activity to a normal level when expressed in N4BP1−/− HEK293T cells (Fig. 5h). When N4BP1 with D424A and D490A mutations was expressed in N4BP1−/− HEK293T cells, it functioned the same as wild-type N4BP1 in reducing IL-1β induced NF-κB driven luciferase activity; in contrast, it diminished the luciferase response induced by TNF as compared to wild-type N4BP1 (Fig. 5i). Taken together, these results suggest that when activated during TRIF-dependent TLR signaling and TNF signaling, caspase-8 cleaves N4BP1 after D424 and/or D490. These cleavage events inactivate N4BP1 and reduce its binding to NEMO, resulting in unrestricted IKK complex formation and NF-κB activation.

N4BP1 limits TRIF-independent NF-κB activation in vivo

To assess the role of N4BP1 as a checkpoint for TRIF-dependent and -independent TLR signaling in vivo, we intravenously injected wild-type and N4bp1−/− mice with CpG or LPS and measured systemic cytokine responses. Three hours after CpG injection, the serum of N4bp1−/− mice had a significantly higher concentration of IL-6 than did wild-type serum (Fig. 6a). However, we did not detect appreciable differences between N4bp1−/− and wild-type mice in serum IL-6 concentration after LPS administration (Fig. 6b). These results were consistent with our previous observation that N4bp1 deficiency enhanced TRIF-independent signaling.

The recognition of herpes simplex virus 1 (HSV1) via TLRs (notably TLR9) and cytosolic receptors21 initiates signaling leading to NF-κB activation. After infection with a non-lethal dose of HSV1, we monitored serum IL-6 in wild-type and N4bp1−/− mice. Six hours after infection, N4bp1−/− mice had significantly elevated serum concentrations of IL-6 compared to wild-type mice (Fig. 6c). Together, these results provide in vivo evidence that N4BP1 attenuates NF-κB activation.