IRAK1-dependent Regnase-1-14-3-3 complex formation controls Regnase-1-mediated mRNA decay

Regnase-1 is an endoribonuclease crucial for controlling inflammation by degrading mRNAs encoding cytokines and inflammatory mediators in mammals. However, it is unclear how Regnase-1-mediated mRNA decay is controlled in interleukin (IL)-1β- or Toll-like receptor (TLR) ligand-stimulated cells. Here, by analyzing the Regnase-1 interactome, we found that IL-1β or TLR stimulus dynamically induced the formation of Regnase-1-β-transducin repeat-containing protein (βTRCP) complex. Importantly, we also uncovered a novel interaction between Regnase-1 and 14-3-3 in both mouse and human cells. In IL-1R/TLR-stimulated cells, the Regnase-1-14-3-3 interaction is mediated by IRAK1 through a previously uncharacterized C-terminal structural domain. Phosphorylation of Regnase-1 at S494 and S513 is critical for Regnase-1-14-3-3 interaction, while a different set of phosphorylation sites of Regnase-1 is known to be required for the recognition by βTRCP and proteasome-mediated degradation. We found that Regnase-1-14-3-3 and Regnase-1-βTRCP interactions are not sequential events. Rather, 14-3-3 protects Regnase-1 from βTRCP-mediated degradation. On the other hand, 14-3-3 abolishes Regnase-1-mediated mRNA decay by inhibiting Regnase-1-mRNA association. In addition, nuclear-cytoplasmic shuttling of Regnase-1 is abrogated by 14-3-3 interaction. Taken together, the results suggest that a novel inflammation-induced interaction of 14-3-3 with Regnase-1 stabilizes inflammatory mRNAs by sequestering Regnase-1 in the cytoplasm to prevent mRNA recognition.


Introduction
The expression of proinflammatory cytokines is the hallmark of innate immune responses against microbial infection. Whereas inflammatory responses are critical for the elimination of invading pathogens, excess and chronic inflammation can culminate in tissue destruction and autoimmune diseases. When innate immune cells encounter pathogen-associated molecular patterns (PAMPs) or damage-associated molecular patterns (DAMPs), they are sensed by pattern-recognition receptors such as Toll-like receptors (TLRs), triggering the transcription of inflammatory genes (Fitzgerald and Kagan, 2020;Takeuchi and Akira, 2010).
The expression of inflammatory genes is also controlled by post-transcriptional mechanisms to facilitate or limit inflammatory responses (Anderson, 2010;Carpenter et al., 2014;Turner and Díaz-Muñoz, 2018). Regnase-1 (also referred to as Mcpip1, Gene name: Zc3h12a), an RNase, is a critical regulator of inflammation. Regnase-1 binds to and degrades inflammatory mRNAs such as Il6 or Il12b by recognizing stem-loop structures present in the 3' untranslated regions (Matsushita et al., 2009;Mino et al., 2015). Zc3h12a-deficient mice exhibit an autoimmune phenotype, indicating its importance as a negative regulator of inflammation (Matsushita et al., 2009;Uehata et al., 2013). Regnase-1 efficiently suppresses the expression of its target genes by degrading CBP80-bound mRNAs during the pioneer-round of translation by associating with ribosome and a helicase protein, UPF1 (Mino et al., 2015;Mino et al., 2019). CBP80 binds to newly synthesized mRNAs in the nucleus and is replaced by eIF4E after the pioneer round of translation following mRNA export from the nucleus (Maquat et al., 2010;Müller-McNicoll and Neugebauer, 2013). Thus, it is possible that Regnase-1 recognizes target mRNAs in the steps leading to the pioneer round of translation.
The stability of cytokine mRNAs is dynamically regulated in innate immune cells under inflammatory conditions (Carpenter et al., 2014;Hao and Baltimore, 2009;Turner and Díaz-Muñoz, 2018). Post-translational control of Regnase-1 in response to inflammatory stimuli contributes to extending half-lives of inflammatory mRNAs. Stimulation of cells with TLR-ligands, IL-1b, or IL-17 results in the activation of IkB kinases (IKKs), which phosphorylate Regnase-1 at S435 and S439, in addition to IkBa (Iwasaki et al., 2011;Kakiuchi et al., 2020;Nanki et al., 2020;Tanaka et al., 2019). Regnase-1, phosphorylated at S435 and S439 is subsequently recognized by bTRCP, one of the components of the SKP1-CUL1-F-box (SCF) complex, which induces K48-linked polyubiquitination of Regnase-1, followed by proteasome-mediated degradation (Iwasaki et al., 2011). On the other hand, these stimuli also induce transcription of Zc3h12a (Iwasaki et al., 2011). Consequently, the protein level of Regnase-1 drastically changes during these stimulations; Regnase-1 levels decrease immediately after the stimulation and then increase to levels higher than its pre-stimulation. However, the post-translational regulatory mechanism of Regnase-1 following inflammatory stimuli is still not fully elucidated.
14-3-3e as well as with HA-tagged bTRCP in HeLa cells in response to IL-1b stimulation ( Figure 1C and Figure 1-figure supplement 1).
Collectively, these results demonstrate that IL-1R/TLR stimulation induces dynamic remodeling of the Regnase-1-associating protein complex from translation machineries to SCF complexes and/or 14-3-3 proteins.
Collectively, these data demonstrate that the IRAK-dependent phosphorylation of Regnase-1 at S494 and S513 is necessary for the association between Regnase-1 and 14-3-3.
These results demonstrate that the binding of Regnase-1 to 14-3-3 and bTRCP occurs independently although IL-1b stimulation simultaneously induces phosphorylation of Regnase-1 at S494 and S513 as well as S435 and S439. In addition, 14-3-3 inhibits the Regnase-1-bTRCP binding.       The S513A mutation destabilizes Regnase-1 protein without affecting target mRNA abundance To evaluate the functional roles of Regnase-1-14-3-3 interaction, we generated Zc3h12a S513A/S513A knock-in mice (Figure 4-figure supplement 1). Zc3h12a S513A/S513A mice did not show gross abnormality, nor did they exhibit alteration in the numbers of T, B cells or macrophages (data not shown). We stimulated mouse embryonic fibroblasts (MEFs) derived from Zc3h12a WT/WT and Zc3h12a S513A/ S513A mice with IL-1b and checked Regnase-1 expression ( Figure 4A). Immunoblot analysis revealed that Regnase-1 was degraded 30 min after stimulation in both WT and S513A mutant MEFs. Following this, Regnase-1 levels increased in WT MEFs at 2 and 4 hr after stimulation ( Figure 4A). Notably, most of the newly synthesized Regnase-1 showed slow migration, consistent with the immunoprecipitation experiment using HeLa cells or RAW264.7 cells shown in Figure 1C and D. On the other hand, the slowly migrating Regnase-1 band did not appear in Zc3h12a S513A/S513A MEFs after IL-1b stimulation. Interestingly, the amount of Regnase-1 at lower bands, which are not the binding target of 14-3-3 ( Figure 2A), was comparable between WT and Zc3h12a S513A/S513A at corresponding time points. Consequently, total Regnase-1 protein expression was severely reduced in Zc3h12a S513A/ S513A MEFs compared with WT after IL-1b stimulation ( Figure 4A). Similar results were also obtained when bone marrow-derived macrophages (BMDMs) and thioglycollate-elicited peritoneal exudate cells (PECs) derived from Zc3h12a WT/WT and Zc3h12a S513A/S513A mice were stimulated with LPS ( Figure 4B-C). Nevertheless, Zc3h12a mRNA levels were comparable between Zc3h12a WT/WT and Zc3h12a S513A/S513A cells ( Figure 4D-F), suggesting that S513A mutation affects the protein stability of Regnase-1. To address this, we examined the kinetics of Regnase-1 degradation following LPS stimulation by treating cells with cycloheximide (CHX). Indeed, Regnase-1-S513A was more rapidly degraded than Regnase-1-WT in PECs after LPS stimulation (
To examine the mechanisms underlying these observations, we developed two mathematical models based on our previous studies (see Materials and methods) (Iwasaki et al., 2011;Mino et al., 2019). The first model (Model 1) assumes that 14-3-3-bound Regnase-1 is unable to degrade its target mRNAs ( Figure 4J). The second model (Model 2) assumes that Regnase-1 binding with 14-3-3 maintains its ability to degrade its target mRNAs to a certain extent ( Figure 4K). Mathematical analysis showed that in Model 2, the abundance of the Il6 mRNAs should be different between Zc3h12a WT/WT and Zc3h12a S513A/S513A cells under the condition that the amount of 14-3-3free Regnase-1 protein (lower bands in Figure 4A-C) is comparable between them. Our observations that the abundance of the target mRNAs did not differ between Zc3h12a WT/WT and Zc3h12a S513A/S513A cells in the late phase of stimulation is inconsistent with Model 2, suggesting that Regnase-1 is inactivated upon binding to 14-3-3.
These results imply that the phosphorylation at S513 and the following association with 14-3-3 nullifies Regnase-1's ability in degrading target mRNAs, although it stabilizes and significantly upregulates the abundance of Regnase-1.

14-3-3 inhibits nuclear import of Regnase-1
We have previously shown that Regnase-1 interacts with CBP80-bound, but not eIF4E-bound, mRNAs (Mino et al., 2019), indicating that Regnase-1 degrades mRNAs immediately after the export from the nucleus to the cytoplasm (Maquat et al., 2010;Müller-McNicoll and Neugebauer, 2013). Although Regnase-1 mainly localizes in the cytoplasm (Mino et al., 2015), we hypothesized Regnase-1 shuttles between the nucleus and the cytoplasm to recognize its target mRNAs in association with their nuclear export. To test this hypothesis, we examined the subcellular localization of Regnase-1 following the treatment with Leptomycin B (LMB), which inhibits CRM1 (also known as Exportin-1)-mediated protein export from the nucleus to the cytoplasm (Yashiroda and Yoshida, 2003). Although Regnase-1 localized in the cytoplasm in the steady state condition, LMB treatment induced rapid accumulation of Regnase-1 in the nucleus within 30 min ( Figure 6A). These results suggest that Regnase-1 dynamically changes its localization between the cytoplasm and the nucleus. Given that Regnase-1 dominantly localizes in the cytoplasm in the steady state conditions, the frequency of its nuclear export seems to be higher than its nuclear import.
CRM1 is known to recognize a nuclear export signal (NES) of a cargo protein for the protein export (Hutten and Kehlenbach, 2007). Thus, we investigated if Regnase-1 harbors a NES. In silico prediction deduced amino acids 433-447 of Regnase-1 as a potential NES with high probability (Xu et al., 2015; Figure 6B-D). Indeed, Regnase-1 lacking 422-451 spontaneously accumulated in the nucleus ( Figure 6E). Since NESs are characterized by hydrophobic residues (la Cour et al., 2003), we also inspected which hydrophobic resides of Regnase-1 were important for the efficient nuclear export of Regnase-1. We found that L440, M444, L447, and W448 of Regnase-1 were critical for the nuclear export of Regnase-1 ( Figure 5E). Noteworthy, all the residues are highly conserved among species ( Figure 5D).
We next examined whether 14-3-3 binding controls the localization of Regnase-1. Interestingly, Regnase-1-ExoSx2 failed to accumulate in the nucleus even after LMB treatment while Regnase-1-WT and -ExoSAAAx2 accumulated in the nucleus by LMB treatment ( Figure 6F). This result indicates that Regnase-1-ExoSx2 is unable to translocate into the nucleus like Regnase-1-WT. Taken together, 14-3-3 inhibits the nuclear import of Regnase-1 as well as its binding to target mRNAs.
While bTRCP regulates the abundance of Regnase-1 through protein degradation, 14-3-3 regulates the activity of Regnase-1. We found that 14-3-3-bound Regnase-1 failed to associate with mRNAs, indicating that 14-3-3 prevents Regnase-1 from recognizing target mRNA. We have previously shown that an RNase domain and an adjacent zinc finger domain play an important role in Regnase-1-RNA binding (Yokogawa et al., 2016). However, the 14-3-3-binding site of Regnase-1 is in the C-terminal part of Regnase-1, which is distant from RNase and zinc finger domains. Therefore, 14-3-3 is unlikely to inhibit Regnase-1-mRNA binding by simple competition between 14-3-3 and mRNAs for the RNA binding domain of Regnase-1. We have previously reported that Regnase-1 interacts with CBP80-bound, but not eIF4E-bound, mRNAs, indicating that Regnase-1 recognizes its target mRNA before or immediately after the nuclear export of the mRNA (Mino et al., 2019). In this study, we found that Regnase-1 shuttles between the nucleus and the cytoplasm while 14-3-3bound Regnase-1 cannot enter the nucleus. Thus, it is tempting to speculate that Regnase-1 recognizes mRNA in the nucleus and induce mRNA decay during pioneer rounds of translation immediately after the nuclear export (Maquat et al., 2010;Müller-McNicoll and Neugebauer, 2013).

Figure 6 continued
Immunofluorescence analysis of HeLa cells transiently expressing FLAG-Regnase-1-WT or indicated mutants treated with LMB (10 ng/ml) for 1 hr. (G) Model of 14-3-3-and bTRCP-mediated regulation of Regnase-1. In the steady state, Regnase-1 shuttles between the nucleus and the cytoplasm and degrades target mRNAs such as Il6. Under IL-1b or TLR-ligands stimulation, two different regulatory mechanisms suppress the activity of Regnase-1 not to disturb proper expression of inflammatory genes; bTRCP induces protein degradation of Regnase-1 and 14-3-3 inhibits nuclear-cytoplasmic shuttling and mRNA recognition of Regnase-1. In (A), (E), and (F), white scale bars indicate 20 mm.
Nevertheless, further investigation is required to clarify the mechanisms of Regnase-1-mediated mRNA decay depending on its nuclear-cytoplasmic shuttling. In addition, further studies are also necessary to clarify the molecular mechanisms how 14-3-3 controls the nuclear-cytoplasmic shuttling of Regnase-1 as well as how it regulates Regnase-1-mRNA binding by exploiting systems beyond the ExoS sequence-mediated interaction with 14-3-3.
Among the molecules involved in MyD88-dependent signaling, we found that IRAK1/2 are potent inducers of the interaction between Regnase-1 and 14-3-3, thereby abrogating Regnase-1-mediated mRNA decay. We also found that kinase-inactive IRAK1 failed to induce the Regnase-1-14-3-3 complex, suggesting that the kinase activity of IRAK1 is required for the phosphorylation of Regnase-1 at S494 and S513. However, previously identified substrate sequence motifs of IRAK1, pSxV, and KxxxpS  do not match the sequence surrounding S494 and S513 of Regnase-1 ( Figure 2F). Although the motif analysis does not exclude the possibility of direct phosphorylation of Regnase-1 by IRAK1, it is possible that kinases activated by IRAK1/2 phosphorylates Regnase-1 at S494 and S513.
In summary, Regnase-1 interactome analysis revealed dynamic 14-3-3-mediated regulation of Regnase-1 in response to IL-1b and TLR stimuli. Since recent studies identified Regnase-1 as a highpotential therapeutic target in various diseases (Kakiuchi et al., 2020;Nanki et al., 2020;Wei et al., 2019), our findings may help maximize the effect of Regnase-1 modulation or provide an alternative way to control the activity of Regnase-1.

Mice
Zc3h12a-deficient mice have been described previously (Matsushita et al., 2009). Zc3h12a S513A/ S513A knock-in mice were generated using CRISPR/Cas9-mediated genome-editing technology as previously described (Fujihara and Ikawa, 2014). Briefly, a pair of complementary DNA oligos was annealed and inserted into pX330 (Addgene plasmid # 42230) (Cong et al., 2013). The plasmid was injected together with the donor single strand oligo into fertilized eggs of C57BL/6J mice. Successful insertion was confirmed by direct sequencing. All mice were grown under specific pathogen-free environments. All animal experiments were conducted in compliance with the guidelines of the Kyoto University animal experimentation committee (Approval number: MedKyo21057).

Reagents
Recombinant cytokines, TLR ligands, and chemical compounds were listed in the key resources table.

Plasmids
For the expression of FLAG-tagged proteins, pFLAG-CMV2 (Sigma) was used as a backbone. For the expression of HA-or Myc-tagged proteins, the FLAG sequence of pFLAG-CMV2 was replaced by HA-or Myc-sequence. Mouse Zc3h12a cDNA was inserted into these vectors as previously described (Matsushita et al., 2009). The coding sequences of 14-3-3 and bTRCP were amplified by using cDNAs derived from HeLa cell as templates and inserted into vectors above using In-Fusion HD Cloning Kit (Takara Bio). For Myc-IRAK1 expression vector, coding sequence of IRAK1 derived from HA-IRAK1 expression vector (Iwasaki et al., 2011) was used. For the mouse Il6 expression vector, the EGFP sequence in pEGFP-C1 was replaced with Il6 gene.
Deletions or point mutations were introduced using the QuikChange Lightning Site-Directed Mutagenesis Kit (Agilent) or In-Fusion HD Cloning Kit.

Plasmid transfection
Plasmids were transfected to HeLa cells or HEK293T cells using Lipofectamine 2000 (Invitrogen) or PEI max (Polysciences) respectively according to manufacturer's instructions.

Generation of doxycycline-inducible FLAG-HA-Regnase-1-expressing HeLa cells
HeLa cells expressing FLAG-HA-Regnase-1 in a doxycycline-dependent manner were generated by lentiviral transduction. To produce lentivirus, HEK293T cells were transfected with pInducer20-puro-FLAG-HA-Regnase1 together with third generation lentiviral packaging vectors. 6 hr after the transfection, the medium was changed to fresh medium and then the cells were incubated at 37˚C for 48 hr. After the incubation, the medium containing lentivirus was harvested and filtrated through 0.45 mm filter. HeLa cells were incubated with the virus-containing medium at 37˚C for 24 hr, followed by 48-hr-incubation with fresh medium. The transduced cells were selected by 0.5 mg/ml of puromycin (InvivoGen). Single clones were picked and evaluated for their expression of FLAG-HA-Regnase-1 in a dox-dependent manner by immunoblotting.

Knockout of IRAK1 and IRAK2
HeLa cells were transfected with two pX459 plasmids which contains gRNA sequence for IRAK1 and IRAK2. As the negative control, empty pX459 plasmid was transfected. Forty-eight hr after the pX459 transfection, puromycin (2 mg/ml) was added to the medium. After 48 hr selection with puromycin, the same number of cells were seeded to new dishes and incubated in fresh media without antibiotics for 48 hr. Knockout efficiency was check by immunoblotting using WCL samples.

DSP-crosslinking
Doxycycline-inducible FLAG-HA-Regnase-1-expressing HeLa cells were treated with doxycycline (1 mg/ml, Sigma) and incubated at 37˚C for 4 hr before the DSP-crosslinking. As a negative control, cells were incubated without doxycycline, and for the IL-1b-stimulated sample, cells were stimulated with human IL-1b (10 ng/ml, R and D Systems) 2 hr before the crosslinking. After the incubation, cells were washed twice with pre-warmed PBS, and then incubated in PBS containing 0.1 mM DSP (TCI) at 37˚C for 30 min. After crosslinking, cells were washed once with pre-warmed PBS and incubated in STOP solution (PBS containing 1 M Tris-HCl pH 7.4) at room temperature for 15 min. Cells were then washed with ice-cold PBS twice, followed by cell lysis and immunoprecipitation.
For DSP-crosslinked samples, cells were lysed in IP buffer (20 mM Tris-HCl pH 7.4, 150 mM NaCl, and 0.5% (vol/vol) NP-40) with cOmplete Mini EDTA-free (Sigma), PhosSTOP (Sigma), and 200 U/ml of Benzonase (Millipore) and incubated on ice for 10 min. The lysates were centrifuged at 15,000 rpm for 5 min and the supernatants were incubated with anti-FLAG-antibody-bound Dynabeads at 4C with rotation for 2 hr. The beads were then washed with IP buffer three times and incubated in FLAG-elution buffer (100 mg/ml FLAG peptides (Sigma), 50 mM Tris-HCl pH7.4, and 150 mM NaCl) at 4˚C with rotation for 10 min twice. Eluted proteins were then immunoprecipitated using anti-HAantibody-bound Dynabeads at 4˚C with rotation for 2 hr. After the second immunoprecipitation, the beads were washed three times with IP buffer and the proteins were eluted in Urea elution buffer (8 M Urea and 50 mM Tris-HCl pH 8.0). The samples were stored at -80˚C until trypsin digestion. Proteins were reduced with 10 mM dithiothreitol (Fujifilm Wako) for 30 min, alkylated with 50 mM iodoacetamide (Fujifilm Wako) for 30 min in the dark, diluted fourfold with 50 mM ammonium bicarbonate (ABC) buffer, and then trypsin digestion was performed. After overnight incubation, digestion was stopped by adding trifluoroacetic acid (TFA) (Fujifilm Wako) to a final concentration of 0.5%. The peptide mixture solution was desalted with SDB-XC StageTips (Rappsilber et al., 2007). The eluates were dried and resuspended in 200 mM 2-[4-2(2-hydroxyethyl)-1-piperazine]ethanesulfonic acid (HEPES) pH 8.5, mixed with 0.1 mg of TMT10-plex labeling reagents (Thermo Fisher Scientific) dissolved in 5 mL acetonitrile (ACN), and incubated for 1 hr at room temperature. The reaction mixtures were quenched by adding hydroxylamine (Sigma) to give a final concentration of 0.33%. After 15 min incubation, the samples were acidified with trifluoroacetic acid, diluted to 5% ACN, and desalted using SDB-XC StageTips. Peptides were dried, resolved in 5 mM ABC buffer and fractionated with a C18-StageTip. Peptides were eluted with 5 mM ABC containing acetonitrile (12.5%, 15%, 17.5%, 20%, 22.5% and 80%) in step gradient manner. Totally six fractions were obtained and analyzed by LC/MS/MS.
For the identification of phosphorylation sites of Regnase-1, HeLa cells expressing FLAG-HA-Regnase-1 or FLAG-Regnase-1 were stimulated with IL-1b (10 ng/ml) or IL-17A (50 ng/ml) respectively for 4 hr. The cells were washed with ice-cold PBS twice and lysed in IP buffer with cOmplete Mini EDTA-free and PhosSTOP. Regnase-1 was immunoprecipitated using anti FLAG antibody as described above and eluted from Dynabeads in SDS sample buffer (50 mM Tris-HCl pH 6.8, 2% (wt/ vol) SDS, 15% (vol/vol) 2-mercaptoethanol, 10% (vol/vol) glycerol and bromophenol blue), followed by incubation at 95˚C for 5 min. Regnase-1was isolated by electrophoresis and the pieces of the gel containing Regnase-1 was stored at 4˚C until trypsin digestion. The gels were de-stained for 30 min with 200 mL of 50 mM ABC / 50% ACN. Then the gels were dehydrated by the addition of 100% ACN. Proteins were reduced with 500 mL of 10 mM dithiothreitol / 50 mM ABC for 30 min, alkylated with 50 mM iodoacetamide / 50 mM ABC for 30 min in the dark. The gels were washed two times with 200 mL of 0.5% acetic acid / 50% methanol. After washing, gels were re-equilibrated with 50 mM ABC, and subsequently dehydrated by the addition of 100% ACN. 10 mL of trypsin solution (10 ng/mL in 50 mM ABC) was added to gel pieces and incubated for 5 min. Another 50 mL of 50 mM ABC buffer was added to gel samples and incubated at 37˚C for overnight. After that, elastase (Promega) (150 ng/mL in water) was added to the final concentration of 7.5 ng/mL and incubated for 30 min at 37˚C (Dau et al., 2020). Digestion was stopped by the addition of 5 mL of 10% TFA. The supernatants were recovered into fresh Eppendorf tubes, and two additional extraction steps were performed with 50% ACN / 0.1% TFA and 80% ACN / 0.1% TFA. The peptides in the supernatants were dried, resuspended in 0.1% TFA, and desalted using SDB-XC StageTips.
For detecting protein-protein binding, cells were lysed in IP Buffer with cOmplete Mini EDTA-free and PhosSTOP and immunoprecipitated as described above using indicated antibodies. The proteins were eluted in the mixture of IP Buffer and SDS sample buffer (2:1) and incubated at 95˚C for 5 min.
For detecting protein-RNA binding, cells were lysed in IP Buffer with cOmplete Mini EDTA-free and RNaseOut (Invitrogen) and immunoprecipitated as described above using indicated antibodies. Some of the precipitates were eluted in the mixture of IP Buffer and SDS sample buffer (2:1) to elute proteins and the others were eluted in TRIzol Reagent (Invitrogen) for RNA isolation.

Database searching and data processing
For DSP-crosslinked samples, peptides were identified with Mascot version 2.6.1 (Matrix Science) against the sequence of Mouse Regnase-1 in addition to the human database from UniprotKB/ Swiss-Prot release 2017/04 and with a precursor ion mass tolerance of 5 ppm and a product ion mass tolerance of 20 ppm. Carbamidomethyl (C), TMT6plex (K) and TMT6plex (N-term) were set as a fixed modification, oxidation (M) was allowed as a variable modification, and up to two missed cleavages are allowed with strict Trypsin/P specificity. Identified peptides were rejected if the Mascot score was below the 95% confidence limit based on the identity score of each peptide. The quantification of peptides was based on the TMT reporter ion intensities in MS2 spectra. Protein quantitative values were calculated by summing the corresponding peptide intensity values. Only proteins with at least two unique peptides were used for further analysis.
For the identification of phosphorylation sites of Regnase-1, peptides were identified with Mascot version 2.7.0 against the sequence of mouse Regnase-1 with a precursor ion mass tolerance of 5 ppm and a product ion mass tolerance of 20 ppm. Carbamidomethyl (C) was set as a fixed modification, oxidation (M) and phosphorylation (STY) were allowed as variable modifications, and up to two missed cleavages are allowed with semitrypsin specificity. Identified peptides were rejected if the Mascot score was below the 99% confidence limit based on the identity score of each peptide. The label-free quantification of peptides was based on the peak area in the extracted ion chromatograms using Skyline-daily software version 21.0.9.118 (MacLean et al., 2010). The peak area ratios between stimulated and non-stimulated samples were calculated, log-scaled, and normalized by the median. For quantitation of phosphosites, the peak area ratios of all monophosphopeptides containing the phosphosites of interest were averaged. Phosphosite localization was evaluated with a sitedetermining ion combination method based on the presence of site-determining y-or b-ions in the peak lists of the fragment ions, which supported the phosphosites unambiguously (Nakagami et al., 2010).
Protein-protein interaction network of the Regnase-1-associating proteins (Log 2 fold change over negative control > 2) was analyzed using STRING database (Szklarczyk et al., 2019) and visualized in Cytoscape (Shannon et al., 2003). Keratins contaminated in the samples were omitted from the analysis.

l-protein phosphatase (lPP) treatment
HeLa cells transiently expressing HA-14-3-3e were stimulated with or without IL-1b (10 ng/ml) for 4 hr and lysed in IP Buffer. Some of the lysates were used in immunoprecipitation as described above. The proteins were eluted using 250 mg/ml of HA peptides as described above. The lysate and the precipitates were treated with Lambda Protein Phosphatase (NEB) according to manufacturer's instructions. For the lPP-negative samples, the same amount of IP Buffer was added instead of lPP.

Immunoblotting
Cells were lysed in IP Buffer or RIPA buffer (1% (vol/vol) NP-40, 0.1% (wt/vol) SDS, 1% (wt/vol) sodium deoxycholate, 150 mM NaCl, 20 mM Tris-HCl pH 8.0, and 10 mM EDTA) with cOmplete Mini EDTA-free and PhosSTOP. The lysates were incubated on ice for 5 min and centrifuged at 15,000 rpm for 5 min. The supernatants were mixed with SDS sample buffer (2:1) and incubated at 95˚C for 5 min. SDS-PAGE was performed using e-PAGEL 7.5% or 5~20% (ATTO) and the proteins were transferred onto 0.2 mm pore size Immun-Blot PVDF membranes (Bio-Rad), followed by blocking with 5% skim milk. The antibodies used in immunoblotting were listed in the key resources table. Luminescence was detected with Amersham Imager 600 (cytiva) and the images were analyzed with Fiji (Schindelin et al., 2012).

RNA isolation and RT-qPCR
Cells were lysed in TRIzol Reagent, and the RNA was isolated according to manufacturer's instructions. For the isolation of the RNA precipitated with Regnase-1, RNA was isolated using RNA Clean and Concentrator-5 (Zymo Research). RNA was reverse transcribed by using ReverTra Ace (TOYOBO). cDNA was amplified by using PowerUp SYBR Green Master Mix (Applied Biosystems) and measured with StepOnePlus Real-Time PCR System (Applied Biosystems). To analyze mRNA expression, each RNA level was normalized with 18S or ACTB. The primers used in qPCR were listed in Supplementary file 1.

RNA sequencing
PECs were harvested from Zc3h12a WT/WT and Zc3h12a S513A/S513A mice as described above. PECs were stimulated with LPS (100 ng/ml) for indicated time and the RNA was collected and isolated using TRIzol Reagent. cDNA library was prepared using NEBNext Ultra RNA Library Prep Kit for Illumina (NEB) and sequenced on NextSeq 500 System (Illumina) according to the manufacturer's instructions. Acquired data was analyzed using Galaxy (Afgan et al., 2018). Briefly, identified reads were mapped on the murine genome (mm10) using HISAT2 (paired end, unstranded) (Galaxy Version 2.1.0), and the mapped reads were counted using featureCounts (Galaxy Version 1.6.3).

ELISA
Cytokine concentration was measured by using IL-6 Mouse Uncoated ELISA Kit (Invitrogen) according to manufacturer's instructions. Luminescence was detected with iMark Microplate Reader (Bio-Rad).
Luciferase assay 5xNF-kB firefly luciferase reporter vector, Renilla luciferase vector, and IRAK1-expressing vector were transfected in HeLa cells and the luciferase activity was measured by using PicaGene Dual Sea Pansy Luminescence Kit (TOYO B-Net). NF-kB activation was calculated by normalizing Firefly luciferase activity with Renilla luciferase activity.

Mathematical model
We developed two dynamical models for the inflammation system regulated by Regnase-1 based on different assumptions of the functions of 14-3-3-bound Regnase-1.

Model 1
In the first model, we assumed that the 14-3-3-bound Regnase-1 does not have the function of degrading its target mRNAs ( Figure 4J). The ordinary differential equations are given as follows: where x 1 , x 2 , x 3 , and x 4 is the abundance of Il6 mRNA, Zc3h12a mRNA, Reg1 Protein, and 14-3-3bound Reg1 Protein, respectively; k 1 and k 2 is the transcription rate constant of Il6, and Zc3h12a, respectively; k 3 is the translation rate constant of Zc3h12a; d 1 and d 2 is the Reg1-induced degradation rate constant of Il6 mRNA and Zc3h12a mRNA, respectively; d 3 , d 4 , and d 5 is the Reg1-independent degradation rate constant of Reg1 protein, Il6 mRNA, and Zc3h12a mRNA, respectively; d 6 is the ubiquitin-dependent degradation rate constant of Reg1 protein; d 7 is the binding rate constant of Reg1 protein to 14-3-3; d 8 is the natural degradation rate constant of 14-3-3-bound Reg1 protein; d 9 is the dissociation rate constant of Reg1 from 14-3-3. signal t ð Þ is the strength of TLR stimulation, which is given as the following form (Mino et al., 2019): signal t ð Þ ¼ s base if 0 t t delay À Á ; s input À s base t raise t À t delay À Á þ s base if t delay t t delay þ t raise À Á ; s input if t delay þ t raise t t delay þ t raise þ t pulse À Á ; s input À s base À Á Â exp À Àtðt delay þ t raise þ t pulse Þ t delay þ sinput ðif t>t delay þ t raise þ t pulse Þ
To determine which model is consistent with the experimental observations, we focus on the experimental findings that there was no difference in the abundance of Il6 mRNA, Zc3h12a mRNA, and Reg1-protein (without 14-3-3 bound) between Zc3h12a WT/WT and Zc3h12a S513A/S513A cells in the late phase of stimulation ( Figure 4A,B,D and E). We will show that in Model 2 (Equation 1.3), the abundance of the Il6 mRNAs should be different between Zc3h12a WT/WT and Zc3h12a S513A/S513A cells under the condition that amount of the 14-3-3-free Reg1 protein is comparable between them.

Analysis of the equilibrium Lemma 1
For Zc3h12a WT/WT cells, there exists only one nonnegative (biologically meaningful) equilibrium of the system (Equation 1.3) if and only if d 3 þ d 6 s input þ d 7 s input À d7d9sinput d7sinputþd9 ! 0. If d 3 þ d 6 s input þ d 7 s input À d7d9sinput d7sinput þd9 <0, there is no equilibrium. For Zc3h12a S513A/S513A cells, there always exists only one nonnegative (biologically meaningful) equilibrium.
It follows from (1.6) that (1.7c) It is easy to see that the quadratic Equation (1.7a) has a nonnegative solution. IfX SA 2 ! 0, it follows from (Equation 1

Consistency with the experiments
The experimental observation shows that there was no difference in the abundance of Reg1 protein between Zc3h12a WT/WT and Zc3h12a S513A/S513A cells at the late phase of stimulation ( Figure (1.9a)