Ubiquitin-binding domain in ABIN1 is critical for regulating cell death and inflammation during development

ABIN1 is a polyubiquitin-binding protein known to regulate NF-κB activation and cell death signaling. Mutations in Abin1 can cause severe immune diseases in human, such as psoriasis, systemic lupus erythematosus, and systemic sclerosis. Here, we generated mice that disrupted the ubiquitin-binding domain of ABIN1 (Abin1UBD/UBD) died during later embryogenesis owing to TNFR1-mediated cell death, similar to Abin1−/− mice. Abin1UBD/UBD cells were rendered sensitive to TNF-α-induced apoptosis and necroptosis as the inhibition of ABIN1UBD and A20 recruitment to the TNF-RSC complex leads to attenuated RIPK1 deubiquitination. Accordingly, the embryonic lethality of Abin1UBD/UBD mice was rescued via crossing with RIPK1 kinase-dead mice (Ripk1K45A/K45A) or the co-deletion of Ripk3 and one allele of Fadd, but not by the loss of Ripk3 or Mlkl alone. Unexpectedly, Abin1UBD/UBD mice with the co-deletion of Ripk3 and both Fadd alleles died at E14.5. This death was caused by spontaneous RIPK1 ubiquitination-dependent multiple inflammatory cytokines over production and could be rescued by the co-deletion of Ripk1 or Tnfr1 combined with Ifnar. Collectively, these data demonstrate the importance of the ABIN1 UBD domain, which mediates the ABIN1-A20 axis, at limiting RIPK1 activation-dependent cell death during embryonic development. Furthermore, our findings reveal a previously unappreciated ubiquitin pathway that regulates RIPK1 ubiquitination by FADD/Casp8 to suppress spontaneous IKKε/TBK1 activation.

INTRODUCTION ABIN1, also referred to as TNF-α-induced protein 3-interacting protein 1, is a polyubiquitin-binding protein that has been implicated in the regulation of cell death, inflammation, and immunity, and has been linked to multiple human inflammatory diseases, including psoriasis, psoriatic arthritis, SLE, and systemic sclerosis [1][2][3][4][5][6]. ABIN1 is a member of the ABIN homology domain (AHD) protein family that mediates A20 binding and polyubiquitin binding. ABIN1 contains a ubiquitin-binding domain (UBD) that binds to polyubiquitin and polyubiquitinated proteins and a NEMO binding domain (NBD) at the C-terminus, both of which have been demonstrated to participate in the inhibition of NF-κB activation [7,8]. In particular, the ABIN1 inhibitory functions on NF-κB activation and antiviral signaling have been reported to rely on its polyubiquitin chain binding ability mediated by the UBD domain, as point mutations in this domain could abolish its inhibitory functions [7,8]. Moreover, ABIN1 knock-in mice harboring a mutation within the UBD domain (D485N, homologous to human D472N) were viable and developed severe inflammation in multiple organs after 5-6 months [9], in sharp contrast to the embryonic lethality observed in Abin1 knockout mice [10]. Given that the binding to K63-or M1-polyubiquitin chains is disrupted by the D485N mutation [9], the polyubiquitin binding ability of ABIN1 is not essential for embryonic development. Therefore, the molecular mechanism by which ABIN1 function is regulated to prevent developmental defects remains unclear.
Besides its role in regulating cell death and NF-kB signaling, ABIN1 has been shown to modulate antiviral signaling pathways by cooperating with A20 [8,[17][18][19]. Previous studies have revealed that ABIN1 can recruit TAX1BP1 and A20 to assemble the A20 regulatory complex by sensing Lys-63-linked polyubiquitin chains via its UBD, and subsequently inhibit type I IFNs production mediated antiviral signaling by disrupting the interactions between TRAF2 and TBK1/IKKε to attenuate TBK1/IKKε polyubiquitination [8]. Furthermore, ABIN1 heterozygous mouse has enhanced antiviral response in vivo by potentially promoting the expression of key viral pattern recognition molecules, which is partially mediated by RIPK1 kinase activity [17]. These studies indicate that ABIN1 plays critical roles in regulating antiviral responses. However, the mechanisms by which ABIN1 regulates the physiological antiviral signaling pathway remains unknown.
In this study, we generated knock-in mice with a disrupted ABIN1 polyubiquitin-binding domain (Abin1 UBD/UBD ). Abin1 UBD/UBD mice resembled Abin1 −/− mice and died at a later embryogenesis stage owing to cell death mediated by TNFR1 and the kinase activity of RIPK1. Such finding indicates that the function of the UBD domain of ABIN1 is indispensable for embryonic development. Furthermore, we found that the embryonic lethality of Abin1 UBD/UBD mice was rescued by the co-deletion of Ripk3 and one allele Fadd, while the deletion of Ripk3 or Mlkl alone had no effect. Surprisingly, the ablation of Ripk3 and two alleles of Fadd from Abin1 UBD/UBD mice resulted in embryonic lethality at E14.5, owing to spontaneous RIPK1 ubiquitination-dependent multiple inflammatory cytokines over production. Remarkably, the lethality of Abin1 UBD/UBD Fadd −/− Ripk3 −/− mice can be rescued by the co-deletion of Tnfr1 and Ifnar. Thus, these results demonstrate the significant contributions of the ABIN UBD domain in regulating cell death and IKKε/TBK1 pathway mediated the production of multiple inflammatory cytokines during development. Fig. 1 The ABIN1 UBD mutants caused embryonic lethality due to aberrant TNFR1-mediated cell death. a Embryo or offspring counts from different periods for Abin1 UBD/+ mice intercrosses. Asterisk symbol (*) Indicated abnormal embryos. b Representative images of embryos with the indicated genotypes from E14.5 to E17.5. c Microscopic images of H&E, cleaved Caspase-3, and TUNEL staining in liver sections with indicated genotypes at E16.5 (scale bar, 110 µm for H&E and cleaved Caspase-3 staining, 50 µm for TUNEL staining). d Western blotting analyses of different lysate fractions of whole embryos with the indicated genotypes from E16.5. e Representative macroscopic images of Tnfr1 −/− and Abin1 UBD/UBD Tnfr1 −/− mice at 2 months. f Offspring counts at weaning from crosses of Abin1 UBD/+ Tnfr1 −/− mice.

RESULTS
The UBD domain of ABIN1 is required for the prevention of TNFR1 signaling-induced embryonic lethality Although ABIN1, a ubiquitin-binding protein, is required to restrict NF-κB activation and regulate cell death upon stimulation with TNFα and TLRs [9,10,15], the specific contributions of the UBD domain of ABIN1 in vivo remain unknown. To determine the physiological role of the ABIN1 UBD domain, we generated Abin1 mutant mice with exon 13 deletion at the transcriptional level using the CRISPR/ Cas9 technology (hereafter termed as Abin1 UBD/UBD mice) (Supplementary Fig. 1a), which caused a deletion of the aa452-478 in the UBD domain ( Supplementary Fig. 1b-d). In contrast to the viable Abin1 D485N/D485N mice expressing ABIN1 with a D485N point mutation in the UBD domain, no Abin1 UBD/UBD mice survived to the weaning age following heterozygous intercrossing (Fig. 1a). Examination of pups obtained from timed mating revealed that Abin1 UBD/UBD mice were alive at E16.5, but died perinatally (Fig. 1b), similar to Abin1 −/− mice. Histological analysis of embryos at E16.5 revealed a reduced number of cells in fetal liver, an increase in the apoptotic marker, cleaved Casp3, and increased TUNEL + cells in Abin1 UBD/UBD embryos (Fig. 1c). Furthermore, immunoblot analysis revealed the presence of cleaved PARP and Casp3, markers of apoptosis, in the NP-40 soluble fraction of the Abin1 UBD/UBD embryo lysate. The necroptosis components containing RIPK1, RIPK3, MLKL, and phosphorylated MLKL were also aggregated in the soluble fraction of 6 M urea (Fig. 1d). These findings demonstrate that Fig. 2 The UBD domain of ABIN1 suppresses RIPK1 activation in TNF-RSC. a, b WT and Abin1 UBD/UBD immortal MEFs were treated with Flag-TNF-α (100 ng/ml) for the indicated time, and the TNFR1 complex (TNF-RSC) was immunoprecipitated using anti-Flag beads. RIPK1 ubiquitination and phosphorylation were detected using western blotting. nec-1 was added to the selected samples, as indicated. c WT, Abin1 UBD/UBD , and Abin1 UBD/UBD Ripk1 K45A/K45A immortal MEFs were stimulated with Flag-TNF-α (100 ng/ml) for the indicated time, and the TNFR1 complex was immunoprecipitated using anti-Flag resin. RIPK1 ubiquitination and phosphorylation were detected by western blotting using anti-RIPK1 and anti-p-RIPK1(Ser166) antibodies. d WT and Abin1 UBD/UBD immortalized MEFs were treated with mouse TNF-α (100 ng/ml) and Zvad (20 µM) for indicated time. Necrosome (complex II) was immunoprecipitated using the anti-RIPK1 antibody, and the components were analyzed by western blotting with the indicated antibodies. For (a-d) asterisk symbol (*) indicates non-specific bands. e The cell survival of WT and Abin1 UBD/UBD immortal MEFs given the indicated stimulations for 12 and 24 h. P values were determined using a two-tailed Student's t test, *P < 0.05; **P < 0.01; ***P < 0.001. f, g WT and Abin1 UBD/UBD immortalized MEFs were stimulated by TS (f) or TSZ (g) at the indicated time, and the apoptosis and necroptosis signaling components were analyzed by western blotting using specific antibodies. Fig. 3 ABIN1 recruits A20 to TNF-RSC by recognizing and binding to the ubiquitinated RIPK1. a, c, e TNF-RSC was pulled down by Flag beads in immortalized MEFs using the indicated genotypes and analyzed by western blot using anti-RIPK1, anti-ABIN1, anti-A20, and anti-SHARPIN. b, d WT MEFs were pretreated with BV6 (0.5 µM) or Nec-1 (30 µM) for 6 h. TNF-RSC was analyzed by western blot using a specific antibody. f TNF-RSC was immunoprecipitated using anti-Flag beads in WT and Abin1 UBD/UBD immortalized MEFs stimulated by flag-TNF-α (100 ng/ml) for the indicated time. g Immortalized WT and Abin1 UBD/UBD MEFs were stimulated with TNF-α (20 ng/ml) for the indicated time. K63linked and M1-linked proteins were immunoprecipitated using K63-TUBEs and M1-TUBEs, respectively. both apoptosis and necroptosis occurred simultaneously in the Abin1 UBD/UBD embryos.
TNF-α signaling has been reported to be required for perinatal death induced by Abin1 knockout [10,20]. To determine whether the perinatal lethality of Abin1 UBD/UBD mice is also mediated by TNFR1 signaling, we crossed Abin1 UBD/UBD mice with Tnfr1 −/− animals. Knocking out Tnfr1 could overcome perinatal lethality in Abin1 UBD/UBD mice (Fig. 1e, f), confirming that TNFR1 signaling is also essential for perinatal death in Abin1 UBD/UBD mice. Therefore, in contrast to Abin1 D485N/D485N mice harboring a D485N point mutation in the UBD domain of ABIN1, which is fertile, the perinatal death of Abin1 UBD/UBD mice was found to be caused by the activation of TNFR1 signaling, demonstrating the essential role of the ABIN1 UBD domain in the regulation of embryonic development.
ABIN1 UBD domain is essential for restricting RIPK1 activation-mediated cell death To study the putative contributions of the ABIN1 UBD domain to different signaling pathways involving ABIN1, we generated Abin1 UBD/UBD mouse embryonic fibroblasts (MEFs) and treated them with TNF/Smac (TS) or TNF/CHX (TC) to induce apoptosis and TNF/Smac/zVAD (TSZ) or TNF/CHX/zVAD (TCZ) to induce necroptosis. Abin1 UBD/UBD MEFs were found to be more sensitive to TS-or TC-induced apoptosis and TSZ-or TCZ-induced necroptosis compared to WT MEFs ( Supplementary Fig. 2a, b). In contrast, Abin1 UBD/UBD MEFs cells displayed similar sensitivity to the Staurosporine (STS)-induced endogenous pathway of apoptosis and NF-κB activation stimulated by TNF-α to WT MEFs (Supplementary Fig. 2c, d). These results suggest that the UBD domain of ABIN1 is critical for restricting TNF-α-mediated apoptosis and necroptosis, but is not essential for NF-κB activation and the endogenous apoptosis pathway. Nec-1, a kinase inhibitor of RIPK1 [21], was also found to strongly inhibit cell death caused by various stimuli, including TNF/Smac, TNF/CHX, TNF/5z7, TNF/TPCA, and those combined with ZVAD in Abin1 UBD/UBD MEFs (Supplementary Fig. 3a-f). Such findings suggest that ABIN1 prevented RIPK1 kinase-mediated cell death through its UBD domain. We characterized the signaling complexes in TNF-α-treated Abin1 UBD/ UBD MEFs. In TNF-α-stimulated Abin1 UBD/UBD MEFs, the ubiquitination of RIPK1 significantly increased and the level of p-Ser166 RIPK1 was elevated at the late time points (Fig. 2a). Furthermore, increased p-Ser166 RIPK1, but not enhanced ubiquitination of RIPK1, was inhibited in the presence of Nec-1 or the genetic kinase inactive mutant, RIPK1 K45A (Fig. 2b, c). Such finding suggests that the ABIN1 UBD domain is required for restricting the ubiquitination of RIPK1, which precedes RIPK1 phosphorylation-dependent activation. We proceeded to determine whether complex II formation and cell death increased in TNF-α-treated Abin1 UBD/UBD MEFs. The components of TNF-α-induced complex II, including p-RIPK1, RIPK1, p-RIPK3, RIPK3, and p-MLKL, were found to be significantly elevated in Abin1 UBD/UBD MEFs compared to those in WT MEFs (Fig. 2d), resulting in the sensitization to TNF-α-induced apoptosis and necroptosis ( Fig. 2e-g). Therefore, the UBD domain of ABIN1 is essential for inhibiting RIPK1 ubiquitination, activation, and complex II formation to prevent apoptosis and necroptosis.
The ABIN1 UBD domain supports the recruitment of A20 to TNF-RSC by recognizing and binding to the ubiquitinated RIPK1 In the absence of RIPK1, ABIN1 could not be recruited into TNF-RSC, resulting in the reduced recruitment of A20 (Fig. 3a). To further investigate how ABIN1 regulates RIPK1 ubiquitination and activation-mediated cell death through its UBD domain, we hypothesized that the recruitment of ABIN1 UBD and A20 to TNF-RSC could be affected when RIPK1 ubiquitination is abolished. CIAP1/2 is a ubiquitin ligase that catalyzes RIPK1 ubiquitination and can be degraded by BV6 treatment [22,23]. As expected, MEFs pretreated with BV6 had reduced levels of ubiquitinated RIPK1 and reduced recruitment of ABIN1 and A20 in TNF-RSC (Fig. 3b), consistent with the results using Smac pretreatment [15]. These findings suggest that ABIN1 recruitment depends on the RIPK1 ubiquitination status. In RIPK1 K376R mutant MEFs, which had defective RIPK1 ubiquitination [24][25][26], we observed the same reductions and further confirmed the effect of RIPK1 ubiquitination on the TNF-RSC recruitment of ABIN1 (Fig. 3c). Treatment with the RIPK1 kinase inhibitor (Nec-1) or genetic kinase-dead mutant of RIPK1 (Ripk1 K45A/K45A ) had no effect on the ubiquitination of RIPK1 and the recruitment of ABIN1 and A20 in TNF-RSC (Fig. 3d, e). Hence, these results suggest that ABIN1 recruits A20 to TNF-RSC by binding to ubiquitinated RIPK1, which are upstream events of RIPK1 kinase activation. To investigate whether the recruitment efficiency of A20 to TNF-RSC was regulated by the ABIN1 UBD, we compared the TNF-RSC components in Abin1 UBD/UBD MEFs to those in WT MEFs after the cells were treated with flag-TNF-α. We found that the mutant ABIN1 UBD could not be recruited to TNF-RSC, resulting in the reduced recruitment of A20 and phosphorylated A20. However, we observed that there was no effect on the recruitment of other TNF-RSC components (including TAK1, TRADD, cIAP1/2, NEMO, and SHARPIN) (Fig. 3f), indicating recruitment of these proteins to the TNF-RSC before ABIN1-A20, which consistent with results of undisturbed NF-κB activation in Abin1 UBD/UBD MEFs upon TNFα stimulation ( Supplementary Fig. 2d). Phosphorylated A20 is the catalytically active form of A20, which hydrolyzes and removes ubiquitin from ubiquitylated RIPK1 in TNF-RSC [13,14,27,28]. Particularly, we further observed that both M1 chains and K63 chains of TNF-RSC showed mildly increased upon TNF stimulation in ABIN1 UBD mutant cells compared to those of control cells ( Supplementary Fig. 2e). Through tandem ubiquitin-binding entity (TUBE)-pull down, we found that the levels of K63 ubiquitin chains and linear (M1-linked) ubiquitin chains on RIPK1 significantly increased in TNF-αstimulated Abin1 UBD/UBD MEFs compared to those in WT MEFs (Fig. 3g), indicating that RIPK1 is the major deubiquitinated target of ABIN1 at the TNF-RSC. Taken together, these results suggested that the ABIN UBD domain is essential for ABIN1's function in recruiting A20 to TNF-RSC, thereby inhibiting RIPK1 activationmediated cell death.
Abin1 UBD/UBD mice died from RIPK1 kinase or FADD/RIPK3dependent apoptosis and necroptosis Although Abin1 deficiency promotes RIPK1 kinase activity mediated both apoptosis and necroptosis in vitro, the perinatal lethality of Abin1-deficient mice was rescued by crossing RIPK1 kinase-dead mutant mice (Ripk1 D138N/D138N ) or crossing RIPK3 knockout mice alone [15]. These findings suggest that blocking RIPK3-dependent necroptosis is sufficient to recover perinatal death of Abin1-deficient mice. As Abin1 UBD/UBD mice died perinatally and resembled Abin1deficient mice, we used a genetic strategy to unravel the mechanisms whereby ABIN1 regulates RIPK1 activation and cell death through its UBD domain. The genetic inactivation of RIPK1 (Ripk1 K45A/K45A ) prevented the lethality of Abin1 UBD/UBD mice (Fig. 4a,  b). Further, the statistical results of the offspring at weaning aligned with Mendel's genetic law (Fig. 4a). Accordingly, Abin1 UBD/ UBD Ripk1 K45A/K45A MEFs were resistant to cell death induced by both apoptotic and necroptotic stimuli ( Fig. 4d-f, Supplementary Fig. 3g-i).

DISCUSSION
Multiple genetic studies have strongly associated the genetic polymorphisms of A20 (TNFAIP3) and its binding partner, ABIN1 (TNIP-1), with the incidence and severity of multiple inflammatory and autoimmune diseases [41][42][43][44]. ABIN1 contains a UBD domain that binds to polyubiquitin and polyubiquitinated proteins and a NEMO binding domain (NBD) at the C-terminus. Recent reports have demonstrated that A20 and ABIN1 synergistically regulate NF-κB signaling and cell death pathway [7,10,15,45]. However, the physiological mechanisms by which the UBD domain of ABIN1 regulates these pathways remains unknown.
Mice expressing ABIN1 mutants D485N within UBD domain were reported to be viable, but displayed features of autoimmunity [9,46,47], in sharp contrast to the embryonic lethality of Abin1 −/− mice. As the functions of the ABIN1 UBD domain might not be entirely destroyed by these mutations, the role of the UBD domain of ABIN1 in embryonic development remains unclear. In this study, we generated an ABIN1 mutant mice with the deletion of the UBD domain at positions 452-478 by scissoring exon 13 using the CRISPR/Cas9 technology. The Abin1 UBD/UBD mice died from excessive fetal liver cell death and hypoplasia, resembling Abin1-deficient mice. However, the mechanism underlying this difference between ABIN1485 and ABIN UBD remains unclear, though both of them could disrupt polyubiquitin-binding ability of ABIN1. Given other functions of ABIN1 including interaction with A20 might also be affected differently by these two mutants, it is likely other components remain to be characterized.
Furthermore, we demonstrated that the UBD domain of ABIN1 is essential for preventing embryonic lethality caused by TNFR1 signaling (Fig. 1e, f). We showed that Abin1 UBD/UBD MEFs were more sensitive to TS-or TC-induced apoptosis and TSZ-or TCZ-induced necroptosis (Supplementary Figs. 2a-c, 3a-d). In addition, we found that the mutant ABIN1 UBD could not be recruited to TNF-RSC, resulting in the reduced recruitment of A20 and phosphorylated A20, which leads to RIPK1 ubiquitination, activation, and complex II formation-mediated apoptosis and necroptosis ( Supplementary Fig. 9b). Thus, the ABIN1 UBD domain is critical for the assembly and recruitment of A20 to TNF-RSC by recognizing and binding to the ubiquitinated RIPK1, thereby preventing RIPK1 activation-mediated apoptosis and necroptosis.
Collectively, our studies uncovered a novel mechanism whereby the ABIN1 UBD domain regulates RIPK1 kinase-mediated cell death and RIPK1 kinase-independent production of multiple cytokines to enable proper development. These new findings reveal an additional mechanism whereby ABIN1 is implicated in the pathogenesis of related human diseases through its UBD domain.

MATERIALS AND METHODS Mice
All mice were housed in a specific pathogen-free (SPF) facility at Shanghai Institute of Nutrition and Health, Chinese Academy of Sciences. Fadd −/− mice were gifted by Dr. Jianke Zhang (Thomas Jefferson University, Philadelphia, USA), and Rip3 −/− mice were provided by Dr. Xiaodong Wang (NIBS, Beijing, China). Ripk1 −/− , Ripk1 K45A/K45A , Mlkl −/− , Tnfr1 −/− , and Ifnar −/− mouse lines have been described previously [24,51]. A20 −/− mice were generated using the CRISPR/Cas9 mutation system (Bioray Laboratories Inc., Shanghai, China). Abin1 UBD/UBD mice were generated by a 110 bp deletion in the coding region of exon 13 and following the intron region using the CRISPR-Cas9 mutation system. Abin1 UBD/UBD mouse genotyping primers (ABIN1-F: 5′-TCAGTGCAGCGGTAATGAGT-3′ and ABIN1-R: 5′-ATACATGCAGGCAGAACACT-3′) amplified 414 bp wild-type and 304 bp ABIN1 UBD DNA fragments. All newly constructed mouse lines were backcrossed onto the C57BL/6 background for more than eight generations. Animal experiments were conducted in accordance with the guidelines of the Institutional Animal Care and Use Committee of the Shanghai Institute of Nutrition and Health, University of Chinese Academy of Sciences, Chinese Academy of Sciences.
For TNF-RSC immunoprecipitation, immortalized MEFs were stimulated with flag-TNF-α (100 ng/ml) for the indicated time. Cells were washed twice with ice-cold 1× PBS and subsequently lysed in NP-40 buffer at 4°C for 1 h with rotation. The lysates were cleared by centrifugation and then incubated with FLAG-tagged beads (Sigma, A2220) for 16 h. Complex II was immunoprecipitated by overnight incubation with RIPK1 (BD, 610459) antibody at 4°C with rotation, followed by 6 h incubation with protein A/G resin (Thermo Scientific, 53133). The beads were washed with ice-cold NP-40 buffer at least three times. Proteins were eluted with 40 µl 2× SDS loading sample buffer.

Cell viability assay
Cells were seeded to 80% confluence in 96-well plates and treated with 20 ng/ ml TNF-α, 30 µM Nec-1, 20 µM zVAD, and other stimuli at the indicated time. Cell viability was measured using the CellTiter-Glo Luminescent Cell Viability Assay kit (Promega, G7572), according to the manufacturer's instructions.

Statistics and reproducibility
The data presented in this article were obtained independently at least thrice. The statistical significance of data was evaluated by unpaired twotailed Student's t test. Statistical calculations and graphs were generated using the GraphPad Prism 8 software.