Main

Members of three families presented with a previously undescribed autoinflammatory disorder characterized by fevers and pronounced lymphadenopathy beginning in early childhood and continuing throughout adulthood (Fig. 1a) From birth or shortly thereafter, all affected individuals experienced fevers usually occurring approximately every 2–4 weeks, lasting 1–7 days, and reaching temperatures as high as 40–41 °C. Some individuals reported extreme chills, severe headaches, and/or hallucinations that coincided with their fevers. These flares were accompanied by intermittent episodes of cervical, axillary, inguinal and/or periaortic lymphadenopathy that often caused pain or discomfort (Fig. 1b, Table 1). Several individuals experienced splenomegaly and/or hepatomegaly, which were generally more prominent early in life, as well as oral ulcers, arthralgia or gastrointestinal symptoms such as abdominal pain, nausea, diarrhoea, constipation, loss of appetite or weight loss (Table 1). Patient 7 (P7) exhibited a more chronic inflammation with acute exacerbation. Study participants often had increased levels of inflammatory markers even during symptom-free periods. In contrast to some more severe autoinflammatory disorders, there were no signs of rash, arthritis, genital ulcers or end-stage organ damage and the condition was not life-threatening in any of the patients (Table 1).

Fig. 1: Heterozygous mutations of the RIPK1 caspase-8 cleavage site cause autoinflammatory disease.
figure 1

a, Affected individuals (filled symbols) in three families carried mutations in RIPK1 D324. Crossed symbol indicates a deceased individual. b, Axial (top) and coronal (bottom) planes of abdominal computerized tomography scans of participant P1 at age 11, after 2 months on tocilizumab but before substantial resolution of symptoms, revealing periaortic lymphadenopathy (arrows), splenomegaly (14 cm craniocaudal length), and liver at upper limit of normal (16 cm craniocaudal length). c, Serum cytokine levels of two P7 samples taken within 1 week, both during infliximab and before tocilizumab treatment, and four unrelated adolescent controls (ctrl). Dots are from technical duplicates for each time point. Graphs show mean. d, RNA sequencing of whole-blood RNA from P7 (two time points, as in c) and two unrelated adolescent unaffected controls, both with technical duplicates. Heat map shows differentially expressed inflammatory response genes (GO: 0006954). For gene names, see Supplementary Fig. 1. e, Response to tocilizumab infusion in P1. Erythrocyte sedimentation rate (ESR), C-reactive protein (CRP), haemoglobin and mean corpuscular volume (MCV) were measured serially before and after the start of tocilizumab treatment (grey shading). Time after the initial evaluation of this subject at age 10 years is depicted on the x axis. Horizontal lines indicate high values (ESR and CRP) or high and low values (haemoglobin and MCV) for the subject age-specific laboratory reference ranges for these markers. RBC, red blood cell count. f, RIPK1 DNA sequence chromatograms show heterozygous single-base substitutions. g, WebLogo demonstrating conservation of the caspase-8 cleavage tetrapeptide motif in RIPK1 (human numbering) in 184 vertebrate species. h, In vitro caspase assays on wild-type (WT) and RIPK1 mutants. Western blots are representative of two independent experiments. For gel source data, see Supplementary Fig. 2.

Full size image
Table 1 Clinical features of patients with CRIA syndrome
Full size table

Lymphocyte counts were normal between flares in the seven affected participants (Extended Data Table 1). However, pro-inflammatory cytokines were increased in the serum from P7 when inflamed but not during a flare (Fig. 1c). Transcriptomic analysis from P7 whole-blood RNA revealed an enrichment of several inflammatory gene signatures (Fig. 1d, Extended Data Fig. 1a, b). Affected members of family 2 had all taken prednisone during flares, with varying degrees of acute relief but without long-term prevention of future episodes (Table 1). Participants P1, P6 and P7 had tonsillitis (Table 1), but tonsillectomy did not improve symptoms. Similarly, the IL-1 receptor antagonist, anakinra, and the TNF antagonist, etanercept, did not suppress inflammation in patients P1, P2, P4 or P7 (Table 1). However, treatment with the IL-6 receptor antagonist tocilizumab markedly, and in some cases severely, reduced the severity and frequency of the symptoms of P1, P2, P3, P6 and P7 (Fig. 1e, Table 1, Extended Data Table 2a). Tocilizumab also provided some initial relief to P4, but P4 reported aggravation of pre-existing oral ulcers, and P6 reported eventual onset of hand pain, and both participants elected to discontinue treatment (Table 1).

Identification of pathogenic mutations in RIPK1

Exome sequencing in P1 and her unaffected parents and all eight members of family 2 revealed that RIPK1 was the only gene in which a variant from both families satisfied filtering criteria. A third mutation in RIPK1 was later discovered in family 3. Affected individuals from the three families had different heterozygous missense mutations at the same crucial aspartate residue required for RIPK1 cleavage by caspase-8 (Fig. 1f). The D324N and D324Y mutations occurred de novo in families 1 (Extended Data Fig. 2) and 3, respectively, whereas D324H was inherited in an autosomal dominant pattern in family 2. These mutations are not reported in variant databases (Extended Data Table 2b), and none of the families had rare co-segregating coding or splice mutations in genes previously implicated in autoimmune lymphoproliferative syndrome (ALPS) or other monogenic autoinflammatory disorders. Mutations in the RIPK1 cleavage site were not found in an additional 554 individuals with sporadic unexplained fever, lymphadenopathy, ALPS or idiopathic Castleman disease that we screened by Sanger or targeted hybrid capture sequencing (Extended Data Table 2c). We therefore designated this condition as cleavage-resistant RIPK1-induced autoinflammatory (CRIA) syndrome.

The optimal caspase-8 cleavage motif is highly conserved in vertebrates (Fig. 1g, Extended Data Table 3). RIPK1 can be cleaved by both caspase-6 and caspase-8, yielding products of similar size, although the caspase-6 cleavage site has not been defined9,10,11. Consistent with these reports, RIPK1 mutants found in the patients—as well as the D324A mutant that has previously6 been shown to prevent RIPK1 cleavage by caspase-8—were resistant to both caspase-6 and caspase-8 cleavage in vitro, which suggests that the cleavage sites of caspase-6 and caspase-8 are the same (Fig. 1h).

Lack of RIPK1 cleavage causes embryonic lethality

To investigate the molecular mechanism for CRIA syndrome and characterize the role of RIPK1 cleavage in vivo, we generated RIPK1 cleavage-resistant mice. Rather than choosing one of the disease-associated variants, we mutated the aspartate to alanine. Although the heterozygous Ripk1D325A/+ mice were viable and grossly normal, the homozygous Ripk1D325A/D325A mice died during mid-embryogenesis; much earlier than the postnatal lethality of the Ripk1−/− mice12,13,14,15 (Fig. 2a, Extended Data Fig. 3a). Ripk1D325A/D325A lethality occurred between embryonic day 10.5 (E10.5) and E11.5, with the embryos showing several sites of mild-to-severe haemorrhage beginning in the cephalic vascular plexus, in the midbrain and hindbrain, but ultimately affecting the entire embryo including the pharyngeal arches and the pericardial space (Fig. 2a). At E11.5, all Ripk1D325A/D325A embryos were dead and displayed major haemorrhage in several locations (Fig. 2a, Extended Data Fig. 3a). E10.5 Ripk1D325A/D325A embryos had endocardial cushion hypoplasia, smaller limbs buds and a thinner neural retina (Fig. 2b). These developmental delays might be due to the defective vasculature, associated with extensive cell death observed in the yolk sac of these embryos (Fig. 2c). This phenotype was reminiscent of several strains of knockout mice with defects in TNF signalling, including Casp8−/− mice8,16,17,18,19,20,21,22. The E10.5 lethality of Casp8−/− mice is TNF-dependent12, and can be prevented by loss of either Ripk3 or Mlkl8,22,23, which suggests that the lethality is due to TNF-induced activation of the necroptotic pathway that is normally inhibited by caspase-8. These findings led to the idea that cleavage of RIPK1 by caspase-8 inhibits necroptosis during embryogenesis8,22. However, Ripk1D325A/D325ARipk3−/− mice were not viable, consistent with a previous report24. Nevertheless, loss of RIPK3 extended survival more than loss of MLKL, which indicates that RIPK3 has a non-necroptotic role in the early embryonic lethality (Fig. 2d, Extended Data Fig. 3b). Combined loss of Casp8 and Ripk3 in these mice prevented the embryonic lethality, which suggests that caspase-8 does more than inhibiting RIPK1/RIPK3/MLKL-induced necroptosis at this embryonic stage (Fig. 2d, Extended Data Fig. 3b). Although loss of Ripk1 ameliorates the ALPS-like disorder observed in Casp8−/−Ripk3−/−mice8,12,13,15, lack of RIPK1 cleavage did not notably affect it (Extended Data Fig. 3c, d), consistent with observations in Fadd−/−Ripk3−/−Ripk1D325A/D325A mice24. Interestingly, inhibition of RIPK1 kinase activity also rescued the embryonic lethality of Ripk1D325A/D325A (Fig. 2d). However, Ripk1D138N,D325A/D138N,D325A mice were runty and did not survive past weaning (Fig. 2d, e). These mice had a multi-organ inflammation presenting with skin hyperplasia, infiltration of leukocytes in the liver and the lung, disorganized splenic architecture and scattered cleaved caspase-3-positive cells in these organs (Extended Data Fig. 3e). Loss of one allele of Ripk3 or Casp8 prolonged the survival of Ripk1D138N,D325A/D138N,D325A mice to 5 weeks of age, and complete loss of Ripk3 rescued the inflammatory phenotype of Ripk1D138N,D325A/D138N,D325A mice (Fig. 2d, e).

Fig. 2: Homozygous mutation of the RIPK1 caspase-8 cleavage site in mice causes early embryonic lethality.
figure 2

a, E10.5 (top) and E11.5 (bottom) embryos, representative of four embryos per genotype. FB, forebrain; HB, hindbrain; He, heart; FL, forelimb; MB, midbrain; PA, pharyngeal arches. Arrows denote sites of haemorrhage. Scale bars, 900 μm (E10.5) and 1,400 μm (E11.5). b, Haematoxylin and eosin (H&E)-stained section of E10.5 embryos, representative of three embryos per genotype. Arrows denote endocardial cushions (top) and neural retina (middle). Scale bars, 200 μm. c, E10.5 yolk sacs stained with anti-PECAM1 (cyan) and anti-cleaved caspase-3 (Cl. CASP3; magenta) antibodies. Images with severely and less severely disrupted vasculature are shown. Scale bars, 50 μm. Images are representative of four embryos per genotype. d, Diagram depicting the extent of viability of different strains of Ripk1D325A mice. e, Representative pictures of three mice per genotype numbered in d.

Full size image

RIPK1 cleavage limits TNF-induced cell death

To explore the function of RIPK1 cleavage in TNF signalling, we tested homozygous Ripk1D325A/D325A mouse embryonic fibroblasts (MEFs) for their response to TNF-induced cell death. Notably, even though TNF is not usually cytotoxic, we found that Ripk1D325A/D325A MEFs were sensitive to TNF alone and this induced increased phosphorylation of RIPK1, as well as activation of caspase-8 when compared to wild-type MEFs (Fig. 3a, b). Although inhibiting caspases or RIPK3 kinase activity did not affect cell death induced by TNF, genetic loss of RIPK3 or RIPK1 kinase activity significantly reduced TNF-induced cell death (Fig. 3a, b). Loss of RIPK3 not only completely abrogated death, but also blocked RIPK1 phosphorylation and caspase activation (Fig. 3a, b).

Fig. 3: Ripk1D325A/D325A and Ripk1D325A/+ cells are hypersensitive to TNF-induced death.
figure 3

a, c, Cell death of MEFs, monitored by time-lapse imaging of propidium iodide (PI) staining over 16 h. I denotes 5 µM caspase-8 inhibitor; N denotes 10 µM necrostatin; NT denotes untreated; R denotes 1 µM RIPK3 inhibitor; S denotes 100 nM SMAC mimetic; T denotes 100 ng ml−1 (a) or 10 ng ml−1 (c) TNF. Graphs are representative of four independent experiments performed with two biological repeats per genotype. b, d, Western blot of MEFs treated as in a for 2 h (b), and as in c for 2 h (d). Results are representative of two independent experiments. p-RIPK1, phosphorylated RIPK1. β-Actin was used as a loading control. For gel source data, see Supplementary Fig. 2.

Full size image

Given that the patients contain RIPK1 mutations in only one allele, we tested the sensitivity of several Ripk1D325A/+ heterozygous cell types to TNF. In contrast to homozygote Ripk1D325A/D325A MEFs, none of the tested Ripk1D325A/+ cell types were sensitive to TNF alone (Extended Data Fig. 4a, b). However, inhibitors that directly activate the cytotoxic activity of RIPK1 (for example, SMAC mimetic, or TAK1, IKK or MK2 inhibitors)1,2,3,4,6,25,26 rapidly sensitized Ripk1D325A/+ MEFs and mouse dermal fibroblasts (MDFs) to low-dose TNF (Fig. 3c, Extended Data Fig. 4a, c). By contrast, only SMAC mimetic and TAK1 inhibitor sensitized Ripk1D325A/+ bone-marrow-derived macrophages (BMDMs) to low-dose TNF (Extended Data Fig. 4b). In Ripk1D325A/D325A MEFs, TNF-induced cell death was more pronounced after the addition of IKK or TAK1 inhibitors or a combination of SMAC mimetic and MK2 inhibitor (Extended Data Fig. 4c). In addition, homozygote and heterozygote Ripk1D325A MEFs and MDFs were slightly more sensitive to apoptosis induced by low-dose TNF and cycloheximide (Extended Data Fig. 4a, c).

Treatment with TNF plus SMAC mimetic induced a strong phosphorylation of RIPK1 and RIPK3, as well as activation of caspase-8 and caspase-3, in Ripk1D325A/+ cells (Extended Data Fig. 4d–f), which was more pronounced in the Ripk1D325A homozygote cells (Fig. 3d, Extended Data Fig. 4f). This increase in cell death induced by TNF plus SMAC mimetic correlated with increased formation of a RIPK1–caspase-8-containing complex 2 (Extended Data Fig. 4g).

Notably, given the increase in caspase-8 activation, loss of RIPK3 markedly delayed cell death induced by TNF plus SMAC mimetic or TAK1, IKK or MK2 inhibitors in both Ripk1D325A homozygote and heterozygote fibroblasts (Fig. 3c, Extended Data Fig. 4a, c). In fibroblasts, loss of RIPK3 correlated with significantly reduced autophosphorylation of RIPK1 and caspase activation after TNF and SMAC mimetic treatment (Fig. 3d, Extended Data Fig. 4d, e). However, inhibition of RIPK3 kinase had little effect on the induction of cell death (Extended Data Fig. 4a–c), which suggests that RIPK3 contributes mostly in a structural capacity to the activation of caspase-8 in Ripk1D325A cells.

We next analysed both Ripk1D138N,D325A homozygote and heterozygote cells and, as expected, genetic loss of RIPK1 kinase activity prevented RIPK1 autophosphorylation (Fig. 3d, Extended Data Fig. 4d). It also provided some protection from cell death and this effect was mirrored by treatment with the RIPK1 inhibitor necrostatin (Fig. 3c, Extended Data Fig. 4a–c). Similar to RIPK3 loss, this correlated with reduced caspase-8 activation (Fig. 3d, Extended Data Fig. 4d). Together, these results indicate that in fibroblasts, RIPK3 promotes caspase-8 activation in a manner that is independent of its kinase activity and mostly independent of RIPK1 kinase activity (Fig. 3a, b), unless RIPK1 is further activated by an activating stimulus, such as SMAC mimetic (Fig. 3c, d).

One surprising observation was that the strong activation of caspase-8 in Ripk1D325A cells led to RIPK1 cleavage (Fig. 3d, Extended Data Figs. 4d, f, 5a). In the case of the heterozygote cells, this was almost certainly due to cleavage of the wild-type protein; however, we also detected a slightly smaller RIPK1 cleavage product in homozygote cells (Fig. 3d, Extended Data Figs. 4f, 5a). This was the result of an alternative cleavage site (D301 in mouse) that is as well-conserved as the canonical site (Extended Data Fig. 5, Extended Data Table 2b). However, possibly owing to the unfavourable hydrophobic amino acid in the P1′ position27, the D301 site was far less efficiently cleaved than the D325 site and only when the canonical site was mutated (Fig. 3d, Extended Data Figs. 4f, 5).

RIPK1 cleavage limits inflammatory responses

Patients with CRIA syndrome have recurrent fevers, so to understand how loss of RIPK1 cleavage might affect the response to inflammatory stimuli, we tested the responsiveness of the Ripk1D325A/+ mice to Toll-like receptor (TLR) ligands. Although there was not a marked difference in levels of IL-6, the levels of TNF and IL-1β were higher in the Ripk1D325A/+ sera after injection of a non-lethal dose of either lipopolysaccharide (LPS) or polyinosinic:polycytidylic acid (poly(I:C)) (Fig. 4a, Extended Data Fig. 6a). Similarly, PBMCs from P7 produced more TNF and IL-1β after LPS or poly(I:C) treatment (Fig. 4b, Extended Data Fig. 6b). Despite these increased levels of cytokines, hypothermia induced by LPS was not life-threatening (Extended Data Fig. 6c), which was also consistent with the symptoms of the patients with CRIA syndrome. BMDMs also produced more TNF after TLR activation (Fig. 4c), which correlated with the amount of cell death induced (Extended Data Fig. 6d).

Fig. 4: RIPK1 cleavage limits inflammation in vivo.
figure 4

a, Serum cytokine levels after 2 h treatment with 2 mg kg−1 LPS. Data are mean ± s.e.m., n = 3 mice for TNF and n = 5 mice for IL-6 and IL-1β. b, Cytokine levels in the supernatant of two unrelated adolescent controls (RIPK1+/+) and P7 RIPK1D324Y/+ PBMCs treated for 3 h with 10 ng ml−1 LPS. Data are mean of triplicates. c, TNF levels in the supernatant of BMDMs treated for 24 h with 25 ng ml−1 LPS or 2.5 μg ml−1 poly(I:C). Data are mean ± s.e.m., n = 3 for Ripk1+/+ and n = 3–4 for Ripk1D325A/+. d, Serum TNF levels in wild-type mice reconstituted with Ripk1D325A/+ haematopoietic cells (left) or Ripk1D325A/+ mice reconstituted with wild-type haematopoietic cells (right), treated for 2 h with 2 mg kg−1 LPS. Data are mean ± s.e.m., n = 3 and 4 Ripk1+/+ → Ripk1+/+, n = 6 Ripk1D325A/+ → Ripk1+/+, n = 3 for Ripk1+/+ → Ripk1D325A/+ mice per genotype. e, Serum cytokines levels after 2 h treatment with 2 mg kg−1 of LPS. Data are mean ± s.e.m., n = 4 for Ripk1D325A/+, n = 5 for the other genotypes. Results in a, c and e are representative of two independent experiments. Each dot in a and c–e represents a mouse. All P values determined by unpaired, two-tailed t-test.

Full size image

To define the contribution of the haematopoietic compartment to the hyper-inflammatory phenotype, we generated bone marrow chimaeras. Notably, both wild-type mice transplanted with Ripk1D325A/+ haematopoietic cells and Ripk1D325A/+ mice transplanted with wild-type bone marrow were hyper-responsive to LPS compared with the controls (Fig. 4d). Although our data suggest that the increased inflammatory response in mice correlates with increased cell death in Ripk1D325A/+ cells, RIPK1 also contributes to the activation of NF-κB and MAPK signalling pathways14,28,29,30. However, loss of RIPK1 cleavage did not affect TNF-induced NF-κB or MAPK activation in either mouse cells or patient-derived dermal fibroblasts (Extended Data Fig. 6e–g). Furthermore, the cytokine increases observed in the Ripk1D325A/+ sera were dependent on RIPK3 and caspase-8, which suggests that cell death is the major contributor to cytokine induction in these mice (Fig. 4e).

RIPK1 has a role in activating NF-κB and MAPK inflammatory pathways, caspase-8-mediated apoptosis and RIPK3-dependent necroptosis. Each of these distinct responses can contribute to inflammatory signalling and it has been difficult to disentangle which pathway causes inflammation in any given physiological situation. We describe a human autoinflammatory disorder caused by heterozygous mutations in RIPK1 seemingly constrained to a single, evolutionarily conserved aspartate residue at the caspase-6/8 cleavage site. Mutation of this key aspartate prevents caspase-6/8 cleavage of RIPK1, sensitizes cells to TNF-induced cell death and causes embryonic lethality in homozygous mice. Several mechanisms inhibit cell death after TNF stimulation1,2,3,4,5,6,7,26,31 and our data emphasize how important this is in limiting an inflammatory response. Pathogens may counter cell-death-mediated inflammation by expressing caspase-8 inhibitors and a cellular defensive mechanism that amplifies the cell death response in the absence of RIPK1 cleavage makes intuitive sense, and may explain why some pathogens also attempt to cleave RIPK132,33. Previously, pathogen inhibition of caspase-8 was thought to unleash the necroptotic pathway; however, RIPK1 cleavage not only limits necroptosis, as previously assumed, but can also limit caspase-8-mediated apoptosis. Furthermore, the kinase activities of RIPK3 and RIPK1 have mainly been thought of as activators of necroptosis. However, the rescue of the postnatal lethal phenotype of the Ripk1D138N,D325A mice by loss of Casp8 or Ripk3 reveals a far more complex interaction between these molecules than previously anticipated. Our data provide support for the concept of a hierarchy of preferred responses to TNF signalling: cell survival, then caspase-8-mediated apoptosis, with necroptosis as a last resort (Extended Data Fig. 7). Notably, despite the fact that most of our knowledge of RIPK1 function comes from analyses of TNF signalling, and that TNF has a pivotal role in many inflammatory diseases, patients with CRIA syndrome responded to the IL-6 inhibitor tocilizumab but did not respond to TNF inhibitors. It will be interesting therefore to determine what role RIPK1 has in IL-6-mediated inflammation.

Methods

Participant enrolment

Families were enrolled and evaluated in the Clinical Center at the National Institutes of Health under a protocol approved by the Institutional Review Board of the National Institute of Diabetes and Digestive and Kidney Diseases and the National Institute of Arthritis and Musculoskeletal and Skin Diseases. Human studies complied with relevant ethical regulations and all participants provided written informed consent. No statistical methods were used to predetermine sample size.

Tocilizumab treatment

P1 was 11 years old at the time of her first intravenous infusion of tocilizumab at a dose of 8 mg kg−1. She initially received medication every 3 weeks but later reduced the frequency to every 4 or 5 weeks because of a busy school schedule. On the less frequent dosing, P1 had more breakthrough symptoms, mainly tender lymphadenopathy. In 2018, when the US FDA approved the use of the injectable form in children with juvenile idiopathic arthritis, P1 was switched to 162 mg by subcutaneous injection every 2 weeks and did very well on this regimen. P2,P3, P4 and P6 received regular self-administered tocilizumab by 162 mg subcutaneous injections starting at every 2 weeks—the standard dose and route of administration for adults. The dose frequency for P3 was gradually increased to every 6 days. P7 received an initial infusion of tocilizumab at 8 mg kg−1 before being switched to the subcutaneous injectable form (162 mg every 2 weeks) for convenience. On this regimen, P7 noted prompt resolution of fevers, abdominal pain and joint pain, and gradual normalization of laboratory testing, including CRP, ESR, haemaglobin, haematological indices and serum iron.

Exome sequencing

Exome capture (Illumina TruSeq v2 for family 1, Roche SeqCap EZ Exome+UTR for family 2, and IDT xGen Exome Research Panel for family 3) and sequencing (Illumina HiSeq 2000, 2500 and NovaSeq 6000) was performed for all available family members at the National Institutes of Health (NIH) Intramural Sequencing Center (NISC) using 2 × 101-, 2 × 126-, and 2 × 151-base-pair (bp) paired-end reads. The data were analysed as follows: alignment with Novoalign; duplicate marking with Picard; re-alignment, re-calibration, and variant calling with GATK; and annotation with Annovar. Variants were filtered to select those that were nonsynonymous or in splice sites within 6 bp of an exon, had less than 1% mutant allele frequency in variant databases, and co-segregated with the phenotype. The mutations were validated by Sanger sequencing in all family members, and to rule out non-paternity, non-maternity or other sample identity errors, genders and relatedness were confirmed by examining heterozygous call rates on the X chromosome, Y chromosome call rates and Mendelian inheritance error rates in the exome data.

In vitro cleavage assays

Unlabelled in vitro transcription and translation of 1 µg of empty pCMV6-Entry control vector (Origene), wild-type RIPK1 cDNA cloned into pCMV6-Entry vector (Origene), p.D324N, p.D324H, p.D324Y and p.D324A mutant RIPK1 constructs (GENEART Site-Directed Mutagenesis System, Invitrogen) was performed in a 50-µl reaction using the TnT T7 Quick Coupled Transcription/Translation System (Promega). We incubated 2 µl of this reaction with either 12 U of purified recombinant caspase-8 (Calbiochem), 12 U of purified recombinant caspase-6 (Calbiochem), or an equal volume of re-suspension buffer, in caspase reaction buffer from the Caspase-8 Fluorometric Assay Kit (Enzo Life Sciences) and 10 mM dithiothreitol (DTT) in a 40 µl final volume at 37 °C for 3 h. These reactions were blotted for RIPK1 using an antibody recognizing a RIPK1 C-terminal antibody (610459, BD Transduction Laboratories).

RNA sequencing

Total RNA was isolated from whole blood collected in PAXgene Blood RNA Tubes using PAXgene Blood RNA Kit (PreAnalytiX) as per the manufacturer’s instructions. Total RNA was used for cDNA library preparation using the TruSeq Stranded mRNA Library Preparation kit for NeoPrep (Illumina). Sequencing was performed on an Illumina HiSeq 3000 System in a 1 × 50-bp single-read mode. Sequenced reads were mapped against the human reference genome (GRCh38) using hisat v.2.2.1.034. Reads mapped to haemoglobin genes were removed from further analysis. Mapped reads were quantified using HTSeq35,36. All the count data were normalized using TCC37 and differentially expressed genes were detected using edgeR38. Gene Ontology enrichment analysis was performed using DAVID36.

Mice

All mouse studies complied with relevant ethical regulations and approved by the Walter and Eliza Hall Institute Animal Ethics Committee. The Ripk1D325A and Ripk1D138N,D325A mice were generated by the MAGEC laboratory (WEHI) on a C57BL/6J background. To generate Ripk1D325A mice, 20 ng μl−1 of Cas9 mRNA, 10 ng μl−1 of sgRNA (ATTTGACCTGCTCGGAGGTA) and 40 ng μl−1 of the oligo donor (tgtcttctcattacagAAAGAGTATCCAGATCAAAGCCCAGTGCTGCAGAGAATGTTTTCACTGCAGCATGCCTGTGTACCATTACCTCCGAGCAGGTCAAATTCAGgtaactcacctattcgttcatttgcatactcgctca) (in which uppercase bases denote exons; lowercase bases denote intron sequences) were injected into the cytoplasm of fertilized one-cell stage embryos generated from wild-type C57BL/6J breeders. To generate Ripk1D138N,D325A mice, 20 ng μl−1 of Cas9 mRNA, 10 ng μl−1 of sgRNA (TGACAAAGGTGTGATACACA) and 40 ng μl−1 of oligo donor (GGATAATCGTGGAGGCCATAGAAGGCATGTGCTACTTACATGACAAAGGTGTGATACACAAGAACCTGAAGCCTGAGAATATCCTCGTTGATCGTGACTTTCACATTAAGgtaatccacaatctg) were injected into the cytoplasm of fertilized one-cell stage embryos generated from Ripk1D325A/D325ARipk3−/−Casp8−/+ breeders. Twenty-four hours later, two-cell stage embryos were transferred into the uteri of pseudo-pregnant female mice. Viable offspring were genotyped by next-generation sequencing. Targeted animals were backcrossed twice to wild-type C57BL/6J to eliminate off-target mutations and to re-integrate Ripk3 and Casp8 genes into Ripk1D138N,D325A mice. The Ripk3−/− mice39, Casp8−/− mice19 and Mlkl−/− mice40 were all previously described. The Ripk3−/− mice were backcrossed to C57BL/6J mice for more than ten generations.

TLR challenge

Eight-to-twelve-week-old male mice received intraperitoneal injection of either 2 mg kg−1 LPS or 50 μg poly(I:C). Calculations to determine group sizes were not performed, mice were not randomized but were grouped according to genotype, and experiments were blinded.

Cells

MEFs were isolated from E10.5 embryos and MDFs were isolated from mouse tails. After SV40 transformation, MEFs and MDFs were tested for mycoplasma. 293T cells (ATCC) used to produce SV40 viruses and in Extended Data Fig. 5b were tested for mycoplasma but not authenticated.

Time-lapse imaging

Percentage cell death was assayed every 30–45 min by time-lapse imaging using the IncuCyte live cell analysis imaging (Essenbioscience) or the Opera Phenix High Content Screening System (PerkinElmer) for 16 h with 5% CO2 and 37 °C climate control. For the IncuCyte and Opera Phenix imaging, dead cells were identified by propidium iodide (0.25 μg ml−1) staining, and for the Opera Phenix imaging, all cells were stained with 250nM of SiR-DNA (Spirochrome). Dyes were added to the cells 2 h before imaging and compounds were added 10 min before the start of imaging. For the Opera Phenix imaging, images were analysed using the server-based Columbus 2.8.0 software (PerkinElmer) to identify nuclei based on SiR-DNA staining and dead cells using propidium iodide staining. Results were exported as counts per well to be processed and graphed using R Studio (https://www.R-project.org/) with the tidyverse package (https://CRAN.R-project.org/package=tidyverse).

Human and mouse cytokines measurement

Human serum and PBMC supernatant cytokine content was measured by enzyme-linked immunosorbent assay (ELISA) (R&D: SLB50, STA00C and S6050) according to the manufacturer’s instructions. The measurements were performed in technical duplicates. Student’s t-test was performed for the statistical analysis. Mouse serum and BMDMs supernatant cytokine content was measured by ELISA (eBioscience for TNF and IL-6 and R&D for IL-1β) according to the manufacturer’s instructions.

Human PBMC ex vivo stimulation

Ficoll-isolated human PBMCs were serum-starved for 20 min and stimulated for 3 h with LPS (Invivogen, tlrl-3pelps) or 6 h with poly(I:C) (Invivogen, tlrl-pic). Cytokines were measured by ELISA as described above.

Reagents

The SMAC mimetic compound A, the caspase inhibitor IDN-6556 (Idun Pharmaceuticals) and the RIPK1 inhibitor necrostatin were synthesized by TetraLogic Pharmaceuticals. The RIPK3 inhibitor GSK’872 was from Calbiochem. The TAK1 inhibitor (5Z)-7-oxozeaenol, the IKK inhibitor IKK-16 and the MK2 inhibitor PF-3644022 were from Tocris Bioscience. Cycloheximide was from Sigma. Recombinant Fc-TNF was produced in house. Ultrapure LPS-EB and poly(I:C) were purchased from Invivogen.

Immunostaining

Embryonic yolk sacs were fixed for 20 min at room temperature in 4% paraformaldehyde, blocked and permeabilized in PBS with 2% normal donkey serum (Jackson ImmunoResearch, 017-000-121) and 0.6% Triton X, probed with primary antibodies, cleaved caspase-3 (9661, CST) and PECAM1 (AF3628, R&D Systems) at 4 °C overnight, then secondary antibodies goat anti-rabbit AF488 (Invitrogen A-11008) and donkey anti-goat cy3 (705-165-147, Jackson ImmunoResearch) at room temperature for 1 h. Samples were cleared in a glycerol gradient (5–80%) overnight, whole-mounted in 80% glycerol and imaged using a DP72 microscope and cellSens Standard software (Olympus).

Immunoprecipitation

Ten million cells were seeded in 10-cm dishes. After the indicated treatments, cells were lysed in DISC lysis buffer (150 mM sodium chloride, 2 mM EDTA, 1% Triton X-100, 10% glycerol, 20 mM Tris, pH 7.5). Proteins were immunoprecipitated with 20 μl of protein G Sepharose plus 1.5 μg of FADD antibody (clone 7A2, in house) with rotation overnight at 4 °C. Beads were washed four times in DISC and samples eluted by boiling in 60 μl 1× SDS loading dye.

Western blotting

Cells lysates were separated on 4–12% gradient SDS–polyacrylamide gels (Biorad), transferred to polyvinylidene fluoride (Millipore) membranes and blotted with indicated antibodies purchased from CST except for phospho-RIPK3 (a gift from Genentech), actin (Sigma) and FADD (clone 7A2, in house). In vitro cleavage assays were blotted with a with an anti-RIPK1 antibody recognizing the C-terminal part (BD Transduction Laboratories, 610459). Cell lysates were blotted with an anti-RIPK1 antibody recognizing the N-terminal part (3493, Cell Signaling Technology).

NF-κB assay in patient-derived cells

NF-κB activation was assessed by measuring nuclear translocation of subunit p65 in fibroblasts derived from a single skin biopsy. Cells were grown overnight in 96-well plates seeded at 16,000 cells per well, and treated for 30 min with TNF (PeproTech) in PBS containing 1 mM CaCl2 and 1 mM MgCl2 (PBS-CM). Cells were pre-fixed for 5 min with 2% paraformaldehyde (PFA) in PBS-CM, then fixed for 10 min with 6% PFA in PBS-CM, and aldehyde groups were quenched with 50 mM NH4Cl in PBS-CM for 15 min. After permeabilization with 0.3% SDS in PBS-CM for 5 min, cells were incubated with donkey serum dilution buffer (DSDB; 16% donkey serum, 0.3% Triton X-100, and 0.3 M NaCl in PBS) for 30 min, followed by overnight incubation at 4 °C with rabbit monoclonal NF-κB subunit p65 antibody (8242, Cell Signaling Technology) diluted at 1:500 in DSDB. Samples were then washed 3 times with permeabilization buffer (0.3% Triton X-100 and 0.1% BSA in PBS) and incubated with a 1:300 dilution of donkey anti-rabbit secondary antibody coupled to Alexa 488 (A21206, Molecular Probes) in DSDB for 1 h. Nuclei were counter-stained with a 1:2,000 dilution of SYTO 59 (Thermo Fisher) for 15 min. Automated field selection and plate imaging were performed with an IncuCyte Zoom incubator-microscopy system (Essen Bioscience) using a 20× objective. Nine fields per well of four wells per participant were pooled for analysis of nuclear p65 signal intensity. Nuclei were marked in red over a phase-contrast image, and p65 immunofluorescence was labelled in green. Overlaying a p65 mask on a nuclear mask showed both positive and negative nuclei, whereas a yellow co-staining mask showed positive nuclei only.

Reporting summary

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