Embryonic Lethality and Host Immunity of RelA-Deficient Mice Are Mediated by Both Apoptosis and Necroptosis

Three signaling pathways, NF-κB, apoptotic, and necroptotic pathways, could be triggered by TNF (1). Upon TNF binding to its cognate receptor, TNFR1, the intracellular domain of TNFR1 recruits multiple proteins to form the membrane-proximal supramolecular complex containing TRADD, RIP1, TRAF2, TRAF5, and cellular inhibitor of apoptosis proteins 1 and 2 (cIAP1/2) (2). RIP1, as a key switch point, tightly regulates TNF-induced cellular NF-κB, apoptotic, and necroptotic pathways (3, 4). RIP1 contains three functional domains: N-terminal kinase domain, intermediate domain, and C-terminal death domain (5). The N-terminal kinase domain and the C-terminal death domain play essential roles in mediating the apoptosis and necroptosis pathways; however, the intermediate domain is required to activate NF-κB signaling (2).

It has been widely reported that NF-κB signaling negatively regulates apoptosis and necroptosis ex vivo. The ubiquitylation of RIP1 in the intermediate domain functions as a docking site for the TAK1/TAB1/TAB2 complex and IKKα/IKKβ/NEMO complex, both of which are activators of the NF-κB pathway (6–8). The products of NF-κB could directly bind and inhibit caspase activities and promote ubiquitylation-dependent survival signals to form a negative feedback to suppress cell death signals (9–13). Cellular deubiquitinases, such as A20 and CYLD, negatively edit the ubiquitylation of RIP1 (14–16). The deubiquitylated RIP1 cannot mediate the NF-κB signaling pathway, but it recruits FADD and caspase 8 to execute apoptosis or further recruits RIP3/MLKL to execute necroptosis if caspase activity is compromised (17–20).

Impaired NF-κB signaling in vivo, such as that via the deletion of IKKβ, NEMO, or RelA, has shown that these mice could not overcome embryonic development due to massive programmed cell death in multiple tissues (21–25). Moreover, mutant mice bearing an alanine instead of serine at position 276 in RelA displayed significantly reduced NF-κB–dependent transcription and died at embryonic day as the result of various developmental abnormalities (26).

In addition to the roles of NF-κB signaling in embryonic development, it has been shown that NF-κB signaling tightly controls the innate and adaptive immune responses. Deletion of IKKβ, NEMO, or RelA in intestinal epithelial cells resulted in severe intestinal epithelial cell loss associated with intestinal inflammatory diseases (27–29). In addition, the lack of IKKβ, NEMO, or RelA in the liver caused hepatocyte or liver parenchymal cell death associated with injury, steatohepatitis, and hepatocellular carcinoma (30–33).

Interestingly, the embryonic lethality of RelA-deficient mice around embryonic day (E)15.5 cannot be rescued by deletion of the Bcl-2–interacting domain, kinase JNK1, or kinase JNK2 or by the double deletion of Bcl-2–interacting domain and JNK2 (34). However, the ablation of TNF or TNFR1 could rescue RelA-deficient mice from embryonic lethality (35–37). Moreover, targeting TNFR1 also prevented the embryonic death of IKKβ-deficient mice (24). Unexpectedly, all RelA^−/−TNF^−/−, RelA^−/−TNFR1^−/−, and IKKβ^−/−TNFR1^−/− mice cannot survive to maturation, suggesting that the downstream signaling of TNF/TNFR1 was responsible for the embryonic lethality in RelA-deficient mice. However, the mechanism by which TNF/TNFR1 causes embryonic lethality in NF-κB–deficient mice is still poorly understood. Moreover, why RelA^−/−TNF^−/−, RelA^−/−TNFR1^−/−, and IKKβ^−/−TNFR1^−/− mice died before maturation remains a mystery.

In this article, we report that RelA deficiency causes apoptosis and necroptosis in vivo. The coablation of Fadd and either Rip3 or Mlkl or the blockade of RIP1 kinase activity rescues the embryonic death of RelA-deficient mice. Although these mice survive for up to 2 wk, they are susceptible to pneumonia caused by bacterial infection. Antibiotic treatment significantly improves postnatal death and extends the lifespan in RelA^−/−Fadd^−/−Rip3^−/− triple-knockout (TKO) mice. This study provides genetic evidence that RelA-mediated NF-κB activation ensures normal embryonic development through suppressing RIP1 kinase–dependent both FADD-mediated apoptosis and RIP3/MLKL-mediated necroptosis signaling, but it maintains host immune homeostasis independent of the FADD/RIP3/MLKL complex and RIP1 kinase activity.

All mouse studies were approved by the Institutional Animal Care and Use Committee at the Institute for Nutritional Sciences, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences. Heterozygous Fadd^+/− mutant mice were provided by Dr. J. Zhang (Thomas Jefferson University, Philadelphia, PA), and Rip3-knockout mice were a gift from Dr. X. Wang (National Institute of Biological Sciences, Beijing China). RelA^Flox/Flox mice, a gift from Dr. Z.-G. Luo (Institute of Neuroscience, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences), were crossed into EIIa-Cre mice to obtain RelA^+/− mice. Rip1^K45A mutant mice were generated by Bioray Laboratory (Shanghai, China). Mlkl^−/− mice were described previously (38). All mice were housed under similar conditions, and that exposure was equivalent among all knockout mice. The morning of vaginal plug detection was considered E0.5.

A broad-spectrum antibiotic mixture containing 200 mg/l ciprofloxacin, 1 g/l ampicillin, 1 g/l metronidazole, and 500 mg/l vancomycin was added to the drinking water of the mother 1 wk before the birth of offspring.

We used mouse TNF-α (aa 80-235; R&D Systems), Z-VAD(OH)-FMK (Z-VAD; 14467-5, Cayman Chemical; and HY-16658, Med Chem Express), Necrostatin-1 (BML-AP309-0100; Enzo Life Sciences), Cycloheximide (C1988; Sigma), CD3-FITC (eBioscience), B220-PE (eBioscience), CD4-allophycocyanin-Cy7 (552051, BD Bioscience), CD8-PerCP (BioLegend), Gr-1 (eBioscience), Mac-1 (eBioscience), F4/80 (eBioscience), RIP1 (610459; BD Biosciences), RIP3 (Prosci), Caspase 8 (ALX-804-447-C100; Enzo Life Sciences), RelA (4764S; Cell Signaling Technology), PARP1 (9542s; Cell Signaling Technology), MLKL (AP14272b; Abgent), GAPDH (G9545; Sigma).

Embryos isolated from 13.5-d pregnant mice were washed with PBS. The head and visceral tissues were dissected, and the remaining bodies were incubated with 4 ml of Trypsin/EDTA solution (Life Technologies) per embryo at 37°C for 40 min. After trypsinization, an equal amount of medium was added and pipetted up and down a few times. All mouse embryonic fibroblasts (MEFs) were cultured in DMEM supplemented with 10% FBS at 37°C with 5% CO₂.

Cell survival was analyzed with a CellTiter-Glo Luminescent Cell Viability Assay (Promega), according to the manufacturer’s instructions.

MEFs were treated with TNF for the indicated times, and cell death was analyzed with an Annexin V-FITC kit (Beyotime Biotechnology, Shanghai, China) using flow cytometry, according to the manufacturer’s instructions.

MEFs were treated with TNF for 1.5 h and washed twice with PBS, and 1 ml of TRIzol Reagent was added for extraction of total RNA. Reverse transcription was performed with a PrimeScript RT Reagent Kit with gDNA Eraser (catalog number RR047A; Takara), according to the manufacturer’s instructions. Quantitative PCR was performed using SYBR Premix Ex Taq (catalog number RR420A; Takara), according to the product manual. The primer sequences used for real-time PCR were as follows: cFLIP_L: 5′-GTGTCTGCCGAGGTCATT-3′ and 5′-CAGCCAGGTTCTCAGTCA-3′; cFLIP-_S: 5′-GTGTCTGCCGAGGTCATT-3′ and 5′-CAGCCAGGTTCTCAGTCA-3′; cIAP1: 5′-CAGAGCACCGCAGACATT-3′ and 5′-GTCCTCAATCGAGCAGAGTG-3′; cIAP2: 5′-CGCAGCCCGTATTAGAAC-3′ and 5′-AGATTCCCAGCACCTCAG-3′; and Actin: 5′-CATCACTATTGGCAACGAGC-3′ and 5′-ACGCAGCTCAGTAACAGTCC-3′.

Cells were collected, washed with ice-cold PBS twice, resuspended in lysis buffer (20 mM Tris-HCl [pH 7.4], 150 mM NaCl, 10% glycerol, 1% Triton X-100, 1 mM Na₃VO₄, 25 mM β-glycerol-phosphate, 0.1 mM PMSF, protease and phosphatase inhibitors), incubated on ice for 20 min, and centrifuged at 12,000 × g for 20 min. For immunoprecipitation of RIP1, 2 μg of Ab was incubated with protein A agarose and cell lysates at 4°C overnight. The next day, beads were washed with lysis buffer three times.

E15.5 and E16.5 embryos were isolated and fixed in 4% paraformaldehyde. After dehydration, embryos were embedded in paraffin. H&E and immunohistochemistry staining for histological analysis was performed using standard protocols.

Lymphocytes were isolated from blood, spleen, thymus, and lymph nodes. RBCs were lysed using ACK lysis buffer. A single-cell suspension was used for staining cell surface markers following standard protocols, and data acquisition was performed using a FACSAria II cytometer (BD). Flow cytometric data were analyzed with FlowJo software.

MEFs were treated with TNF for 1.5 h and washed with PBS twice, and 1 ml of TRIzol Reagent was added for extraction of total RNA. BasePair Biotechnologies (Suzhou, China) performed the RNA sequencing and analyzed the data.

GraphPad Prism 6.0 software was used for all statistical analyses. Data are presented as mean ± SD. The Student t test was performed to assess the means of two groups. Statistically significant differences are indicated in the figures: *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.005, and ****p ≤ 0.001. The log-rank test was used for the comparison of survival curves.

The pleiotropic cytokine TNF can induce various cellular responses, from NF-κB signaling to apoptosis and necroptosis, reflecting an intricate network of signals that is triggered by this molecule upon binding to its cognate receptors. TNF stimulation significantly increased propidium iodide (PI)⁺, annexin V⁺, and PI⁺ annexin V⁺ populations in RelA-deficient MEFs compared with wild-type MEFs (Supplemental Fig. 1A). Furthermore, treatment of RelA-deficient MEFs with TNF resulted in accelerated cell death in a dose- and time-dependent manner (Supplemental Fig. 1B, 1C). Moreover, RelA^−/− MEFs were sensitive to TNF plus Smac mimetics or cycloheximide or to TNF plus pan-caspase inhibitor Z-VAD–induced apoptosis or necroptosis, respectively (Supplemental Fig. 1D, 1E). In addition, we examined the cleavage of poly(ADP)ribose polymerase (PARP) and p-RIP3 in MEFs after TNF plus Smac mimetics and/or Z-VAD treatment. Following treatment for 4 h, cleaved PARP and p-RIP3 were detected in RelA^−/− MEFs (Supplemental Fig. 1F). Together, these results suggested that RelA deficiency caused two forms of cell death in MEFs, apoptosis and necroptosis, after TNF stimulation.

To determine whether apoptosis and/or necroptosis in vivo contributed to embryonic lethality in RelA-deficient mice, we first generated mice lacking both RelA and Fadd or Rip3 or Mlkl. Notably, RelA^−/− mice died as expected around E15.5 (Fig. 1A, 1B). Moreover, the immunofluorescence staining showed enhanced cleaved caspase 3 in RelA^−/− embryos at E15.5 (Supplemental Fig. 1G). These results suggested that RelA deletion promoted cell death in vivo. Interestingly, Fadd deletion accelerated RelA^−/− mice death from E15.5 to E11.5 (Supplemental Fig. 1H, 1I), showing that inhibiting apoptosis could not prevent embryonic lethality of RelA-deficient mice. No RelA^−/−Rip3^−/− or RelA^−/−Mlkl^−/− double-knockout (DKO) mice were observed among the live-born pups (Fig. 1A, Supplemental Fig. 2A), implying that disruption of Rip3 or Mlkl could not rescue the embryonic lethality of RelA-deficient mice. Subsequently, timed mating analysis was performed and showed that RelA^−/−Rip3^−/− and RelA^−/−Mlkl^−/− DKO embryos were present at E13.5 and E15.5 (Fig. 1B, Supplemental Fig. 2B). Gross morphological examination of RelA^−/−Rip3^−/− and RelA^−/−Mlkl^−/− DKO embryos at E15.5 showed blood pooling in the abdomen, and the s.c. vessels were less visible (Fig. 1B, Supplemental Fig. 2B). By E16.5, RelA^−/−Rip3^−/− and RelA^−/−Mlkl^−/− DKO embryos were dead and appeared to have a smaller body size (Fig. 1B, Supplemental Fig. 2B). The most noticeable feature of DKO embryos was massive hemorrhage throughout the entire body, especially in the abdominal region (Fig. 1B, Supplemental Fig. 2B). In most DKO embryos, the formation of blood vessels, including capillaries and other blood vessels in the head, face, tail and footpad, were apparent defects in development (Fig. 1B, Supplemental Fig. 2B). These observations indicated that the disruption of Rip3 or Mlkl could not rescue the embryonic death of RelA-deficient mice.

As reported previously (21), histological H&E staining showed that the cell junctions and cell numbers were greatly reduced in the liver, lung, and skin in RelA^−/− embryos at E15.5 (Fig. 1C). The cell junctions remained unbroken and the cell numbers were relatively unchanged in RelA^−/−Rip3^−/− and RelA^−/−Mlkl^−/− embryo sections at E15.5, but liver sections displayed marked bleeding (Fig. 1C, Supplemental Fig. 2C). However, by E16.5, the cell junctions and numbers in all RelA^−/−Rip3^−/− and RelA^−/−Mlkl^−/− DKO embryos were significantly reduced, as observed in RelA^−/− embryos at E15.5 (Fig. 1D, Supplemental Fig. 2C). Together, these results indicated that ablation of Rip3 or Mlkl partially rescued and extended RelA^−/− embryo survival to a limited degree.

RIP3 and MLKL mainly function in necroptosis but not in apoptosis. Thus, we reasoned that apoptosis still existed in RelA^−/−Rip3^−/− and RelA^−/−Mlkl^−/− MEFs after TNF treatment. TNF stimulation notably increased PI⁺, annexin V⁺, and PI⁺annexin V⁺ populations in RelA^−/−Rip3^−/− and RelA^−/−Mlkl^−/− MEFs compared with control MEFs (Supplemental Fig. 2D, 2E). In addition, DKO MEFs were still sensitive to TNF or TNF plus Smac mimetics–induced cell death, and this could be completely blocked by adding the pan-caspase inhibitor Z-VAD (Supplemental Fig. 2F, 2G). Moreover, cleaved PARP was still detected after TNF or TNF plus Smac mimetic treatment (Supplemental Fig. 3A, 3B). Furthermore, cleaved caspase 3 was markedly enhanced in RelA^−/−Rip3^−/− and RelA^−/−Mlkl^−/− embryos at E16.5 (Fig. 1E, Supplemental Fig. 3C). These results implied that the ablation of Rip3 or Mlkl blocked necroptosis in RelA^−/− embryos, but apoptosis remained, an event that might contribute to the embryonic death of RelA^−/−Rip3^−/− and RelA^−/−Mlkl^−/− embryos. Together, these data indicated a requirement for RelA in preventing RIP3-dependent necroptosis and caspase-dependent apoptosis pathways ex vivo.

To test whether the embryonic lethality in RelA^−/−Rip3^−/− mice was a result of apoptosis in vivo, we intercrossed RelA^+/−Fadd^−/−Rip3^−/− mice to obtain RelA^−/−Fadd^−/−Rip3^−/− TKO mice. Timed mating analysis showed that there was no embryonic developmental abnormality between RelA^−/−Fadd^−/−Rip3^−/− mice and control littermates at E13.5 and E16.5 (Fig. 2A). There were also no obvious differences in body size between RelA^−/−Fadd^−/−Rip3^−/− embryos and control littermates, and no detectable hemorrhage was found in RelA^−/−Fadd^−/−Rip3^−/− embryos at E13.5 or E16.5 (Fig. 2A). Moreover, angiogenesis in the brain, face, tail, and footpad of RelA^−/−Fadd^−/−Rip3^−/− embryos was also normal (Fig. 2A). Histological staining showed that there was no detectable bleeding, cell junctions, or number loss in liver, lung, or skin sections of RelA^−/−Fadd^−/−Rip3^−/− embryos at E16.5 (Fig. 2B). These results suggested that ablation of Fadd and Rip3 prevented massive hemorrhage and the failure of blood vessel formation that were seen in RelA^−/− and RelA^−/−Rip3^−/− and RelA^−/−Mlkl^−/− DKO embryos at E15.5 or E16.5.

Remarkably, we observed live RelA^−/−Fadd^−/−Rip3^−/− neonates at postnatal day (P)1 and P7 (Fig. 2C, 2D, Supplemental Fig. 3D). There were no visible weight differences between RelA^−/−Fadd^−/−Rip3^−/− TKO and littermate mice at P1 and P7 (Fig. 2C, Supplemental Fig. 3D). Histological analysis of RelA^−/−Fadd^−/−Rip3^−/− mice and control littermates at P7 revealed no visible developmental abnormity in the brain, liver, intestine, heart, kidney, spleen, skin, or stomach (Fig. 2E). These results suggested that the deletion of Fadd and Rip3 fully prevented embryonic lethality in RelA-deficient mice, indicating that the interplay among RelA, FADD, and RIP3 was tightly regulated in vivo during embryonic development.

FADD and RIP3 are crucial regulators of the apoptosis and necroptosis pathways. Thus, we reasoned that there was no cell death in RelA^−/−Fadd^−/−Rip3^−/− TKO MEFs. To that end, we isolated primary MEFs from RelA^+/+Fadd^−/−Rip3^−/− and RelA^−/−Fadd^−/−Rip3^−/− embryos at E13.5 and analyzed their responsiveness to various stimuli. As expected (consistent with data obtained in vivo), the sensitivity of RelA^−/− MEFs to apoptosis and necroptosis induced by TNF, TNF plus Smac mimetics, or TNF plus Z-VAD was completely blocked by deleting both Fadd and Rip3 (Supplemental Fig. 3E). These results indicated that the absence of RelA leads to TNF-induced activation of FADD-dependent apoptosis and RIP3-dependent necroptosis.

RelA^−/−Fadd^−/−Rip3^−/− mice were healthy and active during the first days of life. Surprisingly, up to around P10, most of the TKO mice failed to thrive and died (Fig. 3A). Microscopic examination of lung sections from TKO mice at P7 showed normal histological structure; however, massive bacterial infection was observed (Figs. 2E, 3B). The gas exchange areas, including the alveolar duct, alveolar sacs, and bronchi, were completely filled with bacteria at P7, but not at P3, in RelA^−/−Fadd^−/−Rip3^−/− mice (Fig. 3B). Moreover, F4/80⁺, but not Gr-1⁺, myeloid cells increased in the lung of RelA^−/−Fadd^−/−Rip3^−/− mice at P7 (Fig. 3C). These observations implied that the respiratory failure caused by pneumonia might lead to the postnatal death of RelA^−/−Fadd^−/−Rip3^−/− mice around P10.

Therefore, we speculated whether the abnormal production of immune cells in TKO mice might impair host defense responses and cause susceptibility to bacterial infection, leading to premature death. Nevertheless, flow cytometry analysis revealed a comparable frequency of mature B cells and T cells in the spleen and thymus, as well as myeloid cells in the spleen and bone marrow, in RelA^−/−Fadd^−/−Rip3^−/− mice and control littermates at 1 wk (Fig. 3D, 3E), suggesting that the severity of pneumonia in RelA^−/−Fadd^−/−Rip3^−/− mice was not caused by defects in the development of immune cells.

To further clarify the effect of bacterial infection on the premature death of RelA^−/−Fadd^−/−Rip3^−/− mice, we reasoned that postnatal death could be improved by using a mixture of broad-spectrum antibiotics. Antibiotic treatment starting in the mother 1 wk before the birth of offspring significantly improved viability and extended the survival of RelA^−/−Fadd^−/−Rip3^−/− mice from 10 d to >60 d (Fig. 3F). In addition, the bacteria in the lung of RelA^−/−Fadd^−/−Rip3^−/− TKO mice disappeared after antibiotic treatment (Fig. 3G). Moreover, there were no differences in F4/80⁺ and Gr-1⁺ cells in the lung between RelA^−/−Fadd^−/−Rip3^−/− TKO mice and control littermates after antibiotic treatment (Fig. 4A). Interestingly, we did not observe any lymph nodes, such as such as cervical, brachial, inguinal, or mesenteric lymph nodes, in RelA^−/−Fadd^−/−Rip3^−/− TKO mice at 3, 6, or 12 wk (Fig. 4B). There were no differences in the numbers of thymocytes and splenocytes (Fig. 4C–E), or significant changes in B or T cells in blood or thymus, between RelA^−/−Fadd^−/−Rip3^−/− TKO mice and control littermates at 3, 6, or 12 wk (Fig. 5A–C). However, the splenic B220⁺ cell population was significantly reduced in RelA^−/−Fadd^−/−Rip3^−/− TKO mice compared with control littermates at 3 wk of age but not at 6 or 12 wk of age (Fig. 5A–C). There were no visible B or T cells in lymph nodes from RelA^−/−Fadd^−/−Rip3^−/− TKO mice at 3, 6, or 12 wk of age (Fig. 5A–C). Gr-1⁺/Mac-1⁺ myeloid cells in blood decreased slightly in RelA^−/−Fadd^−/−Rip3^−/− TKO mice compared with control littermates at 12 wk of age but not at 3 or 6 wk of age (Fig. 5D–F). However, there were no differences in Gr-1⁺/Mac-1⁺ or F4/80⁺/Mac-1⁺ myeloid cells in the spleen or bone marrow between RelA^−/−Fadd^−/−Rip3^−/− TKO mice and control littermates at 3, 6, and 12 wk (Fig. 5D–F). Collectively, these data suggested that bacterial infection contributes to the death of RelA^−/−Fadd^−/−Rip3^−/− mice at a premature age and that it can be significantly improved by antibiotic treatment. Most interestingly, lymph nodes failed to develop in RelA^−/−Fadd^−/−Rip3^−/− mice.

Pseudokinase MLKL, as an executor of necroptosis, is recruited and phosphorylated by RIP3 during necroptosis (19). Thus, we tested whether the deletion of Fadd and Mlkl also rescued embryonic lethality in RelA-deficient mice. We obtained RelA^−/−Fadd^−/−Mlkl^−/− TKO mice by intercrossing RelA^+/−Fadd^+/−Mlkl^−/− mice (Supplemental Fig. 3F). There were no visible differences in the body weight between RelA^−/−Fadd^−/−Mlkl^−/− TKO mice and control littermates at P11 (Supplemental Fig. 3F). Moreover, histological analysis suggested that no visible abnormalities existed in the examined tissues, including brain, liver, intestine, heart, kidney, spleen, skin, stomach, and lung tissues, in RelA^−/−Fadd^−/−Mlkl^−/− mice (Supplemental Fig. 4A). These results suggested that the ablation of Fadd and Mlkl fully rescued the embryonic lethality of RelA-deficient mice. Remarkably, all of the obtained RelA^−/−Fadd^−/−Mlkl^−/− TKO mice were dead approximately 2 wk after birth, as observed in RelA^−/−Fadd^−/−Rip3^−/− mice.

RIP1 is a key regulator of cell death signals following death receptor ligation through its kinase activity–dependent and/or scaffolding-dependent function. Two kinase-inactive RIP1 mutant mice, RIP1^D138N/D138N (aspartic acid at 138 mutated to asparagine, D138N) and RIP1^K45A/K45A (lysine at 45 mutated to alanine, K45A), were viable and healthy (39, 40), implying that RIP1 kinase activity is dispensable for normal embryonic development. Given that the kinase activity of RIP1 is essential for mediating both apoptosis and necroptosis signaling and that inhibition of RIP1 kinase activity (RIP1^K45A or RIP^D138N) protects against multiple-organ inflammation in cpdm mice and against TNF-induced shock (39–41), we examined cell death in RelA^+/+, RelA^−/−, RelA^+/−RIP1^K45A/K45A, and RelA^−/−RIP1^K45A/K45A primary MEFs with various stimuli in vitro. We found that TNF, TNF plus Smac, or TNF plus Z-VAD treatment for 24 or 48 h triggered increased cell death in RelA^−/− MEFs, but inhibiting RIP1 kinase activity significantly reduced cell death in RelA^−/− MEFs (Supplemental Fig. 4B, 4C). These results demonstrated that RIP1 kinase activity is critical for apoptosis and necroptosis triggered by diverse stimuli in RelA-deficient cells.

To test the role of RIP1 kinase activity in the embryonic lethality of RelA^−/− mice in vivo, we crossed RelA^+/− mice to Rip1^K45A/K45A mice expressing the ATP binding site mutation and intercrossed their offspring. RelA^−/−Rip1^K45A/K45A embryos at E13.5 and E16.5 showed a normal litter size and the formation of blood vessels, including capillaries and other blood vessels in the head, face, tail, and footpad (Fig. 6A). Moreover, histological analysis revealed no visible abnormalities in the liver, lung, or skin of RelA^−/−Rip1^K45A/K45A mice at E16.5 (Fig. 6B). These results suggested that compromising RIP1 kinase activity prevented embryonic lethality in RelA^−/− mice around E15.5. Notably, we obtained RelA^−/−Rip1^K45A/K45A mice postbirth (Fig. 6C, 6D). RelA^−/−Rip1^K45A/K45A mice were born normally, but they exhibited a significant reduction in body weight compared with their littermate controls at 1 wk of age (Supplemental Fig. 4D). These data demonstrated that blocking RIP1 kinase activity (Rip1^K45A/K45A) fully rescued the embryonic lethality of RelA^−/− mice.

Histological analysis revealed no detectable histological differences in multiple examined tissues, including brain, liver, intestine, heart, kidney, and stomach, in RelA^−/−Rip1^K45A/K45A mice at P14 (Fig. 6E). However, all RelA^−/−Rip1^K45A/K45A mice died ~ 2 wk after birth, a finding that was similar to RelA^−/−Fadd^−/−Rip3^−/− TKO mice and RelA^−/−Fadd^−/−Mlkl^−/− TKO mice (Fig. 7A). By P14, RelA^−/−Rip1^K45A/K45A mice showed highly increased signs of pathology in the lung and skin (Fig. 7B). In line with RelA^−/−Fadd^−/−Rip3^−/− TKO mice, RelA^−/−Rip1^K45A/K45A mice displayed serious bacterial infection in the lung at P14 (Fig. 7B). Gas exchange areas in the lung were completely filled with bacteria (Fig. 7B). Immunostaining showed that F4/80⁺, but not Gr-1⁺, myeloid cells were increased in the lung of RelA^−/−Rip1^K45A/K45A mice (Fig. 7C), which was similar to RelA^−/−Fadd^−/−Rip3^−/− TKO mice. However, there were no differences in B or T cells in spleen and thymus, or in Gr-1⁺/Mac-1⁺ myeloid cells in spleen and bone marrow, at 1 wk (Fig. 7D, 7E). These results suggested that bacterial infection in the lung led to the postnatal death of RelA^−/−Rip1^K45A/K45A mice.

Interestingly, RelA^−/−Rip1^K45A/K45A mice developed inflammatory skin lesions (Figs. 6C, 7B), which were not observed in RelA^−/−Fadd^−/−Rip3^−/− or RelA^−/−Fadd^−/−Mlkl^−/− TKO mice. Histological analysis confirmed that 2-wk-old RelA^−/−Rip1^K45A/K45A mice exhibited skin lesions, as reflected by significantly increased epidermal thickness (Fig. 7F). Moreover, the skin of RelA^−/−Rip1^K45A/K45A mice showed increased infiltration of F4/80⁺ myeloid cells (Fig. 7G). Together, these results suggested that RIP1 kinase is critical for embryogenesis of RelA-deficient mice but not for maintaining host defense immunity in RelA-deficient mice.

Due to the similar bacterial pneumonia as observed in RelA^−/−Fadd^−/−Rip3^−/− TKO mice, we treated RelA^−/−Rip1^K45A/K45A mice with antibiotics starting in the mother 1 wk before the birth of offspring. The lifespan of RelA^−/−Rip1^K45A/K45A mice was notably improved by this treatment (Fig. 8A). Moreover, the bacteria in the lung of RelA^−/−Rip1^K45A/K45A mice disappeared (Fig. 8B). In addition, there were no differences in F4/80⁺ or Gr-1⁺ cells in the lung or skin between RelA^−/−Rip1^K45A/K45A mice and littermate controls (Fig. 8C). Interestingly, RelA^−/−Rip1^K45A/K45A mice showed developmental failure in the lymph node, as observed in RelA^−/−Fadd^−/−Rip3^−/− TKO mice (Fig. 8D). Moreover, RelA^−/−Rip1^K45A/K45A mice exhibited moderate splenomegaly compared with control littermates, as reflected by significantly increased spleen weight and splenocyte number (Fig. 8E, 8F). Flow cytometry analysis showed that B and T cells in blood were comparable between RelA^−/−Rip1^K45A/K45A mice and control littermates (Fig. 8G). However, the splenic B cell population in RelA^−/−Rip1^K45A/K45A mice decreased significantly at 3 wk (Fig. 8H), which was similar to RelA^−/−Fadd^−/−Rip3^−/− TKO mice at the same age. The thymic T cell population was comparable in RelA^−/−Rip1^K45A/K45A mice and control littermates (Fig. 9A). In agreement with RelA^−/−Fadd^−/−Rip3^−/− TKO mice, there were no B or T cells in the lymph nodes of RelA^−/−Rip1^K45A/K45A mice (Fig. 9B). Myeloid cells increased in the blood and bone marrow but not in the spleen of RelA^−/−Rip1^K45A/K45A mice compared with control littermates (Fig. 9C). These results suggested that antibiotic treatment improved the bacterial infection in the lung and alleviated inflammatory skin lesions in RelA^−/−Rip1^K45A/K45A mice.

Knowing that RelA is a major component of the NF-κB transcription factor that regulates cell survival and death through increasing prosurvival and/or anti-death gene expression responses to various extracellular stimuli, we hypothesized that RelA contributes to selective gene induction to protect cells from cell death pathways. Whole-genome RNA sequencing analysis was performed using MEFs treated with TNF for 1.5 h. Heat map analysis showed that the transcription of several anti–cell death and cytokine-related genes was impaired in RelA^−/− MEFs compared with RelA^+/+ MEFs (Supplemental Fig. 4E, 4F).

cIAP1/2 is best known for its ability to inhibit caspase activities and regulate cell survival. On the one hand, cIAP1/2 could directly bind to and inhibit caspases (9–11). On the other hand, it could also control the ubiquitin-dependent signaling pathway that is required for activating the NF-κB pathway (12), which, in turn, drives the expression of genes essential for survival, immunity, and inflammation. RNA sequencing data showed that cIAP1/2 mRNA expression was significantly reduced in RelA^-/- MEFs compared with RelA^+/+ MEFs (Supplemental Fig. 4E). We further confirmed this reduction by quantitative PCR analysis (Supplemental Fig. 4G), suggesting that cIAP1/2 might function as a RelA target gene to regulate the cell death pathway.

Cytokines play a pivotal role in the host defense against bacterial infection. Indeed, RNA sequencing data showed notably less mRNA expression of several cytokines in RelA^−/− MEFs than in RelA^+/+ MEFs (Supplemental Fig. 4F). Taken together, the impaired expression of genes involved in cell survival, cytokines, and chemokines may contribute to the susceptibility to cell death under various stimuli in RelA deficiency.

Cell death was significantly increased in all RelA-deficient MEFs. Thus, we assessed formation of the RIP1/FADD/caspase 8/RIP3 death-inducing signaling complex. Immunoprecipitation of RIP1 in MEFs treated with TNF for 12 h showed that interactions between RIP1 and caspase 8, FADD, and RIP3 were significantly increased in RelA^−/− MEFs but not in RelA^+/+ MEFs (Fig. 9D). After TNF treatment for 12 h, we also found that the interaction between RIP1 and FADD/caspase 8 was strengthened in RelA^−/−Rip3^−/− MEFs compared with RelA^+/+Rip3^−/− MEFs (Fig. 9E). In addition, as detected in RelA^−/−Rip3^−/− MEFs, the association between RIP1 and FADD/caspase 8 was also enhanced in RelA^−/−Mlkl^−/− MEFs compared with that in their controls (Fig. 9F). Interestingly, RIP3 was found to be associated with the RIP1/FADD/caspase 8 complex after TNF stimulation for 16 h (Fig. 9F), suggesting that cell death signaling was only transmitted to RIP3 in RelA^−/−Mlkl^−/− MEFs. These observations reflected that RelA tightly controlled apoptosis and necroptosis signaling through inhibiting the formation of the cell death complex comprising RIP1/RIP3/FADD/caspase 8.

Collectively, this study delineates that RelA is not only essential for regulating RIP1 kinase–mediated apoptosis and necroptosis during embryogenesis but is indispensable for host defense immunity postbirth (Fig. 10).

Before the function of RIP3/MLKL in necroptosis was discovered, the experimental results revealed an essential role for RelA in preventing TNF-induced apoptosis in vitro (42, 43). Moreover, genetic deletion of RelA leads to embryonic lethality (21), also demonstrating its indispensable function in maintaining normal embryonic development in vivo. The embryonic death prevented by the ablation of TNF or TNFR1 reminded us that the downstream signaling of TNF was responsible for this embryonic lethality (35–37). However, the mechanism by which TNF/TNFR1 causes embryonic lethality in RelA-deficient mice remains elusive.

This study suggested that treatment of RelA^−/− MEFs with TNF resulted in a significant reduction in viability (Supplemental Fig. 1) and augmented formation of the cell death complex containing RIP1, FADD, caspase 8, and RIP3 (Fig. 9D). Although deletion of Rip3 or Mlkl only partially and improved embryonic survival, and to a limited degree (Fig. 1B, Supplemental Fig. 2B), the coablation of Fadd and either Rip3 or Mlkl fully prevented this embryonic lethality (Fig. 2, Supplemental Fig. 3F). Thus, apoptosis and necroptosis mediate the embryonic lethality of RelA-deficient mice.

RIP1 represents a key switch point in the TNF-induced cellular NF-κB, apoptotic, and necroptotic pathways. RIP1 kinase activity tightly controls cell death signaling (41), as shown by the observation that inhibition of RIP1 kinase activity (RIP1^K45A or RIP^D138N) protected against multiple-organ inflammation in cpdm mice and against TNF-induced shock, respectively (39, 40). Moreover, lack of RIP1 kinase activity prevents hepatocellular carcinoma in NEMO^LPC-KO mice and prevented epithelial cell death, Paneth cell loss, and colitis development in mice with epithelial NEMO deficiency (44, 45). Our results indicated that blocking RIP1 kinase activity inhibited two forms of cell death induced by TNF in RelA-deficient MEFs (Supplemental Fig. 4B, 4C). Moreover, the inhibition of RIP1 kinase rescued the embryonic lethality of RelA-deficient mice in vivo (Fig. 6A–C). Thus, RIP1 functions upstream of FADD and RIP3 to mediate the embryonic death of RelA-deficient mice.

As illustrated in Supplemental Fig. 4C, cell death still occurred in RelA^−/−RIP1^K45A/K45A MEFs treated with TNF for 48 h but was significantly reduced in RelA^−/−RIP1^K45A/K45A MEFs compared with RelA^−/− MEFs. It is uncertain whether another RIP1 kinase site at D138 might compensate for this cell death or an undiscovered function of RIP1 exists. Clarifying RIP1 kinase function in apoptosis and necroptosis by simultaneously mutating the kinase sites at K45 and D138 or identifying unknown roles for RIP1 should be an exciting topic of study in the future.

Notably, all RelA^−/−Fadd^−/−Rip3^−/− TKO and RelA^−/−RIP1^K45A/K45A mice displayed impaired lymph node development, which was similar to RelA^−/−TNFR1^−/− mice (46). Stromal cells played critical roles in the development of lymph nodes (46); however, TNFR1^−/− mice showed normal lymph node and lymphocyte development (47, 48). It implied that RelA deficiency led to the impaired lymph node development. RelA^−/−TNFR1^−/− mice showed severely reduced T cell–dependent Ab responses (46). In addition, RelA and other NF-κB family members were important for functional specialization of B cells (49). Thus, it seems that T and B cell functional deficiency in the RelA^−/−Fadd^−/−Rip3^−/− TKO and RelA^−/−RIP1^K45A/K45A mice caused susceptibility to infection.

Unexpectedly, all RelA^−/−Fadd^−/−Rip3^−/− TKO and RelA^−/−RIP1^K45A/K45A mice died around P10 from pneumonia caused by bacterial infection in the lung (Figs. 3A, 3B, 7A, 7B). It implied that FADD, RIP3, and RIP1 kinase activities were not involved in pneumonia in these mice, because RelA^+/−Fadd^−/−Rip3^−/− and RelA^+/−RIP1^K45A/K45A mice did not exhibit bacterial infection. Loss of epithelial RelA resulted in deregulated intestinal homeostasis and susceptibility to dextran sodium sulfate–induced colitis (29). Moreover, pancreas-specific RelA truncation increased susceptibility to cerulein-induced pancreatitis (50). A heterozygous mutation in human RelA results in autosomal-dominant chronic mucocutaneous ulceration (51). Thus, we speculated that the pneumonia phenotype was caused by RelA deficiency and not by deficiency of Fadd and Rip3 or Mlkl.

Indeed, animals with a loss of RelA function in the lung of alveolar epithelial cells, as well as in myeloid cells, showed defects in cytokine and chemokine responses and displayed sensitivity to Streptococcus pneumoniae (52, 53). Epithelial cell RelA-deficient mice were unable to mediate initial neutrophil recruitment during pneumococcal pneumonia (54); moreover, alveolar macrophage RelA–deficient mice exhibited impaired cytokine expression and neutrophil recruitment (53). In addition, defective neutrophil recruitment could be rescued by the administration of CXCL5 (52, 54). Thus, it implied that RelA-mediated neutrophil recruitment played a critical role in maintaining host defense immunity. Consistent with previous results, we also observed that the expression of cytokines essential for the pulmonary host defense was significantly impaired under RelA deficiency (Supplemental Fig. 4F).

The NLRP3 inflammasome is a multiprotein consisting of NLRP3, caspase 1, ASC, and Nek7 that is best known for its ability to control activation of the proteolytic enzyme caspase 1 and modulate host immune responses against microbial infection (55). Active caspase 1, in turn, regulates the maturation of IL-1β and IL-18 (55). Moreover, pyroptosis regulated by the inflammasome may affect adaptive immunity against the microbial infection through releasing damage-associated molecular patterns, such as IL-1α (56). We observed that the transcription of NLRP3, caspase 1, IL-1α, IL-1β, and Nek7, which is essential for inflammasome activation, was notably reduced in RelA^−/− MEFs upon TNF treatment (Supplemental Fig. 4H). These data implied that inflammasome activation might be compromised in RelA^−/−Fadd^−/−Rip3^−/− and RelA^−/−Rip1^K45A/K45A mice, which might accelerate bacterial pneumonia. Clarifying the mechanism should be an interesting topic for future studies.

We thank Dr. Xiaodong Wang for providing RIP3^−/− mice, Dr. Jianke Zhang for providing Fadd^+/− mice, and Dr. Zhen-Ge Luo for providing RelA^Flox/Flox mice. We also thank the Animal Facility of the Institute for Nutritional Sciences for mouse care.

The authors have no financial conflicts of interest.