ASC- and caspase-8-dependent apoptotic pathway diverges from the NLRC4 inflammasome in macrophages

The NLRC4 inflammasome recognizes bacterial flagellin and components of the type III secretion apparatus. NLRC4 stimulation leads to caspase-1 activation followed by a rapid lytic cell death known as pyroptosis. NLRC4 is linked to pathogen-free auto-inflammatory diseases, suggesting a role for NLRC4 in sterile inflammation. Here, we show that NLRC4 activates an alternative cell death program morphologically similar to apoptosis in caspase-1-deficient BMDMs. By performing an unbiased genome-wide CRISPR/Cas9 screen with subsequent validation studies in gene-targeted mice, we highlight a critical role for caspase-8 and ASC adaptor in an alternative apoptotic pathway downstream of NLRC4. Furthermore, caspase-1 catalytically dead knock-in (Casp1 C284A KI) BMDMs genetically segregate pyroptosis and apoptosis, and confirm that caspase-1 does not functionally compete with ASC for NLRC4 interactions. We show that NLRC4/caspase-8-mediated apoptotic cells eventually undergo plasma cell membrane damage in vitro, suggesting that this pathway can lead to secondary necrosis. Unexpectedly, we found that DFNA5/GSDME, a member of the pore-forming gasdermin family, is dispensable for the secondary necrosis that follows NLRC4-mediated apoptosis in macrophages. Together, our data confirm the existence of an alternative caspase-8 activation pathway diverging from the NLRC4 inflammasome in primary macrophages.

Dysregulated NLRC4 contributes to severe disease beyond the context of microbial infections. Human NLRC4 gain-of-function mutations (H443P, T337A and V341A) are linked to severe autoinflammatory diseases termed NLRC4-MAS (NLRC4 macrophage activation syndrome) or SCAN4 (syndrome of enterocolitis and auto-inflammation associated with mutation in NLRC4) 23,24 . Current treatments for NLRC4-MAS/SCAN4 focus on blocking IL-1, however, some patients respond poorly to IL-1 blockade suggesting that targeting upstream mechanisms of cell death may be a more effective treatment option. NLRC4 has also been implicated in the development of neuroinflammation and ischemic brain injury in pathogen-free conditions 25,26 . Thus, we sought to identify mechanisms of NLRC4-mediated cell death in sterile bacteria-free conditions to gain invaluable insights into the etiology of NLRC4 mediated auto-inflammatory diseases. Several studies have linked NLRC4 to an alternative caspase-8-mediated cell death distinct from caspase-1-dependent pyroptosis in various cell types and conditions [27][28][29][30] . However, genetic evidence of an alternative NLRC4 mediated caspase-8 pathway in primary macrophages with bacterial infection free conditions has not been clearly studied.
In this study, we performed an unbiased CRISPR/Cas9 screen followed by genetic confirmation experiments in primary macrophages from gene-targeted mice to gain a better understanding of the mechanisms involved in NLRC4-mediated cell death. Importantly, we provide genetic data that highlight the critical roles of ASC and apoptotic initiator caspase-8 in an alternative caspase-1-independent NLRC4-mediated cell death.

NLRC4-mediated cell death occurs independently of caspase-1 in macrophages.
To focus solely on NLRC4 inflammasome activation in macrophages, we delivered ultra-purified flagellin into the cytosol of bone marrow derived macrophages (BMDMs) by electroporation. Consistent with previous reports [10][11][12] , flagellin-triggered cell death measured by lactate dehydrogenase (LDH) release was fully dependent on NLRC4 and NAIP5 (Fig. 1a no pre-stimulation). Interestingly, Casp1 −/− Casp11 −/− BMDMs released LDH to levels equivalent to that of wild-type (wt) under no pre-stimulation conditions, suggesting that NLRC4 may engage in an alternative non-pyroptotic cell death signal in the absence of caspase-1 (and caspase-11). To better understand the kinetics of cell death, we performed live cell imaging in the presence of a cell-impermeable fluorescent DNA staining dye (YOYO-1) to identify dead or dying cells. We confirmed that Casp1 −/− Casp11 −/− BMDMs died and became YOYO-1 + in response to flagellin, albeit with slower kinetics compared to wt (Fig. 1b no pre-stimulation). When BMDMs were pre-stimulated with a TLR2-agonist (Pam3CSK4) as a method to mimic the presence of bacteria, Casp1 −/− Casp11 −/− BMDMs became resistant to NAIP5/NLRC4-mediated cell death (Fig. 1a,b Pam3CSK4 pre-stimulation). This implies that TLR2 signaling can somehow block a caspase-1-independent alternative death signal.
To examine possible differences in cell death morphology, we captured transmission electron microscopy images of flagellin-stimulated wt and Casp1 −/− Casp11 −/− BMDMs at 6 h under no pre-stimulation conditions (Fig. 1c). Wt cells exhibited typical necrotic/pyroptotic morphologies, such as plasma membrane rupture and organelle disintegration. Notably, intact free nuclei were frequently observed only in wt cells (Fig. 1c). In contrast, Casp1 −/− Casp11 −/− BMDMs displayed chromatin condensation and maintained a relatively intact plasma cell membrane, characteristic of apoptosis 31 . Collectively, these data indicate that in the absence of caspase-1, NAIP5/ NLRC4 inflammasome has the potential to trigger an alternative death pathway leading to a slower apoptotic-like outcome in myeloid cells.
To determine whether other myeloid cells responded similarly to BMDMs, we tested bone marrow derived dendritic cells (BMDCs) and thioglycollate elicited peritoneal macrophages from Casp1 −/− Casp11 −/− mice. We found these cells also underwent NLRC4 activated caspase-1-independent cell death, which was markedly reduced in absence of Asc (Fig. 3f). Together, we provide genetic and biochemical data which strongly support the model where ASC acts upstream of caspase-8 activation in macrophages undergoing NLRC4-induced apoptotic cell death.
Enzymatically inactive caspase-1 (C284A) fails to engage pyroptosis but does not abrogate caspase-8-dependent apoptosis downstream of NLRC4. Despite evidence of caspase-8 activation in wt cells (Fig. 3e), caspase-1-dependent pyroptosis is the primary mode of cell death in response to NLRC4 activation (Fig. 1). Our results have so far shown that NLRC4 activation skews toward an apoptotic outcome via ASC/caspase-8 in the absence of caspase-1 expression (Casp1 −/− Casp11 −/− , Fig. 1). One hypothesis to explain this observation would be that caspase-1 CARD preferentially interacts with the NLRC4-CARD, thereby blocking the availability of NLRC4 to interact with ASC-CARD. In the absence of caspase-1, NLRC4 would then be free to recruit ASC and proceed with the caspase-8-mediated apoptosis pathway. To test this competition hypothesis, we utilized Casp1 C284A/C284A knock-in mice harboring an enzymatically inactive mutation (C284A) to examine the consequence of full-length caspase-1 protein expression in absence of pyroptosis initiation. If this competition model were true, expression of caspase-1 C284A would block caspase-8-dependent death by competing with ASC-CARD for binding to NLRC4-CARD. We first confirmed that caspase-1 protein expression in Casp1 C284A/ C284A BMDMs is comparable to that of wt (Fig. 5a). BMDMs derived from Casp1 C284A/C284A phenocopied their Casp1 −/− counterparts by exhibiting strong attenuation of pyroptosis and IL-1β release in response to multiple inflammasome stimuli, including ATP, Nigericin, as well as, intracellular dsDNA, flagellin and LPS, following Pam3CSK4 pre-stimulation (Fig. 5b), thus demonstrating that caspase-1 enzymatic activity plays an essential and  (Fig. 5c). Therefore, ASC/caspase-8 recruitment to NLRC4 is likely unperturbed in Casp1 C284A/C284A BMDMs, despite availability of the caspase-1 CARD for interaction. An alternative and more likely hypothesis may be that caspase-1/GSDMD-mediated pyroptosis is simply more rapid and potent, thereby prevailing over the delayed apoptotic signal. DFNA5 is dispensable for secondary necrosis following NLRC4/caspase-8-mediated apoptosis in macrophages. Apoptotic cells can undergo secondary necrosis if they are not effectively cleared by phagocytic cells 40 . As shown earlier, Casp1 −/− Casp11 −/− BMDMs undergo NLRC4-induced caspase-8-dependent apoptosis and eventually release LDH and stain with a cell membrane impermeable dye, both readouts of plasma membrane damage (Figs 1a-c and 3a-d). These data suggest that secondary necrosis may ensue after apoptosis initiation under these conditions in vitro. Caspase-8 initiates apoptotic cell death by activating caspase-3 31,41 . Consistently, we observed caspase-3 activation in wt and Casp1 −/− Casp11 −/− BMDMs following flagellin stimulation as evidenced by the induction of processed caspase-3 (Fig. 3e). Two recent reports demonstrated that caspase-3 cleaves DFNA5/GSDME, a member of the pore-forming gasdermin superfamily, leading to plasma membrane damage and subsequent osmotic burst 42,43 . Thus, we interrogated Casp1 −/− Casp11 −/− iMacs to determine whether DFNA5 contributed to secondary necrosis following NLRC4/caspase-8-mediated apoptosis. To do so, we generated Dfna5 deficient Casp1 −/− Casp11 −/− iMacs using CRISPR/sgRNA and confirmed that knockout of Dfna5 by sgRNA completely abrogated DFNA5 protein expression (Fig. 6a). Deficiency of DFNA5 did not inhibit LDH release nor delay the kinetics of plasma membrane damage after flagellin stimulation when compared to control sgRNA cells (Fig. 6b,c), whereas disrupting Asc by CRISPR/sgRNA completely abrogated cell death in response to flagellin (Fig. 6d,e). In contrast to recent reports 42,43 , we observed that Dfna5 deficient Casp1 −/− Casp11 −/− iMacs still succumbed to plasma membrane damage like wt control cells in response to other known activators of caspase-3 pathway, including FasL (via caspase-8) and cytochrome-c (via caspase-9) (Fig. 6b,c). All stimulants were confirmed to activate caspase-3 and convert the pro-form of DFNA5 into a ~35 kDa processed N-terminal DFNA5 form (with FLAG tag), which corresponds to the active pore-forming DFNA5 N-terminal (1-270) fragment generated by caspase-3, as previously reported 42,43 (Fig. 6f). These observations suggest that while DFNA5 can be activated by caspase-3, it may be redundant for secondary necrosis in macrophages under these conditions, therefore highlighting a distinct DFNA5-independent program for plasma membrane rupture.

Discussion
Pyroptosis constitutes an essential aspect of anti-bacterial innate immune defense. Pyroptosis is a rapid, irreversible and suicidal host defense mechanism for discharging invading intracellular bacteria from the host cell 5,6 . In our present study, we show that NLRC4 is capable of engaging an alternative ASC-and caspase-8-dependent apoptotic pathway, distinct from pyroptosis. NLRC4 has previously been implicated in a caspase-8-driven cell death pathway in several other studies. For example, ASC and caspase-8 in macrophages infected with Salmonella Typhimurium were shown to be co-recruited to the NLRC4 inflammasome and polymerize in the ' ASC speck' 27 . Furthermore, in lung epithelial cell lines that lack caspase-1, forced expression of NLRC4 resulted in caspase-8 activation 28 . More recently, Rauch et al. provided genetic evidence that activation of NLRC4 by flagellin resulted in intestinal epithelial cells (IECs) expulsion activity but not lytic cell death in the absence of caspase-1 and this pathway required caspase-8 and ASC 29 . Another recent study using CRISPR/sgRNA showed that targeting Casp8 in immortalized macrophages attenuated NLRC4-induced apoptosis 30 . Our unbiased CRISPR screening approach identified caspase-8 and ASC as non-redundant factors in NLRC4-mediated apoptotic cell death, which we corroborated with genetic evidence in BMDMs derived from gene-targeted mice to confirm the essential role of caspase-8 and ASC. The model of caspase-8-mediated apoptotic signal downstream of NLRC4/ASC is also in line with two recent independent reports 44,45 that were published during revision of our manuscript.
Unlike other inflammasome sensors such as NLRP3 or AIM2, NLRC4 contains a CARD, which directly interacts with caspase-1 and negates the necessity for ASC as an adaptor protein for interaction with caspase-1 11 . However, we observed that ASC is necessary for NLRC4-mediated caspase-8 activation (Fig. 3e). Current literature supports the model in which ASC interacts with NLRC4 through CARD-CARD interactions and the ASC PYD domain recruits caspase-8 through an unusual heterotypic domain interaction with caspase-8 DED 32 . The biochemical nature of the ASC/caspase-8 interaction after recruitment to NLRC4 remains to be determined. NLRC4-induced apoptosis is most clearly highlighted in cells lacking caspase-1 activity (Figs 1 and 4). Based on our findings, it seems likely that the NLRC4/caspase-1 and NLRC4/ASC/caspase-8 arms are parallel signaling pathways originating from the same NAIP5/NLRC4 signaling platform. We show that caspase-1 does not functionally compete with the ASC/caspase-8 pathway, as BMDMs expressing an enzymatically inactive mutant form of caspase-1 (C284A), which cannot engage in pyroptosis but retains the ability to interact with NLRC4 via CARD interactions, continue to undergo ASC/caspase-8-mediated apoptosis. Therefore, we suggest that the apoptotic outcome is likely overpowered by the rapid and prominent induction of caspase-1/GSDMD-dependent pyroptosis 17,18 . Interestingly, once macrophages sense other bacterial PAMPs recognized by TLRs, the apoptotic cell death pathway can no longer proceed. One mechanism is likely through inhibition of caspase-8 by cFLIP, following TLR/NF-κB activation 45,46 . From a host defense standpoint in the context of an intracellular bacteria infection, initiating pyroptosis over apoptosis would be more beneficial as exposure of "eat-me" signals on apoptotic cells could facilitate engulfment of dying infected cells by phagocytes 47 . The risk of phagocytes engulfing bacteria within apopotic cells could present a second chance for bacteria to replicate inside phagocytic cells. Interestingly, caspase-8 activation was also observed in wt cells undergoing pyroptosis, however, the physiological ramification of this remains unknown. Beyond apoptosis induction, caspase-8 is reported to have non-apoptotic functions including activation of chemokine transcription 48,49 . It may be possible that caspase-8-mediated chemokine induction contributes to anti-bacterial responses by recruiting neutrophils.
NLRC4 gain-of-function mutations are associated with pathogen-free auto-inflammatory diseases. NLRC4-MAS/SCAN4 patients develop serious auto-inflammation, recurrent fever, rash and enterocolitis 23,24 . The existence of an alternative NLRC4-mediated caspase-8-activating pathway in TLR signaling-free conditions may contribute to deleterious tissue-damaging apoptotic signals from gain-of-function NLRC4 mutants in cell types that perhaps do not express caspase-1 or where caspase-1 activity is blocked. In this scenario, uncontrolled or excessive caspase-8-activation could contribute to inflammation if apoptotic cells are not efficiently cleared 50 . This concept will require further exploration.
Downstream of caspase-8, caspase-3 has been reported to directly cleave DFNA5/GSDME and release the pore-forming DFNA5 N-terminal fragment, which leads to osmotic burst of the cell and necrotic cell death termed secondary necrosis 42,43 . DFNA5 is proposed to play a non-redundant role in caspase-3-induced secondary necrosis in BMDMs 42 . However, despite observing DFNA5 cleavage (Fig. 6f), we found that Dfna5 deficiency did not inhibit nor delay secondary necrosis in our cells (Fig. 6b,c), implying that there could be other DFNA5-independent additional mechanisms that play a dominant role in mediating plasma membrane damage in response to flagellin, FasL and cytochrome-c in vitro. Molecular details of DFNA5-independent secondary necrosis remains unclear thus far, however it is unlikely that GSDMD plays a role since caspase-3 can only aberrantly cleave GSDMD to create a short inactive NT fragment (Fig. 3e and ref. 37 ). DFNA5-independent secondary necrosis may represent an interesting and novel inflammatory mode of cell death, which will require further exploration into its mechanistic details and implications in inflammatory conditions in vivo.

Genome-wide CRISPR-Cas9 Screen.
For CRISPR sgRNA library screen, 155,164 guides targeting 19,884 mouse genes were selected by using a custom sgRNA design algorithm and the designated guides were cloned into a pooled lentiviral expression context by Cellecta, Inc. Casp1 −/− Casp11 −/− Cas9 + iMac were infected with a virus library titer to achieve a MOI of 0.3, with sufficient cell numbers plated to obtain a screening depth of 200 cells per sgRNA. After 13 days post-selection, cells were subjected to Neon electroporation with or without flagellin. Genomic DNA isolation was performed using Puregene reagents (Qiagen). Next generation sequencing was performed on MiSeq system (Illumina) with 50 million sequences reads per sample. Screen hits were identified by MAGeCK computational analyses for sgRNA enrichment in flagellin compared to control treatment. Raw counts were filtered to remove low abundance guides using heuristic from the gCrisprTools package that is gRNAs with counts less than 1/16 th of the top 1000 guides for each sample. Counts are log2 transformed and normalized using median scaling. Guide and gene-level statistics comparing flagellin and control samples are calculated using the MAGeCK software with default parameters. Genes with adjusted p-values less than 0.05 and a log2 fold-change greater than 1 were selected for further investigation. Screen samples were performed in triplicate.

Standard (morphology) Transmission Electron Microscopy of Cells.
Flagellin 0.5 μg per 1 × 10 6 cells was delivered into BMDMs by Neon electroporation and harvested at 6 hours. A total of 30E6 cells per sample were fixed in 1.0 ml of 1/2 Karnovsky's fixative (2% paraformaldehyde, 2.5% glutaraldehyde in 0.1 M sodium cacodylate buffer, pH 7.2). Samples were post-fixed in 1% aqueous osmium tetroxide for 2 h, stained with 0.5% uranyl acetate for 1 h and then dehydrated through a series of ethanol (50%, 70%, 90%, 100%) followed by two propylene oxide washes. Samples were embedded in Eponate 12 (Ted Pella, Redding, CA). Curing of the samples was at 65 °C for two days. Semithin (300 nm) and ultrathin (80 nm) sections were obtained with an Ultracut microtome (Leica). The semithin sections were stained with Toluidine Blue and examined by bright field microscopy to obtain overviews of the population of cells. Bright field images were captured with an Axioplan microscope (Zeiss), an AxioCAM MRm digital camera (Zeiss) and oil immersion lenses Plan-Neofluar 40×/1.4 N.A and a Plan-Neofluar 100×/1.4 N.A (Zeiss). Ultrathin sections parallel to the T-Blue sections were collected on electron microscopy grids, counter stained with 0.2% lead citrate and examined in a JEOL JEM-1400 transmission electron microscope (TEM) at 80 kV. Digital images were captured with a GATAN Ultrascan 1000 CCD camera. Data Availability. The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.