Programmed cell death as a defence against infection

Abstract

Eukaryotic cells can die from physical trauma, resulting in necrosis. Alternately, they can die via programmed cell death upon stimulation of specific signalling pathways. Here we discuss the utility of four cell death pathways in innate immune defence against bacterial and viral infection: apoptosis, necroptosis, pyroptosis and NETosis. We describe the interactions that interweave different programmed cell death pathways, which create complex signalling networks that cross-guard each other in the evolutionary arms race with pathogens. Finally, we describe how the resulting cell corpses — apoptotic bodies, pore-induced intracellular traps (PITs) and neutrophil extracellular traps (NETs) — promote clearance of infection.

Introduction

Cell death can be either programmed or accidental. Programmed cell death results in either lytic or non-lytic morphology, depending upon the signalling pathway. For example, apoptosis^1,2 is a non-lytic and typically immunologically silent form of cell death. On the other hand, programmed lytic cell death is highly inflammatory, and includes necroptosis^3–5, pyroptosis⁶, and the rapid release of so-called neutrophil extracellular traps (NETs) in a process known as NETosis⁷. The elimination of host immune cells by programmed cell death can be thought to benefit an infecting pathogen⁸. However, programmed cell death is increasingly understood to benefit the host, for example, by eliminating the intracellular niche of certain pathogens⁹. Furthermore, the resulting cellular corpses coordinate an appropriate innate immune response to promote the resolution of infection. However, since our understanding of innate immune defences is mainly based on work with pathogens that evade these defences, it is often difficult to recognize how programmed cell death functions in response to infections. Indeed, many discoveries of cell death functions are based on experiments with pathogens that have been genetically modified to remove their normal host evasion strategies.

Significant crosstalk exists between different cell death pathways, and as a result, innate immune signalling pathways are well guarded against pathogen attack. The ‘guard hypothesis’ was first developed in plant immunity, where studies have shown that innate immune sensors monitor for perturbation of other innate immune signalling pathways¹⁰. For example, in Arabidopsis thaliana, pathogens often attempt to inhibit key proteins that are involved in innate immune signalling. When pathogens attack one such protein, RPM1-interacting protein 4 (RIN4), the perturbation is detected by two different nucleotide oligomerization domain (NOD)-like receptors (NLRs), which activate programmed cell death — a phenomenon known as the hypersensitive response in plants — to combat the pathogen^11,12. This guard mechanism is echoed in mammalian programmed cell death pathways, which function to guard each other as well as other innate immune signalling pathways.

In this Review, we discuss how the host utilizes programmed cell death to fight infections with a focus on in vivo studies. We also discuss how difficult, but not impossible, it can be for pathogens to evade programmed cell death because of the guard functions built into the signalling pathways. We discuss the role of pyroptosis, necroptosis and apoptosis during infection, and how their signalling pathways are functionally interwoven to make them guard each other. Finally, we discuss the fate of dead cell corpses such as apoptotic bodies, NETs, and pore-induced intracellular traps (PITs), and how their physical properties are beneficial to fight infection.

Inflammasomes trigger pyroptosis

Inflammasomes are cytosolic sensors that activate caspase 1¹³. Once activated, caspase 1 cleaves pro-interleukin-1β (IL-1β) and pro-IL-18 into their mature forms, and cleaves gasdermin D (encoded by Gsdmd) to induce pore opening and pyroptosis. (Figure 1). Inflammasomes are cytoplasmic platforms composed of either NLR, AIM2-like receptor (ALR), or tripartite motif-containing (TRIM) proteins that detect contamination of the cytosol by DNA, bacterial flagellin, the type 3 secretion system (T3SS) needle and rod subunits, toxins and cellular perturbations. Mouse caspase-11 — which is duplicated in humans as caspase-4 and caspase-5 — also induces pyroptosis¹⁴, but does not cleave pro-IL-1β or pro-IL-18 on its own. Caspase-11 directly detects lipopolysaccharide (LPS) when it enters the cytosol^15–17, and thus protects against infection by cytosol-invasive Gram-negative bacteria such as Burkholderia thailandensis^18,19, but can be evaded by pathogens such as Francisella novicida¹⁵ and Shigella flexneri²⁰. Caspases-11, -4 and -5 are often called the non-canonical inflammasome¹⁴. The role of inflammasomes in pyroptosis has been reviewed in detail⁶ and is summarized in Fig. 1.

Figure 1. Pyroptosis.

Pyroptosis is initiated by either caspases 1 or 11. Caspase 1 is activated by one of several inflammasomes, NOD-, LRR- and pyrin domain-containing 3 (NLRP3), AIM2, interferon-γ (IFNγ)-inducible protein 16 (IFI16), pyrin, NOD-, LRR- and CARD-containing 4 (NLRC4) and NLRP1b⁶. NLRP3 responds to numerous agonists, which may converge upon low cellular potassium concentration, although this remains controversial. NLRP3 will not respond to these agonists unless it also receives a priming stimulus via various Toll-like receptors (TLRs) or from tumour necrosis factor (TNF), which trigger post-translational priming (probably de-ubiquitination), as well as boosting NLRP3 sensitivity by transcriptional induction¹⁶². AIM2 detects cytosolic DNA^83,84, and IFI16, which lacks a clear mouse homolog¹⁶⁶, also detects viral nucleic acids⁵⁰. Pyrin detects modulation of Rho-family GTPases by bacterial toxins¹⁶⁷. NLRC4 is activated by one of three bacterial flagellin or type III secretion rod or needle proteins that signal via an upstream NLR in the NAIP family^9,36,168,169. NLRP1b detects the protease activity of the anthrax lethal toxin^170,171. By contrast, caspase 11 itself is the sensor for cytosolic lipopolysaccharide (LPS)^15–17. Similar to NLRP3, caspase 11 requires priming, but priming is by either type I interferon or interferon-γ (IFNγ). Either caspases 1 or 11 independently cleaves gasdermin D, from which the released N-terminal fragment associates with the cell membrane and oligomerizes to form the pyroptotic pore^26–29. The cell then swells, resulting in membrane rupture that is called pyroptosis. In addition, caspase 1 will cleave pro-interleukin-1β (pro-IL-1β) and pro-IL-18 to their mature forms (caspase 11 cannot do this directly)¹⁴. Mature IL-1β and IL-18 can escape through the gasdermin D pore, or be released later by membrane rupture^26,28. Red and blue lines represent initiating and priming events, respectively.

Recently, two research groups independently discovered that caspases-1 and -11 cleave gasdermin D and this cleavage is required to trigger pyroptosis^21,22. This finding was later confirmed by a third research group²³. Interestingly, cleavage of gasdermin D was observed a few years earlier, but its importance remained uninvestigated²⁴. Gasdermin D is a member of the greater gasdermin-domain protein family, consisting of 6 members in humans and 10 in mice²⁵. The function of other gasdermin family members remains to be further elucidated, and while many probably induce pyroptosis²¹, it remains unclear how they are activated.

Gasdermins are bipartite proteins whose N- and C-terminal domains are connected by a linker, and have several proposed roles in immune-related diseases²⁵. Caspases-1 or -11 cleave this linker of gasdermin D, releasing the N-terminal domain^21,22, which then associates with the inner leaflet of the plasma membrane and oligomerizes to form pores. These pores range in size from 10–33 nm^26–29, just large enough to allow passage of mature IL-1β (which is 4.5 nm)^26,28; this could explain how IL-1β can be released from cells that do not lyse, which might occur, hypothetically, if a very low number gasdermin D pores form. Small, but detectable, amounts of IL-1β may escape through these pores, while small amounts of sodium and water enter. The resulting mild cell swelling should be counteracted by regulatory volume decrease mechanisms, preventing membrane rupture³⁰. Later, the cell could remove the pores by normal membrane repair processes³¹. Conversely, if more gasdermin D pores form, significant swelling should overwhelm the regulatory volume decrease, resulting in membrane rupture that we call pyroptosis, and the release of any remaining processed IL-1β from the cytosol. The N-terminal domain of gasdermin D has affinity not only for the inner leaflet of the plasma membrane, but also for other phospholipids and cardiolipin^26–28,32, which suggests that it may also target organelles. In this regard, permeabilization of the endoplasmic reticulum may explain the calcium flux that occurs during pyroptosis^33,34.

Although numerous pathogens trigger pyroptosis in vitro, the role of pyroptosis in vivo is much less studied. Mice deficient in caspase 1 have an increased susceptibility to a variety of pathogens (Table S1). The importance of pyroptosis in vivo is primarily inferred from comparing the resistance to infection of Casp1^−/− mice with Il1b^−/−Il18^−/− mice. This comparison remains to be performed for many pathogens. The recently generated Gsdmd^−/− mice are defective for both pyroptosis and release of IL-1β and IL-18 (at least in vitro)^21–23, so a genetic approach to inhibit pyroptosis while maintaining cytokine secretion remains elusive. Finally, it is important to remember that defence against infection may require IL-1β, IL-18 and pyroptosis in conjunction³⁵.

Pyroptosis defends against intracellular bacteria

The most direct evidence for pyroptosis in clearing infections comes from the investigation of bacterial strains engineered to activate the NLR family, CARD-containing 4 (NLRC4) inflammasome and induce pyroptosis in vivo. Normally, Salmonella enterica subsp. enterica serovar Typhimurium and Listeria monocytogenes efficiently evade inflammasomes during the systemic phase of infection in mice^9,36–39. By contrast, strains engineered to persistently express flagellin are detected by NLRC4 and are rapidly cleared. The observation that mice deficient in IL-1β and IL-18 could still clear these flagellin-engineered bacteria suggests that pyroptosis is the primary mechanism for bacterial clearance^9,36–38. Similarly, caspase 11 provides in vivo defence against S. Typhimurium strains (sifA mutants) that have lost the ability to maintain the integrity of the Salmonella-containing vacuole and thus aberrantly enter the cytosol¹⁸. After pyroptosis in vitro, bacteria are damaged⁴⁰, perhaps by gasdermin D pores²⁸ and are more susceptible to secondary insults such as hydrogen peroxide or antibiotics⁴⁰. However, the physiologic relevance of this observation remains unclear, since pyroptosis and gasdermin D are insufficient to kill intracellular bacteria in vivo⁹. Instead, after pyroptosis in mice, bacteria are transferred to neutrophils, where they are killed by reactive oxygen species (ROS)^9,40.

Pyroptosis also protects against non-engineered, opportunistic microorganisms in vivo, including Chromobacterium violaceum, B. thailandensis, Burkholderia pseudomallei and F. novicida^{18,19,39,41,42}. However, additional supporting data is needed to confirm this protection⁴³. Amongst these bacteria, B. thailandensis and C. violaceum have the strongest phenotypes of all tested pathogens in inflammasome-deficient mice^19,39 (Table S1)⁴⁴. As few as 100 colony forming units (CFUs) of C. violaceum or B. thailandensis are lethal to inflammasome-deficient mice, whereas wild-type mice survive challenges by 1,000,000 to 20,000,000 CFU. The strength of these phenotypes leads us to propose the Red Pawn Hypothesis, a corollary to the Red Queen Hypothesis. We propose that despite their evasion by host-adapted pathogens, inflammasomes may be maintained over evolutionary time due to their critical role in defence against pathogens that are co-evolving with a non-mammalian host that lacks inflammasome defences. Thus, some ubiquitous environmental bacteria have specific virulence traits capable of manipulating a generic eukaryotic cell, making them potentially lethal pathogens. However, they are unprepared for the mammalian innate immune system, including inflammasome-driven responses (Box 1 and further discussion in reference ⁴⁴).

Box 1. Hypothesis to explain inflammasome importance despite evasion by pathogens.

In Lewis Carol’s Through the Looking Glass, Alice finds herself in a race with the Red Queen, in which “it takes all the running you can do, to keep in the same place.” This inspired the evolutionary hypothesis that constant change is required to survive within a constantly changing environmental niche alongside constantly changing competitors. But in the end, each organism tends to maintain its place within the niche. This concept applies to the host-pathogen interaction: as hosts (Alice) evolve new immune responses, pathogens (the Red Queen) evolve virulence factors to overcome them. For example, numerous pathogens evade inflammasome detection¹⁵⁶. Thus, many successful pathogens appear to be perpetually one step ahead in the evolutionary equilibrium between inflammasome detection and evasion. On the other hand, aberrant inflammasome activation can be lethal^15,16,34, which might select for loss of these sensors over time. One must ask, then, why would Alice run with inflammasomes that fail to fight the Red Queen?

Interestingly, there are two pathogens which inflammasomes decisively control in vivo, Burkholderia thailandensis¹⁹ and Chromobacterium violaceum³⁹ (Table S1)⁴⁴. Both employ significant virulence traits that make them lethal to inflammasome-deficient mice and immunocompromised people. But in normal animals, inflammasomes render them harmless. B. thailandensis and C. violaceum are ubiquitous soil/mud bacteria whose natural host remains unknown; perhaps it is a lower eukaryote that lacks inflammasomes, such as insects or nematodes.

This leads us to propose a corollary to the Red Queen Hypothesis. We propose that C. violaceum and B. thailandensis are not running an evolutionary race alongside humans because they are not natural human pathogens. Instead, they run a completely separate race altogether, alongside a non-human host. They may occasionally lose their way and stumble clumsily onto the human racecourse, where they find themselves brushed aside by inflammasomes. Hence, the mammalian innate immune response is perpetually winning the evolutionary race against environmental pathogens that may be dangerous, but that lack sophisticated immune evasion strategies.

This could explain why inflammasomes are maintained over evolutionary time despite the prevalence of evasive pathogens. Inflammasomes may act as a barrier to prevent opportunistic infection. Such a barrier may not be perfect since B. thailandensis has two close relatives that have advanced on the pathway to mammalian pathogenesis – environmental bacteria that have evolved into Red Queens. Burkholderia pseudomallei is broadly pathogenic to mammals and causes melioidosis¹⁵⁷. Further along the adaptation course is B. mallei, which is the highly virulent equine-adapted cause of glanders¹⁵⁸. Further caveats to the this hypothesis are described in ref ⁴⁴.

Our hypothesis thus predicts that certain innate immune sensors provide a near-permanent victory for the host in defence against a large reservoir of potentially deadly environmental pathogens, while host-adapted pathogens must evade these sensors.

Additionally, inflammasomes provide defence against microorganisms that have adapted to mammalian hosts. For example, although S. Typhimurium efficiently evades inflammasomes during systemic infection, they are detected in the gut⁴⁵. Intestinal epithelial cell NLRC4 triggers epithelial cell shedding during infection. Whether this process is a precursor stage during pyroptosis, or is fully independent of pyroptosis requires further study⁴⁵. Caspase-11 appears to play a similar, parallel role in epithelial shedding to eliminate the low frequency but rapidly replicating sub-population of S. Typhimurium that aberrantly enter the cytosol⁴⁶.

Pyroptosis during viral infection

Inflammasomes also induce pyroptosis in vitro following the detection of viruses, such as murine cytomegalovirus (MCMV) via the double-stranded DNA-detecting cytosolic sensor AIM2. Parallel in vivo studies show that Aim2^−/− mice are susceptible to MCMV infection⁴⁷. However, whether this susceptibility arises from pyroptosis, IL-1β or IL-18 remains to be explored. Conversely, a detrimental effect of pyroptosis has been described during HIV infection. Interferon-γ (IFNγ)-inducible protein 16 (IFI16) — which is an AIM2-like receptor that detects viral nucleic acids — induces pyroptosis in CD4⁺ T cells abortively infected by HIV⁴⁸. This is proposed as the major pathway by which HIV depletes CD4⁺ T cells in vivo^49,50. Conversely, IFI16 has also been suggested to play a beneficial role in the defence against HIV infection, although pyroptosis was not examined as a possible mechanism⁵¹.

RIPK3 triggers necroptosis

Necroptosis, like pyroptosis, is a form of programmed lytic cell death. Necroptosis can be triggered by a variety of intricate pathways (discussed below) and perhaps the simplest pathway is through Z-DNA binding protein 1 (ZBP1; also known as DAI and DLM-1).

Induction of necroptosis by ZBP1

ZBP1 binds either Z-DNA or Z-RNA via two Z-DNA binding motifs, and can respond to MCMV (a DNA virus) or Influenza (a RNA virus)^52–54. Unlike the normal B-form of DNA, Z-form DNA results from tosional strain that may occur during rapid DNA synthesis that could occur during viral infection. Once activated, its receptor-interacting protein (RIP) homotypic interaction motif (RHIM) binds the RHIM of RIP kinase 3 (RIPK3)⁵². This interaction stimulates RIPK3 kinase activity, auto-phosphorylation and oligomerization^55–57 and RIPK3 phosphorylates the necroptosis effector mixed lineage kinase domain-like (MLKL)⁵⁸ (Figure 2). Following phosphorylation by RIPK3, the MLKL pseudokinase undergoes a conformational switch⁵⁹, licensing interaction with the inner leaflet of the plasma membrane where MLKL oligomerizes to form the necroptotic pore^60–62, which is analogous to gasdermin D for pyroptosis. Additional steps may also regulate MLKL⁶³. After pore formation, the terminal events in necroptosis are similar to those in pyroptosis; osmotic pressure from ion and water influx results in rupture of the plasma membrane. In our hands pyroptosis and necroptosis (both at 2h post treatment) are morphologically identical in primary macrophages⁴⁰, with the exception of slightly different nuclear morphology (I.J. and E.A.M, unpublished observation). Other research laboratories visualize differences in RAW264.7 macrophages in settings where pyroptosis is fast (2h) while necroptosis is slow (4h)³². The morphologies in comparison deserve further study. Necroptosis can be investigated in vivo by using Ripk3^−/− or Mlkl^−/− mice; however, RIPK3 has been suggested to have non-necroptotic functions during influenza infection⁶⁴ and during inflammatory disease models⁶⁵. Therefore, before attributing a phenotype to necroptosis, it is important to verify that phenotypes observed in Ripk3^−/− mice are also seen in Mlkl^−/− mice.

Figure 2. Apoptosis and Necroptosis.

(Left) Apoptosis can be triggered via intrinsic or extrinsic pathways, including by natural killer (NK) cells and cytotoxic T lymphocyte (CTL) granzymes. These stimulate the cleavage of BID to tBID, which causes BAX and/or BAK to trigger mitochondrial outer-membrane permeabilization (MOMP), resulting in release of cytochrome C (CytC), which binds to apoptotic protease activating factor 1 (APAF1). APAF1 then oligomerizes into the apoptosome, the platform for caspase 9 activation. Either caspases 8 or 9 will cleave the effector caspases 3 and 7 into their active forms, which leads to apoptosis². (Right) Necroptosis can be triggered by DNA-sensing via ZBP1⁵², which activates receptor-interacting protein kinase 3 (RIPK3) followed by phosphorylation of MLKL. Phosphorylated MLKL binds to the inner leaflet of the plasma membrane and forms the necroptotic pore^60–62. (Center) In the more complex necroptotic pathway, tumour necrosis factor receptor 1 (TNFR1), Toll-like receptor 3 (TLR3)–TRIF, or TLR4–TRIF signal via RIPK1 to activate NF-κb (but RIPK1 is not required for the TRIF-type I IFN response^97,98). RIPK1 can be thought of as a finely balanced teeter-totter with two guards: caspase 8 on one side balanced by a duo of RIPK3 and ZBP1 on the other. When TNFR1–TLR3–TLR4 signalling proceed normally, RIPK1 is “balanced” and functions as a scaffolding protein that becomes poly-ubiquitinated, and serving as a platform for signalling complex assembly⁵ driving transcriptional responses. When signalling is inhibited, for example by inhibition of cIAP1/2 pharmacologic inhibitors (with SMAC mimetics), by inhibiting TAK1, by inhibition of protein synthesis, or by blocking RIPK1 ubiquitination ³, or potentially by virulence factors, RIPK1 becomes “unbalanced”. RIPK1 then recruits FADD (a DD-DED adaptor) via DD-DD interactions, and FADD recruits caspase 8 via DED-DED interaction thereby triggering extrinsic apoptosis^172–174. Successful caspase 8 activation provides negative feedback to prevent necroptosis by cleaving RIPK1 and RIPK3 in the kinase domains^175,176. However, inhibition of caspase-8 pharmacologically or by viral virulence factors permits continued RIPK1 kinase activity, and the teeter-totter swings the other direction. RIPK1 recruits RIPK3 via RHIM-RHIM interaction, and phosphorylates it, triggering RIPK3 kinase activity and necroptosis^177–179. Thus, RIPK1, RIPK3, and caspase 8 act as a cross guard system for TNFR1, TLR3, and TLR4. TLRs such as TLR2 that do not signal through TRIF^{76,100,108,114}, unless longer time points are examined⁷⁵ where TLR2-driven TNF could require caspase 8 to cause paracrine TNFR1 stimulation.

Necroptosis defends against viral infections

The clearest phenotypes for necroptosis during in vivo infection highlight an antiviral role. MCMV encodes a viral inhibitor of RIP activation (vIRA), whose RHIM mimic inhibits RIPK3 signalling, and thereby prevents necroptosis⁵². A RHIM domain mutant virus is attenuated, but this attenuation is reversed in Ripk3^−/− and Zbp1^−/− mice⁵². Similarly, human-adapted herpes simplex virus 1 (HSV-1) and HSV-2 express two protein homologs, ICP6 and ICP10, that inhibit human RIPK3^66,67. However, there appears to be some degree of species tropism; human HSV-1 ICP6 interacts with mouse RIPK3, presumably attempting to inhibit it, but due to the species mismatch, RIPK3 is activated, which triggers necroptosis to prevent viral replication^66–68. Accordingly, Ripk3^−/− mice are susceptible to HSV-1 infection⁶⁹. This idea does not hold true for MCMV, which inhibits necroptosis in both human and mouse cells⁶⁶.

The antiviral role for necroptosis is further supported by the observation that Ripk3^−/− mice are susceptible to vaccinia virus infection⁵⁶. The RIPK3 pathway also defends against a mouse adapted influenza infection. Zbp1^−/− and Ripk3^−/− mice have increased mortality during infection influenza A PR8^54,64. However, the susceptibility of Ripk3^−/− mice to influenza virus was not replicated in another study⁷⁰ and a third study shows the reverse, where Zbp1^−/− mice have higher viral titers but reduced tissue damage and mortality⁵³, suggesting a simultaneously beneficial role in defense paired with a detrimental role in immunopathology.

Finally, Ripk3^−/− mice have normal resistance to mouse hepatitis virus⁷¹. Many other mouse viral infection models remain to be examined in Ripk3^−/− and Mlkl^−/− mice. All reported lethality phenotypes of Ripk3^−/− and related necroptosis pathway mice are summarized in Table S1.

Necroptosis during bacterial infections

To date, in vivo data does not support an important role for necroptosis against bacterial infections. The most prominent report proposed that reduced necroptosis leads to increased susceptibility to S. Typhimurium infection in mice lacking Ifnar1 (which encodes interferon α/β-receptor subunit 1)⁷². However, this study did not compare other competing hypotheses, most importantly the hyper-responsiveness of Ifnar1^−/− mice to IFNγ (which combats intracellular bacteria such as S. Typhimurium) due to increased expression of the IFNγ receptor⁷³. Wild-type and Ripk3^−/− mice showed similar susceptibility to S. Typhimurium infection⁷², demonstrating that necroptosis plays little role in IFNAR1-sufficient mice. Similarly, Ripk3^−/− mice and wild-type mice are comparably sensitive to Yersinia pseudotuberculosis⁷⁴ and Citrobacter rodentium infections⁷⁵. Ripk3^−/− mice may have a subtle (but statistically insignificant) susceptibility to an LPS engineered Yersinia pestis strain that cannot evade TLR4⁷⁶ (Table S1). Interestingly, Mlkl^−/− mice are susceptible to Staphylococcus aureus skin infection⁷⁷, but paradoxically Ripk3^−/− mice are resistant to both skin and lung infection^77,78. Inhibitor treatment suggests that necroptosis may be pathologic during Serratia marcescens infection⁷⁹, but this needs further study using knockout mice. Specific inhibition of RIPK3 by bacteria has not yet been described. Many more bacterial infection models remain to be examined in Ripk3^−/− and/or Mlkl^−/− mice, and it seems likely that some will establish a role for necroptosis in defence against specific bacteria.

Apoptosis during infections

Apoptosis is programmed cell death characterized by cytoplasmic shrinking, cell rounding, chromatin condensation, DNA fragmentation and membrane blebbing^1,2. Apoptosis can be initiated by cell extrinsic pathways (which are mediated by death receptors) or cell intrinsic (mitochondrial) pathways, both of which culminate in the activation of the effector caspases 3, 6 and 7 (Figure 2). Apoptosis can also be initiated by cytotoxic T lymphocytes (CTLs) or natural killer (NK) cells that deliver granzymes, which activate apoptotic caspases. Upon completion of apoptosis, cellular contents are encapsulated within membrane bound apoptotic bodies. If these bodies are not cleared, they will undergo secondary necrosis (rupture), releasing cytosolic damage-associated molecular patterns (DAMPs) to the extracellular space⁸⁰. The role of apoptosis against infection is perhaps best appreciated by inference, since numerous viruses and bacteria encode apoptosis inhibitors that are essential for virulence (reviewed in ^81,82). This again raises the conundrums posed in the Red Pawn Hypothesis, if most viruses evade apoptosis, are there pathogens that apoptosis effectively counters, providing fully penetrant innate immunity (Box 1). Here, we focus on some of the recent advances in apoptosis that are relevant to infection.

Guarding cell intrinsic apoptosis with interferons

Intrinsic apoptosis is often driven by cell intrinsic stresses such as DNA damage and growth factor withdrawal, which leads to mitochondrial outer-membrane permeabilization (MOMP) ^83,84. Although its structure remains to be fully elucidated, the MOMP pore appears to be formed by BAX and BAK^83,84. MOMP leads to release of mitochondrial contents, including cytochrome c, into the cytosol where it binds to the NLR protein apoptotic protease activating factor 1 (APAF1), which oligomerizes into a hub to form the apoptosome — an activating platform for the initiator caspase 9. Activated caspase 9 in turn cleaves and activates the effector caspases 3 and 7². Since apoptosis destroys the replicative niche of intracellular pathogens, particularly viruses, those that prevent apoptosis will replicate better as the infected cell lives longer.

Historically, apoptosis is considered immunologically silent owing to the collective efforts of caspases. For example, apoptotic DAMPs that can induce a robust type I interferon-mediated proinflammatory response are dampened by caspases^85,86. During viral infection of Casp9^−/− or Casp3^−/−Casp7^−/− cells, MOMP is initiated, but apoptosis fails to proceed. MOMP eventually releases mitochondrial DNA (mtDNA) into the cytosol, where it is detected by the DNA sensor cGAS, resulting in a type I IFN response^85,86. Thus, pathogens that only block caspases will encounter a cGAS-driven type I IFN response that effectively guards events downstream of MOMP. This interferon response promotes resistance to viral infections including encephalomyocarditis virus (EMCV) and vesicular stomatitis virus (VSV). By contrast, blockade upstream of MOMP — which can be studied in Bax^−/−Bak^−/− mice — does not trigger cGAS because MOMP is not initiated and mitochondria retain their DNA^85,86.

Apoptosis/necroptosis guard innate immune signalling

Cell extrinsic apoptosis is typically initiated by extracellular stress signals that engage a subset of TNF-superfamily of receptors called death receptors². Prototypical extrinsic stimuli include FasL, TNF and TNF-related apoptosis-inducing ligand (TRAIL). Following death receptor engagement, caspase 8 (human caspase 10) are recruited and activated via the death inducing signalling complex (DISC). In certain cell types, caspase 8 can also cleave BID into tBID, initiating MOMP in a positive feedback loop that leads to caspase 9, another initiator caspase, thus ensuring a strong initiation of. Thus, caspase 8 activation connects extrinsic and intrinsic apoptosis (Figure 2).

Additionally, caspase 8 is ensconced within the TNF receptor 1 (TNFR1), TLR3, and TLR4 signalling pathways^3–5,87. Caspase 8 interacts with RIPK1, which is a key scaffolding protein within the signalling cascade^3–5,87. Thus, caspase 8 is often interpreted as a node that regulates the divergence of signalling towards either transcriptional responses or apoptosis. For example, it is frequently stated that caspase 8 regulates TNF-dependent transcriptional responses and TNF-driven apoptosis^3–5,87. In this section, we will present this pathway from the perspective of a pathogen.

TNFR1, TLR3, and TLR4 are important innate innate immune sensors — which respond to TNF, double-stranded RNA and LPS, respectively — that drive potent transcriptional responses (but do not trigger cell death under normal circumstances) ³. Hence, these sensors are attractive targets for pathogens to inhibit, particularly since all three use RIPK1 as a scaffolding node to assemble signalling complexes (Figure 2). However, when signalling at or below RIPK1 is pharmacologically inhibited, this causes an alteration in the flow of signals through the pathway, which triggers a counter response that activates caspase 8^3,5. In other words, if pathogens target these signalling pathways they would be met by a caspase 8 tripwire that triggers apoptosis. If pathogens possessed virulence strategies to simultaneously target RIPK1 and caspase 8, they would encounter a second tripwire in RIPK3, which also interacts with RIPK1. As the first caspase 8 tripwire fails to function, signalling build-up at RIPK1 results in the activation of RIPK3, triggering this second guard pathway that drives the response to infection to necroptosis rather than apoptosis (Figure 2). Thus, caspase 8, RIPK1 and RIPK3 essentially function as a triangular guard system.

This cross-guarded system becomes activated during embryogenesis in Casp8^−/− mice, causing embryonic lethality⁴. In line with the cross-guard model, Casp8^−/− lethality is rescued by simultaneous deletion of Ripk3^88,89. Thus, caspase-8 is guarded by RIPK1-RIPK3 driven necroptosis.

Further, recent evidence suggests that RIPK3 is guarded by caspase 8, essentially reverse guarding the guard. Although Ripk3^−/− mice are viable and healthy, mice engineered to carry some (but a not all) point mutations within the Ripk3 kinase enzymatic site suffer from neonatal lethality as a caspase 8 guard program becomes activated, triggering lethal apoptosis during development^90,91. In this guard system, the RIPK3 protein must be present and abnormal in order to trigger caspase 8, whereas simple absence of RIPK3 does not trip this guard pathway (as it does when caspase 8 is absent. This could explain instances where stimuli cause necroptosis, but Mlkl^−/− cells still die whereas Ripk3^−/− cells live; in Mlkl^−/− cells signalling events may accumulate at RIPK3, triggering the guard function of caspase 8, but this does not happen if Ripk3 is deleted⁵³.

Finally, RIPK1 is guarded by both caspase 8 and RIPK3. Deletion of Ripk1 can only be rescued by simultaneous deletion of both Casp8 and Ripk3^92–94, illustrating redundant guarding. On the necroptosis side, there are two pathways that guard RIPK1. The first is direct, where RIPK3 activates autonomously in response to deleted or abnormal RIPK1. The second is upstream, when RIPK1 is absent or its RHIM domain is mutated, this triggers activation of ZBP1, which normally docks on RIPK1 to hold it in check^95,96. Thus, Ripk1^RHIM mice can be rescued by crossing with either Mlkl^−/−, Ripk3^−/− or Zbp1^−/− mice^95,96. ZBP1 then activates RIPK3. These essentially double guard RIPK1.

To add to the complexity, the caspase 8 guard function is built into the RIPK1 scaffold function; TNFR1–TLR3–TLR4 signalling fails to drive transcription in Casp8^−/− cells^97,98. It seems that RIPK1 fails to function when it is unable to interact with its guard caspase 8. The same may be true for Ripk3^−/− cells but only in specific cell types⁹⁹. In contrast, TLR2 signalling, which does not use TRIF–RIPK1, remains intact^76,100 (although one study did see defects in TLR2 signaling⁷⁵, perhaps due to autocrine TNF signalling; also see Figure 2 legend and Box 2).

Box 2. Compound defects in Casp8^−/− mice.

Investigators generated mice lacking caspase 8 to study cell extrinsic apoptosis in vivo, but these mice turned out not to be viable⁴. A caspase 8 deletion induces necroptosis and destruction of the yolk sac vasculature, and this circulatory failure leads to embryonic death. However, introduction of Ripk3^−/− complements the lethality, and Casp8^−/− Ripk3^−/− mice are viable since cells in these mice cannot undergo necroptosis^88,89. Comparing Casp8^−/− Ripk3^−/− mice to Ripk3^−/− controls allows study of caspase 8 genetic deficiency in vivo. However, these mice must be interpreted with great caution since the Casp8^−/− mutation eliminates multiple signalling pathways: i) it eliminates extrinsic apoptosis driven by TNF, Fas, and TNF-related apoptosis-inducing ligand (TRAIL), which could be beneficial during infection^88,89,159; ii)) it alters homeostasis in adaptive immune cells, resulting in a lymphoproliferative disorder that is apparent at 2 months of age^88,89; iii) it significantly impedes or fully abrogates transcriptional responses via tumour necrosis factor receptor (TNFR), Toll-like receptor 3 (TLR3) and TLR4 due to loss of the caspase 8–RIPK1 scaffolding platform^97,98; iv) its effect on TNF, TLR3 and TLR4 also prevent these receptors from post-translational priming of the NLRP3 inflammasome^{75,76,100,107–109}. Fadd^−/− Ripk3^−/− mice are similarly affected^71,74,75,160. This makes it difficult to assign a precise caspase 8 function to phenotypes observed by comparing Casp8^−/− Ripk3^−/− to Ripk3^−/− mice. In order to do this, one could start by comparing various Ripk3^−/− background mice. For example, the resistance of Casp8^−/− Ripk3^−/− mice to endotoxic shock⁸⁸ may be similar to Tlr4^−/− Ripk3^−/− mice but not Nlrp3^−/− Ripk3^−/− mice. This would begin to ascribe the mechanism to the effect of Casp8 deletion upon TLR4 signalling and not to its influence upon NLRP3.

One way to try to study cell extrinsic apoptosis in vivo is to compare Casp8^−/−Ripk3^−/− mice to Ripk3^−/− control mice. However, given the compound effects of deleting Casp8 (elaborated in Box 2), it is not surprising that Casp8^−/−Ripk3^−/− mice are highly susceptible to multiple infectious challenges. Casp8^−/−Ripk3^−/− mice have more rapid dissemination and/or lethality of fully virulent Y. pestis or Y. pseudotuberculosis^74,76. These mice also have higher stool burdens during C. rodentium infection⁷⁵. Furthermore, they succumb to infections by Y. pestis strains engineered to enhance TLR4 detection ⁷⁶, perhaps due to defective TLR4 signalling. Defective TLR4 signalling would also explain the resistance of Casp8^−/− background mice to endotoxic shock⁸⁸. Fadd^−/−Mlkl^−/− mice (which are similar to Casp8^−/−Ripk3^−/− mice, see Box 2) have increased mortality to influenza A virus infection⁶⁴, perhaps due to defects in TLR3 signalling.

In summary, multiple innate immune defence pathways intersect at the caspase 8–RIPK1–RIPK3 triangle. It seems likely that the intricate cross wiring of these pathways plays an important role in preventing pathogens from inhibiting the function of TLR3, TLR4, TNFR1, extrinsic apoptosis and necroptosis.

Neutrophil NETosis makes NETs

Whereas necroptosis and apoptosis are seen in many cell types, another form of cell death is uniquely observed in neutrophils — NETosis, which primarily functions to extrude NETs¹⁰¹ (Figure 3). NETosis is triggered by various microbial and sterile activators — a process that was recently shown to differentiate between microbes based on their size¹⁰² — or by ligation of specific receptors including complement, antibody, cytokine, and TLRs¹⁰³. During NETosis, ROS-dependent release of elastase and myeloperoxidase from neutrophil granules to the nucleus (though not under all conditions¹⁰⁴) promotes histone degradation and chromatin decondensation¹⁰⁵. The effects of ROS and proteases on the nuclear material leads to extrusion of a meshwork of chromatin fibers dotted with granules containing anti-microbial molecules, which traps and kills extracellular bacteria and fungi.

Figure 3. Apoptotic bodies, neutrophil extracellular traps and pore-induced intracellular traps.

The fate of the cell corpse is dependent on the cell death pathway. Extrinsic or intrinsic apoptosis leads to the formation of apoptotic bodies that are engulfed by a secondary phagocyte, a process that is referred to as efferocytosis. In the absence of efferocytosis, apoptotic bodies can undergo secondary necrosis. Lytic death of cells infected with an intracellular pathogen, including pyroptosis, leads the formation of a pore-induced intracellular trap (PIT), which traps the pathogen within the cell corpse. The PIT also presents ligands that are recognized by neutrophils (and potentially macrophages), which efferocytose the PITs and associated bacteria, ultimately killing the pathogen. Efferocytosis of pathogen-containing apoptotic bodies also eliminates the pathogen. Necroptosis and necrosis also result in PITs in vitro, although the physiologic importance in vivo needs further study. Following detection of extracellular pathogens, neutrophils extrude a meshwork of chromatin dotted with granules loaded with antimicrobial molecules. These neutrophil extracellular traps (NETs) trap and kill extracellular pathogens.

NETosis is accompanied by ruptures in the plasma membrane and neutrophil lysis¹⁰¹. However, a mounting body of evidence demonstrates that neutrophils can release NETs while maintaining an intact plasma membrane¹⁰³; this is paradoxically referred to as ‘vital NETosis’, in contrast to the typical ‘suicidal NETosis’. After vital NETosis, neutrophils continue to function in chemotaxis, phagocytosis and bacterial killing¹⁰⁶. Whether a neutrophil undergoes suicidal or vital NETosis appears to be dependent on the NET-inducing stimulus. It is unknown how NETs are extruded from neutrophils without lysing the cell.

Whether NETosis has antimicrobial functions that are independent of NET extrusion remains unknown, but one could imagine that pathogens within neutrophils during NETosis could be killed by the release of granule contents into the nucleus or cytosol (see REF ⁷ for a comprehensive review on NET signalling mechanisms). How NETs combat infection is described further below in our discussion of consequences of cell death.

Cross-talk between cell death pathways

There are many cross-pathway signalling events between apoptosis, necroptosis and pyroptosis (Figure 4). Most prominent amongst these are NLR family, pyrin domain-containing 3 (NLRP3) interactions with the difference cell death pathways. Although the precise signal detected by NLRP3 is uncertain, NLRP3 can be considered as the “sensor for cellular catastrophe” (Box 3), activating when normal cellular physiology breaks down.

Figure 4. Interactions between apoptotic, necroptotic and pyroptotic pathways.

Cell death pathways can interact with each other, with the caveat that once a cell death pathway runs to completion, the interaction ends. The exception to this is the lysis of apoptotic bodies by secondary necrosis, which releases cytosolic contents, effectively converting apoptosis to lytic cell death. Signalling pathways running through RIPK1 will fail if Casp8 is deleted, affecting priming of NOD-, LRR- and pyrin domain-containing 3 (NLRP3) and caspase 11^{75,76,100,107–109}. Caspase 1 driven IL-18 can have two interactions with other pathways, first by inducing interferon-γ (IFNγ) production to prime caspase 11¹⁹, and second by stimulating NK cytotoxic activity³⁹. The terminal events occurring after the MLKL or gasdermin D pores open are catastrophic, resulting in loss of cellular potassium, which is the trigger for NLRP3 activation^163,164. In the window between potassium loss and membrane rupture, NLRP3 activity will trigger caspase 1-dependent processing of interleukin-1β (IL-1β) and IL-18 processing. Finally, ASC and caspase 1 can activate caspase 8, triggering apoptosis^{22,23,117,118}. Purple lines are interacting pathways.

Box 3. As the sensor of cellular catastrophe, NLRP3 detects the pyroptotic and necroptotic pores.

Although NLRP3 is the most extensively investigated inflammasome, it remains the most enigmatic. NLRP3 is activated in response to numerous insults, including pore-forming toxins, ionophores and phagocytosis of indigestible particulates. As such, NLRP3 detects catastrophic events, including the inability to maintain a cytosolic potassium concentration¹⁶¹ via an unknown biochemical mechanism. Besides loss of cytosolic potassium, other mechanisms have also been proposed to explain NLRP3 activation, including ER stress and mitochondrial dysfunction¹⁶².

NLRP3 is also activated downstream of lipopolysaccharide (LPS) detection by caspase 11¹⁴. When caspase 11 cleaves gasdermin D, it polymerizes into the pyroptotic pore. The resulting loss of cytosolic potassium triggers NLRP3 and consequent interleukin-1β (IL-1β) and IL-18 processing^163,164 (Fig. 3). Thus, NLRP3 has a very short activation window before cellular swelling causes rupture of the plasma membrane. IL-1β secretion downstream of cytosolic LPS is thus much lower than other direct inflammasome agonists. To date there is no in vivo evidence supporting physiologic relevance of NLRP3 activation downstream of caspase 11. In fact, caspase 11-dependent defence against Burkholderia thailandensis does not require NLRP3¹⁹, which argues against an in vivo role, at least in this model system.

Similarly, NLRP3 activation downstream of necroptosis^53,100,110 may be caused by the same mechanism. It is likely that potassium loss after opening of the MLKL pore is detected by NLRP3. This pathway may promote IL-1β secretion in vivo during influenza virus and vesicular stomatitis virus infection¹⁶⁵, but more work is needed to determine whether MLKL is the primary pathway whereby NLRP3 is activated in vivo, or whether NLRP3 is actually responding to some other catastrophic event during infection in vivo.

Role of caspase 8 upstream of inflammasome signalling

Caspase 8 is a component of the TNFR1–TLR3–TLR4 signalling scaffold. Therefore, loss of caspase 8 prevents these receptors from providing both post-translational and transcriptional priming of inflammasome components. For example, pro-IL-1β and NLRP3 transcriptional induction is compromised in Casp8-deficient cells^{75,76,100,107–109}, although this may not hold true for all cell types¹¹⁰. Furthermore, NLRP3 requires TNFR1 or TLR signalling to provide post-translational priming^111–113. Thus, Casp8^−/− cells fail to prime NLRP3 after TNFR1, TLR3, or TLR4 stimulation, but can prime NLRP3 within 10 minutes of TLR2 stimulation^{76,100,108,114}. However, when longer time points (4 hours) are examined, TLR2-driven TNF may require caspase 8 to cause paracrine stimulation⁷⁵. In contrast, Casp8-deficiency does not prevent activation of inflammasomes that do not need priming, such as NLRC4^75,107. The in vivo relevance of these interactions remains to be assessed.

Caspase-8 downstream of inflammasomes ensures death

Caspase 8 can also function downstream of inflammasomes, which makes caspase 8 interactions more complicated. Caspase 8 may guard pyroptosis, and thereby ensure activation of cell death in a hypothetical situation where a pathogen attempt to inhibit pyroptosis. ASC is a small adaptor between most inflammasomes and caspase 1; the ASC pyrin domain (PYD) interacts with the PYD of inflammasomes, while its casapse activation and recruitment domain (CARD) interacts with the CARD of caspase 1¹³. Upon activation by inflammasomes, ASC polymerizes to provide a large surface area of activating CARD domains^115,116. In the absence of caspase 1, cell death still occurs after ASC polymerization, but under slower kinetics and with apoptotic morphology. Here, caspase 8 is proposed to interact with ASC via an unusual heterotypic death domain fold interaction between the ASC CARD and the caspase 8 death effector domain^117–119 (Figure 4). This pathway may be most relevant when a pathogen directly inhibits caspase 1 without preventing ASC polymerization¹²⁰. This situation is mimicked by comparing Asc^−/− to Casp1^−/− mice, which reveal an in vivo effect of cytokine responses via this pathway during a F. novicida infection^117,121, but this has no effect on the F. novicida bacterial load^117,121.

In addition, caspase 8 may be directly activated by caspase 1 in settings where gasdermin D is absent. Although cell death was fully ablated in Gsdmd^−/− cells in one study²¹, a slower form of cell death was observed in another study²² that was later suggested to be apoptosis associated with caspase 8 activation²³. By contrast, this pathway to apoptosis was not observed after activation of caspase 11²² (Figure 4).

Some researchers describe this pathway as “suppression of apoptosis by pyroptosis” ²³, or as a “balance between pyroptosis and apoptosis” that depends on the quantity of agonist¹¹⁸. However, we prefer the interpretation that this is a versatile pathway that can “switch from pyroptosis to apoptosis” in the hypothetical case where a pathogen inhibits pyroptosis, thus ensuring cell death¹¹⁷. Apoptosis could thus be considered as a backup guard pathway for pyroptosis, however the in vivo relevance in controlling pathogen burdens remains to be elucidated.

Finally, caspase 8 has been proposed to substitute for caspase 1 as an pro-IL-1β and pro-IL-18 cleaving enzyme^121–125, although this function warrants further study.

In summary, caspase 8 has been proposed to function upstream of inflammasomes, but also downstream of inflammasomes, which creates a quite confusing picture when one attempts to integrate the multiple functions into a single model. In vivo validation of these interactions will hopefully reveal their physiological importance.

Caspase 1 driven IL-18 primes caspase 11

Mouse caspase 11 activation is strictly dependent upon priming; in vitro this is commonly accomplished by studying cells previously primed by type I IFNs (via TLR3 or TLR4 driven type I IFN secretion)⁶, or by TLR2 stimulation¹⁶. The TLR3 pathway has been used in vivo to enable caspase 11 priming during LPS shock models^15,16. Surprisingly, these pathways were insufficient to prime caspase 11 during infection with the cytosol-invasive bacterium B. thailandensis in vivo¹⁹. Instead, NLRC4-driven caspase 1 activation was required to drive IL-18 production, which in turn stimulated IFNγ within the first day of infection. IFNγ was absolutely essential for caspase 11 priming in this model, while endogenous TLR agonists and type I IFN were insufficient. The IL-18 requirement could be bypassed only under delayed kinetics after 3 days as alternate mediators eventually promoted IFNΓ production¹⁹. Thus, in the context of intracellular bacterial infection in vivo, caspase 1 acts upstream of caspase 11 priming, which was surprising as NLRP3 and caspase 1 are commonly thought to function downstream of caspase 11 (Box 3).

IL-18 drives NK cytotoxicity vs intracellular bacteria

NK-derived IFNΓ is well established to combat intracellular bacterial infection. However, whereas NK cytotoxicity is critical in combating viruses such as MCMV, HSV-1, influenza virus, and cancer, initial in vitro studies suggesting that bacterially infected cells are killed by NK cells have been uniformly disproven by subsequent in vivo studies^126,127. Collectively, this data led to the conclusion that cytokines rather than cytotoxicity is the primary mode of NK cell- mediated protection against intracellular pathogens. However, our group recently showed that NK lysis of infected hepatocytes is crucial for controlling infection by an intracellular bacterial pathogen³⁹. In this model, infected phagocytes detected C. violaceum via NLRC4, which lead to production of IL-18 and activation of liver NK cells. Activated NK cells then directly lysed infected hepatocytes in a perforin-dependent manner. NK cytotoxicity also controlled the inflammasome-evasive L. monocytogenes infection, but only when mice were treated with exogenous IL-18³⁹. This example highlights the importance of direct cytotoxicity against bacteria and crosstalk between two different cell death modalities — inflammasomes and apoptosis — to efficiently fight infection. Similarly, NK cytotoxicity initiated by inflammasome detection of S. Typhimurium in a colitis model was recently shown to promote inflammation¹²⁸ and Prf1^−/− mice were also susceptible to C. rodentium during the innate phase of infection¹²⁹. Interestingly, human, but not mouse, NK cells have additional means to directly kill bacteria, since granulysin delivered upon NK degranulation is directly toxic for both bacteria and parasites^130,131.

How programmed cell death clears infection

Recent studies have examined the fate of pathogens after apoptosis, necroptosis and pyroptosis, and how these cell death mechanisms promote the resolution of infection. Apoptosis converts cells into apoptotic bodies, which release a series of find-me and eat-me signals that promote efferocytosis (phagocytosis of dead cells). How the necroptotic and pyroptotic cells are cleared, is currently under thorough investigation. These mechanisms remove the niche for intracellular viruses and bacteria, but pathogens frequently survive host cell death. Finally, NETosis generates structures that are specifically microbicidal. In this final section we discuss what happens to pathogens after programmed cell death (Figure 3).

Efferocytosis of apoptotic bodies

The resolution phase of apoptosis is another example of how apoptotic caspases maintain immunological silence. Failure to clear apoptotic debris can contribute to inflammatory or autoimmune disease{Poon:2014ik}. During normal development and homeostasis, apoptotic bodies are efferocytosed by a secondary phagocyte, thereby preventing secondary necrosis and maintaining immunological quiescence⁸⁰. Efferocytosis is currently an active area of research in the context of infection.

Until recently, apoptosis was often considered intrinsically bactericidal. For example, apoptosis was thought to kill Mycobacteria spp. in vitro¹³² and in vivo¹³³. However, recent studies suggest that Mycobacteria tuberculosis and Mycobacterium mariunum survive apoptosis, but are subsequently killed after efferocytosis by a secondary macrophage^134,135. This efferocytic phagosome matures, acidifies and kills the bacteria via ROS. Notably, pathogens that resist apoptosis, such as Mycobacterium, also interfere with efferocytosis, which increases their survival¹³⁶. L. monocytogenes, on the other hand, exploits efferocytosis receptors in order to spread from cell to cell¹³⁷. The Listeriolysin O toxin damages the plasma membrane, exposing phosphatidyl serine, which is recognized by TIM4. As L. monocytogenes propels itself with actin comets, TIM4 on the neighbouring cell promotes efferocytosis of the protruding membrane, enabling L. monocytogenes to enter the next cell¹³⁷.

Virus-containing apoptotic bodies can also be efferocytosed; for example, influenza A virus induces apoptosis of infected epithelial cells, and neighbouring macrophages efferocytose apoptotic bodies containing viral progeny, which reduces viral titers¹³⁸. On the other hand, some viruses use mimicry to induce apoptosis in order to spread to neighbouring cells¹³⁹.

NETs trap extracellular microbes

In addition to killing pathogens after phagocytosis, neutrophils can target extracellular pathogens using NETs¹⁰¹. NETs play two distinct anti-microbial roles by trapping and killing pathogens. Extensive intravital and confocal microscopy of neutrophils in infected tissue show in real time how bacterial and fungal pathogens can be trapped in NETs^103,105,140. Moreover, additional pathogens are trapped in the extracellular chromatin in vitro. In fact, systemic DNase treatment in a mouse model of a S. aureus skin infection, which induces NET formation, resulted in the escape of the bacteria from the site of inoculation and increased vascular bacterial loads¹⁰⁶.

The extent to which NETs are directly capable of killing bacterial and fungal pathogens remains controversial, since not all bacterial pathogens trapped in NETs are killed and there are no known specific inhibitors that unlink NET release from NET antimicrobial activity. For example, in vitro DNase treatment of Streptococcus pneumonia and Candida albicans trapped in NETs results in the release of viable bacteria¹⁴¹. This lack of microbial killing may stem from the loss of optimal protease activity of key bactericidal enzymes, such as elastase, under in vitro conditions¹⁴². Moreover, many research groups have clearly demonstrated anti-microbial activity of NETs against a variety of microbes in vitro¹⁴⁰, and trapping is obviously beneficial to the host^103,143. For example, intravital microscopy of Escherichia coli trapped in NETs within liver sinusoids demonstrate that disruption of NETs via platelet depletion or DNase treatment prevents trapping and promotes dissemination¹⁴³. The use of vital DNA-intercalating dyes demonstrate that bacteria trapped within NETs are killed^101,144. A thorough discussion of the controversies regarding the composition and antimicrobial properties of NETs was recently published¹⁴⁵.

Many pathogens directly disrupt NETs — for example group B Streptococcus¹⁴⁶, Vibrio cholera¹⁴⁷, Yersinia enterocolitica¹⁴⁸ and Neisseria gonorrhoeae¹⁴⁹ — by expressing nucleases that degrade the NET-associated chromatin, which enables them to escape from NETs and may result in systemic dissemination and sepsis. Other pathogens, such as Haemophilus influenza¹⁵⁰, Pseudomonas aeruginosa¹⁵¹ and S. pneumoniae¹⁵², evade NETs by incorporating NET-associated DNA into their biofilms, which promotes their replication in vivo. Hence, trapping of some microorganisms by NETs in the absence of bacterial killing could be advantageous to the pathogen. Finally, some pathogens evade killing by NETs by sidestepping the specific antimicrobial activity of NET components by expressing polysaccharide capsules^153,154. A full overview of exotoxins that disrupt NETs and other evasion mechanisms are reviewed elsewhere¹⁵⁵.

Pore-induced intracellular traps

Recent evidence from our group establishes the fate of the pyroptotic cell corpse and its role in orchestrating a downstream innate immune response. Similar to apoptosis, pyroptosis does not kill bacteria^9,40. We initially assumed that pyroptosis released intracellular bacteria to the extracellular space⁹. However, in a follow-up study we show that viable bacteria remain trapped within the cellular debris of pyroptotic macrophages⁴⁰. This trapping appears to be mediated by the pyroptotic cell membrane, which remains largely intact despite containing discreet ruptures, encompassing collapsed organelles and cellular structures. We call this structure the pore-induced intracellular trap (PIT) ^35,40 (Figure 3), which is conceptually parallel to NETs. Both PITs and NETs trap the pathogen and prevent microbial dissemination. In contrast to NETs, which act against extracellular microbes, PITs are trap intracellular microorganisms. In addition to pyroptosis, necroptosis and necrosis also result in formation of a PIT that is efferocytosed by neutrophils and macrophages⁴⁰, although the in vivo relevance of necroptotic and necrotic PITs in the context of infection remains to be elucidated.

PITs coordinate innate immune responses to combat intracellular bacteria; PIT formation leads to complement activation, release of IL-1β and IL-18, and eicosanoid production. These mediators recruit neutrophils, which utilize phagocytic scavenger and complement receptors to efferocytose the PIT and entrapped bacteria^35,40, which is conceptually parallel to efferocytosis of bacteria trapped within apoptotic bodies¹³⁴. Ultimately, this secondary phagocyte kills the bacteria via ROS⁴⁰.

We therefore propose that PITs are highly effective immunologically important structures with two key functions; PITs immobilize the pathogen to prevent dissemination and promote efferocytosis by a secondary phagocyte that kills the pathogen.

Concluding remarks

Much work remains to be done to fully understand how pyroptosis, necroptosis, apoptosis and NETosis defend against infection in vivo. Still, genetic tools to study programmed cell death in vivo remain confounded by effects upon other aspects of host defence. For example, Gsdmd deletion prevents pyroptosis, but also prevents IL-1β and IL-18 secretion; similarly, Casp8 deletion prevents cell extrinsic apoptosis, but also may trigger necroptosis, cause a lymphoproliferative disorder, and inhibit innate immune receptor signalling (Box 2). Despite such difficulties, our understanding of programmed cell death has significantly advanced in recent years, revealing the utility of programmed cell death during infection both to destroy intracellular niches, and to coordinate an appropriate innate immune response thereafter. Finally, if a programmed cell death modality is exquisitely effective against a specific virulence strategy, this exerts strong selective pressure for those pathogens to evade the response. Therefore, when interpreting the results of experimental infections it is important to keep in mind the competing possibilities that the host may be benefiting, the pathogen may be evading, or that host defence and pathogenic evasion may both be partially effective. In this regard, identification of new infectious agents that fail to evade specific cell death pathways can aid our understanding of the true function of programmed cell death during infection.

Online summary.

• Pyroptosis is programmed lytic cell death caused by caspase-1/4/5/11 cleavage of GSDMD. Recent work suggests that pyroptosis defends against infection by vacuolar or cytosol-invasive bacteria.

• Necroptosis is programmed lytic cell death caused by RIPK3 activation of MLKL. Recent work suggests that necroptosis defends against viral infection.

• Apoptosis can be activated by extrinsic and intrinsic signals. It is a critical host defence mechanism, and pathogens have evolved to both evade and co-opt apoptosis.

• Apoptosis and necroptosis guard each other to make it difficult for pathogens to inhibit these programmed cell death mechanisms. Despite this, pathogens such as MCMV seem to inhibit both pathways simultaneously.

• The signaling cascades leading to apoptosis, necroptosis, and pyroptosis have numerous interactions, creating a complex web of programmed cell death mechanisms to defend against infection. In this regard, physiologic relevance for apoptosis and necroptotic cross guarding is established, but most other interactions remain to be fully explored.

• Neutrophil extracellular traps (NETs) and pore-induced intracellular traps (PITs) are mechanism to physically restrain bacteria in the extracellular or intracellular space, respectively. These cellular corpses can be considered as structures that exist in parallel to apoptotic bodies (the corpse of an apoptotic cell).

Glossary

Apoptosis

A common form of cell death, which is also known as intrinsic or programmed cell death. Many physiological and developmental stimuli cause apoptosis, and this mechanism is frequently used to delete unwanted, superfluous or potentially harmful cells, such as those undergoing transformation

Pyroptosis

Programmed lytic cell death via caspase-1/4/5/11 cleavage of GSDMD, and recently expanded to include lytic cell death caused by other gasdermin family proteins

Necroptosis

Programmed lytic cell death via RIPK3 activation of MLKL

Apoptosis

Programmed non-lytic cell death in which initiator caspases-2/8/9/10 activate executioner caspases-3/6/7

NETosis

Programmed neutrophil death that results in formation of NETs

Apoptotic bodies

Apoptotic bodies form after apoptosis

NETs

Neutrophil extracellular traps that form after NETosis

pore-induced intracellular traps (PITs)

Structures that form after pyroptosis and necroptosis to retain organelles and bacteria

Biographies

Dr. Jorgensen received her PhD from Duke University where she studied Chlamydia pathogenesis, and did post-doctoral research with Dr. Miao at the University of North Carolina at Chapel Hill studying how pyroptosis clears intracellular bacteria. She is now a researcher at Oslo University Hospital at Rikshospitalet in Oslo, Norway.

Dr. Rayamajhi received her PhD from the University of Colorado and National Jewish Health, then did post doctoral research with Dr. Miao at the University of North Carolina at Chapel Hill studying apoptotic and inflammatory caspase responses to bacterial infection. She is now a research scientist at Camargo Pharmaceuticals.

Dr. Miao received MD and PhD degrees from the University of Washington, and is now an Associate Professor at the University of North Carolina at Chapel Hill. His laboratory studies the mechanisms whereby the innate immune system differentiates pathogenic from non-pathogenic microbes with a focus on intracellular bacteria. His work considers the concepts within microbial pathogenesis to understand innate immune surveillance.