Effector-triggered immunity and pathogen sensing in metazoans

Abstract

Microbial pathogens possess an arsenal of strategies to invade their hosts, evade immune defences and promote infection. In particular, bacteria use virulence factors, such as secreted toxins and effector proteins, to manipulate host cellular processes and establish a replicative niche. Survival of eukaryotic organisms in the face of such challenge requires host mechanisms to detect and counteract these pathogen-specific virulence strategies. In this Review, we focus on effector-triggered immunity (ETI) in metazoan organisms as a mechanism for pathogen sensing and distinguishing pathogenic from non-pathogenic microorganisms. For the purposes of this Review, we adopt the concept of ETI formulated originally in the context of plant pathogens and their hosts, wherein specific host proteins ‘guard’ central cellular processes and trigger inflammatory responses following pathogen-driven disruption of these processes. While molecular mechanisms of ETI are well-described in plants, our understanding of functionally analogous mechanisms in metazoans is still emerging. In this Review, we present an overview of ETI in metazoans and discuss recently described cellular processes that are guarded by the host. Although all pathogens manipulate host pathways, we focus primarily on bacterial pathogens and highlight pathways of effector-triggered immune defence that sense disruption of core cellular processes by pathogens. Finally, we discuss recent developments in our understanding of how pathogens can evade ETI to overcome these host adaptations.

Metazoan organisms use multiple strategies to protect themselves against invading pathogens. As originally proposed by the late Charles Janeway Jr., germline-encoded pattern recognition receptors (PRRs) sense microbial structures such as cell wall components, termed pathogen-associated molecular patterns (PAMPs)^1,2 (Fig. 1a). PRRs activate signalling pathways that alter the transcription of thousands of immune defence genes, including cytokines, chemokines and interferons (IFNs)³. While PRRs are necessary for innate immune responses to infection and innate instruction of adaptive immunity^4–7, both invasive pathogens and commensal bacteria possess many of the same PAMPs. Added layers of immune recognition must therefore be engaged to distinguish beneficial commensals from harmful pathogens. The host’s ability to gauge the level of threat through multiple layers of sensing allows the immune system to act as a tuneable dial, rather than an on/off switch, that can fine-tune the response according to the threat. This is important because inflammatory responses are not only harmful to the pathogen, but can also be associated with tissue pathology and compromised organ function. Being able to tune the response to appropriate threat levels enables the immune system to limit collateral damage to host tissues and respond to bona fide pathogens.

Fig. 1 |. Host cells possess multiple mechanisms of pathogen detection and immune defence.

a, PRRs represent an evolutionarily conserved mechanism to directly detect microbial products (termed PAMPs) that serve as general structural features of specific categories or classes of microorganisms (i). PAMP sensing by host PRRs triggers release of inflammatory mediators from the host cell. These mediators can, in turn, activate other cells to amplify the immune response. Certain intracellular molecules associated with damaged tissues or cells that are released as a consequence of microbial infection or other kinds of acute stress stimuli also contribute to innate immune activation (ii). These molecules, termed DAMPs, are also sensed by receptors on neighbouring cells and contribute to immune activation. b, ‘Patterns of pathogenesis’ provides an additional framework for understanding immune sensing of microbial pathogens, which manipulate cellular physiology to colonize the host but provide specific signals that enable the host to detect the pathogen. Such patterns include: (1) invasion of pathogens into the host cell, (2) altered vacuole trafficking, (3) vacuole disruption or cytosolic escape of the pathogen, (4) cytosolic replication of the pathogen and (5) disruption of host cellular processes by microbial effectors.

One mechanism of pathogen sensing is to detect host molecules that are present or altered due to pathogen presence. This includes detection of damage-associated molecular patterns (DAMPs) that are normally located intracellularly but are released into the extracellular milieu as a result of microbial infection^8,9. DAMPs include the nuclear protein high-mobility growth Box 1 (HMGB1), ATP and surface exposure of the endoplasmic reticulum (ER) protein calreticulin^10,11 (Fig. 1a). DAMPs can also be released following physical trauma or other forms of tissue damage, independent of overt infection¹². Thus, neither DAMPs nor PAMPs alone signify the presence of a pathogen. Rather, simultaneous detection of altered cellular processes and engagement of a toll-like receptor (TLR) or other PRRs provide the critical combination of signals that could only occur in the presence of an infectious microorganism.

Box 1 |. Effector-triggered immunity in plants.

Nucleotide-binding domain, LRR-containing proteins are present in both plants and animals. NLRs play an important role in innate immunity and have been extensively described in connection with ETI^18,23,24,173. Several NLR proteins serve as receptors that detect pathogenic activity and trigger a downstream immune response. In the plant field, two main strategies for host defence have emerged based on studies on NLR proteins: the ‘gene-for-gene’ model and the ‘guard’ model.

‘Gene-for-gene’ model.

This model proposes that for every ‘avirulence’ (Avr) pathogen gene, there exists a ‘resistance’ (R) gene that confers protection to the host¹⁷⁴. In this model, the R protein directly interacts with the pathogen effector and triggers an immune response. One notable example of plants that use this strategy of recognition is flax plants of the Linum genus. The rust fungus Melampsora lini infects plants such as flax, linseed and linola, causing rust disease¹⁷⁵. Flax plants have several different R proteins that confer protection against the fungal Avr genes. Resistance proteins at the L locus of flax plants directly interact with AvrL567 genes from the rust fungus, thereby triggering a hypersensitive response, resulting in cell death¹⁷⁶. This method of detection relies on direct recognition of the pathogen effector by the host receptor. However, in many instances, the host protein can detect the activity of many effectors, suggesting that there is an indirect mode of recognition, as described below.

‘Guard’ model.

This model proposes that resistance proteins survey host cellular processes, thereby acting as ‘guards’ of host homeostasis^23,177. When pathogen effectors disturb these pathways, host guard proteins detect their activity and initiate an ETI response. Several examples of ETI that support this model have been described in plants. For example, the Arabidopsis surface receptors disease resistance protein RPM1 and disease resistance protein RPS2 both serve as guards of a third protein, RPM1-interacting protein 4 (RIN4). Modifications of RIN4 by any of the Pseudomonas syringae effectors AvrB, AvrRpm1 and AvrRpT2 are detected by RPM1 or RPS2, leading to downstream immune signalling^178–180. Another example of the guard model involves the decoy strategy, whereby a host receptor that resembles the pathogen effector target is itself modified and triggers an immune response. The Arabidopsis serine/threonine protein kinase PBS1 inhibits the plasma membrane NLR RPS5 by interacting with its N-terminal domain. When the P. syringae effector AvrPphB cleaves PBS1, RPS5 is derepressed and becomes activated, leading to downstream cell death¹⁸¹. This example is analogous to the NLRP1B activation strategy seen in murine macrophages in response to anthrax lethal toxin^30,34.

Effector-triggered immunity models in plants.

Several mechanisms of effector-triggered immune responses have been proposed in plants, including two well-studied models. a, In the ‘gene-for-gene’ model, a pathogen avirulence (Avr) protein directly interacts with a host resistance (R) protein. This interaction activates a subsequent immune response from the host. b, In a second model, the guard model proposes that host proteins ‘guard’ cellular processes, and the disruption of these processes by virulence factors can be sensed by the host. For example, a pathogen effector that modifies a host protein can trigger recognition of this modification by the host cell, which then initiates an immune response. Since the activity of the effector, rather than the effector itself, is detected, this model of ETI does not rely on direct effector recognition.

Additional mechanisms of pathogen sensing detect PAMPs or pathogen-specific activities within sites that normally exclude microorganisms. Cytosolic invasion by microorganisms, or cytosolic delivery of their products via pathogen-specific secretion systems, constitutes a ‘violation of cytosolic sanctity’¹³. Such activities, which include disruption of vacuolar trafficking, pathogen replication within the cytosol, disruption of the actin cytoskeleton and cell-intrinsic immune signalling, are ‘patterns of pathogenesis’ (Fig. 1b) that can be detected by host factors¹⁴. Pathogen detection may also occur via sensing of homeostasis-altering molecular processes (HAMPs), including manipulation of the actin cytoskeleton or inappropriate reactive oxygen species (ROS) production, either in the context of microbial infection or sterile inflammation¹⁵. Appropriate clearance of pathogens therefore requires not only sensing of microbial structures but also sensing pathogen-specific activity, allowing the host to assess the threat level and generate appropriate responses^16,17 (Fig. 2).

Fig. 2 |. Microbial threat checkpoints gauge the level of threat posed by a pathogen and fine-tune the host immune response.

Host immune responses are tuned according to the microbial threat level. (1) Low-level threats, such as PAMPs and dead microorganisms, lead to upregulation of cytokines and pro-survival factors. Negative regulators—such as IL-10, suppressor of cytokine signalling (SOCS) proteins and dual specificity phosphatases (DUSPs)—control and prevent excessive responses under such conditions. (2) Intracellular, viable non-pathogens and vita-PAMPs (for example, bacterial mRNA) pose a moderate level threat to the host and therefore lead to limited inflammasome activation (i) and release of proinflammatory cytokines and IL-1 signalling with limited levels of cell death or, in the absence of cell death, hyperactivation. The release of IL-1 cytokines is mediated by GSDMD, which forms pores in the cell membrane. (3) Pathogens that possess secretion systems and toxins that disrupt barrier tissues and membranes, or perturb host cellular processes, lead to robust inflammasome activation (ii), resulting in cell lysis and release of intracellular DAMPs (for example, HMGB1, calreticulin and ATP, together with IL-1). Notably, both pathogenic and sterile events that alter cellular homeostasis can trigger these immune responses. An alternative framework for thinking about these types of responses is as factors that indicate disruption of cellular homeostasis or HAMPs. These checkpoints allow the host to modulate the immune response based on the level of threat posed by a microorganism or other cellular stresses.

Here, we focus on effector-triggered immunity (ETI)^18,19, which senses activities of pathogen virulence factors (effectors) that are secreted during infection by bona fide microbial pathogens²⁰. As these effectors specifically modulate host cellular processes to promote infection and bacterial replication, they can be viewed as a sine qua non of pathogenicity. The term ‘ETI’ refers to sensing of pathogen-specific virulence activities through direct protein–protein interactions between a host target protein and a bacterial effector, or through sensing of pathogen-mediated disruption of cellular activities^21,22. ETI was originally hypothesized to function in plant defence via host resistance proteins that would ‘guard’ cellular processes and detect their disruption by pathogen effectors^18,23,24. Notably, nucleotide-binding domain, leucine-rich repeat (NLR) proteins are evolutionarily conserved resistance proteins that detect such disturbances^18,19,23 (Box 1).

ETI plays an important role in the evolutionary race between host and pathogen, as mechanisms that sense obligate features of pathogenicity, including activities that disrupt host cellular processes, are essential for metazoan survival against host-adapted pathogens. Pathogens, in their turn, are selected by these host mechanisms to evolve countermeasures to evade host detection. In this Review, we seek to integrate the previous concepts of pattern recognition and ETI with recently described mechanisms of pathogen sensing by detection of virulence activities and disruption of cellular homeostasis. We also discuss strategies used by pathogens to evade ETI. A common feature of mammalian ETI responses against pathogens, as distinct from PRR responses to commensal bacteria, is activation of inflammatory cell death, which both eliminates replicative niches and releases DAMPs. Here, we focus on common cellular processes disrupted by pathogens and the host mechanisms that sense this disruption.

Activation of NLRP1B and NLRP3 inflammasomes

Bacterial pathogens produce diverse effectors that have detrimental effects on the host. The host has evolved mechanisms to specifically detect enzymatic activities of these bacterial products, as they are clear signatures of pathogen presence when they occur in the context of TLR signalling. As in plants, nucleotide-binding leucine-rich repeat (NLR) family proteins in the mammalian cell cytosol also serve as sentinels of pathogenic activity. Certain mammalian NLRs form inflammasome complexes that mediate the release of interleukin-1 (IL-1) family cytokines and a type of cell death called pyroptosis²⁵. NLRs serve as guardians of ‘cytosolic sanctity’, as they detect pathogenic insults in the otherwise sterile environment of the host cell cytosol¹³, and inflammatory cell death is a common outcome of activating ETI responses through the NLRs. These responses eliminate infected or damaged cells and release inflammatory mediators that amplify antimicrobial responses in the tissues, analogous to hypersensitive responses triggered by plants in response to the virulence factors of phytopathogens^18,26.

NLRP1B and NLRP3 mediate ETI in response to a number of bacterial toxins. Although their activation mechanisms are different, both NLRs oligomerize to form inflammasomes that recruit and activate the cysteine protease caspase-1 (ref. ²⁷). Caspase-1 cleaves specific substrates, including pro-IL-1 family cytokines and the pore-forming protein gasdermin D (GSDMD), allowing for the subsequent release of IL-1 cytokines and pyroptosis²⁷ (Fig. 3). Other inflammasomes, including the pyrin inflammasome, also mount effector-triggered responses to pathogenic activity and are discussed in separate sections of this Review.

Fig. 3 |. Effector-triggered immunity engages inflammasomes, but can also be targeted by other pathogen virulence factors.

(i) Excessive injection of the Yersinia translocon proteins YopB and YopD damages intracellular endosomal membranes, leading to recruitment of GBPs and downstream activation of the caspase-11 inflammasome. Caspase-11-induced GSDMD pores also trigger K⁺ efflux, thereby secondarily activating the NLRP3 inflammasome. Yersinia suppresses this defence mechanism by preventing hypertranslocation of YopB and YopD via another effector, YopK. (ii) The NLRP1B inflammasome is activated by bacterial effectors such as B. anthracis LeTx or Shigella IpaH7.8, which cleave or degrade its N terminus, respectively. (iii) Effectors, such as YopE and TcdB, modulate the actin cytoskeleton by suppressing Rho GTPases and, consequently, trigger the pyrin inflammasome by inactivating the PKNs, which are sensitive to GTPase activity. Notably, Yersinia YopM overrides this sensing pathway by directly activating PKNs, thereby maintaining pyrin suppression.

Detection of proteolytic activity by NLRP1B

Bacillus anthracis produces a pore-forming toxin called lethal toxin (LeTx). LeTx is a bipartite toxin composed of a protective antigen, which binds to cellular receptors and forms an oligomeric pore within endosomal membranes, and a lethal factor (LF) that is translocated through the pore²⁸. Once inside the host cell, LF cleaves and inactivates mitogen-activated protein kinase (MAPK) kinases, thereby disrupting innate immune signalling pathways and triggering apoptosis²⁸. Additionally, LeTx induces inflammatory cell death in macrophages from certain strains of inbred mice, while other mouse strains are resistant to this effect, suggesting a host-mediated mechanism of cell death²⁹ which was mapped to polymorphisms in the Nlrp1b locus³⁰.

NLRP1B was the first NLR described to assemble an inflammasome complex³¹. NLRP1B is activated in response to cleavage by B. anthracis LeTx, indicating that it acts as a guard to sense this pathogen-specific activity. However, precisely how cleavage by LF enabled NLRP1B activation was enigmatic. NLRP1B contains a function-to-find (FIIND) domain that undergoes autoproteolysis following initial biosynthesis^32–34. This autoproteolysis occurs independently of any stimulus, resulting in a small polypeptide that inhibits NLRP1B activation by remaining non-covalently associated with the remaining C-terminal portion of the NLRP1B protein^35,36. LF-dependent proteolytic cleavage of this auto-inhibitory fragment leads to its recognition by cellular ubiquitin ligases, followed by its degradation via the N-end rule pathway. This releases the C-terminal portion of NLRP1B, which contains the caspase activation and recruitment domain (CARD) domain that recruits and activates caspase-1 (refs. ^35,36) (Fig. 3). Mouse strains containing NLRP1B that is sensitive to LeTx are thus able to induce an inflammasome response to the proteolytic activity of LF.

Interestingly, pathogen effectors that are not directly proteolytic can also activate NLRP1B. The Shigella type III-secreted effector IpaH7.8 is a ubiquitin ligase whose activity triggers assembly of the NLRP1B inflammasome³⁶ (Fig. 3). Given that LF activates the NLRP1B inflammasome by inducing ubiquitin-dependent degradation of an auto-inhibitory fragment, NLRP1B may have evolved to sense at least two activities common to many pathogen virulence factors—proteolysis or ubiquitin ligase activity—either of which can induce degradation of the NLRP1B auto-inhibitory fragment. The protozoan parasite Toxoplasma gondii also activates the NLRP1B inflammasome^37–39, but the factor responsible is unknown. NLRP1B may, therefore, sense a yet-to-be discovered T. gondii ubiquitin ligase or protease.

Mice have two additional NLRP1 homologues, a functional NLRP1A and a putative pseudogene NLRP1C, whereas humans have a single NLRP1 (ref. ³¹). The signals that engage murine NLRP1A and human NLRP1 are unknown. While human NLRP1 is not cleaved or activated by LeTx⁴⁰, in vitro proteolytic cleavage of the human NLRP1 N terminus induces inflammasome activation, suggesting that human NLRP1 is activated via a similar proteolytic mechanism⁴¹. The inability of human NLRP1 to be cleaved by LeTx could, in part, account for the susceptibility of humans to anthrax. As with other NLRs, gain-of-function mutations in human NLRP1 are associated with a variety of autoimmune diseases, suggesting that such mutations may abrogate association of the auto-inhibitory FIIND fragment with the remainder of the protein or enable spontaneous NLRP1 activation in the presence of the FIIND fragment^42,43.

Detection of pore-forming toxins by NLRP3

NLRP3 is activated in response to a diverse array of stimuli, many of which induce changes in cytosolic ion concentrations. A common theme for NLRP3 activation is a drop in intracellular potassium concentration below a critical threshold⁴⁴ and/or mitochondrial dysfunction associated with excessive mitochondrial ROS (mitoROS) production⁴⁵. Whether a specific link exists between mitoROS and ion flux is unclear, although NLRP3-activating stimuli that only induce mitoROS or potassium efflux have been reported^44,46.

Pore formation at the plasma membrane induces potassium efflux, which is a general mechanism of NLRP3 activation⁴⁴. Bacterial pore-forming toxins (PFTs) that permeabilize the plasma membrane, such as aerolysin toxin from Aeromonas spp.⁴⁷ and α-hemolysin from Staphylococcus aureus⁴⁸, cause potassium efflux that triggers the NLRP3 inflammasome (Fig. 3). NLRP3 may thus serve as a guard of cytosolic potassium levels, whereby a drop in potassium levels below a key threshold serves as a HAMP for NLRP3 activation, which mediates ETI by engaging caspase-1 and pyroptotic cell death²⁷.

Detection of intracellular membrane disruption

In addition to PFTs, bacteria employ other virulence factors that have the capacity to activate the NLRP3 inflammasome through disruption of intracellular membranes. This involves guanylate binding proteins (GBPs), although the precise mechanisms remain unclear. GBPs are IFN-inducible GTPases involved in cell-intrinsic immunity against intracellular pathogens, including bacteria, viruses and protozoa^49–53. GBPs are recruited to pathogen-containing vacuoles, where they are thought to deliver a variety of cargo important for host defence⁵⁰. Precisely how GBPs target pathogen-containing vacuoles is unclear, but GBPs are recruited to vacuolar membranes containing secretion system components^54,55 and are thought to mediate inflammasome activation as a result of damage to these membranes.

Gram-negative bacterial secretion system effectors include factors that modulate cellular signalling pathways to evade antibacterial immune defences⁵⁶. However, in doing so, they also activate cellular cytosolic proteins that initiate ETI responses to counteract this activity. Yersinia species possess a type III secretion system that is formed by the Yersinia outer proteins B and D⁵⁷. While these proteins form the translocation complex that mediates the secretion of other effector proteins, YopB and D themselves can also be translocated into the host cell^57–62. Following injection into host cells, YopD colocalizes with markers of damaged membranes and GBP2 (ref. ⁵⁵). Disruption of intracellular membranes by YopB and D leads to activation of the NLRP3 inflammasome, as well as the non-canonical caspase-11 inflammasome that detects bacterial lipopolysaccharide (LPS) in the cytosol (Fig. 3)^55,62,63. GBPs are necessary for this inflammasome response⁵⁵, and they are proposed to activate caspase-11 via a number of mechanisms involving disrupting pathogen-containing vacuoles or directly lysing cytosolic bacteria and releasing LPS into the cytosol^49,64–68. GBPs may, therefore, act both as sensors of intracellular membrane integrity and direct effectors of membrane damage in response to pathogen-induced alteration of vacuolar membranes. Interestingly, Yersinia has evolved to evade this ETI response through the injected effector YopK, which prevents host sensing of pathogen activity by limiting translocation of YopB and D^62,63,69.

Viral proteins that disrupt intracellular membranes also activate NLRP3, again linking NLRP3 activation to membrane disruption. The influenza virus M2 protein forms a proton-specific ion channel that pumps protons out of the Golgi lumen, thereby neutralizing the pH of the trans-Golgi network (TGN)⁷⁰. This activity is sufficient to activate the NLRP3 inflammasome in macrophages and dendritic cells⁷¹, suggesting that NLRP3 may serve as a general guard of ion homeostasis within the cell. Intriguingly, a number of other NLRP3 inflammasome activators induce dispersal of TGN, resulting in recruitment of NLRP3 to dispersed TGN membranes via ionic interactions between a poly-basic region of NLRP3 and negatively-charged phosphatidylinositol 4-phosphotate on TGN membranes⁷². Interestingly, NLRP3 has also been found to bind cardiolipin on the mitochondrial membrane via interactions between cardiolipin and the leucine-rich repeat (LRR) domain⁴⁶. Whether this is the mechanism by which influenza M2 protein triggers NLRP3 activation, and whether GBPs might also be involved in this response to influenza virus, is unknown.

Detection of pathogen interference with immune signalling

The seemingly paradoxical ability to generate a protective immune response against pathogens that suppress innate immune defence is critical to surviving such attacks. Pathogen-driven immune subversion thus also triggers ETI, but how such responses might be engaged and how they function to protect the host has only recently become appreciated. Many of these responses involve the host protein receptor-interacting protein kinase 1 (RIPK1), which is a central component of cell fate decisions involving pro-survival, pro-death and inflammatory gene expression pathways during infection⁷³ (Fig. 4). RIPK1 ubiquitination following activation of TLR or tumour necrosis factor (TNF) receptors (TNFRs) promotes assembly of scaffold complexes that activate nuclear factor-κB (NF-κB) and MAPK signalling as well as cell survival programs^73–77. Conversely, activation of TLR or TNFR signalling while NF-κB signalling is blocked triggers RIPK1 kinase activity, resulting in caspase-8-dependent apoptosis^78–80. Alternatively, if caspase-8 is inhibited or deleted, RIPK1 kinase activity activates RIPK3-dependent necroptosis^81–83. Thus, multiple checkpoints exist to release the brake on cell death if intracellular signalling is disrupted by pathogens. Interestingly, in contrast to the classic view of apoptosis as an immunosuppressive form of cell death, caspase-8 can mediate inflammatory cell death in the setting of NF-κB blockade by cleaving the pore-forming protein GSDMD independently of caspases-1 or −11 (refs. ^84,85). Whether caspase-8-mediated GSDMD cleavage occurs only in the context of NF-κB inhibition is unclear; however, this may be a mechanism to couple apoptotic caspases with inflammation during pathogen inhibition of cell signalling.

Fig. 4 |. Pathogen manipulation of core cellular processes and signalling pathways induces host immune responses.

In addition to inflammasomes, the host has a myriad of other cellular defence pathways activated in response to virulent activity. Many of these pathways trigger signalling cascades that upregulate a subset of innate immune genes that promote host cell defence. Furthermore, other mechanisms, such as autophagy and cell death, serve to eliminate invading pathogens and infected cells. The following are examples of effector-triggered responses in these categories and pathogenic adaptations to evade these host responses. (i) Pathogen-induced amino acid starvation suppresses mTOR, activates autophagy and activates GCN2, leading to a block in host protein synthesis. Inhibiting protein translation may be useful against pathogens that co-opt host cellular processes for their own protein production, such as viral pathogens. Some pathogen effectors, such as Listeria PlcA and PlcB, can block mTOR signalling and autophagy, respectively, in order to evade detection and destruction. (ii) Similarly, pathogens that co-opt the host ER, such as Legionella and Brucella, induce ER stress and consequently trigger a block in protein translation. This activity triggers NF-κB and MAPK signalling and induces expression of subset of proinflammatory cytokines. To counteract this immune response, Legionella also possesses effectors that suppress ER stress and host protein translation, thereby partially masking itself from effector-triggered immunity. (iii) Activation of Rho GTPases by SopE and CNF1 similarly activates the NF-κB pathway to induce inflammatory cytokines and promote cell survival. Several pathogens have evolved to suppress immune signalling pathways in order to limit inflammation. However, in doing so, they activate an ETI pathway mediated by RIPK1 kinase activity, which induces host cell death. Effectors, such as YopJ, that suppress NF-κB and MAPK pathways induce RIPK1-dependent apoptosis. Pathogens that inhibit apoptosis may induce a back-up cell death mechanism, namely RIPK1-dependent necroptosis. Despite these fail-safes, some pathogens have evolved to suppress both of these responses: EPEC inhibits both apoptosis and necroptosis with the effectors NleB, NleF and EspL; the virus murine cytomegalovirus (mCMV) uses its effector viral inhibitor of RIP activation (vIRA) to suppress necroptosis.

Several bacterial virulence factors—including the YopJ/P proteins in Yersinia, the NleE/OspZ proteins from enteropathogenic Escherichia coli and Shigella, and the Shigella effector IpaH9.8—block innate signalling by interfering with key upstream kinases TGF-β-activated kinase 1 (TAK1) and inhibitor of κB kinase (IKK)^78,86,87. YopJ belongs to a family of evolutionarily conserved virulence factors that interfere with activation of MKK family proteins by acetylating a conserved serine residue in the MKK activation domain^88–90. In macrophages, this blockade triggers caspase-8-dependent apoptosis^91,92 (Fig. 4), while in dendritic cells, it disrupts TIR domain-containing adaptor inducing interferon-β (TRIF)-dependent gene expression⁹³. RIPK1 kinase activity is required for apoptosis in response to YopJ activity, and RIPK1-mediated apoptosis protects against Yersinia infection in vivo^78,79,91. Thus, despite YopJ-mediated inhibition of innate immune signalling, RIPK1- and caspase-8-dependent apoptosis triggered by this activity mediates a host-protective immune response. RIPK1 kinase-induced apoptosis is thus intimately tied to the integrity of signalling pathways that RIPK1 itself regulates.

Mechanistically, the switch between RIPK1 pro-death and pro-survival functions is regulated by post-translational modifications that control RIPK1 activity, including phospho-sites on RIPK1 that are targets of the same kinases activated by TLRs or TNFR. Specifically, IKK phosphorylates RIPK1 on Ser25 (refs. ^94,95), and the p38 MAPK-activated kinase, MK2, also phosphorylates RIPK1 at Ser321 and Ser336 (refs. ^96,97). Phosphorylation at these sites prevents RIPK1 from autophosphorylating on Ser166, thus preventing RIPKl-mediated apoptosis and necroptosis^95–97. The lack of phosphorylation at these negative regulatory sites when IKK and MKK signalling are blocked by virulence factors, such as YopJ, thus appears to release the brake on RIPK1-dependent death. RIPK1 therefore functions both as a central regulator and guardian of innate immune signalling.

Evasion of RIPK1-mediated cell death

Given the importance of maintaining cellular integrity for pathogen replication, it is not surprising that pathogens block both apoptosis and necroptosis. For example, human cytomegalovirus (hCMV) protein viral inhibitor of caspase activation (vICA) inhibits apoptosis by preventing caspase-8 activation⁹⁸ (Table 1) and also suppresses necroptosis via the immediate-early 1 (IE1) gene⁹⁹. The murine CMV effector vIRA performs an analogous function and suppresses RIPK3 activity during the caspase-8 blockade to inhibit necroptosis and promote viral replication¹⁰⁰. Similarly, enteropathogenic E. coli (EPEC) possess multiple effectors that block both apoptosis and necroptosis, including NleB, NleF and EspL^101,102 (Fig. 4 and Table 1). The interplay of microbial evasion and host sensing strategies is such that layers of selective pressure on both the host and pathogen have produced a system in which necroptosis is initiated in response to pathogen inhibition of caspase-8 and the pathogen, in turn, blocks necroptosis to maintain its replicative niche and limit release of intracellular DAMPs. Whether an additional layer of host recognition and pathogen evasion exists beyond programmed necrosis remains to be determined.

Table 1 |.

Examples of ETI and pathogen subversion strategies

Pathogenic activity	Host guardee	Host guard	Host ETI response	Pathogen subversion mechanism	References
Inactivation of RhoA by C. difficile TcdB and Yersinia YopE	RhoA	Pyrin	Pyrin-dependent pyroptosis and IL-1 cytokine secretion	Yersinia YopM inhibits pyrin activation by activating PKNs	104,108
Modulation of Rho GTPases by UPEC CNF1, Salmonella SopE and SopB, and Shigella IpgB2 and OspB	Rho GTPase functions	Drosophila IMD, Rho GTPases (guard themselves) and NOD1/NOD2	Drosophila IMD signalling and NF-κB signalling-mediated cytokine secretion		114–121
Activation of NLRP1B by B. anthracis LeTx and Shigella flexneri IpaH7.8	NLRP1B (decoy)	NLRP1B (guards itself)	NLRP1B-dependent pyroptosis and IL-1 cytokine secretion		30,35,162
Activation of NLRP3 inflammasome by Aeromonas aerolysin, S. aureus α-hemolysin, and influenza A virus M2 and PB1-F2 proteins	Cytoplasmic homeostasis (intracellular K+ levels and pH)	NLRP3	NLRP3-dependent pyroptosis and IL-1 cytokine secretion	Measles V protein suppresses NLRP3-mediated IL-1β secretion and influenza A virus NS1 protein suppresses NLRP3 and NF-κB signalling	71,163–166
Activation of caspase-11 and NLRP3 inflammasomes by Yersinia YopB and YopD	Integrity of intracellular membranes	GBPs	Caspase-11 and NLRP3-dependent pyroptosis and IL-1 cytokine secretion	Yersinia YopK prevents hypertranslocation of YopB and YopD, thereby suppressing inflammasome activation	55,62,63,69,167
Amino acid starvation induced by Shigella, Salmonella T3SS and Listeria LLO	Amino acid levels in the host cell	mTOR GCN2	Autophagy Inhibition of host protein translation	Listeria PlcA and PlcB suppress autophagic flux, and Leishmania GP63 and vesicular stomatitis virus M protein suppress mTOR	125–127,168,169
ER stress induced by B. abortus VceC, L. pneumophila and P. aeruginosa exotoxin A	Host protein translation	IRE1α, IκB, NOD1/2 and C. elegans ZIP2	NOD1/2-induced cytokine secretion via NF-κB signalling, NF-κB/MAPK-mediated cytokine secretion and C. elegans infection response gene 1 (irg1)-mediated defence	Legionella Lgt1 and Lgt2 block IRE1-mediated Xbp1 splicing	138,146–148,155,170
Activation of apoptosis by Yersinia YopJ, EPEC NleE and Shigella OspZ	NF-κB/MAPK pathway	RIPK1	RIPK1/caspase-8-dependent apoptosis	EPEC NleB and NleF block caspase-8-mediated apoptosis	78,86,101
Activation of necroptosis by CMV vICA and EPEC NleB	Caspase-8	RIPK3	RIPK3-dependent necroptosis	EPEC NleF blocks necroptosis	98,101
Activation of apoptosis by Aspergillus fumigatus gliotoxin	Cellular attachment to extracellular matrix	RhoA	Activation of anoikis (a form of apoptosis induced by cell detachment)		171
Activation of Toll pathway by Beauveria bassiana PR1	Drosophila persephone protease (decoy)	Drosophila persephone protease (guards itself)	Upregulation of Drosophila Toll pathway-dependent gene expression		172

Detection of Rho GTPase modification by pathogens

Rho GTPases—such as Rac1, RhoA and Cdc42—serve numerous core cellular functions, ranging from actin cytoskeletal dynamics and epithelial barrier integrity to production of ROS and antimicrobial peptides, and are thus targeted by numerous pathogens¹⁰³. Similar to RIPK1, Rho GTPases are centrally positioned to regulate cellular processes and are therefore poised to be both targets and sensors of pathogen virulence activity.

Detection of RhoA inactivation by the pyrin inflammasome

Bacterial inactivation of RhoA can trigger inflammasome responses through the host protein pyrin¹⁰⁴. Pyrin forms an inflammasome that recruits caspase-1, leading to release of IL-1 cytokines and pyroptosis (Fig. 3). Pyrin activation is normally restrained by RhoA through the activity of protein kinase C-related kinases (PKNs, also known as PRKs). When RhoA is active, one of its targets, the PKNs, phosphorylate pyrin, creating binding sites for suppressive 14-3-3 regulatory proteins^105,106. Thus, under basal conditions, the pyrin inflammasome is maintained in an inactive state. Similar to RIPK1 apoptotic activity being kept in check by IKK-dependent phosphorylation, pathogen inactivation of RhoA triggers the pyrin inflammasome by preventing phosphorylation of pyrin. Importantly, a number of pathogens are sensed via the linking of pyrin to RhoA activity.

Burkholderia cenocepacia inactivates RhoA via direct deamidation, thereby disrupting macrophage effector functions^104,107. However, the resulting derepression of pyrin leads to an inflammatory response that controls B. cenocepacia infection¹⁰⁴. Similarly, other RhoA-inactivating toxins, such as Clostridium botulinum C3, Clostridium difficile TcdB, Vibrio parahaemolyticus VopS and Histophilus somni IbpA, also activate the pyrin inflammasome¹⁰⁴. Only catalytically active bacterial effectors trigger this inflammasome response, indicating that pyrin senses the activity of effectors and their disruption of RhoA function¹⁰⁴.

In an elegant interplay of host–pathogen co-evolution strategies, pathogenic Yersinia have evolved to evade the pyrin inflammasome in response to manipulation of RhoA (Fig. 3 and Table 1). Yersinia effectors YopE and YopT inactivate the Rho GTPases RhoA, Rac1 and Cdc42 to prevent phagocytosis of the bacteria⁵⁶, which would trigger pyrin by disrupting its interactions with 14-3-3 proteins^108–110. However, the Yersinia effector, YopM, specifically activates PKN kinases to phosphorylate pyrin, thereby maintaining pyrin’s association with 14-3-3 proteins and masking YopE/T interference with Rho GTPases¹⁰⁸.

Gain-of-function mutations in Mediterranean Fever (MEFV), the gene encoding human pyrin, are associated with Familial Mediterranean fever, an illness characterized by recurrent inflammation in the absence of infection¹¹¹. The high prevalence of mutant MEFV alleles in Mediterranean populations may be the result of plague epidemics within Europe and the possibility that MEFV carriers have increased protection from Y. pestis infection^108,112,113. YopM is among the most polymorphic of Yersinia virulence factors, suggesting that YopM itself may be a target of host selective pressure. There may exist other pathogen virulence factors that similarly overcome host immune defences by targeting other guard proteins.

PRR signalling in response to Rho GTPase disruption

In addition to inflammasome activation, pathogen modulation of Rho GTPases can induce conserved transcriptional responses via NF-κB. The toxin cytotoxic necrotizing factor 1 (CNF1) from uropathogenic E. coli (UPEC) constitutively activates the Drosophila Rho GTPase Rac2 by preventing GTP hydrolysis¹¹⁴. Modified Rac2 interacts with the Drosophila homologue of RIPK1, immune deficiency (IMD), and induces transcription of Drosophila antimicrobial peptides that control bacterial infection¹¹⁴. A constitutively active Rac2 mutant also induces IMD-dependent NF-κB activation and Rac2-deficient flies succumb to UPEC infection¹¹⁴, indicating that one function of Rac2 is to detect the activity of pathogen virulence factors such as CNF1 (ref. ¹¹⁴). Importantly, CNF1-modified human Rac2 interacts with the human homologues of IMD and also induces expression of antimicrobial peptide genes¹¹⁴. These findings imply that detection of bacterial toxins that target Rac2 represents an evolutionarily conserved ETI pathway.

Similarly to CNF1, the Salmonella effectors SopB, SopE and SopE2 activate Rho GTPases, thus triggering innate immune activation^115–117. While SopB/E/E2 are necessary for Salmonella intracellular invasion, their injection activates NF-κB and MAPK signalling in epithelial cells, indicating that inappropriate activation of Rho GTPases engages a conserved immune defence circuit. Ectopic expression of SopE induces NF-κB activation via a complex containing Rho GTPases, NOD1, RIPK2 and heat shock protein 90 (HSP90)¹¹⁵. However, inflammatory responses to these effectors can also occur independently of RIPK2 (refs. ^116,117). Indeed, SopE/E2/B activity can activate p21-activated kinase (PAK1) together with recruitment of TNF receptor-associated factor 6 (TRAF6) and TAK1, resulting in NF-κB and MAPK activation independently of RIPK2 (ref. ¹¹⁷). Distinct signalling platforms may have the capacity to engage NF-κB in response to pathogen manipulation of Rho GTPase activity, perhaps depending on cell type, route of delivery or expression level of pathogen effectors.

Like Salmonella, Shigella interacts intimately with the actin cytoskeleton during invasion of intestinal epithelial cells, but also intracellularly when it employs actin-based motility. Shigella’s manipulation of the actin cytoskeleton also induces NF-κB signalling via NOD1/2 (refs. ^118–120). In unpolarized epithelial cells or macrophages, the phosphatase slingshot 1 (SSH1), which regulates actin depolymerization through cofilin, activates NOD1-dependent NF-κB signalling during Shigella infection, suggesting that disruption of actin filaments triggers the NOD1 pathway¹¹⁸. Consistently, chemical inhibitors of actin polymerization, such as latrunculin B and cytochalasin D, also activate NOD1 (ref. ¹¹⁸). In polarized epithelial cells, cytoskeleton disruption by Shigella effectors IpgB2 and OspB activate NOD1 in a manner dependent on the Rho-associated protein kinase (ROCK)¹¹⁹. Collectively, these and other studies that link modulation of Rho GTPase activity to NOD1/2 signalling^120,121 demonstrate that actin-regulating Rho GTPases are central regulators of ETI responses and serve as important guardians of actin cytoskeleton dynamics.

Detection of pathogen-induced stress responses

Adaptation to environmental stresses is essential for cell survival and growth. The integrated stress response (ISR) is activated by a variety of stimuli, including amino acid starvation, oxidative stress and ER stress, and is required for cell survival¹²². The ISR blocks general mRNA translation while selectively allowing expression of genes that enable specific adaptations that promote cellular recovery¹²². The mammalian target of rapamycin (mTOR) is also an important monitor of nutrient starvation and intersects with the ISR¹²³. Notably, both the ISR and mTOR pathways are also coupled to inflammation and cell death in response to pathogen manipulation of amino acid pools and ER stress¹²², highlighting the link between disruption of core cellular physiologic functions to immune responses during infection.

Detection of pathogen-induced amino acid starvation

Metabolic stress in the form of amino acid starvation is sensed by the cell via two signalling cascades: the mammalian target of rapamycin (mTOR) and eIF2α¹²³. While both pathways sense amino acid availability independently, they intersect with and regulate many other cellular responses, including cell division, gene expression, cell survival and cell death¹²³. Similar to other core processes discussed above, these pathways are both targeted by intracellular pathogens and also play a key role in sensing pathogen-specific disruption of the cell.

The mTOR complex 1 (mTORC1) is normally found on the membrane of late endosomes and lysosomes where it regulates gene transcription via eukaryotic translation initiation factor 4E (eIF4E)-binding protein 1 (4E-P1) and S6 kinase 1 (S6K1), and also negatively regulates autophagy under nutrient-replete conditions¹²³. mTORC1 senses amino acid availability inside lysosomes via the transmembrane protein SLC38A9 and in the cytosol via Sestrins and CASTOR proteins¹²⁴. In response to a drop in amino acid levels, mTORC1 dissociates from the membrane, which initiates autophagy to break down cellular components in order to replenish nutrients¹²³. Therefore, mTORC1 acts as a guard of intracellular amino acid levels and induces an autophagic response when these levels drop.

Intracellular pathogens compete with the host cell for essential building blocks, including amino acids¹²³. Thus, similarly to T3SSs above, an activity necessary for pathogen survival or virulence simultaneously enables host detection and immune defence. Intriguingly, disruption of mTOR signalling during bacterial infection can co-opt the autophagic machinery to engulf the pathogen itself, leading to degradation of the pathogen via xenophagy¹²³. Damage to phagosomal membranes induced by bacteria that invade the cytosol is thought to induce amino acid starvation and activate xenophagy¹²³. In this manner, both the Listeria pore-forming toxin listeriolysin O (LLO) that breaks down the pathogen-vacuole membrane¹²⁵ and the T3SSs from Shigella and Salmonella induce host membrane damage that triggers xenophagy^126,127. Since the amino acid sensor SLC38A9 is necessary for transporting leucine out of lysosomes to interact with and activate mTORC1 (ref. ¹²⁸), pathogen-induced membrane disruption may impede the ability of SLC38A9 to activate mTORC1.

Given the antibacterial nature of this pathway, successful intracellular pathogens must possess mechanisms to either subvert or escape downstream consequences of mTOR signalling. An example of such an activity are the Listeria phospholipases, PlcA and PlcB, which stall maturation of autophagy vesicles¹²⁵. Other pathogens, such as Toxoplasma gondii¹²⁹ and certain viruses^130–132, promote mTOR signalling, likely to maintain this pathway’s function and avoid autophagy.

The eIF2α pathway similarly senses amino acid starvation via the cytosolic kinase general control nonderepressible 2 (GCN2). GCN2 is activated in response to a drop in amino acid concentrations and phosphorylates eIF2α, inducing a block in host translation and the formation of mRNA-containing stress granules¹²³. Like the mTOR pathway, GCN2 is activated during bacterial infection as a result of membrane damage¹²⁶, and both pathways can be activated simultaneously during bacterial infection^125,126. Notably, despite global translational repression induced by inactivation of mTOR and activation of GCN2, expression of key inflammatory genes was selectively upregulated in response to Shigella infection¹²⁶. Thus, both GCN2 and mTOR act as guards of amino acid levels in the host cell and their activation, in response to pathogen virulence activity, promotes anti-microbial responses.

Sensing disruption of protein synthesis and ER homeostasis

ER perturbance also engages the integrated stress response pathway and can be coupled to ETI. Legionella, Chlamydia trachomatis and Brucella spp. all replicate within an ER-derived compartment¹³³. However, these pathogens can be detected via their interactions with the ER stress and unfolded protein response (UPR) pathways^133,134. The UPR engages three different transmembrane receptors, inositol requiring enzyme 1 (IRE1α), double-stranded RNA-activated protein kinase (PKR)-like ER kinase (PERK) and activating transcription factor 6 (ATF6), which sense ER stress and induce downstream responses to alleviate the stress and restore cellular homeostasis¹³⁴. Like GCN2, PERK phosphorylates eIF2α, leading to attenuated protein synthesis¹²³. However, a prolonged UPR response that fails to be resolved, or an acute response that is extreme in amplitude, triggers autophagy or apoptosis of infected cells¹³⁴. Activation of UPR in response to bacterial virulence activity serves as an ETI mechanism whereby receptors of the UPR pathway act as guards of ER homeostasis in the context of infection.

The Legionella pneumophila type IV secretion system (T4SS) injects a large panel of effectors to establish an ER-derived vacuole and replicate within macrophages¹³⁵. Notably, L. pneumophila infection induces UPR in a manner dependent on T4SS activity, suggesting the presence of an ETI response to pathogen disruption of the ER^136,137. However, L. pneumophila also encodes at least 11 effectors that inhibit host translation, perhaps to alleviate ER stress^138–145. Indeed, several of these effectors, particularly Lgt1 and Lgt2, block the ATF6 and IRE1 branches of the UPR^136,137. Nevertheless, in another example of the intimate interactions between host and pathogen, the translational block induced by these Legionella effectors triggers another layer of immune response, as it prevents resynthesis of the NF-κB inhibitor IκB, leading to prolonged NF-κB signalling¹³⁸. The effector-driven translational block also activates p38 and c-Jun N-terminal kinase (JNK) MAPKs¹⁴⁶, and this enhanced NF-κB and MAPK signalling amplifies expression of immune genes, including Il1a and Il1b^{138,143,146–149}. Although infected macrophages produce IL-1α and IL-1β proteins, they are unable to synthesize other key cytokines, including TNF and IL-12, due to the translational block¹⁴⁹. Instead, secretion of IL-1α and IL-1β by infected macrophages is critical for production of TNF and IL-12 by uninfected bystander monocytes and other recruited myeloid cells within the lung¹⁴⁹, indicating that collaboration between infected and uninfected cells is critical for appropriate host defence.

Intriguingly, TLR stimulation coupled with pharmacological inhibition of host translation induces an overlapping set of immune response genes to Legionella infection¹³⁸, indicating that this program represents an immune response disrupting host protein synthesis. How the infected host cell maintains its ability to synthesize a subset of proteins despite the translational block is unclear, but superinduction of a subset of mRNAs is involved^144,148. Collectively, these findings reveal that although Legionella effectors inhibit host translation to alleviate ER stress, this activity triggers an ETI response that promotes NF-κB and MAPK signalling as well as superinduction of immune genes, which allows infected macrophages to synthesize IL-1α and IL-1β. Subsequent release of these DAMPs elicits inflammatory cytokine production from uninfected bystander cells, thus providing a backup mechanism to ensure inflammatory responses. Such a strategy may be a general mechanism employed by the immune system to overcome immunoevasive pathogens that inhibit host translation.

In C. elegans, sensing pathogen disruption of core cellular processes is the primary pathogen detection mechanism, as direct receptors for PAMPs have yet to be identified. Infection with pathogenic bacteria, or pharmacological disruption of translation, cytoskeletal components, mitochondrial function or proteasome activity, triggers a concerted set of cellular protective responses known as ‘surveillance immunity’^150,151. This response includes upregulation of an antimicrobial transcriptional effector program as well as organismal responses, such as avoidance behaviour, reduced ingestion and xenobiotic detoxification¹⁵². The Pseudomonas aeruginosa effector exotoxin A (ToxA) blocks host translation by ADP-ribosylating the translation elongation factor-2 (EF-2)¹⁵³, which paradoxically triggers induction of the transcription factor BZIP domain-containing protein (ZIP-2) and subsequent expression of a suite of infection response genes that mediate anti-bacterial defence^154–156. The ToxA-mediated translational block promoted ZIP-2 protein expression through a mechanism involving an upstream open reading frame (uORF) that normally suppresses ZIP-2 translation in the absence of infection¹⁵⁵. Interestingly, uORF-regulated translation is an evolutionarily conserved mechanism in yeast and mammals that is also employed during non-infectious cellular stress to enable selective translation of genes important for stress adaptation^157–159. Thus, uORF-regulated protein expression may be similarly used in mammalian cells to enable immune defence against pathogens that block host translation.

These studies highlight the role of the ER stress response pathway as a sensor of pathogens that co-opt and replicate in ER-derived compartments. Interestingly, pathogens have evolved to block host protein synthesis as a strategy to prevent the host from mounting an immune response. To counteract this, the host detects this disruption by overcoming the translation block and triggering defence mechanisms.

Concluding remarks and future perspectives

The concept that innate immune systems of eukaryotes can detect the activity of pathogen virulence factors, termed ‘ETI’, was proposed in the early 1980s, with the discovery that plant pathogen virulence (Avr) proteins are sensed by plant innate immune pathways^18,19,23,160. The Pattern Recognition Theory, proposed by Janeway in 1989 and confirmed experimentally in the subsequent decade, provided important insight into innate recognition of microbial pathogens in animal systems^1,3,161. However, metazoans require additional mechanisms to detect pathogens, since commensal microorganisms engage in intimate interactions with their eukaryotic hosts and possess PAMPs that are sensed by PRRs². Thus, pathogen manipulation of key cellular processes must be detected by cellular sensors to enable host survival in the face of such a pathogenic insult. The existence of ETI responses as an additional layer of immune defence that functions in concert with PRR-induced responses is necessitated by the presence of microorganisms that have developed mechanisms to evade, subvert or take advantage of PRR-driven immune defence. In some instances, PRR signalling ‘primes’ host responses to engage ETI in response to a ‘second signal’ that involves manipulation of cell physiology. In this sense, ETI functions as a threat assessment system that detects particularly problematic pathogens, because engagement of ETI only occurs if a pathogen has penetrated other cellular defences. Engagement of ETI, therefore, usually leads to some form of immunologic or inflammatory cell death. Given the potential pathological inflammation that can be elicited by ETI, it makes sense that engaging this pathway would generally be limited to situations in which multiple signals denote pathogen presence.

Combinatorial engagement of PRRs and ETI engages responses that are distinct from those triggered by PAMPs alone, and are therefore particularly important for immune defence against pathogens. Interestingly, as might be expected, given the selective pressure put on pathogens by ETI systems, pathogen mechanisms for evasion of ETI have also emerged. Thus, metazoan-adapted pathogens likely have an array of effector mechanisms that limit or blunt the generation of effective immune responses. This is likely to be the case, particularly in the setting of chronic infections—a class of infection for which we still lack effective antimicrobial treatments or vaccines. Further understanding of ETI pathways and the way in which pathogens specifically subvert these responses may facilitate development of antimicrobials and vaccine strategies.