The pyrin inflammasome and the Yersinia effector interaction

Abstract

Pyrin is a cytosolic pattern-recognition receptor that normally functions as a guard to trigger capase-1 inflammasome assembly in response to bacterial toxins and effectors that inactivate RhoA. The MEFV gene encoding human pyrin is preferentially expressed in phagocytes. Key domains in pyrin include a pyrin domain (PYD), a linker region and a B30.2 domain. Binding of ASC to pyrin by a PYD-PYD interaction triggers inflammasome assembly. Pyrin is held in an inactive conformation by negative regulation mechanisms to avoid premature inflammasome assembly. One mechanism of negative regulation involves phosphorylation of the linker by PRK kinase which in turn is positively regulated by active RhoA. The B30.2 domain also negatively regulates pyrin. Gain of function mutations in MEFV responsible for the autoinflammatory disease Familial Mediterranean Fever (FMF) map to exon 10 encoding the B30.2 domain. Insights into pyrin regulation have come from studies of several Yersinia effectors, which are injected into phagocytes and interact with the RhoA-PRK-pyrin axis during infection. Two effectors, YopE and YopT, inactivate RhoA to disrupt phagocytic signaling. To counteract an effector-triggered immune response, a third effector, YopM, binds to and inhibits pyrin by hijacking PRK and RSK and directing linker phosphorylation. Inhibition of pyrin by YopM is required for virulence of Yersinia pestis, the agent of plague. Recent results from infection studies with human phagocytes and mice producing pyrin B30.2 FMF variants show that gain of function MEFV mutations bypass inhibition by YopM. Population genetic data suggest that MEFV mutations were selected for in individuals of Mediterranean decent during historic plague pandemics. This review discusses current concepts of pyrin regulation and its interaction with Yersinia effectors.

INTRODUCTION

Inflammasomes function to sense danger signals within the cell cytosol and trigger innate immune responses. For example, delivery of pathogen molecules into host cells can trigger inflammasomes leading to innate immune responses ¹. Caspase-1 inflammasome assembly is controlled by distinct pattern recognition receptors (PRRs) or sensors that become activated in response to danger signals. Some activated sensors bind to the adaptor ASC (apoptosis-associated speck-like protein), which oligomerizes into filaments. Caspase-1 is recruited to these filaments, dimerizes and undergoes auto-cleavage into its mature form. Mature caspase-1 cleaves gasdermin-D (GSDMD), releasing a domain which forms oligomeric pores in plasma membranes. Caspase-1 also acts on pro-IL-1β and pro-IL-18 to convert these proinflammatory cytokines into their mature forms, which are released through GSDMD pores ¹. Pore formation by GSDMD can also result in inflammatory cell death termed pyroptosis ^2,3. The production of IL-1β, IL-18 and pyroptosis can be protective or cause immunopathology depending on the context of the host-pathogen interaction.

Pyrin is a unique senor that normally functions to trigger capase-1 inflammasome assembly in response to bacterial toxins or effectors that inactivate the small GTPases RhoA, RhoB or RhoC ^4–6. These three Rho isoforms appear to be functionally redundant in regulating pyrin activity ⁴ and will be referred to as RhoA from here on. Pyrin in humans is encoded by the MEFV gene and is primarily expressed phagocytes such as neutrophils, monocytes and activated macrophages ⁵. Human pyrin consists of four recognized domains: The N-terminal PYD (residues 1–92), the B-box domain (residues 370–412), the coiled-coil (CC) domain (residues 420–582), and the C-terminal B30.2 domain (residues 597–781). In between the PYD and the B-box is a 278 amino acid-long linker region ⁵. This domain structure is shown schematically in Fig. 1, which also presents a speculative model of pyrin regulation. B-box, coiled-coil and B30.2 domains are found in TRIM proteins, like TRIM5a ⁷, and pyrin (TRIM20) is considered a member of this family ⁸. Pyrin has a PYD in place of the RING domain that confers ubiquitin ligase activity in most TRIM proteins. Murine pyrin (encoded by Mefv gene) has the same general structure but has a short undefined C-terminal tail in place of the B30.2 domain. There is evidence that the CC domain of human pyrin mediates formation of homodimers ⁸ or homotrimers ⁹.

Fig. 1.

Regulation of pyrin and the Yersinia effector interaction. The structure and proposed regulation of human pyrin and interactions with Yersinia effectors in an infected phagocyte are shown schematically. Pyrin is shown with pyrin domain (PYD), linker containing two serine residues (S) subject to phosphorylation, B-box domain, coiled-coil (CC) domain, and B30.2 domain indicated. Numbers indicate the following proposed steps in regulation, with solid arrows representing concepts experimentally established in cells, and dashed arrows representing concepts from in vitro data or are hypothetical. Steps 1–6: Negative regulation of pyrin. 1) RhoA cycles between GDP- or GTP-bound conformation, and in the latter activates PKN to phosphorylate S208 and S242 in pyrin linker; 2) 14–3-3 dimer binds to phosphorylated linker; 3) PYD is masked by binding to B-box; 4) CC domain promotes anti-parallel dimer formation; 5) B30.2 domains in a dimer interact with a proline-rich stretch in CC domain of neighboring dimers to form tetramers; 6) Pro-IL-1 β binds to CC-B30.2 domains of pyrin multimers. The dashed circle encompasses possible inactive conformations. Steps 1–3 are shown with pyrin monomers for simplicity but may occur in a multimeric form. Steps 4–6 are derived from in vitro data with a purified CC-B30.2 domain construct (two cross hatches represent truncation of N-termini) and purified IL-1β. Steps 7–10: Positive regulation of pyrin. 7) B30.2 de-repression results in formation of an intermediate parallel trimer, a step that would be constitutive in B30.2 FMF variants. The central pyrin subunit in the trimer is shown with 14–3-3 detached, linker dephosphorylated and PYD unmasked for clarity; 8) Release of 14–3-3 and dephosphorylation of linker by putative protein serine phosphatase (PSP); 9) Unmasking of PYD by B-box via binding of PSTPIP1 PAPA variant is shown as an alternative step; 10) Microtubule (MT) binding results in an active parallel trimer, able to bind ASC via PYD-PYD interaction, triggering ASC filamentation and inflammasome assembly. 11) Colchicine inhibits polymerization of MT, preventing active trimer formation. Steps 12–15: Yersinia effector interaction. 12) YopE acts as a GAP to convert RhoA to GTP bound form; 13) YopT acts as a protease to detach RhoA from the membrane and together with YopE interrupts negative regulation of pyrin by PRK. 14) YopM hijacks PRK and RSK and binds to pyrin to maintain phosphorylation of S208 and S242. Formation of the intermediate trimer is proposed to bypass inhibition of pyrin by YopM. 15) YopJ inhibits expression of IL-1β message.

Gain of function mutations in MEFV, most commonly in the exon corresponding to the B30.2 domain, lead to a monogenic autoinflammatory disease called Familial Mediterranean Fever (FMF) ⁵. FMF is characterized by recurrent episodes of fever and serositis and primarily effects eastern Mediterranean populations ⁵. The dual role of pyrin in regulating immune responses against bacterial pathogens and in causing FMF highlight the importance of research on this caspase-1 inflammasome sensor.

PYD-mediated binding of ASC to pyrin triggers ASC filamentation, inflammasome formation and caspase-1 activation ^9,10. To avoid premature inflammasome assembly, pyrin is subject to multiple mechanisms of negative regulation. Evidence indicates that human pyrin homotrimers can exist in an inactive conformation in which the PYD is masked by the B-box ⁹. The B30.2 domain appears to negatively regulate pyrin, but the mechanism is unclear. More recently, phosphorylation of the linker by a RhoA-protein kinase C-like kinase (PRK; also known as PKN) pathway has been shown to be an important mechanism of negative regulation of pyrin ^11–13. Binding of a 14–3-3 dimer to the phosphorylated linker is proposed to keep pyrin in an inactive conformation ^11–13.

When RhoA is inactivated by a bacterial toxin or effector pyrin activation is likely positively regulated by several mechanisms. After the 14–3-3 dimer dissociates from the linker a phosphatase may act on pyrin ^11,12, but this enzyme has not been identified. Unmasking of the PYD by the B-box would also have to occur ⁹. Finally, microtubules (MTs) regulate pyrin. This is indicated by the inhibitory effect of colchicine and other MT destabilizing drugs on pyrin inflammasome assembly ^9,11,12,14 and the fact that colchicine is a common treatment for FMF patients ⁵. Additionally, pyrin has been shown to co-localize with MT in cells and to bind to paclitaxel stabilized MT in vitro ¹⁵.

Novel insights into the regulation and function of pyrin can be gained through the understanding of its interactions with toxins and effectors of bacterial pathogens ⁶. This may be best exemplified by the interaction of pyrin with Yersinia effectors ^16–19. Yersinia spp. (Yersinia pestis, Yersinia pseudotuberculosis and Yersinia enterocolitica) are bacterial pathogens that use a type III secretion system (T3SS) to inject multiple Yop effectors into phagocytes. YopE and YopT inactivate RhoA as part of a strategy to disarm phagocytic signaling. To counteract an effector-triggered inflammasome response, Yersinia employ the effector YopM. YopM binds to pyrin and hijacks host kinases to maintain inhibitory phosphorylation of the linker ^16,17,19. Inhibition of pyrin by YopM is essential for virulence in Y. pestis, the agent of plague ¹⁷. It has been hypothesized that FMF mutations were selected in the human population as a result of heterozygous advantage against Y. pestis infection during historic plague pandemics ^5,16. Recent population genetic data and experimental results from infection studies with human phagocytes and mice producing pyrin FMF B30.2 variants are consistent with the idea that resistance to plague epidemics was a selective force for MEFV mutations in individuals of Mediterranean decent ¹⁹.

This review expands on the above concepts of the pyrin inflammasome and its interaction with Yersinia effectors at the genetic, biochemical, cellular and immunological levels. The role of pyrin in the context of other host-pathogen interactions has recently been reviewed ⁶. Although not covered here, it is important to point out that pyrin has been implicated in sensing of microbiota metabolites ²⁰ and in other cellular processes such as tumor suppression ²¹ and autophagy ²².

MEFV EXPRESSION

MEFV is located on chromosome 16, spans 15 kb in the 16p13.3 region and contains 10 exons ⁵. The murine homolog Mefv is located on chromosome 16 in mice ²³. MEFV has a restricted pattern of expression, mainly in granulocytes and monocytes ^24,25. Exposure of these cell types as well as macrophages to pro-inflammatory stimuli such as LPS or TNFα can further upregulate MEFV expression ²⁴. Yu et al. observed that production of pyrin and pro-IL-1β is upregulated in THP-1 cells in response to retroviral infection ⁹. TNFα-induced expression of MEFV is dependent on binding of two transcription factors, NF-kB p65 and C/EBPb, to the promoter ²⁵. Studies using mouse bone marrow-derived macrophages (BMDMs) indicate that TNFα signaling is important for TLR ligands such as LPS to upregulate production of pyrin ²⁶. Additionally, TNFα signaling is important for autoinflammatory disease phenotypes in mice with knocked-in B30.2 domain variants ²⁶ (see below).

NEGATIVE REGULATION OF PYRIN

Linker phosphorylation and 14–3-3 binding

Phosphorylation of the pyrin linker creates a binding site for multiple 14–3-3 isoforms, including epsilon (ε) and tau (τ), as first shown by a yeast 2-hybrid approach ²⁷. The linker contains a consensus 14–3-3 binding site sequence ²⁷. Site-directed mutagenesis confirmed that S208 and S242 within this consensus sequence are important for 14–3-3 binding ²⁷. The importance of linker phosphorylation and 14–3-3 binding became evident after three breakthroughs were made in understanding pyrin regulation. The first was the demonstration that the pyrin inflammasome is activated in response to bacterial toxins and effectors that covalently modify RhoA ⁴. The second was the demonstration that the linker is phosphorylated by PRKs ¹¹, which are positively regulated by RhoA, thus connecting pyrin to RhoA. The third was the demonstration that phosphorylation of S208 and S242 (and the corresponding sites in murine pyrin, S205 and S241) is important for negative regulation of pyrin ^11–13. Masters et al. demonstrated that a S242R codon change in MEFV is responsible for a dominantly inherited human autoinflammatory disease, PAAND, that is associated with neutrophilic dermatosis and clinically distinct from FMF ¹³. CD14+ monocytes from PAAND patients that are treated with LPS release a low but significant amount of IL-1β as compared to control monocytes, suggesting that the pyrin S242R variant is constitutively active ¹³. Interestingly, PAAND monocytes treated with LPS and the RhoA toxin TcdB from Clostridium difficile release a much higher amount of IL-1β, indicating that the pyrin S242R variant is not fully constitutively active in unintoxicated cells ¹³. The formation of mixed trimers (see below) containing wild-type pyrin and S242R variants in the heterozygous PAAND monocytes may dampen inflammasome activation.

The negative regulation of pyrin by the RhoA-PRK-pyrin axis occurs in a manner similar to the guard hypothesis observed in plants, whereby a plant disease resistance PRR recognizes modifications to host proteins made by pathogens and induces an innate response ²⁸. In essence, pyrin guards RhoA (the guardee) against bacterial toxins and effectors ⁴. Pyrin does not directly interact with RhoA and is therefore regulated by an indirect guard mechanism ⁴. RhoA functions through a switch like mechanism, cycling between an “on” (RhoA-GTP) and “off” (RhoA-GDP) state. RhoA has numerous roles in phagocytosis, cell cycle and migration and is a common target of bacterial toxins and effectors ²⁹. Membrane-localized RhoA-GTP binds to and activates PRKs, which in turn bind to pyrin and phosphorylate the linker, creating the 14–3-3 binding sites ¹¹(Fig. 1, Steps 1 and 2). Key questions about the mechanism of negative regulation by the RhoA-PRK-pyrin axis remain to be answered. For example, what is the structural basis for the inactive conformation imposed by linker phosphorylation and 14–3-3 binding? One possibility is that linker phosphorylation and 14–3-3 binding stabilizes a conformation in which the PYD is masked by the B-box (see next section).

The PYD and masking by B-box

The PYD is structurally related to the 6-helix bundle death domain-fold family. Other members of this family include the CARDs found in ASC and caspase-1. The PYD is considered the effector domain of pyrin as it binds to the corresponding PYD in ASC to mediate filamentation and trigger inflammasome assembly. Interestingly, the pyrin PYD has three potential binding sites for ASC PYD ³⁰. Molecular docking studies indicate that the pyrin PYD can simultaneously bind to up to three ASC PYDs ³⁰. This has important implications because to avoid premature inflammasome assembly, the pyrin PYD must be kept sequestered in an inactive conformation. How this is achieved in the context of a PYD that has multifaceted binding modes remains an open question. Yu et al. studied pyrin ectopically expressed in HEK293T cells and obtained evidence that the PYD is masked by intramolecular interaction with the B-box ⁹. This is shown schematically using monomeric pyrin in Fig. 1, Step 3. The masking of the PYD could be facilitated by the formation of homotrimers of pyrin as suggested by Yu et al. ⁹. The role of the pyrin B-box in making intramolecular protein contacts is not shared with the B-box in TRIM5a. The main role of the TRIM5a B-box is to form intermolecular trimers of dimers, which allow assembly of a hexagonal lattice on the surface of HIV-1 capsids ⁷. Mechanisms to positively regulate pyrin by unmasking the PYD from the B-box are discussed below.

Dimers, tetramers and pro-IL-1β binding by CC and B30.2 domains

The structure of the pyrin B30.2 domain as determined using protein crystallography revealed a globular domain with a possible substrate-binding cavity ³¹. However, no direct substrate of this cavity has been identified. Major FMF-associated codon changes such as M680I, M694V and V726A correspond to surface exposed residues in the structure, but they are not within the cavity and localize to different sites on the B30.2 domain ³¹.

Weinert et al. used protein crystallography to determine the three-dimensional structure of the CC-B30.2 portion of pyrin ⁸. In the structure, the CC domains mediate dimer formation, with a long alpha helix of one monomer forming an anti-parallel CC with another monomer (Fig. 1, Step 4). This is similar to the main oligomeric state of TRIM5a, which is also a dimer made up of two CC domains packed in an antiparallel manner to form an elongated rod ⁷. The globular B30.2 domains were connected to the CC domains by short flexible hinges and were spaced far apart from each other in the dimers ⁸. This is different from the case in TRIM5a, in which long flexible linkers following the CCs allow B30.2 domains within dimers to pair and bind HIV-1 capsid ⁷. The pyrin B30.2 domains did not form intramolecular contacts within dimers, however tetramers were present in the crystals and tetramer formation was mediated by contacts between B30.2 domains and proline-rich stretches in the CC domains of neighboring dimers (Fig. 1, Step 5). The B30.2 domain residues that contact CC domains in neighboring dimers do not correspond to the putative substrate binding cavity, or to the major FMF-associated codon changes M680I, M694V and V726A ⁸.

Using ectopic overexpression in HEK293T cells and immunoprecipitation Papin et al. obtained evidence that pro-IL-1β binds to full length pyrin as well as subdomains including the B30.2 ³². Following up on this observation, Weinert et al. used purified components to demonstrate that purified pro-IL-1β binds to the CC-B30.2 portion of pyrin at reasonable affinity but not to the individual CC or B30.2 domains ⁸ (Fig. 1, Step 6). Evidence for formation of high molecular weight oligomers of pro-IL-1β and the pyrin CC-B30.2 region was obtained by size exclusion chromatography ⁸. Although B30.2 FMF variants were not tested for pro-IL-1β binding in vitro by Weinert et al. ⁸, binding of pro-IL-1β to full length pyrin or the B30.2 domain in the HEK293T cell ectopic overexpression system was not decreased by a M694V codon change ³².

When truncated pyrin constructs were ectopically overproduced in HEK293T cells along with other inflammasome components, deletion of the B30.2 domain or replacement of the B30.2 with the TRIM5a B30.2 domain did not increase basal inflammasome activation ⁹, indicating that pyrin is not negatively regulated by B30.2 domain in this overexpression system. More recently, Park et al. used human monocytic U937 cells to study the role of the B30.2 domain in regulating human pyrin. U937 cells, which do not natively express MEFV, were retrovirally transduced to ectopically overexpress wild-type pyrin or pyrin lacking the B30.2 domain. Following stimulation with LPS there was more IL-1β released from U937 cells expressing pyrin lacking the B30.2 domain as compared to wild-type, suggesting that the B30.2 domain negatively regulates pyrin in this system ¹⁹.

Although the physiological significance of dimers, tetramers and pro-IL-1β binding by wild-type or FMF variant pyrin remains to be established, the results are intriguing as tetramer formation and pro-IL-1β binding may be the first direct evidence of a protein-protein interaction mediated by the B30.2 domain. One possibility is that dimers, tetramers or oligomeric complexes with pro-IL-1β represent inactive conformations of pyrin, possibly in conjunction with linker phosphorylation and PYD masking by the B-box. If close proximity of PYDs is important to trigger ASC filamentation, this would be prevented by the anti-parallel orientation of pyrin dimers. The possible inactive conformations of pyrin are highlighted in Fig. 1 by the encircling dashed line.

POSITIVE REGULATION OF PYRIN

Trimer formation

The CC-B30.2 portion of pyrin forms anti-parallel dimers in vitro ⁸ but in cells pyrin appears to form trimers ⁹. A crosslinking approach combined with SDS-PAGE showed that pyrin ectopically produced in HEK293T cells, natively produced in THP-1 monocytes, or recombinantly produced in Eschericia coli forms trimer-sized oligomers. Deletion analysis indicated that trimer formation required the CC domain. Results from gel filtration chromatography of E. coli produced protein were consistent with trimer and higher order oligomer formation ⁹. Structural predictions indicate that TRIM protein CC domains can form dimers or trimers and there is evidence that TRIM5a forms trimers in cells ³³. In experiments in which truncated pyrin constructs were ectopically produced in HEK293T cells along with other inflammasome components, the CC domain was required for basal ASC speck formation and cleavage of pro-IL-1β ⁹. Interestingly, a pyrin construct in which the CC domain was replaced with the CC domain of TRIM5a also promoted basal inflammasome activation ⁹.

Fig. 1 shows trimeric forms of pyrin (Intermediate trimer, Active trimer) in which the CC domains are parallel, however trimers could exist in other arrangements. It is possible that pyrin is primarily a dimer in cells but based on available evidence and for the purpose of this discussion we assume pyrin forms trimers. De-repression of the B30.2 domain by the binding of an unknown substrate is proposed to result in the formation of an intermediate trimer (Step 7). In this scheme, the intermediate trimer is kept in an inactive conformation by linker phosphorylation, 14–3-3 binding and masking of the PYD by the B-box but is “primed” to be activated. We hypothesize that pyrin B30.2 FMF variants are constitutively de-repressed for intermediate trimer formation. Trimer formation has important implications for pyrin regulation and in particular for the phenotypic consequences of heterozygous FMF or PAAND MEFV alleles ⁹. For example, in PAAND the probability of a trimer containing three pyrin S242R variant monomers is 12.5% of the total pyrin trimers. The presence of one or two S242R variants in a trimer may not be sufficient to activate ASC filamentation. This is reflected in Fig. 1, which shows that the intermediate trimer remains inactive even though one monomer is dephosphorylated.

Release of 14–3-3 and dephosphorylation of linker

When RhoA is inactivated by a bacterial toxin or effector the negative regulation of pyrin by PRK phosphorylation and 14–3-3 binding is interrupted. It seems likely that a positive regulation step is needed to release 14–3-3 and dephosphorylate the pyrin linker (Fig. 1, Step 8). A conformational change in pyrin could increase release of 14–3-3. It has been suggested that a protein serine phosphatase acts to dephosphorylate the pyrin linker ^11–13. Such a phosphatase could constitutively counter phosphorylation by PRK, or it might be regulated by some mechanism. Identification and characterization of a pyrin phosphatase is a key goal in the field.

Regardless of the mechanism for release of 14–3-3 and dephosphorylation of the pyrin linker, a fundamental unanswered question is how this step is regulated by RhoA inactivation. Given that RhoA molecules normally cycle between active and inactive conformations and on and off membranes in the uninfected host cell, how is inactivation of RhoA by a toxin or effector specifically “sensed” as a danger signal leading to pyrin dephosphorylation? One possibility is that bacterial toxins and effectors that inactivate RhoA trigger dephosphorylation of pyrin when the percentage of active RhoA vs inactive RhoA molecules in a cell drops below a threshold. Park et al. measured the percentage decrease in active RhoA in populations of LPS-primed mouse bone marrow-derived macrophages (BMDMs) treated with Clostridium C3, a toxin that inactivates RhoA and triggers activation of pyrin ¹¹. As compared to unintoxicated controls, there was an ~50% decrease in active RhoA in the C3-treated BMDM populations. This may indicate that a ~50% decrease in active RhoA molecules in a cell is sufficient to trigger pyrin dephosphorylation. Jamilloux et al. studied activation of the pyrin inflammasome in monocytes from FMF patients or healthy controls treated with low doses of TcdB ³⁴. Their data indicate that pyrin B30.2 FMF variants are not constitutively active but can be activated by lower amounts of toxin as compared to controls ³⁴. This may indicate that a less than 50% decrease in the percentage of active RhoA in FMF monocytes will trigger pyrin dephosphorylation.

Magnotti et al. treated monocytes from FMF patients or healthy controls with PKC superfamily kinase inhibitors to mimic the loss of PRK activity when RhoA is inactivated by a bacterial toxin or effector ³⁵. Treatment with staurosporine or two other more selective kinase inhibitors triggered pyrin activation in FMF monocytes as rapid pyroptosis and release of IL-1β were detected ³⁵. However, pyroptosis was delayed and release of IL-1β was reduced in healthy monocytes treated with kinase inhibitors. Evidence that the linker was dephosphorylated in response to kinase inhibitor treatment was obtained using U937 cells ectopically overproducing pyrin. Thus, linker dephosphorylation appears to be sufficient for pyrin B30.2 FMF variants but not wild-type human pyrin to assume an active conformation that can trigger ASC filamentation. The control mechanism of the wild-type B30.2 domain is unknown but could be equivalent to step 7 in Fig. 1. Park et al. observed that staurosporine treatment triggered the pyrin inflammasome in LPS-primed wild-type BMDMs ¹¹. However, treatment of LPS-primed wild-type BMDMs with another PKC superfamily inhibitor more specific for PRK, PKC412, did not trigger the pyrin inflammasome ¹⁶. More work is needed to determine if linker dephosphorylation is sufficient to activate wild-type pyrin in murine phagocytes. Notably, in this context, murine pyrin lacks the B30.2 domain.

Unmasking of PYD by B-box

Pyogenic Arthritis, Pyoderma Gangrenosum, and Acne (PAPA) Syndrome is a dominantly inherited autoinflammatory disease that results from an alternative pathway of pyrin activation ^5,9. PAPA is caused by missense mutations in the gene encoding proline serine threonine phosphatase-interacting protein 1 (PSTPIP1), a known interacting partner of pyrin. PSTPIP1 is a member of the F-BAR family of proteins.

Evidence from ectopic overproduction experiments in THP-1 and HEK293T cells indicate that PSTPIP1-PAPA variants bind to the pyrin B-box, resulting in unmasking of the PYD and inflammasome activation ⁹. Cleavage of caspase-1 in HEK293T cells ectopically producing PSTPIP1-PAPA variants is inhibited by colchicine ⁹, indicating that MTs are required activation of the pyrin inflammasome by this pathway. Activation of the pyrin inflammasome by PSTPIP1-PAPA variants does not require inactivation of RhoA and as a result this regulatory step is shown to occur in parallel to 14–3-3 release and linker dephosphorylation (Fig. 1, Step 9). However, it is possible activation of the pyrin inflammasome by PSTPIP1-PAPA variants does trigger dephosphorylation of the pyrin linker. The role of wild-type PSTPIP1 in regulating pyrin is unknown. Wang et al. generated and characterized mice deficient in PSTPIP1 ³⁶. These PSTPIP1 knockout mice develop normally and did not display autoinflammatory diseases signs ³⁶, which is the phenotype expected for a potential positive regulator of pyrin. Experiments in which PSTPIP1-deficient mouse (and human) phagocytes are tested for pyrin inflammasome activation in response to inactivation of RhoA by a toxin or effector would be informative. Wang et al. also generated mice conditionally expressing human PSTPIP1 PAPA alleles. Although these animals displayed partial embryonic lethality, growth retardation, and elevated level of circulating proinflammatory cytokines, these phenotypes did not require expression of the PSTPIP1 PAPA alleles in hematopoietic cells ³⁶.

Microtubule interaction

FMF patients are commonly treated with the microtubule (MT) depolymerizing drug, colchicine, to control inflammation ⁵. Colchcine as well as various other MT destabilizing agents have been shown to inhibit ASC filamentation, caspase-1 cleavage and IL-1β release when wild-type human or murine pyrin is activated through ectopic production of PSTPIP1 variants ⁹ or toxin-induced inactivation of RhoA ^11,12,14. The inhibitory effect of these MT inhibitors has been shown to be specific for the pyrin inflammasome as compared to other inflammasomes such as NLRC4, NLRP3, and AIM2.

MT destabilizing agents may inhibit pyrin activation by two mechanisms. Park et al. studied the effect of colchicine on pyrin activation in C3-treated BMDMs ¹¹. Evidence was obtained that colchicine increased RhoA-GTP levels by stimulating the release of GEF-H1 from destabilized MTs ¹¹. Other studies suggest that MT destabilizing agents inhibit activation of pyrin downstream of 14–3-3 release and linker dephosphorylation and upstream of ASC speck formation in response to RhoA inactivation ^12,14. The microtubule-stabilizing drug paclitaxel (taxol) had an intermediate ¹² to no inhibitory effect ¹⁴ on pyrin inflammasome activation, in BMDMs and PBMCs, respectively. Given these data and the fact that pyrin has been shown to co-localize with MT in transfected COS-7 cells and to bind to paclitaxel stabilized MT in vitro ¹⁵, Gao et al. suggested that MTs induce a conformational change in pyrin to allow ASC recruitment ¹². This is shown schematically in Step 10 in Fig. 1, indicating that MTs serve as a platform for the interaction of ASC with an active trimer. Inhibition of MT polymerization by colchicine would destabilize the platform needed to induce a conformational change in pyrin to allow ASC recruitment (Fig. 1, Step 11).

MTs are known to undergo cycles of rapid growth and disassembly. The majority of MTs have a half-life of about 5–10 min (dynamic MTs), while a minority are stable up to 20 h (stable MTs) ³⁷. Interestingly, inactivation of RhoA with C3 prevents formation of stable MTs ³⁷. Thus, it is likely that dynamic MTs are required for pyrin inflammasome assembly.

The N-terminal 374 residues (PYD and linker) appears to be sufficient to mediate co-localization of pyrin with MTs as determined by fluorescence microscopy in transfected COS-7 cells ¹⁵. In transfected HEK293 cells endogenous β-Tublin co-immunoprecipitated with full length human pyrin, as well as the individual PYD or a C terminal region (amino acids 375–781) containing the B box, coiled-coil, and B30.2 domains ¹⁴. These results suggest that multiple domains of pyrin may be involved in the interaction with MTs or tubulin subunits.

Interestingly, colchicine does not appear to inhibit the pyrin inflammasome in PBMCs or monocytes of FMF patients treated with C. difficile TcdA ^14,35. This effect seems to be specific for TcdA since colchicine does inhibit the pyrin inflammasome in FMF PBMCs or monocytes treated LPS ¹¹ or with kinase inhibitors ³⁵. Since colchicine inhibits the pyrin inflammasome in PBMCs or monocytes of healthy controls treated with C. difficile TcdA ^14,35, it appears that intoxication with TcdA is somehow bypassing the MT dependent step in the context of pyrin FMF variants.

THE PYRIN-YERSINIA EFFECTOR INTERACTION

Yersinia pathogenesis

The Yersinia genus is comprised of Gram-negative bacteria in the family Enterobacteriaceae. Y. pestis, Y. enterocolitica and Y. pseudotuberculosis are the three species in this genus that cause zoonotic disease in humans. Y. pestis is a pathogen in rodents and fleas that can be transmitted to humans to cause plague, an acute systemic infection with a high rate of mortality ³⁸. Y. pseudotuberculosis and Y. enterocolitica are enteric pathogens that typically cause self-limiting mesenteric lymphadenitis and gastroenteritis. One of the major virulence factors that is present in all pathogenic strains is a plasmid encoding the Ysc-Yop T3SS. The Yop effectors are highly conserved and play similar roles in the pathogenesis of the three Yersinia species. The effectors counteract innate immunity and are required for Yersinia to colonize and replicate as extracellular microcolonies in lymph nodes, the preferred target tissue for these pathogens. For recent reviews on the pathogenesis of Y. pestis, Y. enterocolitica and Y. pseudotuberculosis see ^39–41.

Contact of Yersinia with a host cell allows the T3SS to translocate Yop effectors across the plasma membrane (Fig. 1). Effectors are preferentially translocated into phagocytes ⁴⁰. This target cell specificity is due in part to the fact that neutrophils and inflammatory monocytes are recruited in large numbers to sites of Yersinia infection and make direct contact with the bacteria ^39,40. This cellular recruitment results in the formation of granuloma-like structures, variously characterized histopathologically as microabscesses, necrotic lesions, or pyogranulomas ^42,43. The injectisome components of the T3SS (the YscF needle, LcrV tip protein, YopB and YopD translocators) mediate translocation of the seven Yop effectors ⁴⁴. The effectors can be grouped into three general categories based on the innate defense pathways they inhibit: phagocytic (YopH, YopE, YopT, YpkA), Toll-like receptor (YopJ) and inflammasome (YopM, YopK).

This section focuses on YopE, YopT, YopM and YopJ since they have defined interactions with the pyrin inflammasome. An important breakthrough in the field was the discovery by LaRock et al. that YopM inhibits a caspase-1 inflammasome to promote Yersinia virulence ⁴⁵. Key to the discovery that YopM inhibits activation of a caspase-1 inflammasome was the infection of LPS-primed BMDMs with a Yersinia yopM mutant ⁴⁵, because Mefv is not expressed in naïve BMDMs. This achievement stimulated the discoveries that YopE and YopT cause an effector-triggered immune response via the RhoA-PRK-pyrin axis and that YopM and YopJ counteract this inflammasome pathway to promote Yersinia virulence ^{16,17,19,46,47}.

YopE and YopT activate the pyrin inflammasome

YopE and YopT promote Yersinia virulence by targeting multiple Rho GTPases to inhibit phagocytic signaling pathways. YopE has a C-terminal GTPase-activating protein (GAP) domain ⁴⁸. An arginine finger motif in the GAP domain inserts into the active site of Rho GTPases to accelerate hydrolysis of GTP into GDP. Purified YopE targets Rac1, Rac2, RhoA and Cdc42 in vitro ^48,49. In cultured cells infected with Yersinia YopE preferentially inactivates Rac1, RhoA and Rac2 ^48–50. YopE specificity is controlled by interaction with the divergent switch I and switch II regions of the different GTPases ⁴⁹. YopE targets Rac1 and Rac2 to inhibit phagocytosis and ROS production by phagocytes ⁴⁹. YopE negatively regulates translocation of itself and other Yops into host cells by targeting RhoA ^48–52.

YopT is a cysteine protease that cleaves the C-terminal isoprenoid group from Rho GTPases, detaching them from host membranes ⁵³. YopT appears to preferentially target RhoA and Rac1 and has no preference for the GTP vs GDP bound conformation. GTP-bound Rac1 may localize to the cell nucleus after cleavage by YopT ⁵⁴. YopT and YopE also target RhoG as part of the Yersinia strategy to disrupt phagocytic signaling ⁵⁵.

YopE GAP and YopT protease activities were demonstrated to trigger the pyrin inflammasome by infecting LPS-primed BMDMs with Yersinia strains in which the yopE or yopT genes were mutated in a yopM mutant background ^16,18,47. Introduction of mutations that inactivated the two genes or the catalytic activities of the gene products alone or in combination demonstrated that YopE GAP has a dominant role in triggering activation of pyrin (Fig. 1, Step 13) ^16,18,47. Pyrin activation triggered by YopT protease activity (Fig. 1, Step 14) can only be detected in the absence of YopE in LPS-primed BMDMs infected with Y. pseudotuberculosis ^16,18. No role for YopT in activating pyrin could be demonstrated in LPS-primed BMDMs infected with Y. pestis ⁴⁷. There is a D298G substitution in YopT of Y. pestis relative to Y. pseudotuberculosis, which might explain the lower activity in the former species. It is likely that YopE more efficiently inactivates RhoA as compared to YopT. The dominant role of YopE over YopT in triggering pyrin is reflected in its more important role in virulence. Certain virulent strains of Y. pseudotuberculosis have inactivating deletions in yopT, suggesting that YopT is redundant to YopE ⁵⁶.

The concept that inactivation of RhoA triggers pyrin activation was established using toxins and effectors that introduce covalent modifications in or near the switch 1 region ⁴. Since non-covalent inactivation of RhoA by YopE GAP and detachment of RhoA from plasma membrane by YopT protease also triggers pyrin activation, it is likely that any toxin or effector mechanism that prevents RhoA from activating PRK is sufficient to trigger linker dephosphorylation ^6,16–18.

YpkA is a third Yersinia effector that targets RhoA. YpkA is bi-functional, with an N-terminal serine-threonine kinase domain and a C-terminal Rho GDI-like domain that binds to RhoA and Rac1 ^57,58. The GDI-like domain of YpkA can inhibit nucleotide exchange on RhoA and Rac1, leading to disruption of actin filaments ⁵⁹. Infection of LPS-primed BMDMs with a Yersinia yopEyopTyopM mutant triggers activation of the pyrin inflammasome with delayed kinetics ¹⁸. The possibility that YpkA GDI activity triggers delayed activation of pyrin remains to be investigated.

YopM inhibits the pyrin inflammasome

YopM is member of the leucine-rich repeat (LRR)-containing effector superfamily that can promote virulence by several different mechanisms ⁶⁰. The LRRs are located in the central domain of YopM and the number of repeats can vary between 13 and 21 in different Yersinia strains. Each Yersinia strain produces only one YopM protein. The N-terminal region of YopM contains two α-helices that function in translocation and assists in folding of the LRRs ⁶¹. The C-terminus of YopM consists of a 24 residue unstructured tail ^61,62. Determination of the crystal structure of a Y. pestis 15 LRR YopM isoform revealed that the 20 to 22 amino acid LRRs form loops that stack against each other to form an overall crescent shape protein ⁶¹. The concave surface of the crescent is predicted to be involved in target recognition. Tetramers of YopM were present in the crystals ⁶¹. A purified Y. pseudotuberculosis 15 LRR YopM forms dimers and tetramers in solution ⁶². The biological significance of the multimers seen in vitro is not known.

Biochemical and genetic evidence indicate that counteracting the YopE and YopT effector-triggered immune response of the pyrin inflammasome is a major virulence function of YopM ^16,17,19. All YopM isoforms tested thus far inhibit pyrin by the same general mechanism as discussed below. There may be important differences in the activities of different YopM isoforms. For example, one YopM isoform was reported to activate NLRP3 by ubiquitination ⁶³. Other YopM isoforms may act as general inhibitors of caspase-1 ⁴⁵. For the purpose of this review when discussing mechanisms of pyrin inhibition all isoforms are referred to generically as YopM.

YopM inhibits activation of the pyrin inflammasome in LPS-primed BMDMs, human PBMC, and human monocytes infected with Yersinia ^16,17,19, and in reconstituted HEK293T cells ectopically producing YopM ¹⁷. YopM does not inhibit the NLRP3 or NLRC4 inflammasomes when these pathways were triggered by infection of LPS-primed BMDMs with a Yersinia yopK mutant ^16,17. Additionally, YopM does not inhibit the NLRP3 inflammasome in reconstituted HEK293T cells ¹⁷. Thus, YopM is a specific inhibitor of the pyrin inflammasome and is sufficient for this activity in mouse and human cells.

YopM promotes virulence by counteracting the pyrin inflammasome as demonstrated in mouse infection models ^16,17. The virulence defect of a Y. pseudotuberculosis yopM mutant was fully rescued, as shown by restoration of lethality and bacterial growth in spleen and liver, in Mefv^−/− mice infected by the intravenous route ¹⁶. In addition, the virulence function of YopM became dispensable when wild-type mice were infected by this route with a yopE and yopT mutant that does not activate pyrin ¹⁶. YopM acts redundantly with YopJ (see below) to promote virulence in a Y. pestis subcutaneous infection model of bubonic plague ⁴⁷. The virulence defect of a Y. pestis yopM yopJ mutant is restored in mice knocked out for pyrin, ASC, IL-1β, or IL-18 but not NLRP3 ^17,47. Taken together, these results suggest that YopE and YopT effector-triggered activation of the pyrin inflammasome leads to release of IL-1β and IL-18, which non-redundantly protect against systemic Yersinia infection by reducing bacterial loads, and that YopM (and YopJ in the case of plague) is essential for counteracting this host response.

Clues to the mechanism of YopM came from the discovery that this protein binds to host kinases. McDonald et al. used a proteomics approach to demonstrate that in HEK293T cells YopM forms a complex with PRK and ribosomal protein S6 kinase (RSK) ⁶⁴. Subsequent work using macrophages showed that PRK binds to the LRR region ⁶⁵ while RSK binds to the YopM tail ⁶². YopM can bind to multiple isoforms of PRK (e.g. PRK1 and PRK2) and RSK (e.g. RSK1 and RSK2) ⁶⁶. RSKs are activated by phosphorylation and binding to YopM prevents their dephosphorylation, leading to constitutive activation ^64,66. Activated RSK in turn phosphorylates and activates PRK ⁶⁴. YopM forms high molecular weight oligomers in macrophages infected with Yersinia ⁶². These high molecular weight oligomers contain RSK ⁶². It remains to be determined if PRK and pyrin are also present in these high molecular weight complexes.

The demonstration that YopM can bind to pyrin as well as PRK and RSK lead to the model that these kinases are hijacked to maintain linker phosphorylation ^16,17,19 (Fig. 1, Step 14). This model is supported by several lines of evidence. In vitro assays using purified components demonstrate that YopM can increase pyrin linker phosphorylation by PRK1, PRK2, RSK1, RSK2 or RSK3 ^16,19. The PRKs can mediate basal pyrin phosphorylation in the absence of YopM but phosphorylation of the linker by RSKs requires YopM ¹⁹. Interestingly, YopM-promoted phosphorylation of pyrin by RSKs, especially RSK2, is substantially higher as compared to PRKs ¹⁹. RSK2 can even phosphorylate YopM at unknown sites in the presence or absence of pyrin ¹⁹. No synergistic increase in YopM-promoted pyrin phosphorylation was seen in the presence of both PRK1/2 and RSK1 ¹⁹. Previous work demonstrated that RSKs bind to the C-terminal tail of YopM ⁶² and the use of purified truncated proteins for binding assays suggest that the first two LRRs of YopM are important for binding pyrin ¹⁹. YopM binds to two independent regions in the N-terminal 330 amino acids of pyrin ¹⁹. Although the PYD does not appear to be required for YopM binding, it is necessary for RSK2-mediated phosphorylation of the linker in the presence of YopM ¹⁹.

Results of cell culture infection experiments support the idea that YopM hijacks PKNs and RSKs to maintain linker phosphorylation. Treatment of LPS-primed BMDMs with a PRK inhibitor (PKC412), or with siRNAs targeting Prk1 and Prk2, partially restored cleavage of caspase-1 and IL-1β in response to infection with wild-type Y. pseudotuberculosis ⁶⁷. In addition, treatment of human monocytes with RSK inhibitors increased release of IL-1β in response to Y. pestis infection. Treatment of THP-1 cells with siRNA targeting RSK1/2/3 decreased ASC oligomerization in response to wild-type Y. pestis infection ¹⁹. Results of infections of U937 cells ectopically expressing pyrin S208A and S242A variants indicated that YopM inhibition was bypassed, leading to IL-1β release ¹⁹. Taken together these results suggest that YopM binds directly to the N-terminal region of pyrin, and in an PYD-dependent manner primarily hijacks RSK2 to phosphorylate the linker, while PRKs play a secondary role in counteracting the inflammasome. The dominant role of RSKs in mediating linker phosphorylation likely explains why YopM variants lacking functional C-terminal tails are defective for pyrin inhibition and Yersinia virulence ^17,19,62.

YopJ cooperates with YopM to inhibit the pyrin inflammasome

YopJ is a member of a family of bacterial effectors with acetyl transferase activity ⁶⁸. YopJ acetylates and inactivates TAK1, IKK and MKK proteins, which are important regulators of MAPK and NF-kB pathways activated by TLR signaling. In naïve macrophages infected with Yersinia YopJ inhibits proinflammatory cytokine production and triggers cell death ^69–71. YopJ-dependent apoptosis in Yersinia-infected macrophages is an effector-triggered immune response that occurs as a result of TLR4/TRIF and TFNR1 signaling, which activate caspase-8 and caspase-1 via the kinase activity of RIPK1 ^43,72–74. Caspase-8 in turn activates caspase-3 and caspase-1. Active caspase-1 can cleave GSDMD, resulting in pore formation and a pro-inflammatory form of apoptosis ^75,76. In BMDMs primed with LPS for 5 hours or longer, YopJ does not induce inflammatory apoptosis upon Yersinia infection. Instead, YopJ activity reduces the amount of IL-1β released upon infection with a yopM mutant by ~50% ^17,46,47. In this context YopJ may inhibit infection-induced expression of IL-1β (Fig. 1, Step 15) or other inflammasome components ^46,47, or a pathway important for post-translational regulation of pyrin ⁴⁶. YopJ does not appear to prevent dephosphorylation of pyrin ¹⁸, but may reduce cleavage of caspase-1 ⁴⁶.

In this context, in a Y. pseudotuberculosis intragastric infection model, a yopM mutant is significantly attenuated but a role for YopJ in promoting virulence was evident only in the absence of YopM ⁴⁶. In contrast, in a Y. pestis subcutaneous infection model of bubonic plague, YopJ acts redundantly with YopM to promote virulence ⁴⁷.

MEFV FMF mutations and resistance to Y. pestis

Inhibition of pyrin by YopM is essential for virulence in Y. pestis ¹⁷. It has been hypothesized that FMF mutations were selected in the human population as a result of heterozygous advantage against Y. pestis infection during historic plague pandemics ^5,16. To test this hypothesis, Park et al. infected human phagocytes producing pyrin B30.2 FMF variants with Y. pestis and studied the impact on the inflammasome response. In U937 cells ectopically expressing B30.2 FMF variants and in FMF PBMCs there was increased release of IL-1β upon infection with wild-type Y. pestis as compared to controls ¹⁹. An increased IL-1β release was seen in FMF PBMCs relative to controls infected with wild-type but not yopM mutant Y. pestis. In healthy control monocytes infected with wild-type Y. pestis the presence of YopM resulted in increased pyrin phosphorylation and 14–3-3 binding. However, in FMF monocytes the increased pyrin phosphorylation upon Y. pestis infection is not seen, suggesting that pyrin B30.2 FMF variants are resistant to negative regulation by YopM ¹⁹. No difference in IL-1β release was seen between healthy control and FMF PBMCs infected with Burkholderia cenocepacia ¹⁹. Burkholderia cenocepacia encodes the RhoA-inactivating effector TecA but does not have an effector equivalent to YopM to block the pyrin inflammasome ⁷⁷. CD14+ monocytes heterozygous for FMF alleles also released more IL-1β as compared to controls after infection with WT Y. pestis, suggesting that carriers are also resistant to YopM ¹⁹.

Mice producing “knocked in” pyrin B30.2 FMF variants (FMF mice) ⁷⁸ were used as an in vivo infection model to assess resistance to Y. pestis. These mice display FMF characteristics including autoinflammatory disease dependent upon caspase-1, IL-1β and GSDMD function ^26,79,80. Upon systemic infection of mefv^M680I/M680I mice with Y. pestis, there was increased resistance to lethal infection as compared to control Mefv^B30.2/B30.2 mice ¹⁹. Enhanced resistance to systemic Y. pestis infection was not seen in the FMF mice lacking IL1R, implicating IL-1 signaling in protection ¹⁹.

Finally, Park et al. studied a Turkish population for a genetic imprint of evolutionarily recent natural selection ¹⁹. Results provide evidence that FMF-associated MEFV M694V and V726A mutations have undergone episodic selection in the Turkish population and arose >1800 years ago. The first major plague pandemic, the Justinian plague, began in 541 CE and recurred intermittently for two centuries. The Black Death in 1347 CE began the second pandemic which recurred occasionally to the end of the nineteenth century in the Middle East. The results of Park et al. suggest that modern-day FMF mutation carriers are descended from common ancestors who likely lived in the Middle East before the known historical plague epidemics ¹⁹. Thus, Y. pestis infection and inhibition of pyrin by YopM may have provided the selective pressure for MEFV FMF mutations.

CONCLUSIONS AND OUTSTANDING QUESTIONS

Pyrin is a structurally and biologically unique inflammasome sensor. The B-box, CC and B30.2 domains in pyrin are present in TRIM proteins but not in other inflammasome sensors. MEFV expression is restricted to phagocytic cells unlike other inflammasome sensors that have wider tissue expression. The negative regulation by phosphorylation of the pyrin linker by the RhoA-PRK axis is one of the few examples of a guard-like mechanism for detecting toxins and effectors in mammals. The discovery that Yersinia uses YopM to inhibit pyrin and counteract an effector-triggered immune response has yielded new insights into mechanisms of bacterial pathogenesis and the evolution of MEFV FMF mutations in the human population.

Outstanding questions.

What is the structural basis for the inactive conformation imposed by linker phosphorylation and 14–3-3 binding? Does linker phosphorylation occur in monomers or multimers of pyrin? Is linker phosphorylation and 14–3-3 binding connected to sequestration of the PYD by the B-box?

Given that RhoA molecules normally cycle between active and inactive conformations and on and off membranes in the uninfected host cell, how is inactivation of RhoA by a toxin or effector specifically “sensed” as a danger signal leading to pyrin dephosphorylation?

How does the B30.2 domain negatively regulate pyrin? Why have some animals such as rodents lost the B30.2 domain during evolution, and what impact does this have on pyrin regulation in these contexts? What is the mechanistic basis for the observation that B30.2 FMF variants are not constitutively active, but trigger inflammasome assembly at lower levels of toxin or effector as compared to wild type pyrin? How do pyrin B30.2 FMF variants bypass inhibition by YopH?

How is pyrin activation positively regulated? Is there a phosphatase that dephosphorylates the linker? Does PSTPIP1 regulate sequestration of the PYD by the B-box under normal conditions? Do MTs serve as a platform for ASC nucleation in order to positively regulate pyrin?

Do any effectors or toxins directly modify pyrin?