Design principles for inflammasome inhibition by pyrin-only-proteins

  1. Shuai Wu
  2. Archit Garg
  3. Zachary Mazanek
  4. Gretchen Belotte
  5. Jeffery J Zhou
  6. Christina M Stallings
  7. Jacob Lueck
  8. Aubrey Roland
  9. Michael A Chattergoon
  10. Jungsan Sohn  Is a corresponding author
  1. Department of Biophysics and Biophysical Chemistry, Johns Hopkins University School of Medicine, United States
  2. Division of Infectious Diseases, Johns Hopkins University School of Medicine, United States
  3. Division of Rheumatology, Johns Hopkins University School of Medicine, United States
  4. Department of Oncology, Johns Hopkins University School of Medicine, United States

Abstract

Inflammasomes are filamentous signaling platforms essential for host defense against various intracellular calamities such as pathogen invasion and genotoxic stresses. However, dysregulated inflammasomes cause an array of human diseases including autoinflammatory disorders and cancer. It was recently identified that endogenous pyrin-only-proteins (POPs) regulate inflammasomes by directly inhibiting their filament assembly. Here, by combining Rosetta in silico, in vitro, and in cellulo methods, we investigate the target specificity and inhibition mechanisms of POPs. We find here that POP1 is ineffective in directly inhibiting the central inflammasome adaptor ASC. Instead, POP1 acts as a decoy and targets the assembly of upstream receptor pyrin-domain (PYD) filaments such as those of AIM2, IFI16, NLRP3, and NLRP6. Moreover, not only does POP2 directly suppress the nucleation of ASC, but it can also inhibit the elongation of receptor filaments. In addition to inhibiting the elongation of AIM2 and NLRP6 filaments, POP3 potently suppresses the nucleation of ASC. Our Rosetta analyses and biochemical experiments consistently suggest that a combination of favorable and unfavorable interactions between POPs and PYDs is necessary for effective recognition and inhibition. Together, we reveal the intrinsic target redundancy of POPs and their inhibitory mechanisms.

Editor's evaluation

This important work proposes general principles by which a central class of inflammatory mediators called inflammasomes are regulated by "pyrin only" proteins. The authors employ a convincing combination of in silico and biochemical measurements in support of their model. The work is of broad interest to those interested in the biochemical regulation of inflammatory signaling pathways.

https://doi.org/10.7554/eLife.81918.sa0

Introduction

Inflammasomes are filamentous signaling platforms integral to host innate defense against a wide range of intracellular catastrophes, which include ionizing irradiation, genotoxic chemicals, and pathogen invasion (Broz and Dixit, 2016; Zheng et al., 2020). However, persisting inflammasome activities lead to several human maladies including numerous autoinflammatory disorders, cancer, and even severe COVID-19 (Karki et al., 2017; Tartey and Kanneganti, 2020; Vora et al., 2021). Thus, understanding how inflammasome assemblies are regulated at the molecular level can provide key insights into developing strategies for preventing and treating various diseases (Broz and Dixit, 2016; Karki et al., 2017; Tartey and Kanneganti, 2020; Vora et al., 2021; Zheng et al., 2020).

Inflammasomes transduce signals by sequentially assembling filamentous oligomers, with multiple initial pathways progressively converging at downstream assemblies (Broz and Dixit, 2016; Kagan et al., 2014; Lu and Wu, 2015; Zheng et al., 2020). For instance, an array of molecular signatures arising from various pathogenic conditions induces the oligomerization of inflammasome receptors, resulting in filament assembly by their pyrin-domains (PYDs; e.g. viral nucleic acids, reactive oxygen species, specific lipids from damaged mitochondria, and disruption of the trans-Golgi network) (Andreeva et al., 2021; Fernandes-Alnemri et al., 2009; Hornung et al., 2009; Iyer et al., 2013; Roberts et al., 2009; Zhong et al., 2013); PYDs are ~100 amino acid (a.a.) 6-helix bundle proteins that belong to the death-domain (DD) family often found in apoptotic and inflammatory signaling pathways (Park et al., 2007). The upstream PYD oligomers then induce the filament assembly by the PYD of the central adaptor ASC (ASCPYD), resulting in oligomerization/filamentation of its CARD (ASC: apoptosis-associated speck-forming protein-containing caspase-recruiting domain [CARD]) (Broz and Dixit, 2016; Kagan et al., 2014; Lu et al., 2014; Lu and Wu, 2015). Finally, ASCCARD oligomers recruit pro-caspase-1 and induce its filament assembly, activating the enzyme by proximity-induced auto-proteolysis (Broz and Dixit, 2016; Kagan et al., 2014; Lu et al., 2014; Lu and Wu, 2015). Caspase-1 executes two key innate immune responses, namely the cleavage/maturation of pro-inflammatory cytokines such as interleukin-1β (IL-1β) and IL-18, and the initiation of pyroptosis (Broz and Dixit, 2016; Kagan et al., 2014; Lu et al., 2014; Lu and Wu, 2015).

A hallmark of inflammasome assembly is its binary (on-or-off) nature (Cai et al., 2014; Franklin et al., 2014; Matyszewski et al., 2018; Shen et al., 2021). That is, once assembled, inflammasomes do not dissociate (Franklin et al., 2014; Matyszewski et al., 2018). Moreover, multiple positive feedback loops between upstream receptors and ASC not only bolster the assembly, but also result in prion-like self-perpetuation (Cai et al., 2014; Matyszewski et al., 2018). Such an inherently irreversible assembly mechanism in turn would necessitate extrinsic factors to prevent persistent activities. Indeed, mammalian pyrin-only-proteins (POPs) have emerged as major inhibitors of inflammasomes (de Almeida et al., 2015; Khare et al., 2014; Periasamy et al., 2017; Ratsimandresy et al., 2017), functioning analogous to CARD-only proteins (COPs) that interfere with the oligomerization/activation of pro-caspases (Devi et al., 2020; Indramohan et al., 2018; Lu et al., 2016). It has been proposed that the target specificities of POPs are dictated by their a.a. sequence homologies to inflammasome PYDs (Devi et al., 2020; Indramohan et al., 2018). For example, POP1 is most homologous to ASCPYD (65% sequence identity; Figure 1—figure supplement 1A) and thought to directly inhibit the nucleation of the ASCPYD filament (de Almeida et al., 2015). POP2 shares 68% sequence identity to the PYD of Nod-like receptor containing a PYD-2 (NLRP2PYD; Figure 1—figure supplement 1B); the primary target of POP2 is thought to be ASC, but it is also implicated in inhibiting the absent-in-melanoma-2 (AIM2) receptor (Periasamy et al., 2017; Ratsimandresy et al., 2017). Finally, POP3 is most similar to AIM2PYD (67% sequence identity, Figure 1—figure supplement 1C) and targets AIM2-like receptors (ALRs, e.g. AIM2 and interferon inducible protein 16 [IFI16]) (Khare et al., 2014).

Inflammasome filaments are highly ordered supra-structures that entail at least two distinct steps for assembly: rate-limiting nucleation followed by elongation (Kagan et al., 2014; Lu et al., 2014; Lu and Wu, 2015; Matyszewski et al., 2018). Moreover, although the PYDs of upstream receptors do not display significant a.a. sequence homologies (Kagan et al., 2014; Lu et al., 2014; Lu and Wu, 2015), they all assemble into structurally congruent helical filaments and signal through the common ASC adaptor, suggesting a degenerate-code-like recognition mechanism (Hochheiser et al., 2022a; Kagan et al., 2014; Lu et al., 2014; Lu and Wu, 2015; Matyszewski et al., 2021; Shen et al., 2019). At present, little is known about how POPs selectively target and regulate the assembly of such diverse yet homologous supramolecular structures. This is because the current understanding on the mechanism of inhibition by POPs remains entirely inferred from indirect measurements and phenotypic outcomes (de Almeida et al., 2015; Khare et al., 2014; Periasamy et al., 2017; Ratsimandresy et al., 2017).

Here, by combining in silico, in cellulo, and in vitro methods, we delineate the target specificity and inhibition mechanisms of human POPs. We find that POP1 is a poor inhibitor of ASC and impedes the assembly of upstream receptor filaments instead (e.g. AIM2, IFI16, NLRP3, NLRP6). POP2 not only suppresses the nucleation of ASC, but also interferes with the assembly of multiple upstream receptors. Finally, in addition to potently suppressing the assembly of ALR and NLRP filaments (e.g. elongation of AIM2PYD), POP3 suppresses the nucleation of ASC. Our results indicate that a combination of favorable and strongly unfavorable interactions is necessary for POPs to inhibit the assembly of PYD filaments. Together, we propose that, instead of being dictated by a.a. sequence homology, degenerate-code-like target selection and inhibition mechanisms underpin the regulation of inflammasome assembly by POPs.

Results

Rosetta interface analyses suggest broad target specificities of POPs

Several inflammasome receptor PYDs signal through ASCPYD although their primary a.a. sequences vastly differ (Broz and Dixit, 2016; Kagan et al., 2014; Lu and Wu, 2015). Such a functional redundancy among different PYDs in turn suggests that sequence homology may not dictate the target specificity of POPs. To elucidate how POPs recognize and regulate the assembly of different PYD filaments, we first implemented Rosetta-based in silico approach that we had recently developed to define the directionality of the AIM2-ASC inflammasome (Matyszewski et al., 2021). Briefly, PYDs assemble into helical filaments in which each protomer provides six unique protein-protein interaction surfaces (e.g. Figure 1A, ‘Type’ 1a/b, 2a/b, and 3a/b) (Lu et al., 2014; Lu and Wu, 2015). As we had done before (Matyszewski et al., 2021), we created a honeycomb-like side view of PYD filaments in which the middle protomer makes all six required contacts for filament assembly (Figure 1A). We then calculated Rosetta interface energies (∆Gs) for each PYD filament (e.g. Figure 1B, left), and also determined the ∆Gs for POP•PYD interactions by replacing the center protomer with each POP (e.g. Figure 1B, three honeycombs on the right). Of note, we decided to conduct our in silico and subsequent biochemical experiments on tractable inflammasome components with well-known biological significances such as ASC, AIM2, IFI16, NLRP3, and NLRP6 (Broz and Dixit, 2016; Hochheiser et al., 2022a; Kerur et al., 2011; Lu et al., 2014; Lu and Wu, 2015; Matyszewski et al., 2018; Matyszewski et al., 2021; Morrone et al., 2015; Morrone et al., 2014; Shen et al., 2019).

Figure 1 with 2 supplements see all
Rosetta in silico analyses of putative pyrin-only-protein (POP)•pyrin-domain (PYD) interactions.

(A) The ‘side view’ of the ASCPYD filament (PDB: 3j63). The center magenta protomer makes all six unique contacts with surrounding pink protomers for assembly. Each surrounding protomer is colored in a different shade of pink in the ‘honeycomb’ diagram. (B–F) Rosetta interface energy scores (∆Gs, Rosetta energy score, reu) at individual filament interfaces for homotypic assemblies (left) and putative interactions with POPs (right). Each hexagon represents a PYD or POP monomer. The sum of ∆Gs at the top and bottom half is also listed. The honeycombs were generated based on their respective cryo-EM structures except for IFI16PYD whose filament structure is unknown. We generated a homology model of IFI16PYD filament based on the GFP-tagged AIM2 filament (PDB: 6mb2), which produced more favorable ∆Gs than the one generated from the tagless-AIM2PYD filament (PDB: 7k3r; Figure 1—figure supplement 2).

We found previously that individual AIM2PYD and ASCPYD filaments assemble bidirectionally (i.e. extending from both top and bottom surfaces), with each pair of filament interface types contributing symmetric ∆Gs (e.g. both Type 1a and Type 1b surface show ∆G=–22.1 for ASCPYD•ASCPYD in Figure 1B; Matyszewski et al., 2021). The interface analysis results from other inflammasome PYDs also showed similar symmetric energy landscapes (Figure 1B–F, left), suggesting that bidirectional assembly is universal to all PYD filaments. We next noted that the overall ∆Gs for individual PYD filaments were more favorable than those from any putative POP•PYD interactions, which in turn suggested that excess POPs might be necessary to inhibit the assembly of inflammasome PYDs (e.g. the sum of ASCPYD•ASCPYD interactions on the top half yields ∆G=–35.9, while that of POP1•ASCPYD is –24.7; Figure 1B–F). POP1 showed more favorable overall ∆Gs for ASCPYD compared to POP2 or POP3 (Figure 1B), seemingly supporting the previous report (and sequence homology) that POP1 likely binds ASC (de Almeida et al., 2015). However, although it was reported that POP2 can inhibit ASCPYD more potently than POP1 in vivo (Ratsimandresy et al., 2017), its ∆Gs were less favorable (i.e. Figure 1B: POP1•ASCPYD = −24.7 vs. POP2•ASC = –0.8 on the top half). POP3, on the other hand, appeared to interact with ASC as favorably as POP1 at the bottom interfaces (Figure 1B; ∆G=–34.3 for POP1•ASCPYD vs. ∆G=−34.6 for POP3•ASCPYD), suggesting that it could also inhibit the central adaptor.

For AIM2PYD, all three POPs showed comparable overall ∆Gs on the bottom half (Figure 1C), which suggested that each of them could target AIM2. IFI16 favored POP3 the most (Figure 1D; e.g. ∆G=–9.6 for POP3•IFI16PYD on the top half); however, the other two POPs still showed more favorable ∆Gs than homotypic IFI16PYD•IFI16PYD interactions on at least one individual interface (Figure 1D; e.g. Type 2a for POP1, Type 2b for POP2, and Type 1b for POP3). These results in turn suggested that not only POP3, but POP1 and POP2 might also recognize IFI16. It has been speculated that POPs interfere with the recruitment of ASC by the upstream receptors (de Almeida et al., 2015; Devi et al., 2020; Indramohan et al., 2018; Periasamy et al., 2017; Ratsimandresy et al., 2017); however, it remains unknown whether they do so by directly inhibiting NLRPs, which are the major class of inflammasome receptors. Our Rosetta analyses here suggest that POP1 could interact with NLRP3PYD on the top half (Figure 1E; ∆G=–31.4 for NLRP3PYD•NLRP3PYD vs. ∆G=−30.5 for POP1•NLRP3PYD). However, the interactions between POP2/3 and NLRP3PYD appeared much less favorable (Figure 1E). Similar to NLRP3PYD, POP1 again showed more favorable energy scores toward NLRP6PYD than POP2/3 (Figure 1F). Overall, although these results appear to rationalize some of the proposed target specificities of POPs (Devi et al., 2020; Indramohan et al., 2018), they also suggest confounding interaction and recognition mechanisms, thus warranting further investigations via biochemical approaches.

Mechanisms of ASC inhibition by POPs

Previous investigations on POP•PYD interactions have predominantly relied on in cellulo downstream signaling activities and in vivo phenotypes (de Almeida et al., 2015; Khare et al., 2014; Periasamy et al., 2017; Ratsimandresy et al., 2017). Although establishing the physiological relevance of POPs, these studies have left large voids in understanding their target selection and inhibition mechanisms. Because our in silico analyses did not immediately yield clear explanations for such questions, we set out to test POP•PYD interactions using more direct in cellulo and in vitro methods, focusing on ASC first. When ectopically expressed in HEK293T cells, C-terminally mCherry-tagged ASCPYD forms filaments and full-length ASC (ASCFL) forms puncta (Matyszewski et al., 2021). Importantly, HEK293T cells do not contain any endogenous inflammasome components or POPs, providing an ideal cellular system for directly tracking their interactions (de Almeida et al., 2015; Matyszewski et al., 2021; Shi et al., 2016). Here, we first tested whether co-transfecting C-terminally eGFP-tagged POPs hinders the oligomerization of ASC. Compared to co-transfecting eGFP alone, POP1-eGFP only marginally inhibited the filament assembly of ASCPYD-mCherry (≤20% suppression vs. eGFP control; Figure 2A–B). By contrast, co-transfecting POP2-eGFP or POP3-eGFP significantly reduced the amount of ASCPYD-mCherry filaments, with POP2 being more effective (Figure 2A–B; note more diffused mCherry signals and reduction in linear filaments in the presence of POP2 and POP3 in Figure 2A). Furthermore, POP1 reduced the number of ASCFL puncta by ~30%, yet POP2 and POP3 were again more effective in preventing punctum formation (up to ~80% reduction, Figure 2C–D). Monitoring oligomerization of mCherry-labeled ASC in the presence of untagged POPs also corroborated our observations with eGFP-tagged POPs (Figure 2—figure supplement 1). These results suggest that POP1 may not directly target ASCPYD. Moreover, unlike COPs that co-assemble with CARDs into filaments (Lu et al., 2016), our observations indicate that POPs suppress filament assembly altogether.

Figure 2 with 3 supplements see all
POP1 does not directly inhibit apoptosis-associated speck-forming protein-containing caspase-recruiting domain (ASC).

(A) Sample fluorescent microscope images of HEK293T cells co-transfected with mCherry-tagged ASCPYD (300 ng; crimson) plus eGFP alone or eGFP-tagged pyrin-only-proteins (POPs) (1200 ng; green). Blue: DAPI. (B) The relative amounts of ASCPYD-mCherry filaments (300 ng plasmid) in HEK293T cells when co-transfected with POP-eGFP (+) or eGFP alone (-) (600 and 1200 ng plasmids). n≥4. *: p≤0.05, **: p≤0.01; ***: p≤0.001, hereafter. (C) Sample fluorescent microscope images of HEK293T cells co-transfected with mCherry-tagged ASCFL (300 ng; crimson) plus eGFP alone or eGFP-tagged POPs (1200 ng; green). Blue: DAPI. (D) The relative amounts of ASCFL-mCherry puncta (300 ng plasmid) in HEK293T cells when co-transfected with POP-eGFP (+) or eGFP (-) (600 and 120 ng plasmids). N≥4. (E) Time-dependent increase in Förster resonance energy transfer (FRET) signals of a donor- and acceptor-labeled ASCPYD (2.5 µM total, black circle) was monitored with increasing concentrations POP1 (50 and 150 µM), POP2 (3.3, 6.7, 13.3, 26.7, and 40 µM), or POP3 (3.25, 7.5, 15, 20, and 30 µM); darker shades correspond to increasing POP concentrations. Two- and three-headed arrows indicate the increase in apparent nucleation time (or lack thereof). Arrows pointing upper right directions indicate the change (or lack thereof) in the elongation phase in the presence of the highest POP concentrations used. Data shown are representatives of at least three independent measurements (IC50s are average values of these experiments. N=3). (F) Negative-stain electron microscopy (nsEM) images of ASCPYD filaments (2.5 µM) in the presence and absence of POP1 (150 µM), POP2 (40 µM), or POP3 (30 µM).

Figure 2—source data 1

Data values used in plots from Figure 2 and Figure 2—figure supplements 12 and a folder containing uncropped gel images used in Figure 2—figure supplement 2.

https://cdn.elifesciences.org/articles/81918/elife-81918-fig2-data1-v1.zip

Our observations strongly indicate that POP1 does not directly target ASC assembly. To further test our in cellulo results, we then generated recombinant POPs to investigate their inhibitory mechanisms. POP1 behaved as a monomer without forming filaments or higher-order species in our hands (Figure 2—figure supplement 2A–B). On the other hand, recombinant POP2 and POP3 were prone to aggregation/degradation during purification and required an N-terminal maltose-binding protein (MBP) tag to obtain intact proteins (Figure 2—figure supplement 2A). Cleaving MBP via tobacco etch virus protease (TEVp) indicated that POP2 and POP3 form elongated oligomers with undefined structures (Figure 2—figure supplement 2B).

We incorporated recombinant POPs into our well-established polymerization assay in which we track the Förster resonance energy transfer (FRET) ratio between a 1:1 mixture of donor- and acceptor-labeled PYDs (Matyszewski et al., 2018; Matyszewski et al., 2021; Mazanek and Sohn, 2019); the auto-assembly of individual PYDs is suppressed by an N-terminal MBP tag and polymerization is triggered by cleaving MBP via TEVp. Of note, our assay displays two distinct phases of filament assembly, namely the rate-limiting nucleation (initial lag; double/triple-headed arrows in Figure 2E) followed by elongation (exponential/linear phase; single-headed arrows pointing to the upper right-hand corner in Figure 2E; Matyszewski et al., 2018; Matyszewski et al., 2021; Mazanek and Sohn, 2019).

Consistent with our cellular imaging assays (Figure 2A–B), POP1 did not affect the oligomerization kinetics or maximum FRET efficiency (amplitude) of ASCPYD up to the highest concentration we could achieve (Figure 2E; 150 µM POP1 vs. 2.5 µM ASCPYD). Moreover, no FRET signals were detected between donor-labeled POP1 and acceptor-labeled ASCPYD (Figure 2—figure supplement 2C). Additionally, the presence of POP1 did not affect the formation of ASCPYD filaments when visualized with negative-stain electron microscopy (nsEM) (Figure 2F). By contrast, both POP2 and POP3 prolonged the initial lag phase of ASCPYD polymerization in a dose-dependent manner (up to ~15 min delay in nucleation; Figure 2E), while also moderately affecting the elongation phase (≤20% reduction in the slope; Figure 2E). Estimating the polymerization half-times at different POP concentrations indicated that POP2 and POP3 are similarly effective in suppressing the oligomerization of ASCPYD (Figure 2E, IC50s). When visualized via nsEM, no ASCPYD filaments were detected in the presence of POP2, and only a few filaments were seen with POP3 (Figure 2F). These results suggest that even if ASCPYDs form oligomers (rise in FRET signals in Figure 2E), most of them fail to assemble into intact filaments in the presence of POP2/3. Moreover, the complete absence of any ASCPYD filaments in the presence of POP2 is consistent with our cellular experiments in which POP2 was most potent in inhibiting the central adaptor (Figure 2B and F). Together, our in cellulo and in vitro experiments consistently indicate that POP1 is only marginally effective in directly suppressing the polymerization of ASC. We also find that both POP2 and POP3 impede the nucleation of the ASCPYD filament.

Re-examining the Rosetta analyses in light of biochemical experiments

Our biochemical experiments indicated that excess POPs are required to inhibit the polymerization of ASC (Figure 2E), which is in agreement with the Rosetta analyses in which no POP•PYD interactions were more favorable than homotypic PYD•PYD interactions (Figures 1 and 2). However, although our in silico analyses suggested that POP1 should interact most favorably with ASCPYD (Figure 1B), our in vitro and in cellulo experiments consistently showed that POP1 is only marginally inhibitory (Figure 2). We thus re-examined our Rosetta results in light of our biochemical experimental results, and noted that the interface energy profiles of POP2•ASCPYD and POP3•ASCPYD are different from that of POP1•ASCPYD. For instance, all three POPs contain favorable protein•protein interaction surfaces for ASCPYD (∆∆G = ∆GPYD•PYD- ∆GPOP•PYD ≤3.5; arbitrarily determined, marked as blue dots in Figure 2—figure supplement 3). However, although both POP2 and POP3 show multiple interfaces with less favorable ∆Gs than ASCPYD•ASPYD interactions, POP1•ASCPYD interfaces lack such negative interactions (∆∆G≥10.0; marked as red dots in Figure 2—figure supplement 3). These observations in turn raised the hypothesis that a combination of favorable (recognition) and unfavorable interfaces (repulsion) is necessary for POPs to interfere with the assembly of inflammasome PYDs.

Mechanisms of ALR inhibition by POPs

We then set out to test our amended interpretation of Rosetta analyses on other POP•PYD interactions such as those with AIM2 and IFI16. Here, we noted a mixture of both favorable and unfavorable interactions between all three POPs and both ALRs (Figure 3—figure supplement 1A–B), raising the possibility that not only POP3, but the other two POPs might also inhibit the oligomerization of ALRs. To test this, we monitored the filament assembly of AIM2PYD-mCherry and IFI16PYD-mCherry in HEK293T cells (Figure 3A–D). Compared to co-transfecting with eGFP alone, POP1-eGFP reduced the number of AIM2PYD and IFI16PYD filaments, apparently more effective than against ASCPYD (Figure 2B vs. Figure 3B and D; e.g. at 1200 ng POP1, AIM2PYD and IFI16PYD assemblies were inhibited ~60%, while ASCPYD assembly was suppressed ~20%). On the other hand, co-transfecting POP2 or POP3 essentially obliterated the filament assembly of AIM2PYD and IFI16PYD (Figure 3A–B and C–D). In our FRET assays tracking AIM2PYD polymerization, all three POPs decreased the slope of the linear phase in a dose-dependent manner without affecting the initial lag phase (Figure 3E; recombinant IFI16PYD does not form filaments in our hands; Morrone et al., 2014). Our observations indicate that all three POPs can interfere with the elongation of the AIM2PYD filament, with POP3 being most effective (Figure 3E, IC50s). Moreover, imaging AIM2PYD filaments using nsEM in the presence of POP1 revealed that the filaments are shorter and fewer, and the presence of POP2/3 abrogated filament formation (Figure 3F). As seen from ASCPYD, the dearth of filaments in the presence of POP2/3 in our nsEM and in cellulo experiments (Figure 3B, D and F) indicated that AIM2PYD oligomers rarely progressed into functional filaments (i.e. rise in FRET signals in Figure 3E vs. the lack of filaments in Figure 3F).

Figure 3 with 3 supplements see all
Inhibition of AIM2-like receptor (ALR) assembly by pyrin-only-proteins (POPs).

(A) Sample fluorescent microscope images of HEK293T cells co-transfected with mCherry-tagged AIM2PYD (300 ng; crimson) plus eGFP alone or POP-eGFP (1200 ng; green). Blue: DAPI. (B) The relative amounts of AIM2PYD-mCherry filaments (300 ng plasmid) in HEK293T cells when co-transfected with POP-eGFP (+) or eGFP alone (-) (600 and 1200 ng plasmids). N≥4. (C) Sample fluorescent microscope images of HEK293T cells co-transfected with mCherry-tagged IFI16PYD (300 ng; crimson) plus eGFP alone or eGFP-tagged POPs (1200 ng; green). Blue: DAPI. (D) The relative amount of IFI16PYD-mCherry filaments (300 ng plasmid) in HEK293T cells when co-transfected with POP-eGFP (+) or eGFP alone (-) (600 and 1200 ng plasmids). N≥4. (E) Time-dependent increase in Förster resonance energy transfer (FRET) signals of a donor- and acceptor-labeled AIM2PYD (2.5 µM, black) was monitored with increasing concentrations of POP1 (25, 50, 100, and 150 µM), POP2 (12.5, 25, 50, and 75 µM), or POP3 (7.5, 15, 30, 40, and 50 µM); darker shades correspond to increasing POP concentrations. Arrows pointing upper right directions indicate the change (or lack thereof) in the elongation phase in the presence of the highest POP concentrations used. Data shown are representatives of at least three independent measurements (IC50s are average values of these experiments. n=3). (F) Negative-stain electron microscopy (nsEM) images of AIM2PYD filaments (2.5 µM) in the presence and absence of POP1 (100 µM), POP2 (50 µM), or POP3 (40 µM).

It is noteworthy that the oligomerization of PYD is important for stable dsDNA binding by ALRs (Morrone et al., 2015; Morrone et al., 2014). Conversely, although isolated AIM2PYD can form filaments by mass action (i.e. high concentrations; Morrone et al., 2015), dsDNA provides a one-dimensional diffusion scaffold to facilitate the assembly of full-length ALRs at significantly lower concentrations (Morrone et al., 2015; Morrone et al., 2014). AIM2FL forms punctum-like oligomers when transfected in HEK293T cells (Matyszewski et al., 2021), and we found that POP1 slightly reduced the number of AIM2FL puncta, while POP2 and POP3 were more effective (Figure 3—figure supplement 1C–D). IFI16FL localizes in the nucleus (Antiochos et al., 2018; Kerur et al., 2011; Li et al., 2012), precluding our investigation with cytosolic POPs (Figure 3—figure supplement 1E). Consistent with the lack of significant inhibition in cells (Figure 3—figure supplement 1C–D), POP1 failed to interfere with dsDNA-binding/oligomerization of recombinant AIM2FL (Figure 3—figure supplement 2A). However, POP2 and POP3 still inhibited the dsDNA binding of AIM2, while only POP3 was inhibitory toward the dsDNA binding of recombinant IFI16FL (Figure 3—figure supplement 2B). Additionally, imaging experiments with untagged POPs again corroborated our observations using eGFP-tagged proteins (Figure 3—figure supplement 3). Overall, our results indicate that POP3 directly inhibits ALR assembly (elongation in particular for AIM2PYD). We also find that POP1 and POP2 can inhibit the assembly of ALR filaments, with POP2 being more effective than the former; the presence of activating ligands can diminish the inhibitory effect of POPs (Figure 3—figure supplement 2). Furthermore, these results are consistent with our amended interpretation of Rosetta analyses in which a combination of favorable and unfavorable interfaces allow POPs to target and inhibit PYD filament assembly.

POP1 likely targets upstream receptors instead of ASC

Although it has been speculated that POP1 and POP2 would interfere with the recruitment of ASC by NLRPs (de Almeida et al., 2015; Devi et al., 2020; Indramohan et al., 2018; Periasamy et al., 2017; Ratsimandresy et al., 2017), it remains unknown whether either POP can directly suppress the filament assembly of NLRPPYDs. Of note, our investigations here revealed that POP1 is ineffective in inhibiting the oligomerization of ASC (Figure 2). Moreover, albeit less inhibitory than POP2 or POP3, POP1 was more effective in suppressing the assembly of AIM2PYD and IFI16PYD filaments than that of ASCPYD (Figure 3). These observations strongly suggest that the role of POP1 is to interfere with the assembly of upstream receptors rather than directly inhibiting ASC (i.e. a ‘decoy’ ASC; targeting ASC or multiple upstream receptors would result in the same phenotype). Indeed, our Rosetta analyses indicated that POP1 can make a combination of favorable and unfavorable interactions with both NLRP3PYD and NLRP6PYD (Figure 4—figure supplement 1). Furthermore, POP2 and POP3 also showed favorable and unfavorable interactions with NLRP3 (Figure 4—figure supplement 1A); although the ∆Gs between NLRP6 and POP2/3 were largely unfavorable, the Type 3a surface showed an energy score that might allow the two POPs to recognize NLRP6 if present at high enough concentrations (∆G ~ –9; Figure 4—figure supplement 1B, marked as light pink); our reasoning is based on ∆Gs ~ –9 seen from native PYD•PYD interactions (e.g. NLRP3PYD•NLRP3PYD shows ∆Gs of ~–8 and –10 on Type 2 and Type 3 interfaces, Figure 1 and Figure 4—figure supplement 1A).

The activation mechanisms of NLRPs are complex and involve different types of active and inactive oligomers (Andreeva et al., 2021; Gong et al., 2021; Hochheiser et al., 2022b; Lu et al., 2014; Ohto et al., 2022; Sharif et al., 2019; Sharma and de Alba, 2021; Shen et al., 2021; Shen et al., 2019); we thus monitored whether POPs impede the filament assembly using the isolated PYDs of NLRP3 and NLRP6, which formed filaments when ectopically expressed in HEK293T cells (Figure 4A and C). Of note, NLRP2PYD-mCherry did not form filaments when expressed in HEK293T cells (Figure 4—figure supplement 1C), precluding further investigations despite its high sequence similarity to POP2 (Indramohan et al., 2018; Periasamy et al., 2017; Ratsimandresy et al., 2017). When co-transfected, POP1 was more effective in suppressing the filament assembly by both NLRP3PYD and NLRP6PYD than that of ASCPYD (Figure 4A–D). For example, with 1200 ng POP1, ASCPYD assembly was only suppressed by ~20%, but the filament assembly by NLRP3PYD and NLRP6PYD was suppressed ~60% (Figure 2B vs. Figure 4B and D). On the other hand, POP2 was less effective in inhibiting the polymerization of NLRPPYDs than that of ASCPYD (e.g. at 600 ng POP2, ASCPYD assembly was abolished, but the assembly of NLRP3PYD and NLRP6PYD was minimally suppressed; Figure 2B vs. Figure 4B and D). POP3 was almost equally effective in suppressing the polymerization of NLRPPYDs and ASCPYD, but not nearly as effective as against ALRs (Figure 4A–D). For example, at 600 ng POP3, AIM2PYD assembly was suppressed 70%, but those of ASCPYD and NLRPPYDs were reduced by 40–50%; Figure 3B vs. Figures 2B, 4B and D. As with other PYDs, cellular imaging experiments using untagged POPs corroborated our observations (Figure 4—figure supplement 2A–D).

Figure 4 with 2 supplements see all
Inhibition of NLRPPYD assembly by pyrin-only-proteins (POPs).

(A) Sample fluorescent microscope images of HEK293T cells co-transfected with mCherry-tagged NLRP3PYD (600 ng; crimson) plus eGFP alone or eGFP-tagged POPs (1200 ng; green). Blue: DAPI. (B) The relative amounts of NLRP3PYD-mCherry filaments (600 ng plasmid) in HEK293T cells when co-transfected with POP-eGFP (+) or eGFP alone (-) (600 and 120 ng plasmids). N≥4. (C) Sample fluorescent microscope images of HEK293T cells co-transfected with mCherry-tagged NLRP6PYD (300 ng; crimson) plus eGFP alone or eGFP-tagged POPs (1200 ng; green). Blue: DAPI. (D) The relative amounts of NLRP6PYD-mCherry filaments (300 ng plasmid) in HEK293T cells when co-transfected with POP-eGFP (+) or eGFP alone (-) (600 and 120 ng plasmids). N≥4. (E) Time-dependent increase in Förster resonance energy transfer (FRET) signals of a donor- and acceptor-labeled NLRP6PYD (2.5 µM, black) was monitored with increasing concentrations of POP1 (25, 50, and 100 µM). POP2 (30, 60, and 120 µM). POP3 (15, 30, and 60 µM); darker color shades correspond to increasing POP concentrations. Two- and three-headed arrows indicate the increase in apparent nucleation time. Arrows pointing upper right directions indicate the change (or lack thereof) in the elongation phase in the presence of the highest POP concentrations used. Data shown are representatives of at least three independent measurements (IC50s are average values of these experiments. N=3). (F) Negative-stain electron microscopy (nsEM) images of NLRP6PYD filaments (5 µM) in the presence and absence of POP1 (100 µM), POP2 (30 µM), or POP3 (30 µM).

Next, using FRET donor and acceptor-labeled NLRP6PYDs, we then monitored whether POPs suppress the nucleation and/or elongation (recombinant NLRP3PYD does not auto-assemble into filaments in our hands [e.g. Bae and Park, 2011]). POP1 predominantly inhibited the elongation of NLRP6PYD filament, and POP2 appeared to interfere with its nucleation. Although POP3 mostly reduced the elongation kinetics of NLRP6PYD, it also seemed to interfere with nucleation (Figure 4E). Consistent with these observations, the number and length of NLRP6PYD filaments were reduced in the presence of POPs (Figure 4E). The lack of filaments (Figure 4F) despite the increase in FRET signals (Figure 4E) again indicates that NLRP6PYD oligomers fail to form intact filaments. Finally, we tested the efficacy of POPs in suppressing the downstream signaling activities of inflammasomes by reconstituting NLRP3-dependent IL-18 release in HEK293T cells. Consistent with our imaging and polymerization assays, here, POP2 was most effective in suppressing IL-18 release, POP3 was the second, and POP1 was least effective (Figure 4—figure supplement 2E). Our results are consistent with the idea that POP1 acts as a decoy ASC, moderately inhibiting the assembly of upstream PYDs; POP2 is the most potent inhibitor of inflammasomes; POP3 can inhibit inflammasomes beyond AIM2.

Introducing deleterious mutations for self-assembly can reprogram PYDs into POP-like inhibitors

Next, to further test our working model for understanding the inhibition mechanism of POPs, we generated AIM2PYD mutants defective in self-oligomerization and tested whether they can inhibit the assembly of WT (i.e. such mutant proteins contain both favorable and unfavorable interactions for WT). We chose D23K and N73L (Figure 5A), because we previously found that not only are these mutations deleterious for self-assembly, but resulting mutant AIM2FL•dsDNA oligomers were defective in promoting the polymerization of WT-AIM2PYD (Matyszewski et al., 2021). Both N73L-AIM2PYD and D23K-AIM2PYD were impaired in filament assembly (Figure 5B, Figure 5—figure supplement 1A). Imaging WT-AIM2PYD filaments using nsEM in the presence of either mutant showed fewer filaments (N73L) and mesh-like non-filamentous aggregates (D23K; Figure 5B). The presence of unlabeled N73L-AIM2PYD or D23K-AIM2PYD also impeded the polymerization of WT-AIM2PYD in our FRET assay, with the former being more effective (Figure 5C). Moreover, when co-transfected in HEK293T cells, eGFP-tagged AIM2PYD mutants diminished the number of mCherry-tagged WT filaments, again N73L being more effective than D23K (Figure 5D–E, Figure 5—figure supplement 1B–C). Overall, these results consistently support our hypothesis that a combination of favorable and unfavorable interactions underpins the target selection and inhibition by POPs. Our results also provide a proof of principle that PYDs can be reprogrammed into POP-like inhibitors even by a single mutation.

Figure 5 with 1 supplement see all
Reprogramming AIM2PYD into a pyrin-only-protein (POP)-like inhibitor.

(A) AIM2PYD filament ‘honeycomb’ showing the location of mutations. (B) Negative-stain electron microscopy (nsEM) images of AIM2PYD variants. WT: 2.7 µM, N73L: 3 µM, D23K: 30 µM. (C) Time-dependent increase in Förster resonance energy transfer (FRET) signals of a donor- and acceptor-labeled WT-AIM2PYD (2.5 µM, black) was monitored with increasing concentrations of AIM2PYD mutants (1 (brown), 2 (orange), and 3 (mustard) µM for N73L; 15 (light green) and 30 (dark green) µM for D23K). Data shown are representatives of at least three independent measurements. (D) Sample fluorescent microscope images of HEK293T cells co-transfected with mCherry-tagged AIM2PYD (300 ng plasmid; crimson) plus eGFP-tagged N73L- or D23K-AIM2PYD (1200 ng plasmids; green). Blue: DAPI. (E) The relative amount of AIM2PYD-mCherry filaments (300 ng plasmid) in HEK293T cells when co-transfected with N73L or D23K mutants (600 and 1200 ng plasmids). n≥4.

Discussion

Redefining the target specificity and inhibitory mechanisms of POPs

A hallmark of inflammasomes is their exceptional stability. For example, ASC promotes ‘solidification’ of inflammasomes in a prion-like manner, allowing them to perpetuate even after cells undergo pyroptosis (Cai et al., 2014; Franklin et al., 2014; Shen et al., 2021). Moreover, AIM2 and IFI16 filaments also persist and are even stigmatized as autoantigens in debilitating autoimmune disorders such as systemic lupus erythematosus and Sjögren’s syndrome (Antiochos et al., 2018; Antiochos et al., 2022; Baer et al., 2016). Indeed, persisting inflammasome oligomers and their aberrant activities are implicated in a wide range of human diseases including COVID-19 (Karki et al., 2017; Tartey and Kanneganti, 2020; Vora et al., 2021). It is thus critical for the host to carefully modulate the assembly of inflammasomes at the onset, as it would be much more difficult to demolish such hyper-stable supra-structures. POPs have emerged as major endogenous regulators of inflammasomes by directly interfering with the assembly of PYD filaments, functioning analogous to COPs that target the oligomerization of pro-caspases (Devi et al., 2020; Indramohan et al., 2018). Although the biological significances of POPs are well established (de Almeida et al., 2015; Devi et al., 2020; Indramohan et al., 2018; Khare et al., 2014; Periasamy et al., 2017; Ratsimandresy et al., 2017), their intrinsic target specificities and inhibition mechanisms have remained speculative.

Our investigations here reveal that POPs interfere with the polymerization (nucleation and/or elongation) of various inflammasome filaments without co-assembling, which is different from COPs that co-assemble into filaments with CARD of caspase-1 (Lu et al., 2016). Moreover, also unlike COPs that can inhibit the assembly of pro-caspases at sub-stoichiometric concentrations (Lu et al., 2016), excess POPs were necessary to inhibit inflammasome PYDs, especially when an activating ligand was present (dsDNA for AIM2 and IFI16; Figure 3—figure supplement 2). Additionally, POP2/3 largely suppressed the nucleation of ASC (e.g. prolonged lags in Figure 2E), yet these POPs also interfered with the elongation of upstream receptors (e.g. Figures 3E and 4E). Although often considered as a harmful phenomenon, inflammation is integral to host innate defense and survival (Bennett et al., 2018; Meizlish et al., 2021). We reason that being able to modulate two key assembly steps (nucleation and elongation) while requiring excess POPs is well suited for attenuating inflammasome activities without shutting them down altogether.

Although previous studies showed that POP1 inhibits ASC-dependent inflammasomes (de Almeida et al., 2022; de Almeida et al., 2015), its target selection and inhibition mechanism remains poorly understood. Here, we provide new insights into the inhibitory mechanism of POP1. That is, POP1 is only marginally effective in inhibiting the polymerization of ASCPYD. However, it is likely to modulate inflammasomes largely through interfering with the oligomerization of upstream PYDs, most notably by halting their elongation. It is noteworthy that the expression of POP1 is predominantly induced by IL-1β (de Almeida et al., 2015), a major final product of inflammasome cascades (Broz and Dixit, 2016; Zheng et al., 2020). Thus, it is likely that POP1 is part of a negative feedback loop for attenuating excessive inflammasome activities by preventing perpetuation of upstream filaments (Figure 6).

Figure 6 with 1 supplement see all
Target selection and mode of inhibition of pyrin-only-proteins (POPs).

A cartoon summarizing the refined intrinsic target specificity of POPs. Solid red lines indicate the most likely primary inhibitory targets for each POP. Solid gray lines indicate an additional target that each POP could also directly inhibit. Dotted gray lines indicate other possible targets for each POP.

POP2 is thought to function similarly to POP1 by suppressing the activation of ASC (Periasamy et al., 2017; Ratsimandresy et al., 2017). Interestingly, POP2 is reportedly more effective than POP1 in preventing spurious inflammasome activities in vivo (Periasamy et al., 2017; Ratsimandresy et al., 2017). Nevertheless, it has remained unclear whether it directly suppresses ASC assembly or those of upstream receptors. We find here that POP2 can indeed potently inhibit the nucleation of ASCPYD (Figure 2) while moderately suppressing the oligomerization of select upstream receptors such as AIM2. It is noteworthy that the expression of POP2 is readily induced by a wide variety of pro- and anti-inflammatory cytokines (Ratsimandresy et al., 2017). Considering that it primarily targets ASC, we postulate that POP2 would act as the pan-inflammasome inhibitor essential for preventing spurious innate immune responses (Figure 6; i.e. decoy receptor as it is most similar to upstream NLRP2PYD, Figure 1—figure supplement 1B).

POP3 was identified as a selective inflammasome inhibitor targeting ALRs (Khare et al., 2014). In addition to inhibiting the elongation of the AIM2PYD filament, we find here that POP3 can robustly inhibit the nucleation of ASCPYD (Figure 2), revealing its dual targeting function similar to POP2. We also find that, in principle, POP3 can inhibit the assembly of NLRPPYDs (Figure 4). Unlike the other two POPs, POP3 is exclusively induced by interferons (IFNs) (Khare et al., 2014), a major cytokine family that counteracts IL-1β (Mayer-Barber and Yan, 2017). Considering that IFNs also drive the expression of ALRs (Antiochos et al., 2018; Khare et al., 2014), we postulate that although POP3 is intrinsically capable of inhibiting various inflammasome receptors, the contextual expression would dictate its in vivo targets (Figure 6).

Design principles for inflammasome inhibition by POPs

We envision that the nonequilibrium assembly of inflammasome filaments plays a major role in defining the mechanism of regulation by POPs (i.e. kinetically driven without having conceivable off-rates) (Cai et al., 2014; Matyszewski et al., 2018). For instance, considering that all POP and PYD protomers share the same overall structure (shape complementarity), we reason that any (semi) favorable protein•protein interaction interfaces would allow POPs and PYDs to associate at least transiently (i.e. classic reversible protein•protein interaction equilibrium). However, the more favorable homotypic PYD•PYD interactions would readily outcompete such meta-stable interactions especially when only basal amounts of POPs are present (Figure 6—figure supplement 1A). Moreover, as indicated from the failure to inhibit ASCPYD by POP1, if POPs do not contain any strong unfavorable interactions against target PYDs, homotypic PYD•PYD interactions would still outcompete even excess POPs and lock themselves into irreversible filament assembly (Figure 6—figure supplement 1B). However, when such meta-stable POP•PYD complexes contain at least one very unfavorable interface, excess POPs would then expose a multitude of adverse protein•protein interaction surfaces that would hamper filament assembly (Figure 6—figure supplement 1C). Also of note, given that POP2 and POP3 can form oligomers (Figure 2—figure supplement 2A–B), it is highly likely that multimeric POPs are more effective in preventing the association of inflammasome PYDs (Figure 6—figure supplement 1C). We found previously that the recognition between AIM2 and ASC occurs when at least one is filamentous (Matyszewski et al., 2021). Thus, it is also possible that POPs might preferentially interact with oligomeric PYDs that are not yet fully filamentous (e.g. (pseudo)-nucleation unit), trapping them into nonfunctional states (Figure 6—figure supplement 1C).

In closing, a possible caveat of our study is that although our experiments using AIM2FL, IFI16FL, and NLRP3FL are consistent with those from isolated PYDs (Figure 3—figure supplements 1C, 2 and Figure 3—figure supplement 3E; Figure 4—figure supplement 2E), considering that inflammasome receptors contain multiple domains, there could be yet another layer of complexity in POP•inflammasome interactions beyond what we have reported here. It will be also interesting to see to what extent our findings for POP•PYD interactions can be applied to other DD family proteins such as COPs and CARDs. Overall, our multi-disciplinary approach provides an example of how to use in silico predictions judiciously for investigating multipartite protein-protein interactions.

Materials and methods

Rosetta simulation

Request a detailed protocol

The InterfaceAnalyzer script in Rosetta was used to determine the interaction energy (Rosetta energy units, reu) at individual interfaces of the honeycomb (Matyszewski et al., 2021). We used the cryo-EM structures of ASCPYD (PDB: 3j63; Lu et al., 2014), AIM2PYD (PDB: 7k3r; Matyszewski et al., 2021), NLRP3PYD (PDB: 7pdz; Hochheiser et al., 2022a), and NLRP6PYD (PDB: 6ncv; Shen et al., 2019) filaments to generate corresponding honeycombs. Because the structure of the IFI16PYD filament is unknown, we used the eGFP-AIM2PYD filament (PDB: 6mb2; Lu et al., 2015) that shows a pentameric filament base as a template; using the untagged AIM2PYD filament (PDB: 7k3r; Matyszewski et al., 2021), which shows a hexameric filament base, as a template resulted in largely unfavorable energy scores (Figure 1—figure supplement 1C). For POPs, we used the crystal structure of POP1 (PDB: 4qob), and generated homology models of POP2 and POP3 based on monomeric NLRP3PYD (PDB: 7pdz) and AIM2PYD (PDB: 7k3r), respectively.

Cell culture and imaging

Request a detailed protocol

Each protein was cloned into pCMV6 vector containing C-terminal mCherry (inflammasome PYDs) or eGFP (POPs). HEK293T cells (ATCC, CRL-11268) were seeded into 12-well plate (0.1×106 per well) with round cover glass (20 mm). All cells were authenticated via STR profiling and free from mycoplasma. eGFP (or vector) alone or POP-eGFP plasmids (or tagless POPs; 600 and 1200 ng) were co-transfected with inflammasome-mCherry plasmids (300 ng, except NLRP3PYD [600 ng]) at 70% confluence using Lipofectamine 2000 (Invivogen). After 16 hr, cells were washed twice with 1× phosphate-buffered saline, fixed with 4% paraformaldehyde, then mounted on glass slides. Images were taken using the Cytation 5 multi-functional reader equipped with a fluorescent microscope (BioTek) and analyzed via the Gen5 software (BioTek). All paired two-tailed t-tests were performed using Excel (*: p≤0.05, **: p≤0.01; ***: p≤0.001). Source Data are appended for each figure.

IL-18 release assay

Request a detailed protocol

Plasmids encoding full-length NLRP3 (2 µg), ASC (1 ng), pro-caspase-1 (42 ng), and pro-IL-18 (360 ng) were co-transfected into HEK293T cells in the presence and absence of increasing POPs (six-well plate, 0.75×106 per well). After 16 hr, 0.5 µMf nigericin was added to activate NLRP3 and incubated for 4 hr. Supernatant from each well was then collected and the presence of mature IL-18 was monitored by human IL-18 ELISA kit (Invitrogen).

Recombinant proteins

Request a detailed protocol

Inflammasome proteins were generated and labeled with fluorophores when appropriate as previously described (Matyszewski et al., 2018; Matyszewski et al., 2021; Morrone et al., 2015; Morrone et al., 2014). POP1 was cloned into the pET21b vector, and POP2 and POP3 were cloned into the pET28b vector containing an N-terminal His6-MBP tag flanked by a cleavage site for TEVp. All recombinant proteins were expressed in Escherichia coli BL21 Rosetta2DE3 cells and purified using Ni2+-NTA followed by size-exclusion chromatography (SEC) (storage buffer: 40 mM HEPES-NaOH at pH 7.4, 400 mM NaCl, 2 mM dithiothreitol, 0.5 mM EDTA, and 10% glycerol). Proteins were then concentrated and stored at –80°C.

Biochemical assays

Request a detailed protocol

FRET and FA-based quantitative assays were conducted as described previously (Matyszewski et al., 2018; Matyszewski et al., 2021; Mazanek and Sohn, 2019; Morrone et al., 2014). For example, the polymerization of indicated amounts of FRET-labeled MBP-PYD constructs was triggered by adding TEVp in the presence of increasing amounts of POPs. Half-times for polymerization (t1/2s) and the concentration of each POP needed for decreasing the 1/(t1/2)s by 50% (IC50) were calculated as described in Matyszewski et al., 2018; Matyszewski et al., 2021. The IC50s for inhibiting the dsDNA-binding activities of ALRs were determined with increasing concentrations of POPs as described previously (Matyszewski et al., 2018; Matyszewski et al., 2021); the MBP tag was pre-cleaved in these experiments via TEVp for 30 min. Source Data are appended for each figure.

nsEM

Request a detailed protocol

Each PYD was incubated with TEVp for 30 min to remove MBP tag and promote polymerization in the presence or absence of POPs. Samples were then applied to carbon-coated grids and imaged as described previously (Matyszewski et al., 2018; Matyszewski and Sohn, 2019; Morrone et al., 2015).

Data availability

Source data filles for Figures 2, 3 and 4 contain the numerical data used to generate figures.

References

Decision letter

  1. Russell E Vance
    Reviewing Editor; University of California, Berkeley, United States
  2. Carla V Rothlin
    Senior Editor; Yale University, United States

Our editorial process produces two outputs: i) public reviews designed to be posted alongside the preprint for the benefit of readers; ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Decision letter after peer review:

Thank you for submitting your article "Design Principles for Inflammasome Inhibition by Pyrin-Only-Proteins" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Carla Rothlin as the Senior Editor. The reviewers have opted to remain anonymous.

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted a consensus set of essential revisions. Overall our judgement is that the paper is of potential enough importance to be considered for eLife, but significant revisions including new experiments are required.

In this manuscript, Mazanek and colleagues combine computational analysis and in vitro experiments to develop a comprehensive analysis of the ability of pyrin-only proteins (POPs) to inhibit inflammasome assembly. The pyrin domain (PYD) of each POP has a total of six interaction surfaces that can bind to the PYD of different inflammasome components. The results lead the authors to propose that a mixture of favorable and unfavorable interaction surfaces is required for a POP to inhibit a given inflammasome component. The results are potentially important as they may provide a generalizable mechanism by which POPs inhibit inflammasome activation. However, additional experimentation is required to fully justify the authors' model. Indeed, their 'mixed interface' model is not directly tested and the possibility that POPs could inhibit inflammasome signaling by being incorporated into filaments is not ruled out. A limitation of the study is that the authors only measure PYD filament formation and do not actually measure downstream inflammasome activation.

Essential revisions:

1. The authors show that the MBP tag affects the oligomerization of POPs. The POPs used in Figures 2A, 3A, and 4A contain a GFP tag which may change the inhibitory effect of POPs on ASC filament formation. Experiments with untagged POPs are therefore required to validate the results.

2. The authors take the reduction of PYD filamentation as an indication of inhibition, but it is not clear how they ruled out the possibility that POP1 co-assembles into the ASCPYD filaments and inhibits inflammasome formation by repressing the recruitment of Caspase-1 (as POP1 lacks the CARD the effector domain). Thus, some functional assays measuring downstream Caspase-1 activation are required. In addition, the possibility that POP1 and ASC co-assemble could be tested directly with FRET experiments in which one protein is the donor and the other is the acceptor. Without such experiments, the statement in the Discussion "Our investigations here reveal that POPs interfere with the polymerization (nucleation and/or elongation) of various inflammasome filaments without co-assembling, … " does not appear to be justified.

3. Further computational analysis should be performed to determine if the theory that a combination of favorable and unfavorable interactions is generally applicable. Does this theory account for other PYD/PYD interactions and CARD/CARD interactions? For example, for the AIM2PYD/ASCPYD interface, do they see only a favorable interface or a mixture? How about two unrelated PYDs, such as between AIM2PYD and NLRP3PYD? How about for COPs? How about the homotypic interfaces between the POPs themselves? In addition, as raised by reviewer 3, the authors do not consider at all in this manuscript that there is a seventh interface described for PYDs besides the hexagonal assembly in the filament. This is the homodimer interface seen, e.g., for NLRP3, in the crystal structure of the PYD and in size exclusion chromatography (Bae and Park, 2011; 3QF2). For completeness, the energy scores should be also calculated for this interface. It might well be that POPs associate in this binding mode as heterodimeric assemblies to monomeric PYDs of NLRs or ALRs to regulate their activities. This would be somehow reminiscent of profilin binding to actin, regulating the pool of free G-actin for filament assembly.

4. To more directly test the mixed-interaction model, the authors should use their Rosetta structural predictions as a guide to introduce mutations into the various POP1/ASC-PYD interfaces to reduce the binding energies of those specific interactions and then test whether the introduction of a single or multiple, weak interactions then allows POP1 to restrict ASC-PYD oligomerization. To further elucidate their mixed interface model, the authors should also address whether the weak interactions need to be on the same 'half' of the interface, e.g., does weakening the 1b and 2b interfaces lead to better disruption of PYD filamentation than a 1b/1a combination mutant?

Reviewer #1 (Recommendations for the authors):

1. Figures 2A, 3A, and 4A could be complemented with untagged POPs, which may limit the observation of POP expression, but could still tell how they affect ASC filamentation in their physiological form by observing the mCherry signal.

2. Figure 2-4, panel Es and panel Fs: It should be described in the figure legend that the MBP-tagged ASC protein is used to measure the FRET signal upon TEV cleavage. It should also be noted whether the MBP-tagged POPs are used in these assays, or the proteins w/o tag.

3. To test the co-assembly model, the FRET assay should be performed with either POP1 or PYD as the donor, and the other protein as the receptor.

4. An inflammasome functional assay should be performed to test if POP1 inhibits inflammasome activation w/o disrupting ASC filament formation.

5. If the authors can calculate the AIM2PYD /ASCPYD interface, do they see only a favorable interface or both? How about two unrelated PYDs, such as between AIM2PYD and NLRP3PYD? How about COPs?

Reviewer #2 (Recommendations for the authors):

I believe that this is a strong manuscript with a careful and detailed analysis of the ability of the POPs to disrupt PYD filament formation. However, I believe that there are two critical points that the authors should address.

First, the authors propose that the mix of strong and weak interactions between the POPs and the various PYDs regulate the ability of the POPs to block PYD filamentation. For example, the authors propose that the reason POP1 fails to restrict ASC-PYD filamentation is due to a lack of weak interactions, whereas both POP2 and POP3 are able to interfere with ASC-PYD oligomerization through a mix of strong and weak interactions. This is an interesting hypothesis, however I believe that it needs to be tested directly.

In particular, I would like to see the authors use their Rosetta structural predictions as a guide to introduce mutations into the various POP1/ASC-PYD interfaces that reduce the binding energies of those specific interactions and then test whether the introduction of a single, or multiple, weak interactions then allows POP1 to restrict ASC-PYD oligomerization. Additionally, I believe that it may be prudent to generate combinations of mutants along the different interfaces to uncover whether the weak interfaces are required to be on the same facing to disrupt PYD oligomerization – for example, would weakening the 1b and 2b interfaces lead to better disruption of PYD filamentation than a 1b/1a combination mutant? I believe these experiments are necessary for the authors to demonstrate that a combination of strong and weak interfaces is necessary to disrupt PYD filamentation as they propose. Additional mutagenesis of the other POP/PYD interactions may be an appropriate alternative to directly test the proposed mechanism.

A potential area for clarification on this point is the mechanism through which POP1 fails to regulate ASC PYD oligomerization. Looking at the del(G) calculated for POP1/ASC-PYD interaction the del-del(G) compared with ASC-PYD/ASC-PYD is only ~1. Compared with the POP2/NLRP3-PYD interaction in which POP2 fails to prevent NLRP3-PYD oligomerization and where the del-del(G) is nearly 30, this seems unexpected. Is POP1 becoming incorporated within the ASC-PYD filament? If the authors mix individually labeled POP1 and ASC-PYD protein in their FRET assay do they observe co-oligomerization and or co-localization in their spec assays?

The second issue I believe that the authors need to address is whether the POPs truly disrupt inflammasome signaling through the mechanisms they propose. While I appreciate the authors' microscopy demonstrating a lack of PYD spec formation in their reconstituted system I believe that they need to demonstrate that the POPs are able to restrict inflammasome activation in a more natural setting. For example, this could be accomplished by transducing the POPs (and/or mutants that are able or unable to restrict PYD oligomerization) into THP1s, prime, and activating the NLRP3 inflammasome and showing a loss of cell death and/or cytokine processing. Of course, it would be outside the scope of this manuscript to expect this for all the inflammasomes and POPs in combination, however, I believe that showing POP-mediated inhibition for at least one inflammasome/POP pair, along with POP mutants to demonstrate the loss of specific interface interactions proposed by the Rosetta calculations are required for the POPs to block inflammasome activity.

Reviewer #3 (Recommendations for the authors):

It seems to this reviewer that the authors have written the manuscript in chronological order as they have performed the experiments, which might, however, not be the best way to present their data. I propose to rewrite and reorganize the manuscript, to better make the points. The Discussion is a bit repetitive and seems overwritten. Some shortening, despite adding new ideas and considerations (see below), might be reasonable.

This reviewer does not understand the scientific rationale regarding the PYD sequence identity and similarity of POPS to either NLRs or ALRs as a measure of their potential regulatory/inhibitory function in the respective inflammasome formation. Think of a protein with a single-point mutation that would abrogate filament formation. It would have the exact same deltaGs in five interfaces but only one is different. If this protein would be added, e.g., in a 1:1000 ratio to wt protein that has been pre-incubated to form filaments, it might stop elongation. This is, for example, how some capping proteins work in the actin cytoskeleton. Thus, the immanent interpretation of the Rosetta calculations of the ASC-POP1 interfaces in Figure 1 showing that POP1 is so similar to ASC-ASC interfaces would be that POP1 could be easily integrated into the ASC filament as a surrogate. I, therefore, disagree with the authors about the statement of the "counterintuitive"-ness: "Moreover, although somewhat counterintuitive at first, further examining Rosetta in silico results in light of our initial biochemical experiments suggested that a combination of favorable and unfavorable interactions underpin the target specificity of POPs." [Discussion section, page 22 middle, and similar in the Results]. A more balanced consideration here would be favorable.

In addition to this comment, from the sequence similarity of the PYDs of NLRP3, AIM2, or NLRP6 to ASC-PYD we learn that the degree of sequence homology does not determine or correlate to the ability for homotypic transition and filament elongation.

Figure 1 is not a good start, particularly, as it is repeated and reversed later in Figure 2 Supplement 2, Figure 3 Supplement 1, and Figure 4 Supplement 1. I suggest putting the entire present Figure 1 into the supplement and rather start with something similar to Figure 2, Supplements 2, and so on, which is more informative. I also wonder if it wouldn't be easier to show the subtractions of the deltaGs (so the δ(deltaGs)) to better visualize the differences in the Rosetta calculations. The idea with the blue and red dots is good, the boundaries (<3.5 dG and >10 dG) seem reasonable, so I would include them from the beginning. Thinking of it, an ideal inhibitor might have 5 blue dots and one red. This would be preferentially in interface I (the largest) on the b-side (the growing direction) and it would ensure assembly but prevent elongation. I suggest to revise the text in light of this consideration.

The abundance of proteins is a major determinant in the law of mass action, besides the thermodynamic energies of the direct contacts made. The transcriptional control and the protein levels of the POPs might be therefore an important determinant similar to the yields of NLRs and ALRs. The authors reflect on this in the Discussion ("…, excess POPs were necessary to inhibit inflammasome PYDs, especially when an activating ligand was present, …) but the more than 1 stoichiometry required is puzzling and rather points towards a "chaperone-type" function for POPs.

The authors do not consider at all in this manuscript, that there is a seventh interface described for PYDs besides the hexagonal assembly in the filament. This is the homodimer interface seen, e.g., for NLRP3, in the crystal structure of the PYD and in size exclusion chromatography (Bae and Park, 2011; 3QF2). For completeness, the energy scores should be also calculated for this interface. It might well be that POPs associate in this binding mode as heterodimeric assemblies to monomeric PYDs of NLRs or ALRs to regulate their activities. This would be somehow reminiscent of profilin binding to actin, regulating the pool of free G-actin for filament assembly. This might be indeed an effective way to regulate effector signaling.

Discussion: "Our investigations here reveal that POPs interfere with the polymerization (nucleation and/or elongation) of various inflammasome filaments without co-assembling, … " Where is this shown, is it from the "%filament" in Figures 2B, D or the level of saturation in Figure 2E? Further: "…, excess POPs were necessary to inhibit inflammasome PYDs, especially when an activating ligand was present …". Isn't this a strong argument to consider also other interfaces/binding modes than the filament assembly. See the previous comment on profilin-G-actin binding.

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Design Principles for Inflammasome Inhibition by Pyrin-Only-Proteins" for further consideration by eLife. Your revised article has been evaluated by Carla Rothlin (Senior Editor) and a Reviewing Editor and all three of the original reviewers.

The reviewers remain interested in the work and consider that it is potentially appropriate for eLife. The reviewers are satisfied with many of the revisions and in particular agree that you should not be made to extend your observations to CARD-CARD interfaces. However, because you failed to address all the "required" revisions, significant issues remain.

In particular, after substantive discussion among the reviewers, there remains considerable skepticism about the fundamental model that is being proposed. This skepticism is summarized by one reviewer as follows: "The authors propose that a mixture of favorable and unfavorable interactions is critical in mediating inhibition of filament formation, however, it is unclear if this prediction holds true. For example, looking at Figure 4 and the associated supplemental material the authors find that POP1, POP2, and POP3 can inhibit NLRP6(PYD) filamentation. However, the IC50 varies by nearly 6-fold and more confusingly looking at the energy calculations in Figure S4.1 there does not seem to be a significant difference to explain why this would be. Additionally, if I compare the δ-deltaGs in the NLRP6 calculation with those from AIM2 calculation I cannot see any clear reason why POP3 is the strongest inhibitor of NLRP6 when every interaction is energetically unfavorable compared with the other POP/NLRP6 or NLRP6 homo-interactions. Yet, looking at the inhibition of AIM2, POP3 has the fewest energetically unfavorable interactions but is still the strongest inhibitor. The calculations for POP homotypic interactions supplied in the reviewer response further confound this issue – POP1 shows favorable interactions to form filaments similar to the free energies calculated for ASC and NLRP3, whereas POP2 does not – and yet as the authors acknowledge, POP2, but not POP1 forms filaments. In short, I am not convinced that there is sufficient evidence that the calculated energies are predictive of the behavior of the different POPs… without any experiments that directly address whether the energy calculations are predictive of POP-mediated inhibition, any connection between the calculated Rosetta energies and inhibition are correlative." The other reviewers tend to concur with this assessment. In order for your paper to be considered for publication we require the following:

1. You need to explicitly recognize/discuss/acknowledge that NLRs and ALRs contain additional domains beyond the PYD that can modify the energy landscape to create oligomers, and it is vital to consider this aspect as a caveat to your study which focuses on dissociated PYD interactions.

2. The consensus of the reviewers remains that you need to test your model by employing site-directed mutagenesis. This was a clear requirement in the previous decision letter (point 4), and the reviewers were not convinced by your argument that this is not necessary. One reviewer suggests: "One could easily imagine replacing a favorable interaction with alanine to gently weaken the interaction or with glutamate or other charged residue to break the interface. By taking POP1 and introducing mutations to disrupt one or multiple surfaces the authors could show that these mutants do or do not inhibit filament formation ASC. Alternatively, the authors could introduce mutations into the ASC PYD and determine whether the mutant protein could now act like a POP and block WT ASC filament formation. While I appreciate that doing such a comparison for every POP and PYD in the paper would be excessive, I strongly believe that the authors need to experimentally test their model with at least some site-directed mutants".

https://doi.org/10.7554/eLife.81918.sa1

Author response

Essential revisions:

1. The authors show that the MBP tag affects the oligomerization of POPs. The POPs used in Figures 2A, 3A, and 4A contain a GFP tag which may change the inhibitory effect of POPs on ASC filament formation. Experiments with untagged POPs are therefore required to validate the results.

We added new results monitoring the oligomerization of mCherry-tagged PYDs, ASCFL, and AIM2FL in the presence of untagged POPs, which remain consistent with our existing results (Figure supplements following Figures 2-4).

2. The authors take the reduction of PYD filamentation as an indication of inhibition, but it is not clear how they ruled out the possibility that POP1 co-assembles into the ASCPYD filaments and inhibits inflammasome formation by repressing the recruitment of Caspase-1 (as POP1 lacks the CARD the effector domain). Thus, some functional assays measuring downstream Caspase-1 activation are required. In addition, the possibility that POP1 and ASC co-assemble could be tested directly with FRET experiments in which one protein is the donor and the other is the acceptor. Without such experiments, the statement in the Discussion "Our investigations here reveal that POPs interfere with the polymerization (nucleation and/or elongation) of various inflammasome filaments without co-assembling, … " does not appear to be justified.

We included data showing that POP1 is least effective in inhibiting the NLRP3/nigericin-mediated release of mature interleukin-18 (Figure 4—figure supplement 2E; i.e., a functional assay requested by the reviewer). Here, we also find that POP2 is most effective, likely by directly suppressing ASC polymerization. Our observation here is also consistent with the report by Ratsimandresy et al. reporting that POP2 is more effective in suppressing ASC/inflammasomes than POP1.

We also generated fluor-labeled recombinant POP1 and confirmed that there are no discernable FRET signals between donor-labeled POP1 and acceptor-labeled ASCPYD, again supporting that POP1 and ASCPYD do not intermix (Figure 2, Figure Supplement 2C).

3. Further computational analysis should be performed to determine if the theory that a combination of favorable and unfavorable interactions is generally applicable. Does this theory account for other PYD/PYD interactions and CARD/CARD interactions? For example, for the AIM2PYD/ASCPYD interface, do they see only a favorable interface or a mixture? How about two unrelated PYDs, such as between AIM2PYD and NLRP3PYD? How about for COPs? How about the homotypic interfaces between the POPs themselves?

We are afraid that the reviewers are asking for future orthogonal studies significantly beyond the scope of the current work. This present manuscript focuses on PYD•POP interactions by studying more than a dozen pairwise interactions using three different methods (in silico, recombinant proteins, in cell imaging). We strongly believe that any additional in silico predictions need to be validated and (re)interpreted in light of biochemical experiments. Thus, investigating different systems such as CARD•CARD and NLRP•ASC interactions with our rigorous approach would require years’ worth of additional work, especially if they have unique patterns (rules) for themselves. Also of note, there are reports on the co-activation of different inflammasome receptors (e.g., Han et al., Sci. Immun., 2021). We do not yet know whether they directly interact with/regulate one another via PYDs or only communicate with ASC. We refrain from suggesting any interactions or lack thereof without biological/biochemical studies (i.e., beyond the scope of the present manuscript).

In our recent work (Matyzewski et al., Nat Com, 2021), we did not see any strong unfavorable energy scores between AIM2 and ASC, and our in silico approaches helped in identifying the directional interaction between AIM2PYD and ASCPYD filaments. Of note, we do not treat Rosetta energy scores as absolute free energy terms (as they are not), we use them as relative-yet-quantitative measures to guide our investigations. Importantly, our conclusion then was strictly based on our subjects at hand, as is here (any potential differences found in investigating NLRP•ASC interactions would not validate or invalidate our working model for AIM2•ASC).

We ran Rosetta interface energy analyses on hypothetical homotypic interactions for each POP. Although we refrain from making any claims without conducting extensive biochemical studies, it appears that POP1 lacks the symmetric landscape for “top” and “bottom” halves seen from PYD filaments, which like reflect the lack of filament formation.

On the other hand, POP2 and POP3 show significantly unfavorable energy scores vs. PYD filaments such as AIM2PYD and ASCPYD. Of note, we do not yet understand how POP2 and POP3 form oligomers. For instance, we do not know whether they oligomerize via different interfaces than those mediate filament assembly in PYDs (Type 1-3). We believe that delineating how POP2 and POP3 form oligomers is beyond the scope of our current manuscript and requires extensive combinations of in silico and biochemical experiments.

Author response image 1

We believe our work here opens a door for testing to what extent our findings reported here are applicable to other death-domain (DD) proteins (or other filamentous assemblies).We look forward to such future studies to compare and contrast, improve, and even modify our understanding of how DD proteins interact with one another. To further clarify our stance, we added the following sentence at the end of the Discussion:

“Future investigations using molecular dynamics simulations and extensive mutagenesis will further delineate the complexity of oligomerization mechanisms and target specificities of POPs in more detail. It will be also interesting to see to what extent our findings for POP•PYD interactions can be applied to other DD family proteins such as COPs and CARDs. Overall, our multi-disciplinary approach provides an example of how to use in silico predictions judiciously for investigating multipartite protein-protein interactions.”

In addition, as raised by reviewer 3, the authors do not consider at all in this manuscript that there is a seventh interface described for PYDs besides the hexagonal assembly in the filament. This is the homodimer interface seen, e.g., for NLRP3, in the crystal structure of the PYD and in size exclusion chromatography (Bae and Park, 2011; 3QF2). For completeness, the energy scores should be also calculated for this interface. It might well be that POPs associate in this binding mode as heterodimeric assemblies to monomeric PYDs of NLRs or ALRs to regulate their activities. This would be somehow reminiscent of profilin binding to actin, regulating the pool of free G-actin for filament assembly.

In the report by Bae and Park, a minor population of the NLRP3PYD dimer was observed in the SEC/MALS analysis using very high protein concentrations under an acidic buffer system without any reducing agents (~600-800 µM protein at pH 5.0; Bae and Park, JBC, 2011). Considering the conditions, we are afraid that such dimer formation is unlikely physiologically important. Indeed, we regret that we do not find any biological relevance for NLRP3 dimers in the literature (full-length or PYD), and even Bae and Park did not report any functional relevance of this dimer in their paper. Although this 2011 study revealed the structure of NLRP3PYD monomer, it predates the groundbreaking discovery by Hao Wu, Ed Egelman, and colleagues showing that inflammasomes form filaments (Lu et al., Cell, 2014).

Nevertheless, we generated dimer models of PYDs and POP•PYDs based on the NLRP3PYD dimer crystal structure and conducted Rosetta interface analyses. We regret that we do not see any compelling signs indicating that such hypothetical interactions would play a major role in the target specificity of POPs:

Author response table 1
Rosetta interface energy scores between PYD•PYD and PYD•POP interaction on the putative dimer seen from the crystal structure of NLRP3PYD.
SelfPOP1POP2POP3
ASCPYD-6.6-5.6-5.2-4.4
AIM2PYD-4.02.11.30.8
IFI16PYD46.6-5.62.610.8
NLRP3PYD-17.2-8.6-3.7-15.1
NLRP6PYD-6.60.1-5.6-3.3

Also importantly, the putative dimer interface is part of the filament interface we have already included in our analyses (Author response image 2; type 3a/1b). We thus respectfully disagree that this is a uniquely important seventh interface.

Author response image 2
The NLRP3PYD filament.

The colored region is thought to mediate dimerization.

4. To more directly test the mixed-interaction model, the authors should use their Rosetta structural predictions as a guide to introduce mutations into the various POP1/ASC-PYD interfaces to reduce the binding energies of those specific interactions and then test whether the introduction of a single or multiple, weak interactions then allows POP1 to restrict ASC-PYD oligomerization. To further elucidate their mixed interface model, the authors should also address whether the weak interactions need to be on the same 'half' of the interface, e.g., does weakening the 1b and 2b interfaces lead to better disruption of PYD filamentation than a 1b/1a combination mutant?

We considered such mutagenesis approaches early on but decided against them. For example, one can transplant POP2 residues on POP1, and vice versa. However, this would simply make POP1 more like POP2 and vice versa, which, in our view, do not provide significant new insights.

Instead, we are developing a new approach of combining the Monte Carlo simulation with Rosetta to further delineate how POPs and PYDs interact (this project has been inspired by our prior work reported in Matyzewski et al., PNAS, 2018). Briefly, our approach will test the probability of assembling the filament base (and inhibition of) by various PYD-POP pairs with different number of favorable and unfavorable interfaces (we will also give different weights to each interface and introduce in silico mutations). As with the current manuscript and our prior studies, any in silico predictions will be tested using biochemical methods employing recombinant proteins and cellular assays, which will take more than a year to complete. This follow-up study will help in further improving our understanding (model) of how POPs regulate PYDs. Moreover, it could also allow us to generate “designer” POPs that can target a wide variety of PYDs. We look forward to reporting our findings in the future.

We would also like to stress that our manuscript presents a comprehensive approach for investigating the interaction between POPs and PYDs. Moreover, the present manuscript marks only the beginning of our long-term goal of elucidating the “interaction codes” that underpin the specificity and interaction mechanisms of DD proteins.

Reviewer #3 (Recommendations for the authors):

It seems to this reviewer that the authors have written the manuscript in chronological order as they have performed the experiments, which might, however, not be the best way to present their data. I propose to rewrite and reorganize the manuscript, to better make the points. The Discussion is a bit repetitive and seems overwritten. Some shortening, despite adding new ideas and considerations (see below), might be reasonable.

Thanks for the suggestion. We shortened the Discussion.

Additionally, our apologies for the oversight, we improved the resolution of Figure 1—figure supplement 1.

This reviewer does not understand the scientific rationale regarding the PYD sequence identity and similarity of POPS to either NLRs or ALRs as a measure of their potential regulatory/inhibitory function in the respective inflammasome formation.

We do not intend to support or refute the rationale behind this previously proposed notion by Devi et al., Indramohan et al., de Almeida et al., and others (cited throughout the manuscript); however, our observations here indicate it’s indeed more complex.

[Editors’ note: what follows is the authors’ response to the second round of review.]

1. You need to explicitly recognize/discuss/acknowledge that NLRs and ALRs contain additional domains beyond the PYD that can modify the energy landscape to create oligomers, and it is vital to consider this aspect as a caveat to your study which focuses on dissociated PYD interactions.

We added such a disclaimer at the end of the paper.

2. The consensus of the reviewers remains that you need to test your model by employing site-directed mutagenesis. This was a clear requirement in the previous decision letter (point 4), and the reviewers were not convinced by your argument that this is not necessary. One reviewer suggests: "One could easily imagine replacing a favorable interaction with alanine to gently weaken the interaction or with glutamate or other charged residue to break the interface. By taking POP1 and introducing mutations to disrupt one or multiple surfaces the authors could show that these mutants do or do not inhibit filament formation ASC. Alternatively, the authors could introduce mutations into the ASC PYD and determine whether the mutant protein could now act like a POP and block WT ASC filament formation. While I appreciate that doing such a comparison for every POP and PYD in the paper would be excessive, I strongly believe that the authors need to experimentally test their model with at least some site-directed mutants".

We added new mutagenesis data at the end of the Results section. Briefly, we introduced mutations that hamper the self-assembly of AIM2PYD (i.e., unfavorable) and found that resulting mutant proteins can inhibit the polymerization of WT-AIM2PYD.

https://doi.org/10.7554/eLife.81918.sa2

Article and author information

Author details

  1. Shuai Wu

    Department of Biophysics and Biophysical Chemistry, Johns Hopkins University School of Medicine, Baltimore, United States
    Contribution
    Data curation, Formal analysis, Validation, Investigation, Methodology, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-1677-0221
  2. Archit Garg

    Department of Biophysics and Biophysical Chemistry, Johns Hopkins University School of Medicine, Baltimore, United States
    Contribution
    Data curation, Investigation, Writing – review and editing
    Contributed equally with
    Zachary Mazanek
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-5931-2522
  3. Zachary Mazanek

    Department of Biophysics and Biophysical Chemistry, Johns Hopkins University School of Medicine, Baltimore, United States
    Present address
    Georgiamune, Gaithersburg, United States
    Contribution
    Data curation, Validation, Methodology
    Contributed equally with
    Archit Garg
    Competing interests
    Affiliated with Georgiamune. The author has no financial interests to declare
  4. Gretchen Belotte

    Department of Biophysics and Biophysical Chemistry, Johns Hopkins University School of Medicine, Baltimore, United States
    Contribution
    Data curation, Formal analysis, Investigation
    Competing interests
    No competing interests declared
  5. Jeffery J Zhou

    Department of Biophysics and Biophysical Chemistry, Johns Hopkins University School of Medicine, Baltimore, United States
    Contribution
    Data curation, Formal analysis, Investigation
    Competing interests
    No competing interests declared
  6. Christina M Stallings

    Department of Biophysics and Biophysical Chemistry, Johns Hopkins University School of Medicine, Baltimore, United States
    Contribution
    Data curation, Investigation, Writing – review and editing
    Competing interests
    No competing interests declared
  7. Jacob Lueck

    Department of Biophysics and Biophysical Chemistry, Johns Hopkins University School of Medicine, Baltimore, United States
    Contribution
    Data curation, Investigation, Writing – review and editing
    Competing interests
    No competing interests declared
  8. Aubrey Roland

    Division of Infectious Diseases, Johns Hopkins University School of Medicine, Baltimore, United States
    Contribution
    Investigation
    Competing interests
    No competing interests declared
  9. Michael A Chattergoon

    Division of Infectious Diseases, Johns Hopkins University School of Medicine, Baltimore, United States
    Present address
    Vir Biotechnology, San Francisco, United States
    Contribution
    Formal analysis, Project administration
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-2220-1116
  10. Jungsan Sohn

    1. Department of Biophysics and Biophysical Chemistry, Johns Hopkins University School of Medicine, Baltimore, United States
    2. Division of Rheumatology, Johns Hopkins University School of Medicine, Baltimore, United States
    3. Department of Oncology, Johns Hopkins University School of Medicine, Baltimore, United States
    Contribution
    Conceptualization, Supervision, Funding acquisition, Investigation, Writing – original draft, Writing – review and editing
    For correspondence
    jsohn@jhmi.edu
    Competing interests
    Reviewing editor, eLife
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-9570-2544

Funding

National Institute of General Medical Sciences (R01GM129342)

  • Shuai Wu

National Institute of General Medical Sciences (R35GM145363)

  • Shuai Wu

National Science Foundation (MCB1845003)

  • Archit Garg

National Institute of Allergy and Infectious Diseases (K08AI102696)

  • Aubrey Roland

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank Dr. Mariusz Matyzewski and Naveen Mohideen for assistance in Rosetta experiments. Fundings from NIH R01GM129342, R35GM145363, NSF MCB1845003 awards to JS NIH K08 AI102696 to MC are acknowledged. Computational Resources were provided by Maryland Advanced Research Computing Center at Johns Hopkins University.

Senior Editor

  1. Carla V Rothlin, Yale University, United States

Reviewing Editor

  1. Russell E Vance, University of California, Berkeley, United States

Version history

  1. Received: July 15, 2022
  2. Preprint posted: August 2, 2022 (view preprint)
  3. Accepted: January 2, 2024
  4. Version of Record published: January 22, 2024 (version 1)

Copyright

© 2024, Wu et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 249
    Page views
  • 45
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Shuai Wu
  2. Archit Garg
  3. Zachary Mazanek
  4. Gretchen Belotte
  5. Jeffery J Zhou
  6. Christina M Stallings
  7. Jacob Lueck
  8. Aubrey Roland
  9. Michael A Chattergoon
  10. Jungsan Sohn
(2024)
Design principles for inflammasome inhibition by pyrin-only-proteins
eLife 13:e81918.
https://doi.org/10.7554/eLife.81918

Share this article

https://doi.org/10.7554/eLife.81918

Further reading

    1. Biochemistry and Chemical Biology
    2. Plant Biology
    Pradeep Kumar, Ankit Roy ... Rajan Sankaranarayanan
    Research Article

    Aldehydes, being an integral part of carbon metabolism, energy generation, and signalling pathways, are ingrained in plant physiology. Land plants have developed intricate metabolic pathways which involve production of reactive aldehydes and its detoxification to survive harsh terrestrial environments. Here, we show that physiologically produced aldehydes, i.e., formaldehyde and methylglyoxal in addition to acetaldehyde, generate adducts with aminoacyl-tRNAs, a substrate for protein synthesis. Plants are unique in possessing two distinct chiral proofreading systems, D-aminoacyl-tRNA deacylase1 (DTD1) and DTD2, of bacterial and archaeal origins, respectively. Extensive biochemical analysis revealed that only archaeal DTD2 can remove the stable D-aminoacyl adducts on tRNA thereby shielding archaea and plants from these system-generated aldehydes. Using Arabidopsis as a model system, we have shown that the loss of DTD2 gene renders plants susceptible to these toxic aldehydes as they generate stable alkyl modification on D-aminoacyl-tRNAs, which are recycled only by DTD2. Bioinformatic analysis identifies the expansion of aldehyde metabolising repertoire in land plant ancestors which strongly correlates with the recruitment of archaeal DTD2. Finally, we demonstrate that the overexpression of DTD2 offers better protection against aldehydes than in wild type Arabidopsis highlighting its role as a multi-aldehyde detoxifier that can be explored as a transgenic crop development strategy.

    1. Biochemistry and Chemical Biology
    2. Structural Biology and Molecular Biophysics
    Greg J Dodge, Alyssa J Anderson ... Barbara Imperiali
    Research Article

    Bacterial cell surface glycoconjugates are critical for cell survival and for interactions between bacteria and their hosts. Consequently, the pathways responsible for their biosynthesis have untapped potential as therapeutic targets. The localization of many glycoconjugate biosynthesis enzymes to the membrane represents a significant challenge for expressing, purifying, and characterizing these enzymes. Here, we leverage cutting-edge detergent-free methods to stabilize, purify, and structurally characterize WbaP, a phosphoglycosyl transferase (PGT) from the Salmonella enterica (LT2) O-antigen biosynthesis. From a functional perspective, these studies establish WbaP as a homodimer, reveal the structural elements responsible for dimerization, shed light on the regulatory role of a domain of unknown function embedded within WbaP, and identify conserved structural motifs between PGTs and functionally unrelated UDP-sugar dehydratases. From a technological perspective, the strategy developed here is generalizable and provides a toolkit for studying other classes of small membrane proteins embedded in liponanoparticles beyond PGTs.