Crystal Structure of the Human NLRP9 Pyrin Domain Suggests a Distinct Mode of Inﬂammasome Assembly Michael Marleaux1, Kanchan Anand1, Eicke Latz2 and Matthias Geyer1

Crystal structure of the human NLRP9 pyrin domain suggests a distinct mode of inﬂammasome assembly Michael Marleaux1, Kanchan Anand1, Eicke Latz2 and Matthias Geyer1

1 Institute of Structural Biology, University of Bonn, Bonn, Germany 2 Institute of Innate Immunity, University of Bonn, Bonn, Germany

Correspondence Inflammasomes are cytosolic multimeric signaling complexes of the innate M. Geyer, Institute of Structural Biology, immune system that induce activation of caspases. The NOD-like receptor Biomedical Center (BMZ), University of NLRP9 recruits the adaptor protein ASC to form an ASC-dependent inflam- Bonn, Venusberg Campus 1, 53127 Bonn, masome to limit rotaviral replication in intestinal epithelial cells, but only lit- Germany Tel: +49 228 287 51400 tle is known about the molecular mechanisms regulating and driving its E-mail: [email protected] assembly. Here, we present the crystal structure of the human NLRP9 pyrin domain (PYD). We show that NLRP9PYD is not able to self-polymerize nor (Received 19 April 2020, revised 17 May to nucleate ASC specks in HEK293T cells. A comparison with filament-form- 2020, accepted 1 June 2020, available ing PYDs revealed that NLRP9PYD adopts a conformation compatible with online 6 July 2020) filament formation, but several charge inversions of interfacing residues might cause repulsive effects that prohibit self-oligomerization. These results pro- doi:10.1002/1873-3468.13865 pose that inflammasome assembly of NLRP9 might differ largely from what Edited by Roosmarijn Vandenbroucke we know of other inflammasomes.

Keywords: ASC specks; ﬁlament; inﬂammasomes; NLRP9; PYD; structure

The innate immune system comprises germline-en- amplify signals from the PRR via an adaptor protein coded pattern recognition receptors (PRRs) that have to activate the downstream effector cysteine protease evolved to directly recognize a wide range of conserved caspase-1, which in turn processes the highly proin- pathogen-associated molecular patterns, danger-associ- flammatory cytokines pro-IL-1b and pro-IL-18 into ated molecular patterns, and homeostasis-altering their mature forms [4,5]. Active caspase-1 also cleaves molecular processes [1]. They fulfill diverse functions gasdermin D to induce pore formation in the host cell with importance not only in pathogen recognition but for the release of cytokines and pyroptotic cell death also in tissue homeostasis, apoptosis, graft-versus-host [5,6]. Notably, signal amplification is facilitated by uni- disease, early development, and even negative regula- fied nucleation-driven polymerization of the individual tion of inflammatory responses [2]. In the human inflammasome components [7–9]. As an example, immune system, PRRs group into five families, includ- NLRP3 is an inflammasomal sensor protein that ing Toll-like receptors, RIG-I-like receptors, C-type belongs to a subfamily of NLRs (termed NLRPs) con- lectin receptors, AIM2-like receptors (ALRs), and taining an N-terminal pyrin domain (PYD). After an NOD-like receptors (NLRs). Of these PRRs, ALRs initial priming step and activation by a wide range of and certain NLRs were found to oligomerize upon diverse stimuli, it recruits the bipartite protein ‘apopto- ligand stimulation and participate in the formation of sis-associated speck-like protein containing a caspase cytosolic multimeric signaling complexes, called inflam- recruitment domain’ (ASC) as an adaptor for caspase- masomes [3]. Canonical inflammasomes translate and 1 activation [8,10–12]. NLRP3 polymerizes with its

Abbreviations GST, glutathione S-transferase; PYD, pyrin domain.

FEBS Letters 594 (2020) 2383–2395 ª 2020 The Authors. FEBS Letters published by John Wiley & Sons Ltd 2383 on behalf of Federation of European Biochemical Societies This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modiﬁcations or adaptations are made. PYD structure of NLRP9 M. Marleaux et al.

PYD to form the nucleation seed for ASC filament Materials and methods assembly using three major types of asymmetric interfaces, which has also been described for NLRP6 and Cloning, expression, and purification of the AIM2 [7,13–15]. ASC filaments in turn nucleate the human NLRP9 pyrin domain formation of large caspase-1 filaments by caspase activation and recruitment domain (CARD)/CARD inter- The coding sequence of a human NLRP9 cDNA clone (Gen- Bank accession NM_176820) corresponding to residues 1–97 actions, which finally leads to proximity-induced was amplified by PCR with specific oligonucleotides encom- autoproteolytic activation of caspase-1 [7]. Therefore, passing NcoI/EcoRI restriction sites at the 50 and 30 end, respec- inflammasome assembly significantly relies on homo- tively. The DNA product was cloned into a modified pGEX- typic interactions between PYD and CARD domains 4T-1 vector carrying an N-terminal glutathione S-transferase of the respective components. (GST) tag for affinity purification followed by a tobacco etch A less studied member of the NLRP subfamily is virus (TEV) protease cleavage site. The expression construct NLRP9, which was first described as a maternal was confirmed by DNA sequencing and transformed into – effect gene in bovine [16 19]. Upon recognition of Escherichia coli BL21(DE3) cells. Expression cultures were ini- low molecular weight dsRNA by the endogenous tially grown at 37 °C in LB medium, supplemented with 1 RNA helicase DHX9, its mouse analogue NLRP9b 100 mgL ampicillin. When OD600 reached 0.8, cells were was described to form an inflammasome with ASC cooled to 16 °C before protein expression was induced by addi- and caspase-1 [20,21]. This inflammasomal complex tion of 0.55 mM IPTG. Bacteria were further grown overnight mediates pyroptosis of intestinal epithelial cells at 16 °C, harvested by centrifugation, and stored at 80 °C. (IECs), thereby limiting rotaviral replication in mice For purification, cells were resuspended in buffer A [20,21]. Additionally, NLRP9b deficiency resulted in (50 mM HEPES pH 7.5, 150 mM NaCl, and 10 mM bME) less pulmonary histopathological alterations in a containing 1 mM phenylmethylsulfonyl fluoride and mouse model of acute lung injury and markedly ele- 1mgL 1 DNAse I, and lysed by the addition of lysozyme vated proinflammatory cytokine and chemokine levels, and further sonication. The lysate was cleared from cell deb- as well as ROS production, in a LPS-activated mouse ris by centrifugation at 70 000 g for 30 min. The protein epithelial cell line [22]. Also, expression of NLRP9 in containing an N-terminal GST tag was captured using a human IECs and DHX9-dependent association with GSTrap column (GE Healthcare, Freiburg, Germany), and ASC upon stimulation with rotavirus could be shown the unbound fraction was washed off with buffer B (50 mM b [21]. Moreover, upregulation of expression was found HEPES pH 7.5, 300 mM NaCl, 10 mM ME), followed by upon inflammatory stimulus ROS and interferon-c in buffer A. After elution with buffer A containing 10 mM reduced glutathione, the GST tag was cleaved off overnight brain pericytes and cerebral endothelial cells, respec- at 4 °C by the addition of 1 : 100 w/w TEV protease. Protein tively [23,24]. However, the role of human NLRP9 in was further purified using a Superdex 75 gel filtration col- innate immunity is not yet clarified and elucidation of umn 16/600 (GE Healthcare) that was connected to a its inflammasome function would be valuable. This GSTrap column for prolonged retention of the cleaved GST holds particularly true, as alterations in NLRP9 tag and pre-equilibrated with buffer C (50 mM HEPES pH expression were linked to several inflammatory dis- 7.5, 150 mM NaCl, 1 mM TCEP). Peak fractions containing eases, including urothelial carcinoma, multiple sclero- NLRP9PYD were pooled, concentrated to 31.7 mgmL1, sis, juvenile idiopathic arthritis, familial late-onset snap-frozen in liquid nitrogen, and stored at 80 °C. Alzheimer disease, and infection with Helicobacter pylori [25–31]. To gain first insights into NLRP9 inflammasome Crystallization and diffraction data collection assembly, we expressed and purified recombinant Protein solution was crystallized by the hanging-drop vapor- PYD human NLRP9 and studied its capability to self- diffusion method at 15 °C. Initial crystals were obtained by polymerize into filaments and nucleate ASC speck for- JCSG + screen [32] (0.2 M potassium formate, 20% PYD mation. We found that NLRP9 exists as a mono- PEG3350). The optimal hexagonal-shaped crystals were mer under near-physiological conditions in vitro and in grown with 1 : 1 ratio of protein and precipitant solution by PYD cells. In contrast to NLRP3 , overexpression of using 0.1 M Bicine pH 9.0, 0.2 M sodium thiocyanate, 0.1 M murine or human NLRP9PYD did not nucleate ASC sodium formate, and 25–27% v/w PEG3350. Crystals were speck formation in HEK293T cells. We determined the cryoprotected with 15% ethylene glycol in the mother liquor crystal structure of human NLRP9PYD to study the and flash-cooled in liquid nitrogen. molecular differences from filament-forming PYDs Data were collected at the synchrotron beamlines of and discuss its implications for the assembly of the Swiss Light Source (SLS) in Villigen, Switzerland, and NLRP9 inflammasome. Deutsches Elektronen-Synchrotron (DESY) in Hamburg,

2384 FEBS Letters 594 (2020) 2383–2395 ª 2020 The Authors. FEBS Letters published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies M. Marleaux et al. PYD structure of NLRP9

Germany. Presented diffraction data up to 1.95 A resolu- Table 1. Crystallographic data collection and refinement statistics. tion were collected from native crystals at 100K on the X- Data collection ray diffraction PX1 beamline using Eiger detector at SLS. Space group P6522 a, b, c (A) 33.33, 33.33, 311.09 Resolution (A)a 28.74–1.95 (2.02–1.95) Data processing, structure determination, and Rmerge (%) 8.25 (98.49) refinement I/r (I) 18.58 (1.84) Completeness (%) 99.52 (99.23) Data were processed and scaled using the XDS program CC1/2 0.998 (0.916) package [33]. The phase problem was solved by molecular Redundancy 22.3 (16.7) replacement method [34] using program PHASER [35] and the Refinement PYD coordinates of NLRP4 (4EWI) [36] as a search model for Resolution (A) 28.74–1.95 the phase solution. The model was refined in alternating No. of reflections 14 097 b cycles of refinement with PHENIX [37]. Manual rebuilding and Rwork/Rfree 0.215/0.222 visual comparisons were made using the graphical program No. of atoms COOT [38]. The stereochemical quality of the model was veri- Protein 772 fied using the Ramachandran plot. Molecular diagrams were Water 17 B drawn using the PyMOL molecular graphics suite [39]. The factors Protein 68.96 final model contains residues 9–97 and has been refined to Water 75.60 R and R values of 21.5% and 22.2%, respectively. work free Rmsd Details of the diffraction data collection, quality, and refine- Bond lengths (A) 0.006 ment statistics are given in Table 1. The atomic coordinates Bond angles (°) 0.74 and structure factor amplitudes have been deposited in the Residues in Ramachandran Protein Data Bank with accession code 6Z2G. Favored regions (%) 98.86 Allowed regions (%) 1.14 PDB Dynamic Light Scattering Accession number 6Z2G

1 The protein was diluted to 2 mgmL in buffer C and pre- aValues in parentheses are for the highest resolution shell.; b pared for dynamic light scattering (DLS) by centrifugation Rfree value is equivalent to the R value but is calculated for 5% of at 18 000 g for 10 min. The supernatant was incubated for the reflections chosen at random and omitted from the refinement 11.5 h at 25 °C, and the size distribution of the sample was process. measured every 30 min using a DynaPro NanoStar (Wyatt Technology, Dernbach, Germany) set to an acquisition time of 3 s. Each data point was averaged from 20 DLS were seeded into 96-well plates and grown for 48 h. Per well, acquisitions. The cumulant analysis and calculation of the 200 ng of plasmid DNA coding for the indicated fusion pro- hydrodynamic radius (rH) and molecular weight of the proteins or control was transfected with Lipofectamine 2000 tein were automatically performed by the software supplied according to the manufacturer’s instructions. Twenty-four with the instrument (DYNAMICS 7.8.3, Wyatt Technology). hours after transfection, cells were fixed and simultaneously stained for nuclei using 4% formaldehyde plus 10 µM Draq5 Analytical gel filtration diluted in PBS. Cells were imaged using a (ZEISS, Jena, Ger- many) Observer.Z1 epifluorescence microscope, equipped Fifty microgram of the protein that was prepared for DLS with 209 objective (dry, Plan Apochromat, NA 0.8), Axio- ° was loaded before and after incubation for 12 h at 25 C cam 506 mono, and software (ZEN 2.3 PRO, ZEISS). For visu- onto a 1260 Infinity II LC system connected to a Superdex alization, brightness and contrast were adjusted. Six images 75 analytical gel filtration column 3.2/300 (GE Healthcare) per experiment and condition were analyzed for filament or that was pre-equilibrated with buffer C. Gel filtration stan- ASC speck formation using program CELLPROFILER [40].In dard (Bio-Rad) was used according to the manufacturer’s total, 3338, 938, 2398, and 2794 HEK293T cells and 4549, instructions to calculate a standard curve and determine 1628, 3593, and 5350 HEK293TASC-BFP cells positive for the molecular weight of the protein species. expression of either hNLRP3PYD-mCitrine, hNLRP9PYD- mCitrine, mNLRP9bPYD-mCitrine, or the mCitrine vector Analysis of pyrin domain filament and ASC speck control, respectively, were analyzed for filament and ASC speck formation. Plotting of graphs and calculation of formation in cells statistics were performed using (GRAPHPAD PRISM version HEK293T cells or HEK293T cells stably expressing ASC- 7.0e, GraphPad Software, San Diego, CA, USA) for TagBFP under the CMV promoter (HEK293TASC-BFP) Mac.

FEBS Letters 594 (2020) 2383–2395 ª 2020 The Authors. FEBS Letters published by John Wiley & Sons Ltd 2385 on behalf of Federation of European Biochemical Societies PYD structure of NLRP9 M. Marleaux et al.

Results of a1, the C terminus of a6, and within the a2-a3 loop (Fig. 1D). In contrast, the overall fold of NLRP9PYD Structure of the NLRP9 pyrin domain is substantially stabilized by a central hydrophobic core with lower flexibility, which is formed by residues NLRP9 consists of an N-terminal PYD and a central Leu15 of a1, Leu18 in the a1-a2 loop, Phe26 and NACHT domain similar to all NLRP proteins. The Leu30 of a2, Ile42 in the a2-a3 loop, Leu47 of a3, NACHT domain is directly followed by a C-terminal Ala50 in the a3-a4 loop, Val55 and Leu59 of a4, leucine-rich repeat (LRR) domain that can be divided Leu75 and Phe76 of a5, Ile79 in the a5-a6 loop, and into a transition and a canonical LRR (Fig. 1A). Leu84 of a6 (Fig. 1D,E). Hydrophobicity of the While the interplay of all different domains might be majority of those residues positioned in the central important for active inflammasome formation, PYDs core of NLRP9PYD was found to be conserved among of NLRP3, NLRP6, and AIM2 were found to form other PYDs (Fig. 3A), confirming their importance for filaments and connect the inflammasome sensor with domain stability. Similarly, the a2-a3 loop is stabilized the adaptor protein ASC for caspase recruitment and by a second hydrophobic, but more flexible, cluster activation [13–15]. To study the inflammasome assem- PYD formed by residues Leu34, Phe37, and Leu39 in the bly of NLRP9, we expressed human NLRP9 as a a2-a3 loop, Tyr63 in the a4-a5 loop, and Val71 of a5 GST-fusion protein in E. coli. The protein was purified (Fig. 1E). The determined distribution of regions with to homogeneity by affinity chromatography followed lower and higher flexibility is equivalent for PYDs of by tag cleavage and size-exclusion chromatography at – PYD other NLRPs [42 44]. However, the overall flexibility near-physiological buffer conditions. NLRP9 was was found to be significantly higher in NLRP9PYD, 1 readily concentrated up to 120 mg mL without pre- which prompted our further investigations. cipitation or aggregation. Consequently, we crystallized NLRP9PYD for investigation of its structural PYD features. Hexagonal-shaped crystals appeared in 0.1 M NLRP9 is monomeric and does not self- Bicine pH 9.0, 0.2 M sodium thiocyanate, 0.1 M polymerize in vitro sodium formate, and 25–27% PEG3350 and diffracted Since PYDs of NLRP3, NLRP6, and AIM2 are able up to 1.95 A resolution using the PX1 beamline and to form filaments [13–15], we asked whether Eiger detector at SLS. Initial phases were obtained by NLRP9PYD would also feature such ability. Our initial molecular replacement using the coordinates of the PYD investigations by electron microscopy revealed that the NLRP4 crystal structure [36] with 59.6% sequence protein did not tend to form filamentous structures identity as search model. The protein structure was under the tested conditions in vitro. Therefore, we set refined to Rwork of 21.5% and Rfree of 22.2% with out to analyze its size distribution and experimental excellent stereochemistry (Table 1). In the protein crys- PYD molecular weight during a 12-h incubation period at tal, one monomer of NLRP9 formed the asymmet- 25 °C using DLS and analytical gel filtration (Fig. 2). = ric unit (unit cell dimensions a 33.33 A, We found that during incubation, NLRP9PYD particles = = b 33.33 A, and c 311.09 A (Table 1)). In the final have a stable average hydrodynamic radius of 1.7 nm, PYD – model of NLRP9 , residues 9 97 could be built into corresponding to a molecular weight of approximately the electron density map (Fig. 1B). The N terminus 12 kDa (Fig. 2A). Consistently, analytical gel filtration region of the molecule that contributes to about 10% revealed an average molecular weight of 11.93 kDa for of the structure and several sidechains facing toward NLRP9PYD independent of the incubation time solvent channels are disordered and could not be (Fig. 2B,C). Since the theoretical molecular weight of resolved by electron density, leading to a higher over- PYD 2 monomeric NLRP9 is 11.96 kDa, we concluded all temperature factor of 68.96 A (Table 1, Fig. 1B). that NLRP9PYD is monomeric in solution and does As other members of the death domain superfamily, not self-polymerize in vitro. NLRP9PYD displayed six antiparallel helices, connected by five loops and arranged in the typical death Comparison of NLRP9PYD with filament-forming domain fold [41] (Fig. 1B). The helices are formed by pyrin domains residues 10–15 (a1), 20–31 (a2), 44–47 (a3), 52–62 (a4), 66–78 (a5), and 82–96 (a6). Of note, a3 is a short Using our structural data, we next tried to understand PYD PYD 310-helix. NLRP9 contains polar residues at the the inability of NLRP9 to form filaments on a surface, which were found to be of mainly basic char- molecular basis. To compare residues at positions that acter (Fig. 1C). Residues with highest flexibility of are typically involved in filament formation, we per- backbone and sidechains were found at the N terminus formed structure-based sequence alignments of the

2386 FEBS Letters 594 (2020) 2383–2395 ª 2020 The Authors. FEBS Letters published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies M. Marleaux et al. PYD structure of NLRP9

Fig. 1. Crystal structure of the human NLRP9 PYD. (A) Cartoon of the human NLRP9 domain architecture. (B) Cartoon representation of the NLRP9PYD crystal structure with six antiparallel helices and long a2-a3 loop, surrounded by detailed views of the respective residues ﬁtted PYD into the electron density map (2Fo-Fc map contoured at 1.0 sigma). (C) Color-coded electrostatic surface representation of NLRP9 PYD generated with APBS [49]. Unit of the scale is kBT/ec. (D) Plot of individual B factors for each residue of NLRP9 . Values for backbone (blue) and sidechains (red) were calculated separately. Sidechains without occupancy were excluded. Secondary structure elements of NLRP9PYD are plotted above the corresponding residue numbers. (E) Bottom view of NLRP9PYD (magniﬁed and rotated by 90° about the x- axis relative to B). Hydrophobic core residues stabilizing the six-helical fold in NLRP9PYD are shown as pink sticks. A second hydrophobic cluster stabilizing the a2-a3 loop is shown as green sticks.

PYD of NLRP9 and the ﬁlament-forming PYDs of in such interfaces and matching by charge or hydropho- NLRP3, NLRP6, AIM2, and ASC (Fig. 3A). Assem- bicity in NLRP3, NLRP6, AIM2, or ASC PYDs might bly of PYD ﬁlaments is dependent on the formation of be mismatching in NLRP9PYD. Indeed, we found sev- three asymmetric interfaces, of which the type I inter- eral charge inversions in the proposed interfaces of face mediates intrastrand interactions and type II and NLRP9PYD compared with the other four PYDs, such type III interfaces mediate interstrand interactions as Asp8, Lys16, Arg19, Lys20, Glu46, Lys48, Lys57, [7,13,15] (Fig. 3B). We supposed that residues located and Lys61 as well as charge insertions of otherwise

FEBS Letters 594 (2020) 2383–2395 ª 2020 The Authors. FEBS Letters published by John Wiley & Sons Ltd 2387 on behalf of Federation of European Biochemical Societies PYD structure of NLRP9 M. Marleaux et al.

hydrophobic residues, such as Glu28 and Glu53. Addi- tionally, Phe9 and Trp24 were found as large hydrophobic insertions (Fig. 3A). Although the presumed mismatches could compensate each other in the interfaces, we speculate that some of them might disrupt NLRP9PYD filament formation. Of note, NLRP3PYD, NLRP6PYD, AIM2PYD, and ASCPYD displayed sequence identities of 28.7%, 26.4%, 25.0%, and 21.8% with NLRP9PYD, respectively (residues 8-94; Fig. 3A). Given that PYDs can undergo large conformational changes upon formation of the filament [7,13,15], also residues outside of binding interfaces have to be consid- ered for further analysis, since they might limit NLRP9PYD to conformations incompatible with filament formation. In PYDs of NLRP6 and ASC, the position and rotation of the flexible a2-a3 loop together with its connecting a2- and a3-helices were found to undergo most prominent changes on filament formation [7,15]. Therefore, we analyzed those regions in the NLRP9PYD structure and asked whether they would be able to adopt the filament conformation without steric hindrance. For this purpose, we overlaid NLRP9PYD with monomeric and filamentous NLRP6PYD (Fig. 3C). Unexpectedly, we found NLRP9PYD to superimpose better with the structure of NLRP6PYD from the filament (RMSD of 1.2 A) as if compared with monomeric NLRP6PYD (RMSD of 1.6 A). Indeed, especially the conformation of the a2-a3 loop region of NLRP9PYD is highly similar to the filamentous NLRP6PYD (Fig. 3C). To further validate this finding, we aligned NLRP9PYD with the less related monomeric and filamentous AIM2PYD (Fig. 3D) and ASCPYD (Fig. 3E). They superimposed to RMSD values of 1.1 and 1.7 A for AIM2PYD, and to 1.3 and 1.5 A for ASCPYD, respectively. However, intuitively the appearance of the a2-a3 loop region of NLRP9PYD seems to resemble the conformations of filamentous AIM2PYD and ASCPYD. Taken together, we assume that monomeric NLRP9PYD already exists in a conformation compatible with fila- PYD Fig. 2. Molecular dispersion of recombinant NLRP9PYD. (A) ment formation. Since NLRP3 showed the highest PYD Hydrodynamic radius and corresponding molecular weight of sequence identity with NLRP9 , we also compared NLRP9PYD as determined by DLS. Sample was measured every the conformations of both monomers (Fig. 3F). Inter- 30 min during 11.5 h of incubation at 25 °C. Data points are estingly, they superimposed to RMSD of 1.0 A and the representative of three independent experiments. n = 3 SEM. structure of NLRP3PYD closely resembled NLRP9PYD. (B) Representative elution profile from analytical size-exclusion Therefore, NLRP9PYD and NLRP3PYD are not only chromatography of NLRP9PYD before and after 12 h of incubation related by sequence but also related by conformation. time at 25 °C. The void volume (v0) and total volume (vt) of the column are depicted as arrows. (C) Calculation of the molecular PYD weight of NLRP9 from the peak retention volume obtained by NLRP9PYD does not self-polymerize nor nucleate analytical size-exclusion chromatography. The calibration curve was ASC speck formation in HEK293T cells plotted using the partition coefficient (Kav) versus the logarithm of the molecular weight of a standard. Data points are representative As our structural analysis did not reveal any character- n = of three independent experiments. 3 SEM. istic of NLRP9PYD conclusively explaining its inability

2388 FEBS Letters 594 (2020) 2383–2395 ª 2020 The Authors. FEBS Letters published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies M. Marleaux et al. PYD structure of NLRP9

Fig. 3. Structural comparison of NLRP9PYD with filament-forming PYDs. (A) Structure-based sequence alignment of the PYD of NLRP9 and the filament-forming PYDs of NLRP3, NLRP6, AIM2, and ASC. Asp8 was included in the alignment, since density of the mainchain could be seen at lower sigma level. For NLRP6, AIM2, and ASC, the filament structures are known and the residues forming the asymmetric interfaces are highlighted with the indicated colors. Residues that form the hydrophobic core of NLRP9PYD are shown in yellow boxes. Secondary structure elements of NLRP9PYD are plotted above the corresponding sequence. (B) Chart of three flanking subunits in a typical PYD filament. They form unified type I, type II, and type III interfaces. Subunits are labeled light blue, light green, and light magenta. (C) Overlay of the NLRP9PYD crystal structure with the MBP-NLRP6PYD crystal structure (6NDJ, chain A) and the NLRP6PYD filament structure (6NCV). (D) Overlay of NLRP9PYD with the MBP-AIM2PYD crystal structure (3VD8) and the GFP-AIM2PYD filament structure (6MB2). (E) Overlay of NLRP9PYD with the monomeric ASCPYD structure (1UCP) and the ASCPYD filament structure (3J63). (F) Structural comparison of the NLRP9PYD crystal structure with the NLRP3PYD crystal structure (3QF2, chain A).

FEBS Letters 594 (2020) 2383–2395 ª 2020 The Authors. FEBS Letters published by John Wiley & Sons Ltd 2389 on behalf of Federation of European Biochemical Societies PYD structure of NLRP9 M. Marleaux et al. to form filaments in vitro, we set out to validate its To further investigate the differences between behavior in the more physiological setting of NLRP3 and NLRP9 and their impacts on filament HEK293T cells by overexpressing human NLRP9PYD formation and nucleation of ASC specks, we modeled as mCitrine fusion protein. For comparison, we also one ring of the human NLRP3PYD filament and a overexpressed NLRP3PYD fusion protein as the closest hypothetical filament slice of human NLRP9PYD on relative forming filaments. Similar to what was demon- the basis of the NLRP6PYD filament structure [15] and strated by Stutz et al. [14], in average 58.5 4.7% of used these to calculate APBS [49]-generated surface cells overexpressing NLRP3PYD-mCitrine displayed electrostatics (Fig. 4D). Although it is hardly possible large filaments with ring-shaped morphology (Fig. 4A, to predict behavior from the electrostatic surfaces of B). In contrast, protein was found to diffusively locate proteins, it is apparent that many charged clusters in within cells overexpressing NLRP9PYD-mCitrine or NLRP3PYD are found to be inversed in NLRP9PYD mCitrine control under otherwise identical experimen- (Fig. 4D). As assumed previously, such charge inver- tal conditions (Fig. 4A). Significantly, no cells with fil- sions or insertions might disrupt intra- and interstrand aments could be detected, thereby supporting our interactions, since clusters of similar charge would in vitro findings (Figs 2 and 4B). Since NLRP9-depen- come close in a potential NLRP9PYD filament dent inflammasomes have been studied more exten- (Fig. 4D, NLRP9PYD top and bottom views). Addi- sively in the murine system [20–22], we also tested tionally, Lys48 and Lys49 of different subunits would filament formation for murine NLRP9bPYD fusion center at the inner core ring in the hypothetical protein overexpressed in HEK293T cells. Given its rel- NLRP9PYD filament (Fig. 4D), which might cause atively low sequence identity of 52.1% with human strong repulsive forces, that are potentially not over- NLRP9PYD (murine 1-91 versus human 1-94), we come by intra- and interstrand interactions and there- found it reasonable to anticipate murine NLRP9PYD fore would prohibit formation of a filament. Since we to potentially form filaments. Interestingly, we could do not know the exact molecular mechanism of how not detect any filaments when overexpressing murine nucleation of ASC specks is facilitated by PYD fila- NLRP9PYD in HEK293T cells, similarly as observed ments, hypotheses about NLRP9-dependent ASC for human NLRP9PYD (Fig. 4A,B). speck formation are highly speculative. But consider- In contrast, many ASC-dependent inflammasomes ing our model would resemble an effectively formed studied so far displayed filament formation of the pyrin NLRP9PYD filament, we were unable to find charge effector domain, [13–15] and for NLRP3, it was also matching surface patches at interfacing positions shown that mutations in its PYD abrogated not only fila- (Fig. 4D), thereby not excluding, for example, poten- ment formation but also nucleation of ASC specks [14]. tial hydrophobic interactions between NLRP9 and However, the ASC-dependent pyrin [45] or NLRP12 [46– ASC PYDs. However, those findings are of rather low 48] inflammasomes are examples where no hints for fila- evidence and further research revealing the molecular ment formation of their respective effector domains were mechanisms that effectively prevent filament formation found. We therefore investigated the ability of and nucleation of ASC specks would be needed for NLRP9PYD to nucleate ASC speck formation regardless definite conclusions. of filament formation. To this end, the fusion proteins were overexpressed in HEK293T cells stably expressing Discussion ASC-BFP (HEK293TASC-BFP). First, we analyzed whether the presence of ASC has any effects on filament At present, there is only limited knowledge about the formation and found no significant difference in the num- details of NLRP inflammasome assembly. Evidenced ber of cells with filaments between HEK293T and by structural studies of the NAIP/NLRC4 inflamma- HEK293TASC-BFP cells (Fig. 4A,B). But interestingly, some [50–53], the topologically conserved NACHT NLRP3PYD-mCitrine filaments changed to a more star- and LRR domains are presumed to adopt a closed like morphology with its centers colocalizing with ASC and inactive conformation and transit upon activation specks (Fig. 4A). Indeed, the number of cells with fila- into an open conformation allowing for oligomeriza- ments (65.1 7.4%, Fig. 4B) clearly correlated with the tion and successive signal transduction. Such oligomer- number of cells that formed ASC specks (64.6 4.8%, ization events might include homotypic filament Fig. 4C), which differed significantly from mCitrine conformation of then accessible PYDs, as demonstrated trol cells (2.1 0.3%, Fig. 4C). Importantly, when over- by cellular and structural studies of NLRP3 [14] and expressing human NLRP9PYD or murine NLRP9bPYD, NLRP6 [15]. For NLRP6, it was additionally shown the amounts of specking cells did not change significantly that such filaments in turn were sufficient to nucleate from the mCitrine control (Fig. 4C). ASC polymerization [15]. We found that also in case

2390 FEBS Letters 594 (2020) 2383–2395 ª 2020 The Authors. FEBS Letters published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies M. Marleaux et al. PYD structure of NLRP9

Fig. 4. Analysis of PYD polymerization and ASC speck formation in cells. (A) Images of indicated mCitrine fusion proteins overexpressed in HEK293T or HEK293TASC-BFP cells. (*) Plasmid coding for mCitrine-HA was transfected as a control and due to IRES expression had to be imaged with higher exposure time. Bar, 10 µm. Images are representative of three independent experiments. (B) Quantification of mCitrine- positive cells with filaments. Bars are representative of three independent experiments. n = 3 SEM; ****P < 0.0001 (two-way ANOVA followed by the Holm–Sidak’s multiple comparisons test). (C) Quantification of mCitrine-positive cells with ASC speck. Bars are representative of three independent experiments. n = 3 SEM; ****P < 0.0001 (ordinary one-way ANOVA followed by the Holm–Sidak’s multiple comparisons test). (D) Color-coded electrostatic surface representation of subunits of NLRP9PYD and filament slices of NLRP3PYD, ASCPYD, and NLRP9PYD generated with APBS [49]. The filament slice of ASCPYD was adapted from the ASCPYD filament structure (3J63). Crystal structures of monomeric NLRP3PYD (3QF2) and NLRP9PYD were used to model slice representations of a NLRP3PYD filament and a hypothetical NLRP9PYD filament. Modeling was based on the filament structure of NLRP6PYD (6NCV). Bottom views represent a superjacent slice of the respective filament that was rotated by 180° about the y-axis and mirrored horizontally to show interfacing regions in the

ﬁlament at the same x/y coordinates of top and bottom views. Unit of the scale is kBT/ec.

FEBS Letters 594 (2020) 2383–2395 ª 2020 The Authors. FEBS Letters published by John Wiley & Sons Ltd 2391 on behalf of Federation of European Biochemical Societies PYD structure of NLRP9 M. Marleaux et al. of NLRP3, PYD filaments are sufficient nucleators of monomer-to-dimer transition of the PYD. Our obser- ASC specks (Fig. 4A). Yet, in contrast to NLRP3, our vation of a completely monomeric behavior of data also suggest that murine and human NLRP9PYD NLRP9PYD is in accordance with these descriptions does not self-polymerize into filaments, nor nucleate and the stability of the PYD domain, although it does ASC speck formation in cells. This finding is contra- not explain its inability to form filaments. dictory with a previous study, showing NLRP9 to Post-translational modifications (PTMs) were found form an active inflammasome upon infection with to be a common mechanism for the regulation of rotavirus in the murine and human system [21]. inflammasomes [56] and other PRRs [57]. The ability Instead, our findings are reminiscent of a recent study of NLRP3PYD to form filaments is regulated by a sin- on NLRP11, which was also shown to not interact gle phosphorylation on Ser5, which prevents its with ASC but rather to repress NF-jB and type I oligomerization [14]. In turn, phosphorylation of interferon responses, two key innate immune pathways Ser198 in NLRP3 was shown to be a prerequisite for involved in inflammation [54]. activation [58], indicating the delicate balance of inhi- Using X-ray crystallography, we determined the bition and activation by PTM regulation. For structure of human NLRP9PYD and analyzed its char- NLRP9PYD, a single phosphorylation could turn the acteristics on a molecular basis. We identified basic, repulsive core of the estimated filament into an hydrophobic core residues to substantially stabilize the associative interaction. Likewise, regulative interaction typical death domain fold of NLRP9PYD, which are, partners of NLRP9 might be missing in our experi- in terms of hydrophobicity, conserved among the mental setup using recombinant proteins. For instance, NLRP subfamily (Fig. 3A). Additionally, we found a Zhu et al. [21] described the RNA helicase DHX9 as a second hydrophobic cluster stabilizing the flexible a2- potential interaction partner of NLRP9 during infec- a3 loop region and therefore partially determining the tion of IECs with rotavirus, yet without investigating conformation of NLRP9PYD. Upon filament forma- the detailed interacting domains. We used HEK293T tion, NLRP6PYD was shown to undergo large confor- cells for our experiments, which are known to express mational changes especially in the a2-a3 loop region DHX9, but potentially cotransfection of certain RNA [15]. Our data instead support the hypothesis that the might be necessary to stimulate the interaction with conformation of monomeric NLRP9PYD and NLRP9PYD. To exclude the need of further interaction NLRP3PYD is already compatible with filament forma- partners not expressed in HEK293T cells, filament for- tion with respect to the a2-a3 loop and the conforma- mation of NLRP9PYD could be tested in organoid cul- tion of the C-terminal helix. Therefore, we assume that tures of IECs. It is also possible that the PYD of the inability of NLRP9PYD to form filaments might NLRP9 is not involved in oligomerization and active not result from conformational restraints. Several resi- inflammasome formation. Alternatively, synergism dues were identified in comparison with filament-form- with the NACHT or LRR domains might be needed ing PYDs, which are charge-inversed in NLRP9PYD to effectively form filaments. However, previous stud- and typically positioned on otherwise interface sur- ies on different inflammasome effector domains faces. The primarily basic character of the hypothetical [13,9,15] and our own findings on NLRP3PYD showed NLRP9PYD filament inner core might cause repulsive their sufficiency in filament formation and inflamma- forces that are not compensated by interfacing residues some activation and thus argue against this hypothe- matching in charge, such that they most likely disrupt sis. filament formation (Fig. 4D). To validate this hypoth- In summary, the mechanisms of NLRP9 inflamma- esis, extensive mutational studies would be needed that some formation remain largely unknown. If filament will conclusively show this relationship. formation of effector domains is a hallmark of active In a previous study of the NLRP14PYD, a cluster of inflammasomes, it must be tightly regulated in NLRP9 three sites on the PYD was identified, described as the and further research elucidating its molecular mecha- Glu–Arg–Asp charge relay, that stabilizes the assembly nisms would be valuable to enable strategies for thera- between helices a2 and a6 and the loop preceding a6 peutic intervention in diseases associated with its through salt bridge formation [55]. This pattern of dysregulation. charged residues is contained in the PYDs of NLRP3 – – and NLRP9 (E22 R81 D83) but not in NLRP14 Acknowledgements where the central arginine is found to be a leucine. Accordingly, the loose coordination of a6in We would like to thank the beamline scientists at SLS NLRP14PYD allows for the formation of a long a5-a6 (Villigen) and DESY (Hamburg) for support. We also stem helix that leads to a concentration-dependent thank Dr. Gabor Horvath from the core facility

2392 FEBS Letters 594 (2020) 2383–2395 ª 2020 The Authors. FEBS Letters published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies M. Marleaux et al. PYD structure of NLRP9 microscopy in Bonn and Tomasz Prochnicki for sup- 11 Munoz-Planillo~ R, Kuffa P, Martınez-Colon G, Smith port with imaging of cells. MG and EL are funded by BL, Rajendiran TM and Nunez~ G (2013) K+ Efflux is the Deutsche Forschungsgemeinschaft (DFG, German the common trigger of NLRP3 inflammasome Research Foundation) under Germany’s Excellence activation by bacterial toxins and particulate matter. Strategy—EXC2151–390873048. This work was sup- Immunity 38, 1142–1153. ported by a grant from the Else Kroner-Fresenius-Stif-€ 12 Amores-Iniesta J, Barbera-Cremades M, Martınez CM, tung to MG (2014_A203). Pons JA, Revilla-Nuin B, Martınez-Alarcon L, Di Virgilio F, Parrilla P, Baroja-Mazo A and Pelegrın P (2017) Extracellular ATP activates the NLRP3 Author contributions inflammasome and is an early danger signal of skin allograft rejection. Cell Rep 21, 3414–3426. MM and KA performed experiments and analyzed the 13 Lu A, Li Y, Yin Q, Ruan J, Yu X, Egelman E and Wu data; EL and MG conceived and supervised the study. H (2015) Plasticity in PYD assembly revealed by cryo- MM and MG wrote the manuscript with input from EM structure of the PYD filament of AIM2. Cell all authors. Discov 1, 15013. 14 Stutz A, Kolbe C-C, Stahl R, Horvath GL, Franklin References BS, van Ray O, Brinkschulte R, Geyer M, Meissner F and Latz E (2017) NLRP3 inflammasome assembly is 1 Xue Y, Enosi Tuipulotu D, Tan WH, Kay C and Man regulated by phosphorylation of the pyrin domain. J SM (2019) Emerging activators and regulators of Exp Med 214, 1725–1736. inflammasomes and pyroptosis. Trends Immunol 40, 15 Shen C, Lu A, Jun Xie W, Ruan J, Negro R, Egelman – 1035 1052. EH, Fu T-M and Wu H (2019) Molecular mechanism 2 Kufer TA and Sansonetti PJ (2011) NLR functions beyond for NLRP6 inflammasome assembly and activation. – pathogen recognition. Nat Immunol 12,121 128. Proc Natl Acad Sci USA 116, 2052–2057. 3 Yin Q, Fu T-M, Li J and Wu H (2015) Structural 16 Dalbies-Tran R, Papillier P, Pennetier S, Uzbekova S biology of innate immunity. Annu Rev Immunol 33, and Monget P (2005) Bovine mater-like NALP9 is an – 393 416. oocyte marker gene. Mol Reprod Dev 71, 414–421. 4 Mariathasan S, Newton K, Monack DM, Vucic D, 17 Ponsuksili S, Brunner RM, Goldammer T, Kuhn€ C, French DM, Lee WP, Roose-Girma M, Erickson S and Walz C, Chomdej S, Tesfaye D, Schellander K, Dixit VM (2004) Differential activation of the Wimmers K and Schwerin M (2006) Bovine NALP5, inflammasome by caspase-1 adaptors ASC and Ipaf. NALP8, and NALP9 genes: assignment to a QTL – Nature 430, 213 218. region and the expression in adult tissues, oocytes, 5 Man SM and Kanneganti T-D (2016) Converging roles and preimplantation embryos. Biol Reprod 74, 577– of caspases in inflammasome activation, cell death and 584. – innate immunity. Nat Rev Immunol 16,7 21. 18 Thelie A, Papillier P, Pennetier S, Perreau C, Traverso 6 Shi J, Zhao Y, Wang K, Shi X, Wang Y, Huang H, J, Uzbekova S, Mermillod P, Joly C, Humblot P and Zhuang Y, Cai T, Wang F and Shao F (2015) Cleavage Dalbies-Tran R (2007) Differential regulation of of GSDMD by inflammatory caspases determines abundance and deadenylation of maternal transcripts – pyroptotic cell death. Nature 526, 660 665. during bovine oocyte maturation in vitro and in vivo. 7 Lu A, Magupalli VG, Ruan J, Yin Q, Atianand MK, BMC Dev Biol 7, 125. € Vos MR, Schroder GF, Fitzgerald KA, Wu H and 19 Romar R, De Santis T, Papillier P, Perreau C, Thelie Egelman EH (2014) Unified polymerization mechanism A, Dell’Aquila ME, Mermillod P and Dalbies-Tran R for the assembly of ASC-dependent inflammasomes. (2011) Expression of maternal transcripts during bovine – Cell 156, 1193 1206. oocyte in vitro maturation is affected by donor age: 8 Lu A and Wu H (2015) Structural mechanisms of gene expression in different bovine competence models. – inflammasome assembly. FEBS J 282, 435 444. Reprod Domest Anim 46, e23–e30. 9 Li Y, Fu T and Wu H (2018) Cryo-EM structures of 20 Ngo C and Man SM (2017) NLRP9b: a novel RNA- ASC and NLRC4 CARD filaments reveal a unified sensing inflammasome complex. Cell Res 27, 1302–1303. mechanism of nucleation and activation of caspase-1. 21 Zhu S, Ding S, Wang P, Wei Z, Pan W, Palm NW, – Proc Natl Acad Sci USA 115, 10845 10852. Yang Y, Yu H, Li H-B, Wang G et al. (2017) Nlrp9b 10 Heid ME, Keyel PA, Kamga C, Shiva S, Watkins SC inflammasome restricts rotavirus infection in intestinal and Salter RD (2013) Mitochondrial reactive oxygen epithelial cells. Nature 546, 667–670. species induces NLRP3-dependent lysosomal damage 22 Yanling Q, Xiaoning C, Fei B, Liyun F, Huizhong H – and inflammasome activation. J Immunol 191, 5230 and Daqing S (2018) Inhibition of NLRP9b attenuates 5238. acute lung injury through suppressing inflammation,

FEBS Letters 594 (2020) 2383–2395 ª 2020 The Authors. FEBS Letters published by John Wiley & Sons Ltd 2393 on behalf of Federation of European Biochemical Societies PYD structure of NLRP9 M. Marleaux et al.

apoptosis and oxidative stress in murine and cell the Thermotoga maritima proteome. Acta Crystallogr D models. Biochem Biophys Res Commun 503, 436–443. Biol Crystallogr 59, 1028–1037. 23 Nagyoszi} P, Nyul-T oth A, Fazakas C, Wilhelm I, 33 Kabsch W (1993) Automatic processing of rotation Kozma M, Molnar J, Hasko J and Krizbai IA (2015) diffraction data from crystals of initially unknown Regulation of NOD-like receptors and inflammasome symmetry and cell constants. J Appl Crystallogr 26, activation in cerebral endothelial cells. J Neurochem 795–800. 135, 551–564. 34 Rossmann MG (1990) The molecular replacement 24 Nyul-T oth A, Kozma M, Nagyoszi} P, Nagy K, method. Acta Crystallogr A 46,73–82. Fazakas C, Hasko J, Molnar K, Farkas AE, Vegh AG, 35 Read RJ (2001) Pushing the boundaries of molecular Var oG et al. (2017) Expression of pattern recognition replacement with maximum likelihood. Acta Crystallogr receptors and activation of the non-canonical D Biol Crystallogr 57, 1373–1382. inflammasome pathway in brain pericytes. Brain Behav 36 Eibl C, Grigoriu S, Hessenberger M, Wenger J, Immun 64, 220–231. Puehringer S, Pinheiro AS, Wagner RN, Proell M, 25 Tadaki H, Saitsu H, Nishimura-Tadaki A, Imagawa T, Reed JC, Page R et al. (2012) Structural and functional Kikuchi M, Hara R, Kaneko U, Kishi T, Miyamae T, analysis of the NLRP4 pyrin domain. Biochemistry 51, Miyake N et al. (2011) De novo 19q13.42 duplications 7330–7341. involving NLRP gene cluster in a patient with systemic- 37 Adams PD, Grosse-Kunstleve RW, Hung L-W, Ioerger onset juvenile idiopathic arthritis. J Hum Genet 56, TR, McCoy AJ, Moriarty NW, Read RJ, Sacchettini 343–347. JC, Sauter NK and Terwilliger TC (2002) PHENIX: 26 Castano-Rodr~ ıguez N, Kaakoush NO, Goh K-L, Fock building new software for automated crystallographic KM and Mitchell HM (2014) The NOD-like receptor structure determination. Acta Crystallogr D Biol signalling pathway in Helicobacter pylori infection and Crystallogr 58, 1948–1954. related gastric cancer: a case-control study and gene 38 Emsley P and Cowtan K (2004) Coot : model-building expression analyses. PLoS One 9, e98899. tools for molecular graphics. Acta Crystallogr D Biol 27 Poli G, Brancorsini S, Cochetti G, Barillaro F, Egidi Crystallogr 60, 2126–2132. MG and Mearini E (2015) Expression of 39 DeLano WL (2002) The PyMOL Molecular Graphics inflammasome-related genes in bladder cancer and their System. DeLano Scientific, San Carlos, CA. association with cytokeratin 20 messenger RNA. Urol 40 McQuin C, Goodman A, Chernyshev V, Kamentsky L, Oncol Semin Orig Investig 33, 505.e1–505.e7. Cimini BA, Karhohs KW, Doan M, Ding L, Rafelski 28 Mearini E, Poli G, Cochetti G, Boni A, Egidi MG and SM and Thirstrup D (2018) Cell Profiler 3.0: Next- Brancorsini S (2017) Expression of urinary miRNAs generation image processing for biology. PLOS Biol 16, targeting NLRs inflammasomes in bladder cancer. Onco e2005970. Targets Ther 10, 2665–2673. 41 Ferrao R and Wu H (2012) Helical assembly in the 29 Poli G, Cochetti G, Boni A, Egidi MG, Brancorsini S death domain (DD) superfamily. Curr Opin Struct Biol and Mearini E (2017) Characterization of 22, 241–247. inflammasome-related genes in urine sediments of 42 Hiller S, Kohl A, Fiorito F, Herrmann T, Wider G, patients receiving intravesical BCG therapy. Urol Oncol Tschopp J, Grutter€ MG and Wuthrich€ K (2003) NMR Semin Orig Investig 35, 674.e19–674.e24. structure of the apoptosis- and inflammation-related 30 Fernandez MV, Budde J, Del-Aguila JL, Ibanez~ L, NALP1 pyrin domain. Structure 11, 1199–1205. Deming Y, Harari O, Norton J, Morris JC, Goate AM, 43 Bae JY and Park HH (2011) Crystal structure of Cruchaga C et al. (2018) Evaluation of gene-based NALP3 Protein Pyrin Domain (PYD) and its family-based methods to detect novel genes associated implications in inflammasome assembly. J Biol Chem with familial late onset alzheimer disease. Front 286, 39528–39536. Neurosci 12, 209. 44 Pinheiro AS, Proell M, Eibl C, Page R, 31 Gil-Varea E, Urcelay E, Vilarino-G~ uell€ C, Costa C, Schwarzenbacher R and Peti W (2010) Three- Midaglia L, Matesanz F, Rodrıguez-Antiguedad€ A, dimensional Structure of the NLRP7 Pyrin domain: Oksenberg J, Espino-Paisan L, Dessa Sadovnick A insight into Pyrin-Pyrin-mediated effector domain et al. (2018) Exome sequencing study in signaling in innate immunity. J Biol Chem 285, 27402– patients with multiple sclerosis reveals variants 27410. associated with disease course. J Neuroinflammation 45 Schnappauf O, Chae JJ, Kastner DL and Aksentijevich 15, 265. I (2019) The pyrin inflammasome in health and disease. 32 Page R, Grzechnik SK, Canaves JM, Spraggon G, Front Immunol 10, 1745. Kreusch A, Kuhn P, Stevens RC and Lesley SA (2003) 46 Wang L, Manji GA, Grenier JM, Al-Garawi A, Shotgun crystallization strategy for structural genomics: Merriam S, Lora JM, Geddes BJ, Briskin M, DiStefano an optimized two-tiered crystallization screen against PS and Bertin J (2002) PYPAF7, a Novel PYRIN-

2394 FEBS Letters 594 (2020) 2383–2395 ª 2020 The Authors. FEBS Letters published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies M. Marleaux et al. PYD structure of NLRP9

containing Apaf1-like protein that regulates activation Structural and biochemical basis for induced self- of NF-jB and Caspase-1-dependent cytokine propagation of NLRC4. Science 350, 399–404. processing. J Biol Chem 277, 29874–29880. 53 Zhang L, Chen S, Ruan J, Wu J, Tong Ab, Yin Q, Li 47 Vladimer GI, Weng D, Paquette SWM, Vanaja SK, Y, David L, Lu A, Wang Wl et al. (2015) Cryo-EM Rathinam VAK, Aune MH, Conlon JE, Burbage JJ, structure of the activated NAIP2-NLRC4 Proulx MK, Liu Q et al. (2012) The NLRP12 inflammasome reveals nucleated polymerization. Science inflammasome recognizes yersinia pestis. Immunity 37, 350, 404–409. 96–107. 54 Ellwanger K, Becker E, Kienes I, Sowa A, Postma Y, 48 Jin T, Huang M, Jiang J, Smith P and Xiao TS (2018) Gloria YC, Weber ANR and Kufer TA (2018) The Crystal structure of human NLRP12 PYD domain and NLR family pyrin domain–containing 11 protein implication in homotypic interaction. PLoS One 13, contributes to the regulation of inflammatory signaling. e0190547. J Biol Chem 293, 2701–2710. 49 Jurrus E, Engel D, Star K, Monson K, Brandi J, 55 Eibl C, Hessenberger M, Wenger J and Brandstetter H Felberg LE, Brookes DH, Wilson L, Chen J and Liles (2014) Structures of the NLRP14 pyrin domain reveal a K (2018) Improvements to the APBS biomolecular conformational switch mechanism regulating its solvation software suite: improvements to the APBS molecular interactions. Acta Crystallogr D Biol Software Suite. Protein Sci 27, 112–128. Crystallogr 70, 2007–2018. 50 Hu Z, Yan C, Liu P, Huang Z, Ma R, Zhang C, Wang 56 Yang J, Liu Z and Xiao TS (2017) Post-translational R, Zhang Y, Martinon F, Miao D et al. (2013) Crystal regulation of inflammasomes. Cell Mol Immunol 14,65– structure of NLRC4 reveals its autoinhibition 79. mechanism. Science 341, 172–175. 57 Liu J, Qian C and Cao X (2016) Post-translational 51 Diebolder CA, Halff EF, Koster AJ, Huizinga EG and modification control of innate immunity. Immunity 45, Koning RI (2015) Cryoelectron tomography of the 15–30. NAIP5/NLRC4 inflammasome: implications for NLR 58 Song N, Liu Z-S, Xue W, Bai Z-F, Wang Q-Y, Dai J, activation. Structure 23, 2349–2357. Liu X, Huang Y-J, Cai H, Zhan X-Y et al. (2017) 52 Hu Z, Zhou Q, Zhang C, Fan S, Cheng W, Zhao Y, NLRP3 phosphorylation is an essential priming event Shao F, Wang H-w, Sui S-f and Chai J (2015) for inflammasome activation. Mol Cell 68, 185–197.

FEBS Letters 594 (2020) 2383–2395 ª 2020 The Authors. FEBS Letters published by John Wiley & Sons Ltd 2395 on behalf of Federation of European Biochemical Societies