The Yersinia pestis effector YopM binds the human pyrin death domain to inhibit inflammasome activation and effector-triggered immunity

Abstract

Bacteria in the genus Yersinia use a type III secretion system to inject several effectors into host cells to disrupt multiple signaling processes. Two effectors disrupt host signaling by inactivating RhoA GTPases. Inactivation of RhoA triggers pyrin inflammasome assembly in infected phagocytes leading to a protective immune response. A third protein, YopM, is an essential virulence factor that counteracts effector-triggered immunity by inactivating pyrin. YopM has a leucine-rich repeat (LRR) domain and hijacks host kinases to inactivate pyrin by phosphorylation. Previously, the LRR domain of Yersinia pestis YopM was implicated in binding to an N-terminal region of human pyrin, which includes the eponymous pyrin domain (PYD), a member of the death domain family. However, the interaction mechanism was undefined. Using a bacterial two-hybrid assay, protein biochemistry and X-ray crystallography, we determined that the concave surface of the YopM LRR domain binds to the PYD. Identification of critical residues in the interaction revealed that YopM binds PYD R42, an amino acid important for positive and negative regulation of pyrin. YopM binding to the PYD was reduced by R42W, a codon change associated with the autoinflammatory disease Familial Mediterranean Fever (FMF). Furthermore, we show that YopM codon substitution variants defective for PYD binding fail to inhibit the pyrin inflammasome in human monocytes infected with Yersinia. In addition, we found that PYD binding is dispensable for YopM to inhibit the inflammasome in murine macrophages, suggesting this effector uses a distinct mechanism to target mouse pyrin. These results define how Y. pestis YopM binds the human PYD and provide insights into how this interaction likely selected for pyrin gain of function variants resulting in FMF.

INTRODUCTION

Pathogens use a variety of strategies to evade detection and elimination by host immune responses. One method employed by numerous pathogenic bacteria involves injecting effector proteins directly into eukaryotic cells. These effectors can be injected via the type III secretion system (T3SS) which is employed by Gram-negative bacteria that infect both animals and plants (1). The T3SS is a sophisticated protein secretion machinery which spans from the bacterial membrane through the host plasma membrane (2). Human pathogenic species of Yersinia, Shigella and Salmonella among others use a contact-dependent T3SS to secrete effectors, regulate host cell responses and promote pathogenesis (2). Host cells can sense assault by the T3SS and or effectors and mount immune responses (3). In macrophages infected with bacterial pathogens, perturbations induced by T3SSs and/or effectors can result in the assembly of inflammasomes (4). Canonical inflammasomes are multiprotein assemblies that promote the recruitment and maturation of caspase-1 into its active form (4). Active caspase-1 cleaves gasdermin-D (GSDMD), pro-IL-18 and pro-IL-1β (5, 6). The N-terminal domain of GSDMD forms pores in the plasma membrane, promoting release of mature IL-18 and IL-1β (5, 6). GSDMD pores can also result in the release of additional pro-inflammatory molecules and intracellular bacteria as a result of cell death, termed pyroptosis (6, 7). Pyroptosis downstream of GSDMD pore formation requires plasma membrane rupture mediated by the ninjurin 1 protein (8). To counteract protective immune responses mediated by IL-1β, IL-18 and pyroptosis, virulent bacterial pathogens have evolved the ability to inhibit caspase-1 inflammasomes in host cells (4, 9).

The genus Yersinia contains several Gram-negative bacterial pathogens that cause invasive human infections ranging in severity from self-limiting mesenteric lymphadenitis (Y. pseudotuberculosis) to lethal plague (Y. pestis). Yersinia colonize lymphoid tissues and replicate as extracellular microcolonies in phagocyte-rich lesions known as pyogranulomas and a plasmid-encoded T3SS is required for this infection process (10, 11). The T3SS delivers seven effectors named Yersinia outer proteins (Yops) into the cytosol of host cells in contact with the bacteria (12). The effectors inhibit key host cell processes, such as proinflammatory cytokine production and phagocytosis (12). YopE and YopT are two Yersinia effectors that contribute to the anti-signaling functions of the T3SS by targeting Rho GTPases (13, 14). YopE has a GAP domain that accelerates conversion of GTP to GDP, locking the GTPase in an inactive form (13). YopT is a cysteine protease which cleaves off a C-terminal amino acid and lipid anchor from the GTPase, disrupting its association on host membranes and preventing interaction with downstream targets (14). YopM is a leucine rich repeat (LRR) effector that inhibits the pyrin inflammasome in activated macrophages (15, 16). YopJ is a member of the acetyltransferase effector family (17) that inhibits MAPK and NF-kB signaling pathways to inhibit proinflammatory cytokine production (18, 19). YopJ also cooperates by an unknown mechanism with YopM to inhibit the pyrin inflammasome in activated macrophages (20).

Inactivation of Rho GTPases (RhoA, -B or -C, hereafter referred to as RhoA) by bacterial toxins or effectors triggers assembly of the pyrin inflammasome by an indirect guard mechanism in phagocytes (21, 22). YopE or YopT triggers assembly of the pyrin inflammasome in activated macrophages infected by Yersinia (15, 16, 23, 24). Pyrin, encoded by the Mediterranean fever (MEFV) gene, is a cytosolic pattern recognition receptor (PRR) which is preferentially expressed in phagocytes such as macrophages and neutrophils (25, 26). Human pyrin is comprised of the following domains: An N-terminal PYD followed by a linker region, B-box domain, coiled coil domain (CC) and a C-terminal B30.2 domain (Fig. S1A). Murine pyrin is structurally like human pyrin except that it lacks the C- terminal B30.2 domain, which is replaced by a short unstructured tail (Fig. S1A) (23, 25, 26). Pyrin is also known as TRIM20 as it belongs to the TRIM family of proteins (27) and forms dimers via the CC domain (Fig. S1B) (28). By analogy to TRIM5 (29), the pyrin B-box potentially mediates the formation of higher order oligomers such as dimers of dimers (30). Pyrin is negatively and positively regulated by multiple mechanisms (21, 23, 30–35). Pyrin is negatively regulated by phosphorylation via the protein kinase C-related kinase (PRK, also known as PKN, protein kinase N) which in turn is activated by RhoA. Human pyrin is phosphorylated on serine residues 208 and 242 (205 and 241 on mouse pyrin) on the linker region creating a binding site for 14-3-3 protein dimers (32, 33). Pyrin is also negatively regulated by binding of the PYD to the B-box (31). A model posits that phosphorylated pyrin is rendered inactive as the 14-3-3-bound linker is bent into a hairpin (36) and the PYD is sequestered by the B-box (Fig. S1B), preventing interaction with apoptosis-associated speck-like protein containing a CARD (ASC) (23). In Yersinia-infected cells pyrin should be activated when YopE or YopT inactivate RhoA, and a positive regulation step leads to dephosphorylation by a phosphoprotein phosphatase (23, 24, 30). According to the model, when pyrin is dephosphorylated, the PYD is released from the B-box and exposed, ASC is recruited, caspase-1 is activated, GSDMD pores are formed and active IL-1β and IL-18 are secreted (23, 30). However, pyrin inflammasome activation in phagocytes infected with Yersinia is blocked due to the activity of YopM (15, 16). YopM hijacks the host kinases PRK and ribosomal S6 kinase (RSK) (37, 38) to phosphorylate pyrin, locking it in an inactive state (15, 23, 39). YopM is crucial for Yersinia virulence in mouse models of infection (16, 40, 41) and yet a Y. pseudotuberculosis ΔyopM mutant is fully virulent in Mefv^−/− mice, demonstrating that pyrin is the major host target of YopM (15).

YopM is a member of the bacterial subfamily of leucine rich repeat (LRR) effector proteins (42). Pathogenic species and strains of Yersinia encode for a variety of different YopM proteins, referred to here as isoforms, with varying numbers of LRRs. Despite these differences in the number and composition of LRRs, all YopM isoforms studied so far have been shown to inhibit the pyrin inflammasome (15, 16). The structure of YopM from Y. pestis 195/P has been determined and adopts an overall horse-shoe shaped fold comprised of two N-terminal α-helices followed by an extended LRR region with short β-strands (43). The two α-helices provide a nucleation site that facilitates the folding of the LRRs (43). The N- and C-termini of YopM are unstructured and yet highly conserved among YopM isoforms. The N-terminus contains the secretion signal required for YopM translocation into host cells (44), and the C-terminal tail is a binding site for the N-terminal kinase domain of host kinase RSK (Fig. S1A) (40, 41, 45).

During Yersinia infection of host phagocytes, YopM acts as a scaffold connecting the host kinases RSK and PRK (37, 38) to allow for pyrin phosphorylation and inactivation (15, 16, 39). Binding to YopM results in autophosphorylation and activation of RSK, which in turn phosphorylates and activates PRK (Fig. S1C) (37). Inflammasome activation was increased when murine macrophages infected with Y. pseudotuberculosis were treated with a PRK inhibitor or with siRNAs that reduced expression of Prk, suggesting that PRK activity is important for YopM to inhibit pyrin in mouse phagocytes (15). Similarly, inflammasome activation was increased when human monocytes infected with Y. pestis were treated with RSK inhibitors or with siRNAs that reduced expression of Rsk, suggesting that RSK activity is important for YopM to inhibit pyrin in human phagocytes (39). Park et al. used in vitro kinase assays with purified proteins to show that Y. pestis YopM could hijack either PRK or RSK to phosphorylate the human pyrin linker, but RSK had much higher activity, and there was no additive effect when both kinases were present (39). These data suggest that both kinases are important, but RSK may be critical for Y. pestis YopM to inhibit human pyrin.

Like other proteins in this family, the concave surface of the YopM LRR domain contains exposed amino-acid sidechains (43) and is likely important for its scaffolding function in binding host proteins. The kinase domain of PRK binds to YopM (Fig. S1A) (37), and the interaction site appears to be within LRRs 6-15 of the Y. pestis isoform (40). Deletion analysis with purified Y. pestis YopM variants demonstrated an importance for LRRs 1 and 2 for pyrin binding (39). However, a previous study showed low expression in Yersinia of a YopM ΔLRR2-3 variant (41), suggesting that this deletion disrupted YopM stability. Therefore, it’s possible that the reduced binding of YopM ΔLRR 1-2 to pyrin (39) was a result of structural instability caused by the deletion, and the pyrin binding site on YopM is yet to be established. Additionally, it has been shown that YopM binds to the N-terminus of human pyrin encompassing the PYD and linker (39). Removal of the 90 amino acid PYD prevented phosphorylation of human pyrin by YopM-hijacked RSK in an in vitro kinase reaction (39), revealing a key role for this N-terminal domain, but the molecular mechanism underscoring this interaction is not fully understood.

Here, we provide evidence of direct binding between Y. pestis YopM isoforms and the human PYD. We determined two co-crystal structures of YopM bound to human PYD and identified key residues on each that are involved in their interaction. We also identified YopM variants with alanine substitutions in LRR 5 that are defective for PYD binding in vitro and used human monocyte infection assays to demonstrate that PYD binding is critical for YopM to target and inhibit the pyrin inflammasome in vivo. Our data also indicate that Y. pestis YopM specifically binds a key regulatory site in the PYD, which has implications for the evolutionary host-pathogen arms race that resulted in FMF (39) and may explain why a distinct mechanism appears to be used to target the mouse pyrin inflammasome.

RESULTS

YopM Binds to the PYD of Human Pyrin

We tested for direct interactions between Y. pestis YopM isoforms and N-terminal domains of human or murine pyrin. We first used the bacterial adenylate cyclase-based two hybrid assay (BACTH) to test for protein-protein interactions. Two different Y. pestis YopM isoforms were studied, the 13-LRR version from Pestoides A (YopM^PestA) and the 15-LRR version from KIM (YopM^KIM) (46), which have been shown to inhibit the pyrin inflammasome (15, 16). YopM^PestA and YopM^KIM were expressed as N-terminal fusions with the T25 domain and the human (h) or murine (m) PYD domains as C-terminal fusions to T18 in Escherichia coli. The hPYD and mPYD share 76% identity at the primary sequence level. YopM^PestA and YopM^KIM showed similar strong binding to hPYD (Fig. 1A, Fig. S2A). Surprisingly, YopM^PestA showed no detectable binding to mPYD (Fig. 1B). We further investigated these interactions using GST pulldowns, whereby GST-tagged YopM^KIM was co-expressed in E. coli with His-tagged mPYD or hPYD. Samples of clarified E. coli input lysates or pull downs were analyzed by SDS-PAGE and blue staining or anti-His immunoblotting. As compared to the GST-alone control, there was low but detectable pull down of mPYD with GST-YopM^KIM (Fig. S3). However, as compared to mPYD, there was increased soluble hPYD in the input lysates, and as a result, more hPYD was pulled down with GST-YopM^KIM (Fig. S3). Overall, these results suggest that YopM^PestA and YopM^KIM strongly bind to hPYD, increasing its solubility, and have a lower affinity for mPYD (Fig. S1A).

Fig. 1. Analysis of YopM-PYD interactions using BACTH.

The indicated YopM proteins were N-terminally tagged with T25 and hPYDs (A, B) or mPYDs (B) were C-terminally tagged with T18. Plasmids encoding the fused partners, ZIP positive controls or empty vector (EV) controls were co-transformed into E. coli to test for protein-protein interactions. Measured Miller units of β-galactosidase activity indicating protein interaction in E. coli are indicated on the y-axis. Results are averages and standard deviations with data pooled from at least 3 independent experiments. Significance determined by one-way ANOVA and Tukey’s post hoc test in comparison to PestA + hPYD.****p<0.0001; **p=0.01; *p=0.05; ns, not significant.

Next, we followed up on previous experiments that identified the human pyrin linker as a YopM^KIM binding site (39). We tested a fragment of the human pyrin linker (amino acids 186-326) which was shown to bind to YopM^KIM in GST pull-downs (39). However, we found that this linker fragment (hPyrin186-326) had non-specific binding to GST (Fig. S4A). In addition, we found that YopM^KIM did not bind to hPyrin186-326 in BACTH (Fig. S4B). In summary, these results indicate that the PYD is the primary targeting domain for Y. pestis YopM isoforms in the human pyrin N-terminal region (Fig. S1A). Since YopM^KIM and YopM^PestA inhibit mouse pyrin in macrophages and in vivo (15, 39), there may be a different mechanism of inflammasome targeting by this effector in murine macrophages, including possible multidentate interactions with the PYD and the linker or other domains, or the use of hijacked PRK as an adapter.

In-Solution State of YopM^PestA

A crystal structure of the 15-LRR YopM protein from Yersinia pestis 195/P has been determined (43). In that study, the authors found that YopM^195/P formed superhelical tetramers in the crystal lattice. The biological relevance of YopM tetramers is unknown, and it was hypothesized to be a crystallographic artifact given that a tetrameric state was not observed in solution (43). However, other studies have shown that YopM forms higher-order oligomers when analyzed by native PAGE in purified form (dimers and tetramers) (41) or in lysates of Yersinia-infected macrophages (30, 41), and in the latter case YopM also appeared to interact with dimers and higher-order multimers of the host target protein pyrin (30). We ultimately aimed to determine the structure of the smaller 13-LRR YopM^PestA isoform in complex with the hPYD, but first sought to determine the oligomeric state of YopM^PestA in solution. We performed analytical size exclusion chromatography (SEC) of purified full-length YopM^PestA. We observed that YopM^PestA elutes at a faster rate than expected for a monomer on SEC compared to globular standard proteins, implying that YopM^PestA could be a dimer in solution (~90kDa) (Fig. S5A,B), consistent with prior work. Notably, a previous study reported that a 20-LRR YopM from Y. enterocolitica was a dimer in solution, based on its elution from SEC (47). However, when calibrated against globular standards, SEC retention volume does not accurately predict the molecular mass of non-globular proteins or of proteins which interact with the column matrix (48). Given that YopM is an elongated protein, its migration during SEC may well be unusual compared to the globular protein standards. Therefore, we performed further analysis using SEC coupled with multi-angle light scattering (SEC-MALS) which permits a shape-independent analysis of protein size (48, 49). Data from SEC-MALS indicated that YopM^PestA is a monomer (~42kDa) at neutral pH (7.4) in solution (Fig. 2A). We further utilized velocity analytical ultracentrifugation (vAUC), another shapeindependent method of calculating protein size in solution, to analyze purified YopM^PestA. vAUC data also suggested that YopM^PestA is a monomer in neutral pH (Fig. 2B). Interestingly, the sedimentation coefficient of YopM^PestA shifted in low pH (5.0) buffer conditions, consistent with a higher molecular mass (~148 kDa, Fig. 2B). Given that the calculated molecular mass of the trimeric and tetrameric states are ~126 kDa and ~168 kDa, respectively, we cannot unambiguously assign an exact stoichiometry for assembly. However, a tetramer is consistent with previously reported crystal structures (43). Taken together with previous observations, our data suggest that YopM^PestA is monomeric at neutral pH, and transitions to an oligomeric form under mildly acidic conditions in solution.

Fig. 2. Characterization of YopM^PestA in solution by SEC-MALs or vAUC.

A) SEC-MALS analysis of YopM^PestA at pH 7.4. The experimental relative molar mass across the peak width is plotted in magenta (right-hand axis), overlaid with the differential refractive index (left-hand axis) of the chromatogram. The estimated relative molar mass is 42.9 kDa. B) Sedimentation coefficient distribution of YopM^PestA by vAUC at pH 7.4 (Blue; main peak, corresponding to an estimated Mr of 43.5 kDa) and pH 5.0 (Black; main peak, corresponding to an estimated Mr of 148.2kDa).

Structures of YopM^PestA-hPYD Complexes

Although structural studies of hPYD and closely related YopM isoforms have been conducted in isolation, information pertaining to their complex formation is lacking. Therefore, we sought to determine the structure of a YopM^PestA-hPYD complex to establish the stereochemical basis for the scaffolding function of YopM. Crystallization trials using the full-length YopM^PestA co-purified with hPYD yielded crystals that were difficult to reproduce and optimize (Fig. S6A). Concurrently, a truncation construct of YopM^PestA lacking the unstructured first 33 and last 23 residues (43) (amino acids 34-344), was designed to minimize interference with lattice packing by the disordered termini. The resulting protein, henceforth referred to as YopM^ΔNΔC, has a predicted monomer molecular mass of 35.0 kDa, was produced at high purity, and exhibited excellent biochemical behavior. Importantly, YopM^ΔNΔC retained binding to recombinant hPYD as demonstrated by SEC co-elution (Fig. S7). Interestingly, co-elution with hPYD did not significantly alter the SEC retention volume of the complex compared to that of YopM^ΔNΔC on its own. In consideration of the aforementioned structural data, we hypothesized that hPYD could ‘tuck’ into the concave surface of YopM^ΔNΔC without substantially increasing the hydrodynamic radius of the nascent complex (Fig. S7A,C). Having verified the integrity of the binding interaction, recombinant hPYD was then mixed with YopM^ΔNΔC in a 4:1 molar ratio for use in crystallization trials, which led to a condition that produced suitable crystals (Fig. S6B).

In total, X-ray diffraction data were obtained from crystals produced using both strategies and the resulting complexes were found to crystallize in different space groups: YopM^PestA co-purified with hPYD crystallized in space group C 2, and the YopM^ΔNΔC-hPYD complex crystallized in space group P 3₂21 (Table 1). Although the former produced quality diffraction to 2.1 Å, it was apparent from the electron density map that the C-terminal portion of YopM^PestA beginning in the middle of LRR 9, as well as the sixth α-helix of hPYD, were missing. Careful inspection of the crystal lattice led us to conclude that these missing regions could not be included without significant steric overlap with symmetry-related molecules. Based on these observations, we hypothesize that the fragments captured in this crystal were the result of an unanticipated limited proteolysis event, which is consistent with the difficulties in reproducing this crystal form. This complex is henceforth referred to as YopM^PestA*-hPYD*. In light of these observations, the co-crystal structure of YopM^ΔNΔC-hPYD, which diffracted to 3.7 Å, was also refined and used to authenticate the limited proteolysis structure, which possessed the resolution necessary to confidently model the side-chain interactions at the binding interface.

Table 1.

Summary of data collection, reduction, and refinement statistics

	YopMPestA*-hPYD*	YopMΔNΔC-hPYD
Data Collection
Beamline	NSLS-II 17-ID-1	NSLS-II 17-ID-1
Wavelength (Å)	0.97933	0.920105
Space Group	C 2	P 3221
Unit cell parameters
a, b, c (Å)	166.2, 41.6, 115.0	141.6, 141.6, 61.6
α, β, γ (°)	90, 132.7, 90	90, 90, 120
Resolutiona (Å)	39.4 - 2.10 (2.18 - 2.10)	30.7 - 3.70 (3.83 - 3.70)
Rmeasb (%)	5.3 (30.8)	8.4 (83.9)
Rp.i.m c. (%)	2.9 (18.0)	4.6 (47.1)
CC1/2d (%)	99.9 (94.8)	99.8 (43.4)
I/σI	14.0 (3.3)	8.9 (1.4)
Completeness (%)	98.9 (94.6)	96.6 (98.7)
Redundancy	3.3 (2.6)	3.1 (2.9)
Refinement
#Total reflections	33962 (3229)	7525 (774)
Test-set reflections	1664 (159)	407 (43)
Rworke/Rfreef (%)	18.2/21.2	23.9/25.0
Number of atoms:
Protein	4601	3140
Water	240	0
Ramachandran Plotg (%)	95.5/4.5/0.0	94.8/5.2/0.0
Baverage (Å2)	50.3	145.5
Protein	50.6	145.5
Water	44.4
Bond length RMSD (Å)	0.007	0.003
Bond angle RMSD (°)	0.82	0.71
PDB ID	pdb_00009oyu	pdb_000010dr

^aValues in parentheses correspond to the highest resolution shell.

^bR_meas: the redundancy independent R-factor, described in reference (77).

^cR_p.i.m: the precision-indicated merging R-factor described in (78).

^dCC_½: the percentage of correlation between intensities from random half-datasets (79).

^eR_work = Σ_h |F_obs(h) – F_calc(h)|/Σ_h F_obs(h), h ∈ [working set]

^fR_free = Σ_h |F_obs(h) - F_calc(h)|/Σ_h F_obs(h), h ∈ [test set]

^gFavored/allowed/outliers

With the exception of the truncations described above, the quaternary structures of both complexes are highly similar. The complex forms in a 1:1 stoichiometry with hPYD* nestled into the concave surface of YopM^PestA* (Fig. 3A) and is consistent with the SEC co-elution profiles. Overall, the two YopM^PestA*-hPYD* molecules in the asymmetric unit are similar to each other (C_α RMSD = 0.6 Å), but hPYD α2, which lies near the binding interface, was observed in slightly different positions (Fig. S8A). This difference could be attributed to lattice packing, and alignment to the YopM^ΔNΔC-hPYD complex identified “Molecule 1” as being most similar (compare Fig. S8B, C_α RMSD = 0.5 Å to Fig. S8C, C_α RMSD = 0.8 Å). Therefore, we conclude that the purported proteolytic truncation bears no influence on complex formation, and the biologically relevant interface is best interpreted from YopM^PestA*-hPYD* “Molecule 1,” which was used for subsequent analyses as detailed below.

Fig. 3. hPYD* binds to the concave surface of YopM^PestA*.

The co-crystal structure of YopM^PestA*-hPYD* reveals a binding interface composed of the negatively charged concave surface of YopM^PestA* spanning LRRs 1–9 and the positively charged surface of hPYD* involving α-helices 2–4. A) YopM^PestA* (pink) and hPYD* (green) are depicted as ribbon outlines, with interfacial residues depicted as stick outlines. The YopM^PestA* residues selected for mutation (YopM Ala2) and cognate hPYD* residues are highlighted as pink and green sticks, respectively. B and C) The electrostatic surfaces of YopM^PestA* (B) and hPYD* (C) are shown in orthogonal views. As a visual aid, the second protein of the complex is shown as a black-and-white outline in the side-on view (left-hand panel). D-F) Focused views of interactions highlighted in (A) between YopM^PestA* and hPYD* α3 that are disrupted in the YopM Ala2 mutant. YopM^PestA* D159 (D) and D161 (E) are the foci of hydrogen-bond networks with hPYD* α3 that involve structural water WAT409 and salt-bridges with hPYD* R39 and R42. YopM^PestA* F177 (F) packs against hPYD* α3, becoming shielded from solvent by apolar and polar atoms of hPYD*. Non-carbon atoms are colored by type: oxygen, red; nitrogen, blue.

The co-crystallized complex reveals an extensive region of van der Waals contact, distributed across YopM^PestA* LRRs 1-9. The interacting residues of hPYD* are localized to α-helices 2-4, with a3 serving as a focal point (Fig. 3A). Detailed analysis of the contact surfaces suggests that electrostatic attraction is a driving feature of the YopM^PestA*-hPYD* interaction. The inner, concave surface of YopM^PestA* carries a uniform, negative charge (Fig. 3B). While the electrostatic surface potential of hPYD* is marked by both positive and negative patches, an equatorial and continuous band of positive charge is found on hPYD* that interfaces with the negatively charged, concave surface of YopM^PestA* (Fig. 3C). In agreement with a binding interface that favors electrostatic complementarity, several extensive hydrogen-bond networks and salt bridges are observed (Fig. S9).

Identification of YopM Residues Required for PYD Binding and Inhibition of the Human Pyrin Inflammasome

To identify residues on YopM^PestA that are critical for hPYD binding, positions predicted by the complex structure to be important for interaction were subjected to alanine substitution in groups of 3, and the resulting variants were named Ala1-3 (Table 2 and Fig. S10A). The variants were tested for hPYD binding by the BACTH assay, which indicated that Ala2 had significantly diminished binding whereas Ala1 and Ala3 retained binding to PYD at WT levels (Fig. S2B, Fig. S10B, Fig. 4A). Of the positions targeted in Ala2, D159 and D161 are in LRR 5, while F177 is in LRR 6 (Fig. S10A). An analysis of the contact frequency between YopM^PestA* and hPYD* for each interfacial residue is summarized in Fig. S9. Although each YopM^PestA Ala variant contains residues that make substantial contact with hPYD (Fig. S9A,B), YopM^PestA* D159 and D161 also form a dense electrostatic network with hPYD* a3 that involves multiple hydrogen bonds, two salt bridges, and coordination of a structural water (Fig. 3D,E and Fig. S9B). The interactions made by YopM^PestA* F177 are entirely different, with the aromatic side chain becoming shielded from bulk solvent by packing against hPYD* α3 during binding (Fig. 3F). We purified the Ala2 variant of YopM^ΔNΔC (YopM^ΔNΔC Ala2) from E. coli at similar yields as YopM^ΔNΔC. Analysis by SEC showed that YopM^ΔNΔC Ala2 eluted similarly to YopM^ΔNΔC (Fig. S11A,B vs. Fig. S7A,B). As expected, when incubated at an equimolar ratio with purified hPYD and run on SEC, YopM^ΔNΔC Ala2 did not co-elute with hPYD (Fig. S11C,D).

Table 2.

Vectors Name

	Bacterial host strain	Description or insert	Reference
pUT18	BTH101	pUT18 empty vector allowing fusion to the N-terminus of the cya T18 fragment	Euromedex
pKT25	BTH101	pKT25 empty vector allowing fusion to the C-terminus of the cya T25 fragment	Euromedex
pKT25_YopMPestA	BTH101	Full-length YopMPestA CDS	This study
pKT25_YopMPestA Ala1	BTH101	YopMPestA Ala1 (N120A, Y135A, and N140A)	This study
pKT25_YopMPestA Ala2	BTH101	YopMPestA Ala2 (D159A, D161A, and F177A)	This study
pKT25_YopMPestA Ala3	BTH101	YopMPestA Ala3 (Y201A, N221A, and E279A)	This study
pKT25_YopMKIM	BTH101	Full-length YopMKIM CDS	This study
pKT25_YopMKIM Ala2	BTH101	YopMKIM Ala2 (D159A, D161A, and F177A)	This study
pUT18_hPYD	BTH101	Human pyrin aa 1-90	This study
pUT18_hPYD R42W	BTH101	Human pyrin aa 1-90 R42W	This study
pUT18_mPYD	BTH101	Mouse pyrin aa 1-90	This study
pUT18_hPyrin186-326	BTH101	Human pyrin aa 186-326	This study
pKT25_ZIP	BTH101	Leucine zipper	Euromedex
pUT18C_ZIP	BTH101	Leucine zipper	Euromedex
pGEX_2T	BL21	pGEX_2T empty vector allowing fusion to the C-terminus of the GST tag	Chung et al 2016
pGEX_2T_YopMKIM	BL21	Full-length YopMKIM CDS	Chung et al 2016
pACYCDuet_hPYD	BL21	Human pyrin aa 1-90	This study
pACYCDuet_mPYD	BL21	Mouse pyrin aa 1-90	This study
pGEX_KT_YopMPestA	BL21	Full-length YopMPestA CDS in a modified pGEX_KT vector with a 6XHIS and TEV site downstream of the thrombin site	This study
pET_6XHIS_GST_TEV_YopMΔNΔC	BL21	YopMPestA aa 34-344	This study
pET28A_ hPYD _8xHIS	BL21	Human pyrin aa 1-90 in a modified pET28A vector with an 8X HIS tag	This study
pMMB67EH_YopMPestA WT	Y. pseudotuberculosis mJΔyopM	Full-length YopMPestA CDS	This study
pMMB67EH_YopMPestA Ala1	Y. pseudotuberculosis mJΔyopM	YopMPestA Ala1 (N120A, Y135A, and N140A)	This study
pMMB67EH_YopMPestA Ala2	Y. pseudotuberculosis mJΔyopM	YopMPestA Ala2 (D159A, D161A, and F177A)	This study
pMMB67EH_YopMPestA D159A	Y. pseudotuberculosis mJΔyopM	YopMPestA (D159A)	This study
pMMB67EH_YopMPestA D161A	Y. pseudotuberculosis mJΔyopM	YopMPestA (D161A)	This study
pMMB67EH_YopMPestA F177A	Y. pseudotuberculosis mJΔyopM	YopMPestA (F177A)	This study
pMMB67EH_YopMPestA Ala3	Y. pseudotuberculosis mJΔyopM	YopMPestA Ala3 (Y201A, N221A, and E279A)	This study
pMMB67EH_YopMKIM WT	Y. pseudotuberculosis mJΔyopM	Full-length YopMKIM CDS	This study
pMMB67EH_YopMKIM Ala2	Y. pseudotuberculosis mJΔyopM	YopM KIM Ala2 (D159A, D161A, and F177A)	This study

Bacterial expression vectors used in this study are listed by name, bacterial host strain, description or insert and reference to indicate source.

Fig. 4. Analysis of YopM variants by BACTH or THP-1 infection assays.

A,B) The indicated YopM proteins were N-terminally tagged with T25, and hPYDs were C-terminally tagged with T18. Plasmids encoding the fused partners, ZIP positive controls or empty vector (EV) controls were co-transformed into E. coli to test for BACTH interactions as in Fig.1. Significance determined by one-way ANOVA and Tukey’s post hoc test in comparison to PestA (A) or KIM (B) + hPYD. C,D) THP-1 cells were left uninfected (UI) or infected for 4 h at MOI 30 with Y. pseudotuberculosis yopJ^C172AΔyopM complemented with empty vector (EV) or the indicated YopM isoforms or variants. C) Immunoblot analysis of THP-1 lysates with the indicated antibodies. Band intensity signals show the ratio of phospho 242/total pyrin. Results are representative of two independent experiments. D) ELISA of IL-1β released from THP-1 cells. Results are averages and standard deviations with data pooled from at least 3 independent experiments. Significance determined by oneway ANOVA and Tukey’s post hoc test. Comparisons of each group to EV are shown. ****p<0.0001; ***p<0.001; ns, not significant.

We also tested the impact of the Ala2 D159, D161 and F177 substitutions in the 15-LRR YopM^KIM isoform. The sequence alignment of YopM^KIM and YopM^PestA shown in Fig. S12 indicates that all three positions are conserved. The Ala2 variant of YopM^KIM displayed abolished binding to the hPYD in BACTH (Fig. 4B, Fig. S2C).

We next expressed WT or Ala2 variants of YopM^KIM and YopM^PestA in Yersinia for functional studies. A mutant Y. pseudotuberculosis strain (yopJ^C172AΔyopM) was complemented with plasmids producing the different YopM proteins under control of native promoters. The resulting strains were tested by Yop secretion assay, which confirmed that the YopM Ala2 variants were released by the T3SS at levels similar to WT, indicating that the amino-acid substitutions did not interfere with protein folding or stability (Fig. S13A lanes 2-6). THP-1 human monocyte cells were infected with the Y. pseudotuberculosis strains. We analyzed YopM translocation into the cells by immunoblotting, and confirmed that the amino-acid substitutions did not impact injection by the T3SS (Fig. S13B, lanes 1-6). Pyrin inflammasome activation was measured in THP-1 cells infected with the Y. pseudotuberculosis strains using immunoblotting and ELISA. We monitored the phosphorylation level of pyrin at S242 by immunoblotting and found that the WT YopM proteins maintained total pyrin and S242 phosphorylation, whereas in THP-1 cells infected with Y. pseudotuberculosis strains producing the Ala2 variants there was reduced total pyrin and pyrin S242 phosphorylation levels (Fig. 4C, lanes 1-6). As compared to the WT, both YopM^KIM Ala2 and YopM^PestA Ala2 were deficient for pyrin inflammasome inhibition as measured by release of IL-1β from the THP-1 cells (Fig. 4D). Results of colony-forming unit (CFU) assays performed on bacteria before and after infection indicated that the THP-1 cells were not bactericidal (Fig. S14).

We then tested the contribution of individual side-chain changes in YopM^PestA Ala2 on pyrin inflammasome inhibition. We generated Y. pseudotuberculosis yopJ^C172AΔyopM strains producing YopM^PestA variants containing alanine substitutions of each of the three residues individually (D159A, D161A or F177A) in Ala2. The strains were tested for YopM^PestA secretion as above, and normal levels were confirmed (Fig. S13A, lanes 7-9 vs. lane 5). In the THP-1 infection assay, the YopM^PestA variants with single amino-acid changes were translocated at WT levels (Fig. S13B, lanes 7-9 vs lane 5). Mutation of D159 had the strongest impact on the pyrin inhibiting activity of YopM^PestA, as pyrin dephosphorylation and IL-1β release in THP-1 cells infected with this variant was similar to Ala2 or the empty vector control (Fig. 4C, lane 7, and 4D). The D161 variant had an intermediate phenotype, while mutation of F177 did not have a major impact on YopM^PestA pyrin inhibiting activity (Fig. 4C, lanes 8 and 9, and 4D). These data indicate that D159 is essential and D161 is important for YopM^PestA to inhibit the human pyrin inflammasome.

Ala2 variants of YopM^KIM and YopM^PestA Inhibit the Mouse Pyrin Inflammasome

Results of BACTH and GST pull down assays suggested that YopM^PestA and YopM^KIM bind with lower affinity to mPYD than hPYD (Fig. 1B, Fig. S3). Based on this result, we hypothesized that binding to mPYD is not required for YopM to target and inhibit mouse pyrin. Therefore, we tested whether YopM Ala2 variants inhibit the pyrin inflammasome during Y. pseudotuberculosis infection of murine bone marrow derived macrophages (BMDMs). YopM^KIM and YopM^PestA Ala2 variants were translocated into infected BMDMs at WT levels (Fig. S13C). YopM^KIM and YopM^PestA Ala2 variants had only a subtle defect in inhibiting IL-1β release from infected BMDMs, a phenotype that was not statistically significant when compared to WT controls (Fig. S15). Fig. 3D and E and Fig. S9C and D show that in hPYD, R39, R42 and S43 contribute to most of the interactions with YopM^PestA. The alignment of hPYD with mPYD reveals that while R42 is conserved, position R39 is K39 (conservative change) and S43 is G43 (non-conservative change) in mPYD (Fig. S16A). These and other differences in the binding interface residues between hPYD and mPYD (Fig. S16A, B) are likely responsible for the weak interaction of the latter with YopM.

The hPYD R42W variant is defective for YopM binding

Amino-acid changes R39G or R42W in hPYD have been implicated in causing FMF, although neither mutation has been validated to be disease causing (50). FMF is an autoinflammatory disease resulting from gain of function mutations in the pyrin gene MEFV (25). People with FMF experience recurring episodes of fever with inflammation of serosal membranes in the abdomen, lungs and heart (25). Results of population genetic analysis are consistent with the hypothesis that historic plague epidemics in the Mediterranean selected for FMF gain of function codon changes in the B30.2 domain (39). Since hPYD R42 makes substantial contacts with YopM^PestA (Fig. 3D and E) we tested the impact of the R42W codon change in the BACTH. As shown in Fig. 5A, hPYD R42W showed significantly reduced interaction with YopM^PestA and YopM^KIM. Immunoblotting of soluble and insoluble fractions of the E. coli strains used in the BACTH was performed to confirm production of hPYD R42W. The hPYD R42W-T18 fusion protein was present in the insoluble samples at comparable amounts to hPYD-T18 (Fig. 5B, lower panel). Lower amounts of hPYD R42W-T18 was seen in the soluble samples as compared to hPYD-T18, consistent with reduced solubility of the PYD domain in E. coli in the absence of YopM binding (Fig. S3). Thus, results with the non-conservative R42W change confirms that YopM makes a key interaction with this position, which has been implicated in FMF and positive and negative regulation of the hPYD (51).

Fig. 5. Analysis of hPYD variants by BACTH assay.

A) The indicated YopM proteins were N-terminally tagged with T25, and hPYDs were C-terminally tagged with T18. Plasmids encoding the fused partners, ZIP positive controls or empty vector (EV) controls were co-transformed into E. coli to test for BACTH interactions as in Fig.1. Significance determined by one-way ANOVA followed by a Dunnett’s multiple comparisons test in comparison to PestA + hPYD. ***p<0.001; ns, not significant. B) The E. coli strains analyzed in (A) were lysed and separated into soluble and insoluble fractions. Immunoblot analysis of the fractions was performed with anti-T18 antibody. Results are representative of three independent experiments.

Discussion

Here, we aimed to investigate the molecular mechanism of the YopM-pyrin interaction. Previous work showed that Y. pestis YopM binds human pyrin on the N-terminus that includes both the PYD and linker region (39). We found that Y. pestis YopM specifically binds the hPYD, and that the previously identified interaction with part of the linker (39) may have been affected by non-specific binding to GST. Y. pestis YopM binding to the hPYD was demonstrated in vitro using BACTH, GST pulldowns, and purified proteins. We also determined two co-crystal structures of YopM^PestA bound to hPYD and identified residues on both proteins that participate in binding. We show that Y. pestis YopM variants which lose binding to hPYD in vitro can no longer inhibit the pyrin inflammasome in human monocytes infected with Yersinia. Taken together, these data support a model in which Y. pestis YopM hijacks host kinases RSK and PRK, binds to the hPYD as a targeting site and promotes pyrin inactivation by phosphorylation of the linker (Fig. S1C).

The YopM^PestA-hPYD complex is an additional notable structure of a virulence factor targeting a member of the death domain fold (DDF) family of proteins (52). Enteropathogenic Escherichia coli encodes T3SS effector NleB, which attaches N-acetylglucosamine to the death domain of the Fas-associated death domain (FADD) or Tumor necrosis factor receptor type 1-associated death domain (TRADD), preventing interaction with death receptors and assembly of death receptor complexes (52).

The DDF family is comprised of four sub-groups; death domain (DD), death effector domain (DED), caspase activation and recruitment domain (CARD) and pyrin domain (PYD) (53). Members of this family play roles in cytosolic innate immune and cell death pathways (53). All DDF proteins participate in homotypic protein-protein interactions resulting in the assembly of higher-order oligomers (53). DDF proteins have a conserved Greek-key motif comprising of six a-helices with variations in the helices defining the sub-families (53). PYDs stand out in this superfamily because they have the shortest helix 3, which is also preceded by a long unstructured α2-α3 loop (51, 54). Helix 3 and the unstructured α2-α3 loop have been hypothesized to play a role in protein-protein interactions of PYDs (51, 54). Intriguingly, R42, which is located in α3, is predicted to be a key residue involved in YopM binding, alongside R39 (Fig. 3DE), which is located on the α2-α3 loop. R42 was proposed to act as a conformational switch due to the unstructured nature of α3 and the proximal α2-α3 loop (54).

The complex structure presented here reveals a strong charge complementarity between YopM^PestA* and hPYD*. This is not surprising, given that the concave surface of YopM^PestA* is negatively charged (Fig. 3B) and the hPYD* interacts via a positively charged surface. Three negatively charged residues on YopM^PestA* (D159, D161 and D203) were predicted from the complex to contribute to hPYD* binding (Fig. S9B). YopM^PestA* D159 and D161 are located in LRR 5, while D203 is in LRR 7 and was not subjected to detailed analysis in this study. Substitution of YopM^PestA D159 had a strong negative impact on hPYD binding, and this could be because it forms hydrogen bonds with R42 and S43 and is also facilitating hydrogen bonds with one of the structural waters observed at the YopM^PestA*-hPYD* interface (Fig. 3D). Surprisingly, YopM^PestA* D161 makes extensive interactions with hPYD*, forming two salt bridges with hPYD* R39 and R42 (Fig 3E), yet the impact of its mutation was intermediate. This might be because the same hPYD* residues make more interactions with other YopM* residues, hence only dampening rather than abolishing the interface. Substitution of YopM F177 had minimal impact on the inhibitory activity of YopM and is unsurprising given the interaction is restricted to more superficial van der Waals contacts with hPYD* (Fig 3F). Three residues on YopM^PestA* (N120, Y135 and N140) are also predicted to contribute extensive van der Waals interactions and hydrogen bonds yet simultaneously mutating them (Ala1) had no impact on hPYD binding in BACTH. These results could indicate that the overall impact of losing these interactions in the Ala1 variant is minimal, and that YopM relies more on the centrally located acidic residues targeted in Ala2.

Our results indicate that most of the concave surface of the YopM LRR domain is sterically occluded when bound to hPYD. YopM LRRs 6-15 have been implicated in binding to PRK (40). If PRK also binds to the concave surface of the YopM LRR domain, binding to the PRK kinase domain and hPYD would be mutually exclusive. As shown in Fig. S1C, PRK may need to be released to allow YopM to target the hPYD and use hijacked RSK to phosphorylate the human pyrin linker. Additional studies are needed to understand if binding of PRK and hPYD to YopM are mutually exclusive.

PYDs are known to interact through electrostatic interactions (51, 55). Critical to assembly of the pyrin inflammasome is an interaction between the pyrin and ASC PYDs (51) which then allows a recruitment of pro-caspase-1 (56). The two PYDs interact using complementary positive and negatively charged sites (51). Additionally, because these PYDs have multiple charged surfaces, they can simultaneously self-associate and participate in inter-molecular interactions, leading to the assembly of a multi-protein complex (51). Vajjhala et al. identified three and two binding sites on pyrin and ASC PYDs, respectively (51). R42 of pyrin PYD and R41 of ASC PYD, which are in α3, and in Site 2 for each protein, were identified to be important for binding to each other (51). The finding that YopM specifically binds to the positively charged patch of hPYD containing R42 and representing Site 2 is intriguing, given that it overlaps with an ASC binding site and the B-box binding site (51). This might indicate that YopM evolved to target a specific binding region on hPYD essential for its function. Additionally, YopM may have to compete with ASC and the B-box to bind hPYD. Future studies that define hPYD interactions with the B-box, ASC and YopM will be informative in testing this hypothesis.

The hPYD R39G and R42W variants have been implicated in FMF, and we found that the latter codon change significantly reduced YopM binding (Fig. 5). The impact of the R42W substitution on hPYD structure and function has been studied, and it was found to increase the stability of the short helix α3 (51). While counterintuitive, this increased structural stability might be deleterious to the function of hPYD, due to the importance of flexibility in this region. The R42W substitution appeared to cause a greater reduction in binding of the B-box to hPYD as compared to ASC (51). The R42W substitution may reduce negative regulation by the B-box (31) which in turn could increase activation of pyrin (51). Recently, Bronnec et al. carried out a high-throughput analysis of MEFV variants and reported that neither R39G nor R42W could be validated as pathogenic (57). If the the R42W substitution was selected to reduce YopM binding to hPYD, it is unlikely to have gain of function activity as an added benefit. Most pathogenic FMF codon changes are located on the C-terminal B30.2 domain (25, 50). FMF amino-acid acid changes in the B30.2 domain may overcome YopM inhibition by decreasing the activation threshold of the pyrin inflammasome (58). A second class of gain of function MEFV mutations are those that prevent phosphorylation of the linker, such as S242R, resulting in the disease pyrin-associated autoinflammation with neutrophilic dermatosis (59). It is possible that by targeting the key R42-containing site of positive regulation in hPYD YopM forced MEFV to evolve gain of function mutations elsewhere, i.e. in the linker or B30.2 domain to counteract plague infection. Additional experiments are needed to understand how the R42W and R39G substitutions in hPYD impact YopM and B-box binding, pyrin activation and Yersinia pathogenesis.

The finding that Y. pestis YopM does not strongly bind to mPYD could mean that this domain has evolved reduced affinity for recognition. This is possible, given that rodents have likely been hosts for Y. pestis longer than humans. Human and mouse PYDs have a sequence identity of 76% and aminoacid differences between the two could indicate sites of positive selection on mPYD. hPYD R39, R42, and S43 have the most interactions with YopM, and two of these sites have conservative (R39 to K39) or non-conservative (S43 to G43) changes in mPYD. These substitutions could be sufficient to disrupt YopM binding. The R47M change in mPYD could also interfere with the band of positively charged surface residues seen in hPYD. Given that there are only three positively charged residues in the YopM binding region (R39, R42 and R47), the loss of one of them might have a disproportionate impact, especially because the PYD is small. Additionally, because the three substituted residues are located on α3, they might cause structural changes that decrease recognition by YopM. These three residues might be the hotspots of mPYD evolution, but a more detailed analysis of positive selection in mouse pyrin would be informative in testing this hypothesis. There is evidence that a small number of residue substitutions can have a significant impact on the interactions of host and pathogen proteins (60). These residues function as hotspots in the evolutionary host-pathogen arms race (60).

Despite the lack of stand-alone mPYD binding, we found that Y. pestis YopM Ala2 variants retain the ability to inhibit the pyrin inflammasome in infected mouse macrophages. These results suggest that YopM targets and inhibits murine pyrin independent of the PYD. PRK was shown to be important for YopM to inhibit the pyrin inflammasome in mouse macrophages infected with Y. pseudotuberculosis (15). PRK uses its kinase domain to target and phosphorylate pyrin (33, 39). One possibility is that YopM uses hijacked PRK to target and phosphorylate murine pyrin (Fig. S1C). Further studies that focus on how YopM targets and inhibits mouse pyrin are needed to better understand differences in this key mechanism of pathogenesis in murine vs human cells.

MATERIALS AND METHODS

Protein Expression Vectors

The vectors and bacterial host strains used are provided in Table 2. In brief, YopM^PestA constructs were cloned as N-terminal hexahistidine-GST (6xHisGST) fusions. In addition to the full-length coding sequence (6xHisGST-YopM^PestA) that was cloned into a modified pGEX_KT vector, a truncated construct lacking the disordered N- and C-terminal extensions (residues 34-344, YopM^ΔNΔC) was cloned into a pET28-derived expression vector. These extensions were excluded from the final construct to mitigate potential interference during crystallization trials. The CDS of the human PYD (residues 1-90, herein referred to as hPYD) was cloned as either an N-terminal hexahistidine fusion in a pDuet vector, or as a C-terminal octahistidine fusion in a pET28a vector. All inserts were first PCR amplified then inserted into their respective vectors by either restriction cloning using BamHI and EcoRI sites (pGEX_KT YopM^PestA), EcoRI and HindIII (pACYCDuet_hPYD) or using the NEB Gibson Assembly kit according to the manufacturer’s recommendations.

Bacterial Two Hybrid Assay

PYDs and pyrin linker constructs were expressed from pUT18 with C-terminal fusions of T18, and YopM isoforms and variants were expressed from pKT25 with N-terminal fusions to T25. A Quickchange kit (Agilent) was used to introduce codon changes into the BACTH vectors, followed by sequence validation. The BACTH assay was performed as previously described (61, 62). Briefly, electrocompetent BTH101 E. coli cells were co-transformed with each vector or empty vector controls and recovered in 250 μl SOC media at 37°C. Cells were plated on an Ampicillin (Amp) or Carbencillin (Carb) + Kanamycin (Kan) plate for 2 days at 30°C and resulting colonies streaked onto Amp/Carb + Kan + isopropyl β-D-1-thiogalactopyranoside (IPTG) plates with and without X-Gal. After 1 day at 30°C, the cells on plates without X-Gal were scraped off and resuspended in Phosphate Buffered Saline (PBS; 140 mM NaCl, 2.68 mM KCl, 8.1 mM Na₂HPO₄, 1.47 mM KH₂PO₄) and normalized to 0.D600 ~0.5. The cells were then incubated at 30°C for 5 min for lysis with 25 μl of 0.1% SDS, 50 μl chloroform in Z buffer (60 mM Na₂HPO₄, 40 mM NaH₂PO₄, 10 mM KCl, 1 mM MgSO₄) with 50 mM β-mercaptoethanol (β-ME). 200 μl of 4 mg/mL ortho-nitrophenyl-β-D-galactopyranoside (ONPG) in Z buffer was added to the cells and the reaction performed for 5 minutes at 30°C. A 500 μl solution of 1 M sodium bicarbonate was added to stop the reaction and the absorbance was measured at 420 and 550 nm. The β-galactosidase activity was then calculated using the following formula as described by Miller:

$$\text{Miller Units}=1000*[({\mathrm{OD}}_{420}-{1.75}^{*}{\mathrm{OD}}_{550})/({\mathrm{time\; in\; minutes}}^{*}{{\mathrm{OD}}_{600}}^{*}\mathrm{volume\; of\; scraped\; cultures})]$$

For immunoblotting bacteria recovered from plates as above were lysed with Bugbuster Protein Extraction Reagent (Sigma-Aldrich) and separated into soluble and insoluble fractions by centrifugation. Immunoblotting of soluble and insoluble samples was performed as described below using Cya A (3D1) mouse monoclonal (1:10,000, Santa Cruz Biotechnology).

GST Pulldown Assay

GST-YopM and His-tagged murine or human PYDs were co-expressed in E. coli BL21 cells. The bacterial lysates were prepared with BugBuster^® Protein Extraction Reagent (Millipore) in PBS and supplemented with lysonase (Millipore) and Roche complete mini protease inhibitor. Lysates were then incubated with GST resin (GST.Bind, Millipore) overnight. The resin was then washed and proteins eluted using 1X sample buffer and analyzed on SDS-PAGE gels followed by blue staining or α-His immunoblotting. Immunoblotting procedures are described below.

Protein Expression and Purification

All recombinant proteins used in this manuscript were expressed in BL21(DE3) cells. Proteins were either co-expressed and co-purified for crystallization of the YopM^PestA*-hPYD* structure or individually prepared for the YopM^ΔNΔC-hPYD complex. In all preparations, cells were grown in Terrific Broth (TB) at 37°C for 2-3 hours until an OD₆₀₀ of ~1 was achieved. Cells were induced with 0.1 mM IPTG and protein expression was allowed to proceed at 18°C for 16-18 h. For YopM, induced cells were allowed to express protein at 37°C for 4 h. Cells were harvested by centrifugation at 4,200 x g at 4°C for 15 min. Cells were resuspended in lysis buffer (150 mM NaCl, 1 mM Tris(2-carboxyethyl)phosphine hydrochloride [TCEP], and 20 mM Tris, pH 7.4) supplemented with Roche complete mini EDTA (ethylenediaminetetraacetic acid) protease inhibitor and Lysonase Bioprocessing Reagent from Millipore Sigma and lysed using an LM10 microfluidizer. The insoluble fraction was separated by centrifugation using a Ti 45 Rotor in a Beckman Ultima L-70 ultracentrifuge at 40,000 rpm for 1 h at 4°C. For both the YopM^PestA only construct and YopM^PestA co-expressed with hPYD, recombinant protein was captured on GST resin and eluted by on-column proteolysis using the Tobacco Etch Virus (TEV) protease in buffer composed of 150 mM NaCl, 1 mM TCEP, and 20 mM Tris, pH 7.4. For YopM co-purified with hPYD, the hexahistidine tag on hPYD was cleaved using the HRV-3C protease during dialysis against buffer containing 150 mM NaCl, 1 mM TCEP, and 20 mM Tris, pH 7.4; cleavage and dialysis were allowed to proceed overnight at 4°C. In both cases, recombinant protein was further purified by size exclusion chromatography (SEC) using a HiLoad Superdex 200 26/600 PG column and fractions containing purified protein, as assessed by reducing SDS-PAGE, were pooled and concentrated.

For the YopM^ΔNΔC-hPYD complex structure, proteins were individually expressed and purified. The bacterial pellets were expressed, lysed, and clarified as described above except using a 20 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), pH 7.4 buffer system in place of Tris. Clarified lysates were applied to a 5 mL His-TRAP column then washed with buffer composed of 500 mM NaCl and 20 mM Tris, pH 8.0 supplemented with imidazole concentrations ranging from 10-50 mM. Proteins were eluted by gradient elution ranging from 10-250 mM imidazole. For the hPYD preparation, fractions containing protein were pooled, concentrated and run on a HiLoad Superdex 200 26/600 PG column in buffer composed of 150 mM NaCl, 1mM TCEP, and 20 mM HEPES, pH 7.4. For YopM^ΔNΔC, fractions containing recombinant protein were pooled and TEV cleavage reactions were set up during simultaneous dialysis against buffer composed of 150 mM NaCl, 1 mM TCEP, and 20 mM HEPES, pH 7.4 at 4°C, for 16-18 h. Cleaved protein was further purified by cation exchange using SP Sepharose and IMAC using Ni-NTA resin. Finally, samples were purified by SEC using a HiLoad Superdex 200 26/600 PG column.

Analytical SEC and SEC-MALS

Analytical SEC was performed on a Superdex 200 10/300 column attached to an ÄKTA pure^™ protein purification system (Cytiva). Samples were analyzed in 150 mM NaCl, 1 mM TCEP, and 20 mM Tris, pH 7.4 for GST resin purified YopM^PestA and YopM^PestA-hPYD. YopM^ΔNΔC and YopM^ΔNΔC-hPYD complex were analyzed in 150 mM NaCl, 1 mM TCEP, and 20 mM HEPES, pH 7.4. Analytical SEC was performed at 0.75 mL/min. The column was calibrated using the 29,000-700,000 Da gel filtration markers kit from Millipore Sigma (MWG1000F) containing the following molecular mass standards: Blue Dextran (2,000 kDa), Apoferritin (440 kDa), β-Amylase (200 kDa), Alcohol Dehydrogenase (150 kDa), Albumin (66 kDa), Carbonic Anhydrase (29 kDa).

For SEC-MALS, an ÄKTA pure^™ instrument (Cytiva) coupled to an Optilab refractive index detector (Wyatt) and light-scattering detector (Wyatt MiniDawn Tristar) were used. SEC was performed using a Superose 10/300 Column at 0.5 mL/min and the laser wavelength was set to 690 nm. The molar mass was determined using the ASTRA software package (Wyatt).

Velocity Ultracentrifugation

Analytical ultracentrifugation was conducted using a Beckman Proteomelab XL-A and an AN-60 rotor. For sedimentation velocity analytical ultracentrifugation, YopM^PestA in 150 mM NaCl, 1 mM TCEP, with either 20 mM Tris, pH 7.4 or 20 mM MES pH 5.0 was centrifuged at 35,000 rpm with monitoring at 280 nm. Data were analyzed using Sedfit (63) to determine sedimentation coefficient, frictional ratio, and apparent mass, as described elsewhere (64). The sedimentation coefficient reported is that of the major peak (at least 80% of the total analyzed mass) at OD₂₈₀ for each sample.

Crystallization

The YopM^PestA*-hPYD* complex was crystallized using the co-purified YopM-hPYD. The complex was crystallized by vapor diffusion in 4 μL sitting drops composed of well solution (12% [w/v] PEG 4K, 40 mM MES, pH 5.0) and co-purified YopM^PestA-hPYD at a concentration of 6.5 mg/mL as determined by the Bradford assay; drops were set at a 1:1 ratio. Prior to harvesting, crystals were cryoprotected in well solution supplemented with 20% (v/v) glycerol then flash cooled in liquid nitrogen.

For co-crystallization of the YopM^ΔNΔC-hPYD complex, the purified YopM^ΔNΔC was first combined with purified hPYD in a 1:4 molar ratio at a final total protein concentration of 1.2 mg/mL. Suitable crystals were obtained by vapor diffusion in 4 μL hanging drops composed of well solution (19% [w/v] PEG 3350, 300 mM NaSCN, and 100 mM Bis-Tris Propane, pH 6.5) and protein set at a 1:1 drop ratio. Crystals were harvested and flash cooled as described above.

Structure Determination and Refinement

For both co-crystal structures, data were collected at the National Synchrotron Light Source II on the AMX beamlines (17-ID-1) using a Dectris 9M pixel detector. For the full-length co-purified YopM^PestA*-hPYD* complex, 180° of rotation X-ray diffraction data were collected using 0.1° per frame. For the YopM^ΔNΔC-hPYD complex, 120° of rotation X-ray diffraction data were collected using 0.2° per frame. The diffraction images were reduced using XDS and scaled using XSCALE (65). The R_free set was generated from 5% of reflections in thin resolution shells using Phenix reflection file editor. The Matthews coefficient was determined using the CCP4 supported program MATTHEWS_COEF (66). In both cases, initial phases were determined by molecular replacement using PHASER-MR and iterative automated and manual refinements were carried out using phenix.refine and Coot, respectively. Specific details pertaining to the phasing and refinement of each structure follows.

For the YopM^PestA*-hPYD* complex, initial phase estimates were obtained using the crystal structure of NLRP9 PYD (pdb_00006z2g) as a search model. For the 13-LRR YopM^PestA isoform used in these studies, a composite model was initially created using the crystal structure of a related 15-LRR Y. pestis YopM isoform (pdb_00001g9u) in PyMol (67) to omit additional leucine rich repeats in the central LRR domain. Following molecular replacement, negative density peaks were observed in the mFo - dFc map along the entirety of the C-terminal portion of YopM. After omission of the region spanning residues 248-344 followed by iterative automated and manual refinements using phenix.refine and Coot, respectively, it was determined that a YopM^PestA truncation was captured in the crystal lattice. We posit a limited proteolysis event occurred in the mother liquor preceding nucleation that resulted in a shortened YopM^PestA in the co-purified YopM^PestA-hPYD construct represented in the deposited coordinates. This conclusion is supported by the inability of the missing YopM^PestA sequence to be included in either molecule of the asymmetric unit without incurring severe steric overlap with molecules related by both crystallographic and non-crystallographic symmetry. A region spanning residues 79-90 and encompassing the last α-helix of hPYD is also absent in this crystal structure and is presumed to be another target of errant proteolytic activity. The stable complex fragments resolved in this co-crystal structure are denoted as YopM^PestA*-hPYD*. The CCP4 supported programs PEAKMAX and WATPEAK (68) were used to identify unmodeled positive density peaks and evaluate them for waters, which were modeled as appropriate following manual inspection. Additionally, a > 4 sigma positive density peak adjacent to the sulfhydryl group of Cys100 of Chain B was observed that exhibited good fitment to a carboxymethyl cysteine. However, LC-MS/MS analysis of crystals rehydrated from neighboring drops with similar well-solution compositions failed to corroborate the observations in the crystallographic data despite evidence for low-abundance carboxymethylation at other cysteines. The positive density peak was therefore left unmodeled. Anisotropic refinement of protein atoms was carried out using one TLS group per chain; subsequent refinements transpired without incident.

For the co-crystal structure of the YopM^ΔNΔC-hPYD complex, initial phase estimates were obtained using a model composed of the refined YopM^PestA*-hPYD* structure described above with a graft of the C-terminal portion of the aforementioned Y. pestis YopM isoform (pdb_00001g9u, residues 286-396). The graft was created by performing a least-squares superposition of the preceding two LRRs with the corresponding region of the refined YopM^PestA*-hPYD* complex. Due to the low resolution and map quality, automated refinements were performed using secondary structure restraints. The refined coordinates of the YopM^PestA*-hPYD* complex were also used to impose reference model restraints during automated refinement but were lifted for the final stages.

Accession Numbers

The coordinate files and structure factor files for both co-crystal structures solved herein are available at the Protein Data Bank (69) using the following codes: pdb_00009oyu (YopM^PestA*-hPYD*) and pdb_000010dr (YopM^ΔNΔC-hPYD). Rotation X-ray diffraction image files were uploaded to the SBGrid Data Bank and are searchable by PDB codes.

Structural Analyses and Visualization

Structural analyses were performed using PyMol (67) and the CCP4 Suite (68) as part of the SBGrid software environment (70). Structural alignments were performed in PyMol using all C_α of the YopM chains for each complex. The Adaptive Poisson-Boltzmann Solver (71) (APBS) plugin was used to calculate electrostatic surface potentials for YopM^PestA* and hPYD* in PyMol. All structural figures were rendered using PyMol. Residues at the binding interface were identified using the CCP4 supported program CONTACTS (68) with a 5 Å cutoff, and histograms displaying per-residue contacts were created using R (72) with the following packages: tidyverse (73), ggprsim (74), and ggplot2 (75).

Construction of Y. pseudotuberculosis Strains

Y. pseudotuberculosis strains expressing different YopM isoforms and variants (see Table 2) were generated by conjugating donor E. coli S17 lambda-pir containing the pMMB67EH plasmid plus YopM isoforms/variants with a Y. pseudotuberculosis strain lacking the yopM gene and harboring the catalytic mutation in YopJ (mJΔyopM).

Yop Secretion Assay

Y. pseudotuberculosis yopJ^C172AΔyopM strains were inoculated into LB Amp and grown with shaking at 28°C. Overnight cultures of Yersinia were diluted 1:40 in LB media containing 20 mM sodium oxalate and 20 mM MgCl₂ and grown for 2 h at 28°C then shifted to 37°C for 4 h. Supernatants were precipitated using trichloroacetic acid (TCA) overnight. Precipitated proteins were centrifuged at 14,000 rpm for 30 min, and the resulting pellet washed with acetone. The protein pellets were resuspended in 100 μl 1X NuPAGE LDS Sample Buffer and Sample Reducing Agent and 10 μl of the samples were run on 4–12% NuPAGE Bis-Tris SDS-PAGE gels (Invitrogen by ThermoFisher Scientific). The gels were stained using GelCode Blue Stain Reagent (ThermoFisher) or One-Step Blue Protein Gel Stain (Biotium) and imaged on an iBright FL1500 imager (ThermoFisher Scientific).

THP-1 Cells Tissue Culture and Infections

THP-1 human monocyte cells (ATCC TIB-202) were grown in filter-sterilized RPMI-1640 media with 10% heat-inactivated fetal bovine serum (FBS), 0.05 mM β-ME and 1% pen-strep (10,000 U/mL Penicillin + 10,000 μg/mL streptomycin, Gibco) in 25 cm² flasks. Once the cells reached confluency, they were split into 75 cm² flasks. For infections, 8x10⁵ total cells per well were seeded in 1.5 mL serum-free media in 6-well plates. Overnight cultures of Yersinia were diluted 1:40 in LB media containing 20 mM sodium oxalate and 20 mM MgCl₂ and grown for 1 h at 28°C, then shifted to 37°C for 2 h. Cultures were centrifuged and the bacterial pellets washed in 1X PBS. Resuspended bacteria were then mixed with cells at a multiplicity of infection (MOI) of 30 in serum-free media then plated on 6-well plates. After infection for 4 h at 37°C in 5% CO₂, the cell contents were collected and centrifuged for 15 min at 14,000 rpm. The supernatants were collected for IL-1β analysis by ELISA (see below). The cell pellet was then washed with 1X Hanks balanced salt solution (HBSS) and lysed for immunoblot analysis as described below. Samples of bacteria inoculated into each well at t=0 h were serially diluted and plated for CFU quantification. Total bacterial-monocyte co-cultures were solubilized in 0.1% NP-40 detergent (t=4 h) followed by serial dilutions and plating to determine CFU recovery post-infection.

BMDM Cells Tissue Culture and Infections

BMDMs were obtained from femurs and tibia of 8-week-old WT C57BL/6 mice (Jackson Laboratories) and cultured as previously described (76). After differentiation for 7 days, the BMDMs were seeded in 3 mL MGM 10/10 containing Dulbecco’s modified Eagle Medium (DMEM) plus GlutaMAX (Gibco), 10% (v/v) heat-inactivated FBS (GE HealthCare), 10% (v/v) L929 cell-conditioned medium, 1 mM sodium pyruvate (Gibco), and 10 mM HEPES (Gibco) at a density of 0.8x10⁶ cells/well. The cells were allowed to settle for at least 1 h then primed overnight by addition of 100 ng/mL O26:B6 E. coli LPS (Sigma) in 1X PBS. Yersinia strains were grown as described above and added to serum-free MGM 10/10. MGM 10/10 with or without bacteria at an MOI of 30 was added to the wells of primed macrophages. Plates were centrifuged for 5 min at 112 x g to facilitate Yersinia contact with cells and incubated for 90 min at 37°C in 5% CO₂. Following the infection, cell supernatants were collected for IL-1β analysis by ELISA and cells lysed for immunoblot analysis (see below).

THP-1 and BMDM Lysates and Immunoblotting

THP-1 and BMDM cells were lysed in mammalian protein extraction reagent (MPER, Thermo Scientific) supplemented with protease inhibitor cocktail and phosphatase inhibitor (Roche). The lysates were spun down to separate out soluble from insoluble fractions. Protein concentration levels were measured using the bicinchoninic acid assay (BCA, Thermo Fisher Scientific). 5-10 μg total soluble protein lysate was run on 4-12% NuPAGE Bis-Tris SDS-PAGE gels (Invitrogen by ThermoFisher Scientific) and transferred to polyvinylidene difluoride membranes (ThermoFisher Scientific) using an iBlot 3 gel transfer device (Life Technologies). Membranes were blocked using 5% (w/v) nonfat dairy milk (Lab Scientific bioKEMIX) for 1 h at room temperature before incubation with primary antibody overnight. Primary antibodies used in this study include rabbit anti-mouse phospho-serine 241 monoclonal antibody (1:1,000 dilution, Ab200420 ; Abcam), rabbit anti-human pyrin monoclonal (1:1,000, Cell Signaling #58525 and #40649), rabbit-anti-mouse/human polyclonal β-actin (1:1,000 # 4967; Cell Signaling), rabbit anti-Y. pestis YopM polyclonal (1:10,000 dilution, antibody provided by Susan Straley) and mouse anti-HIS tag monoclonal (1:5,000 dilution, # MMS-156P BioLegend). Horseradish peroxidase-conjugated anti-rabbit or anti-mouse (Jackson Laboratory) were used as secondary antibodies. Protein bands bound by antibodies were visualized using chemiluminescent detection reagent (GE HealthCare) on an iBright FL1500 imager (ThermoFisher Scientific). Band quantification was also performed on the iBright FL1500 imager.

IL-1β Quantification

Levels of IL-1β in THP-1 and BMDM supernatants were quantified using human (RandD Systems) and mouse (MLB00C, R&D Systems) ELISA kits respectively, following the manufacturer’s instructions.

Statistical Analysis

GraphPad Prism was used to perform statistical analyses. Significance was determined by one-way ANOVA and Tukey’s post hoc tests performed for BACTH and ELISA experiments.

SIGNIFICANCE.

Pyrin is an inflammasome sensor encoded by the MEFV gene that is notable for its role in the autoinflammatory disease Familial Mediterranean Fever (FMF) and immunity to the plague agent Yersinia pestis. Y. pestis normally prevents inflammasome activation using the virulence factor YopM which binds pyrin and inhibits effector-triggered immunity. Gain of function mutations in MEFV that cause FMF were likely selected during historic plague pandemics to counteract YopM. Here, we defined the molecular mechanism of the YopM-pyrin interaction, which revealed that the effector binds to a key site of positive and negative regulation on the N-terminal death domain. These findings have important implications for understanding how YopM promotes pathogenesis and likely selected for gain of function mutations in MEFV.