Inhibition of NLRP3 inflammasome activation by cell-permeable stapled peptides

Interleukin-1β (IL-1β) is a major cytokine that initiates and enhances inflammatory responses. Excessive IL-1β production is a characteristic of most chronic inflammatory diseases, including atherosclerosis, type 2 diabetes, and obesity, which affect a large proportion of the global population. The production of bioactive IL-1β is mediated by a caspase-1-activating complex known as an ‘inflammasome’. The NLRP3 inflammasome has been associated with several human inflammatory and autoimmune diseases and represents a potential therapeutic target for disrupting IL-1β production. We used molecular modeling guided by molecular dynamics simulations to design α-helical stapled peptides targeting the pyrin domain of the adaptor protein ASC to interrupt the development of its filament, which is crucial for NLRP3 inflammasome formation. The peptides were effectively internalized by human monocytic cells and efficiently suppressed the release of the inflammasome-regulated cytokines IL-1β and IL-18, following exogenous activation of the NLRP3 inflammasome. The peptides reduced ASC speck formation and caspase-1 processing thereby suppressing pro-IL-1β processing and release of active IL-1β. This is the first demonstration of the successful use of stapled peptides designed to target the adaptor protein ASC, and can be extended to other inflammatory pathways to disrupt excessive IL-1β production.

Docking simulations confirm that the designed peptide binds to the target binding site. To explore whether charge complementation between the peptide (cationic) and the ASC PYD binding surface (type-Ia, anionic, Fig. 2A) was sufficient for site-specific binding of the peptides, we performed a docking search using simulations of coarse-grained models of the protein and peptide where their interaction was modeled solely by electrostatic interactions between the charged residues. Coarse-grained simulations access time scales that are long enough to achieve the sampling of the binding process, which are not yet possible by all-atom simulations. Simulations were started by placing the representative peptide (ms_ASC PYD #5, Table 1) ~35 Å away from the protein (See Methods). As the numbers and the positions of the positively charged residues were similar in all the peptides, we assumed that all other peptides would show similar binding behavior in the coarse-grained simulations. The peptides congregated at the ASC PYD type-Ia binding surface, and their conformations matched the all-atom initial model (Fig. 2B), thus demonstrating that electrostatics is indeed the primary driver of recognition and the designed peptides should successfully target the ASC PYD type-Ia interface. Furthermore, docking simulations between ASC CARD and ms_ASC PYD #5 clearly showed that the peptide does not bind to ASC CARD .
Design, mutation, and characterization of I, I + 7 hydrocarbon-stapled peptide inhibitors. The conformation of helix 2 ( 16 TAEELKKFKLKLLSV 30 ) bound to ASC PYD (Fig. 1A) was used as the starting structure for the peptide-protein interaction modeling. Following I,I + 7 stapling scheme, we inserted a hydrocarbon staple across positions 5 and 12 along the peptide and then introduced mutations to produce a set of stapled peptides. Two variants of the template peptide were designed -a short version containing 13 residues (ms_ASC PYD #6) and a long version containing 15 residues (ms_ASC PYD #1-5) with two additional residues at the N-terminus The type-I dimeric interface formed by two ASC PYD monomers is shown, where the negatively charged surface (Type-Ib interface, shown in surface representation) of one monomer interacts with the two positively charged helices 2 and 3 (Type-Ib interface) from the other monomer. Interface residues of the first monomer are marked and colored according to their charges (red = negative, blue = positive, and yellow = hydrophobic). For the second monomer, helices 2 and 3 are in green and sticks depict the interacting residues. (B) Sequence and structure of helix 2, the template peptide (wt_ASC PYD ) used for designing the other peptides is shown in green. Interacting residues are depicted as sticks and are marked. (C) Three of the designed stapled peptides are shown with their sequences labeled. The interacting residues are depicted as sticks. The mutation sites are shown in orange. The staple positions are marked as '*' in the sequence, and the 'hydrocarbon staple' is depicted as orange sticks.
www.nature.com/scientificreports www.nature.com/scientificreports/ (Table 1). These two additional residues were inserted to increase the length of the peptide, thus potentially enabling sufficient conformational flexibility for the region involving the critical Arg3 to remain helical. A similar approach was taken to introduce the staple in the ms_NLRP3 PYD peptide. As determined from Circular Dichroism (CD) spectra, inserting the I,I + 7 staple had the intended effect of increasing helicity of the peptides (Fig. 3). The wild-type unstapled NLRP3 PYD peptide showed no helicity, which was increased to 48% when stapling was introduced. Percentage of helicity was much higher for ASC PYD stapled peptides #5 and #6, 95% and 70% respectively at 100 μM peptide concentration. Percentage of helicity for the peptides were estimated from the web server K2D3 36 using CD data.
To optimize peptide stability and binding, we introduced up to six further mutations into the wild-type sequence (Table 1). Binding energy contributions for the type-Ib interface residues was calculated from simulations of the ASC PYD :ASC PYD type-I complex structure (prepared from 3J63.pdb), which identified the most appropriate residue positions to be mutated (Table S1). At the N terminus, we mutated Glu3 and Glu4 to Arg and Ala, respectively to gain a positive charge and further stabilize the electrostatic interactions and van der Waals contacts of the peptide with ASC PYD (Fig. 4 and Table S2). The Phe8 side chain faces away from ASC PYD and does not interact with the protein; it was mutated to Ala to reduce thermodynamic penalty due to solvating the hydrophobic side chain. We mutated the Leu10 side chain (that inserts into a hydrophobic pocket on ASC PYD ) to Phe or Ile to increase the hydrophobic packing (Figs 4 and 5). We also mutated Lys7 to Leu in ms_ASC PYD #5, to improve packing against the residues of the hydrophobic pocket. At the C-terminus, we mutated Leu13 to Glu to introduce polar interactions with Arg5 of ASC PYD . Although Lys9 faces away from the interface and does not interact with the protein, we maintained this site to gain an extra positive charge, which should aid cell permeability. Using a similar approach, we mutated two positions in the ms_NLRP3 PYD peptide to improve the binding energy (Table 1). We noted that all the ASC PYD stapled peptides showed more favorable binding energies compared to the NLRP3 PYD stapled peptide, possibly due to less helicity and a missing positive charge at position 11 for the later peptide.
Validation of peptide binding stability by molecular dynamics (MD) simulations. All-atom MD simulations of the stapled peptide:ASC PYD complexes suggested that all peptides bound similarly to the interactions seen in the type-I interface of the ASC PYD filament (Figs 1A and 4). The Arg1, Lys4, Lys5, Ile8, Lys9, Ser12, and Val13 residues of the short peptide ms_ASC PYD #6 (corresponding residues from the ms_ASC PYD #5 peptide are Arg3, Lys6, Leu7, Ile10, Lys11, Ser14, and Val15) made a significant energetic contribution to their interactions with ASC PYD (Fig. 4 and Table S2). Arg1/Arg3 is involved in stable salt-bridge interactions with both Glu13 t (t denotes residues from the target protein ASC PYD ) and Asp48 t (Fig. S1), and in van der Waals interactions with Leu50 t , thus contributing the highest binding energy among all the peptide residues. This effect helps anchor the peptide N-terminal to the protein surface. Lys4/Lys6 (corresponds to ms_ASC PYD #6 and #5, respectively) forms less stable salt-bridges with one of the negative side chains of Asp51 t or Asp54 t from helix 4, compared to a more stable salt-bridge between Lys9/Lys11 and Asp6 t of helix 1 (Fig. S1). In the case of the shorter peptide ms_ASC PYD #6, an additional electrostatic interaction was observed between Lys5 and Glu13 t (Fig. S1). At the C-terminus, non-specific hydrogen bonds were observed in the simulations between the main-chain of Ser12/Ser14 or Val13/ Val15 and the main-chain of Met1 t , Gly2 t or Arg5 t . Another major binding energy contribution comes from Ile8/ Ile10 where it engages in van der Waals contacts with a hydrophobic pocket formed on the ASC PYD surface. This pocket was observed at the middle of the type-I interface in the ASC PYD filament structure (Fig. 5A) where it is System Peptide sequence 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 ΔGcalc. (kcal/mol) ASC PYD -ASC PYD complex -−35.1 ± 11.8 Peptide template: ASC PYD helix 2 Peptide template: NLRP3 PYD helix 2 www.nature.com/scientificreports www.nature.com/scientificreports/ formed by the solvent-exposed side chains of Leu9 and Leu50 from the type-Ia interface into which Leu25 from the type-Ib interface interacts (Fig. 1A).
To investigate whether this hydrophobic binding pocket on the ASC PYD surface is flexible and can accommodate larger hydrophobic side-chains such as Ile or Phe, an in-house 'ligand-mapping' approach was used where benzene probes were used to tease out cryptic regions of the protein, as previously described 37 . The resulting MD trajectories of the ASC PYD monomer structure showed high benzene occupancy in this pocket, indicating that this hydrophobic pocket is exposed in the absence of its binding partner (Fig. 5B). Indeed, the Ile8/Ile10 side chain was well placed (Phe10 in the case of the peptide ms_ASC PYD #4) in the area occupied by the benzenes (Fig. 5C,D); however, the pocket was not large enough to accommodate bigger side chains such as Trp. In the designed peptides, Ile8/Ile10 (in ms_ASC PYD #6/ ms_ASC PYD #5), and Phe10 (in ms_ASC PYD #2, #3 and #4) are located at the center of this pocket and make multiple van der Waals contacts with the side chains of Arg5 t , Leu9 t  1A) to judge their binding accuracy. The similarity measure (D 1 and D 2 ) quantifies the structural similarity between the predicted peptide-protein binding interface in the docking simulation and the type-I ASC PYD interface structure, where a lower value of D 1 or D 2 corresponds to a higher degree of similarity (see Methods). The colored bar indicates the free energy of different binding conformations of the predicted peptide-protein complex. Representative docked peptide conformations (green helix) from three regions of the map are shown in the right panel. ASC PYD is shown in gray. The position of helix 2 (blue helix) in the type-I interface is also shown to judge the binding accuracy of the docked peptide. Close positioning of the green helix to the blue helix represents a predicted conformation (top) similar to the experimental structure. Furthermore, the highest population (blue regions) is found closer to min(D 1 , D 2 ) on the map which indicates that peptides docked at the correct binding site in the type-I interface. . Circular dichroism (CD) spectra of (A) NLRP3 PYD and (B) ASC PYD peptides to measure secondary structures. CD spectra were measured at room temperature in water at a peptide concentration of 100 μM. ms_ ASC PYD #5 and ms_ASC PYD #6 showed typical wavelength minima for α-helices at 208 nm and 222 nm, whereas ms_NLRP3 PYD showed deviation from a helical trend.
www.nature.com/scientificreports www.nature.com/scientificreports/ and Leu50 t (Fig. S2), thus anchoring the peptide to the ASC PYD surface. A more favorable binding energy for all the stapled peptides designed from the ASC PYD template compared to that of the ASC PYD :ASC PYD association (Table 1) suggested that they should successfully compete with ASC PYD or even NLRP3 PYD monomers to bind the target ASC PYD type-Ia interface.
ASC peptides suppress production of IL-1β from human monocytic cells. We next examined whether the ASC peptides could inhibit IL-1β production, which is regulated by the inflammasome. We used THP-1 human monocytic cells to study inflammasome activation as they express high levels of NLRP3, ASC, and pro-casp-1. THP-1 cells were incubated with ms_ASC PYD #1-6 or ms_NLRP3 PYD peptides at increasing concentrations for 4 hours and were then stimulated with LPS for 4 hours, followed by the NLRP3 activator nigericin for 1 hour. IL-1β levels were measured in the culture supernatant. LPS/nigericin treatment induced a robust IL-1β release (Fig. 6A). The ms_ASC PYD peptides reduced IL-1β production from THP-1 cells in response to LPS/ nigericin stimulation to variable degrees (Fig. 6A). Similarly, THP-1 cells pre-incubated with the ms_NLRP3 PYD peptide produced less IL-1β in a dose-dependent manner (Fig. 6B), indicating that NLRP3 peptides also can efficiently inhibit IL-1β release from human monocytic cells. We decided to continue investigating in detail the inhibitory effects of ms_ASC PYD #5 and #6 peptides, as their estimated binding energies matched well with the experimental data of suppression of ASC-dependent IL-1 release in a dose-dependent manner. Since ms_ ASCPYD #6 is the shorter version of ms_ASCPYD #5, this selection also enabled us to verify the effect of varying peptide lengths. Peptide residues are marked with a prime symbol (´). At the peptide N-terminus, the Arg residue (R'3 or R'1) is involved in stable salt bridge interactions with both Glu13 and Asp48 of ASC PYD . Two Lys residues (K'6 or K'4; K'11 or K'9) form frequent salt-bridges with residues Asp51, Asp54 and Asp6 of ASC PYD (see Fig. S1). Phe10 or Ile8 (F'10 or I'8) from the central parts of the peptides are involved in hydrophobic contacts with the side chains of Leu9, Leu50, and Arg5 of ASC PYD (details in Fig. 5 and S2). Non-specific hydrogen bonds were observed in the simulations between the main-chain of Ser or Val (S'14 or S'12; V'15 or V'13) from the peptide C-terminal and the main-chain of Met1, Gly2 or Arg5 from ASC PYD . (C and D) Energetic analysis of the MD simulations of the stapled peptide:ASC PYD complex for two representative peptides ms_ASC PYD #4 (C) and ms_ASC PYD #6. (D) Residue-wise binding energy contributions are shown.
www.nature.com/scientificreports www.nature.com/scientificreports/ We note a crucial difference between ms_ASC PYD #2 and ms_ASC PYD #5 (or ms_ASC PYD #6) is the presence of a negatively charged Glu at position 4 which would restrict the positioning of the crucial Arg (position 3) to interact with the protein. Indeed, docking simulations for ms_ASC PYD #2 predicted a lower number of native-like conformations compared to ms_ASC PYD #5 (28% vs. 38%, Fig. S3), whereas residue-wise binding energy analysis predicted unfavorable (positive) energy contribution for that Glu (Fig. S3) indicating that the negative charge is responsible to impede the binding of ms_ASC PYD #2. Although ms_ASC PYD #2 was not designed as a negative control, it serves as such due to its inability to suppress IL-1β release.
ASC peptides effectively penetrate human monocytic cells. As NLRP3 and ASC are cytosolic proteins, we examined whether the designed peptides could be taken up by monocytic cells to exert their biological effects. To assess peptide internalization, ms_NLRP3 PYD , ms_ASC PYD #5 and #6 peptides were labeled with the FAM fluorophore at the N-terminus. Internalization of the FAM-labeled peptides was measured as the mean fluorescence intensity (MFI) by fluorescence-assisted cell sorting (FACS). THP-1 cells were incubated with FAM-labeled ms_NLRP3 PYD (12.5, 25 and 50 μM), ms_ASC PYD #5 and #6 peptides (50 and 100 μM) for 4, 8 and 24 hours. The internalization of the three FAM-labelled peptides was concentration dependent (Fig. 7A,B). When THP-1 cells were incubated with FAM-ms_NLRP3 PYD and FAM-ms_ASC PYD #6 peptides at 37 °C, the fluorescent signal increased over time and reached its maximum intensity at 24 hours (Fig. 7A,B). Although the ms_ASC PYD #5 peptide penetrated THP-1 cells to a similar extent at 4 hours, the fluorescent intensity subsequently decreased over-time (Fig. 7B). We also examined the internalization of FAM-labeled peptides in THP-1 cells by confocal The small hydrophobic pocket between residues Leu9, Leu50 and Arg5 is observed at the ASC PYD type-I binding interface in the experimental structure (3J63.pdb). (B) Presence of the pocket was confirmed by benzene-mapping simulation of the ASC PYD monomer, where benzene probes sampled the region (orange grid). The benzene probability grids were generated by using the PTRAJ module of AMBER12 to bin carbon atoms of benzene molecules into 1 Å × 1 Å × 1 Å grid cells from a benzene-mapping MD trajectory. For the visualization of benzene occupancy in the grids, the cutoff isocontour value used was five times the threshold bulk value, which was defined as the highest isovalue at which benzenes were detected in the bulk solvent. This was chosen to filter out most of the spurious binding sites, and highlight those that are likely to be legitimate. (C) Phe10 of the peptide ms_ASC PYD #4 fits into the hydrophobic mesh when the benzene-mapped simulated structure of ASC PYD with the hydrophobic mesh is superposed onto the simulated structure of the ms_ASC PYD #4:ASC PYD complex. (D) Ile8 of the peptide ms_ASC PYD #6 also fits well into the hydrophobic mesh. The hydrophobic interaction shown markedly contributes to the net binding energy (Fig. 4).
To examine whether ASC peptides interfere with NLRP3 inflammasome activation, we first assessed processing of pro-IL-1β. The amount of processed IL-1β was inhibited in supernatants from ms_ASC PYD #5 and ms_ASC PYD #6 peptide treated THP-1 cells, in a dose-dependent manner (Fig. 8C). Similarly, treatment with ms_ASC PYD #5 and ms_ASC PYD #6 peptides reduced caspase-1 activation in a dose-dependent manner, evaluated as caspase-1 p20 in supernatants from THP-1 cells stimulated with LPS/nigericin (Fig. 8D). These data support that ASC peptides inhibit caspase-1 activation by NLRP3. The ASC peptides did not consistently affect pro-IL-1β or pro-caspase-1 expression in cell lysates (Fig. 8C,D), indicating that ASC peptides did not interfere with the priming step of NLRP3 inflammasome activation.
We next examined the effect of ASC peptides on NLRP3-dependent ASC oligomerization -a pivotal event in NLRP3 inflammasome activation. Upon NLRP3 activation by LPS/nigericin, ASC condensed into a large speck in each cell, but pre-treatment with ms_ASC PYD #5 and ms_ASC PYD #6 significantly inhibited speck formation (Fig. 8E). Collectively, these results indicate that the ASC peptides designed in our study efficiently inhibit NLRP3 inflammasome activation by interfering with NLRP3-induced ASC oligomerization, thereby reducing caspase-1 activation and IL-1β and IL-18 production from human monocytes/macrophages.

Discussion
Many inflammasome receptors require the adaptor protein ASC to form oligomeric filaments, which is an essential intermediate step to activate caspase-1. We, therefore, proposed that ASC may be a suitable target for ASC-dependent inflammatory pathways to disrupt excessive IL-1β production. From a structural view, spatial assemblies and monomer subunits of ASC PYD in the filament structure are conserved between human and mouse, implying the functional importance of its polymerization mechanism. The ASC PYD monomer is an all-α protein composed of six helices, suggesting that helical peptides would be potent binders. The flat and charged binding surface of ASC PYD also renders helical peptides more suitable to bind compared to small molecule-based inhibitors. Structural and mutational studies showed that electrostatic interactions are a dominant driving force for ASC PYD filament formation 32,34,38 . The ASC mature filaments are stabilized by three types of interactions between the PYD subunits, with the type I (H1-H4 helices of one PYD binding to H2-H3 of the other PYD) as dominant and found both in the cryo-electron microscopy structures 32,34 and NMR analysis 39 , specifying the reactivity of the type I surface. Utilizing a charge complementation approach, we designed peptides composed of positively charged residues to target the negatively charged type I surface of ASC PYD . We introduced a larger hydrophobic www.nature.com/scientificreports www.nature.com/scientificreports/ side chain (Leu to Ile/Phe) to increase the peptide affinity for the flat surface; this modification optimized packing inside the hydrophobic pocket on the ASC PYD surface, thus anchoring the peptides. This small hydrophobic pocket had an important role in the interaction with the peptides. This agrees with previous findings where the size and shape of the ASC filament were altered by the L25A mutation and recovered back by the L25M mutation 38 , suggesting a key function of these hydrophobic interactions at the core of the type-I interface in ASC filament formation. Major issues dealing with linear helical peptides include loss of their helical conformation in solution, faster proteolytic degradation, and reduced cell permeability. Stapling at the I,I + 7 positions ensures that the unbound peptide retains its helical conformation in solution thus minimizing the loss in conformational entropy upon binding. The presence of positively charged residues is advantageous when considering cell membrane permeability of the peptides 40 ; stapling itself is also helpful in increasing cellular uptake of peptides 40 .
We designed and experimentally validated a set of stapled peptide inhibitors targeting the PYD domain of ASC with the aim to block inflammasome formation and inhibit IL-1β and IL-18 release. These peptides were based on the ASC PYD helix 2 sequence and were developed to competitively target the protein-protein interaction between ASC PYD molecules. Specific structural factors such as charge distribution, the location of the hydrophobic patch and the different lengths of certain α-helices put together make the type I surface more reactive www.nature.com/scientificreports www.nature.com/scientificreports/ which should facilitate these positively charged stapled peptides to specifically bind to the negatively charged type I surface. As such, measuring peptide binding to ASC PYD by binding assays is difficult due to the highly aggregating features of ASC 39,41 . Starting from the ASC PYD -peptide structural model built from the ASC PYD filament structure, we confirmed by coarse-grained docking simulations that the designed peptides could bind to www.nature.com/scientificreports www.nature.com/scientificreports/ the target binding site and did not bind to other similar death domains such as ASC CARD . The complex structure was analyzed by atomistic MD simulation to guide the design of a set of stapled peptides with improved binding affinities. In contrast to the unstapled wild-type counterparts, the stapled peptides retained their helical conformations in solution. Notably, the prior organization of the peptides into conformations similar to the ones they adopt upon binding to their targets results in lowering the entropic costs associated with embedding the peptide free in solution onto a surface. Tested stapled peptides were cell-permeable, and were able to suppress the IL-1β and IL-18 production by human monocytic cells (THP-1 cells and blood monocytes). We discovered that ASC peptides blocked NLRP3 inflammasome activation by preventing ASC oligomerization, which in turn suppressed casp-1 activation and thereby IL-1β release. The stapled peptides did not interfere with the priming (signal 1) of the NLRP3 inflammasome, as pro-IL-1β expression remained unaffected. Peptides did not induce significant cell death compared to THP-1 cells treated by the vehicle control treated cells (data not shown).
Our proof-of-concept study is the first to use stapled peptides to block the NLRP3 inflammasome. Peptide inhibitors designed to mimic known sequences of caspase substrates effectively suppress caspase-dependent effects in various cells. In particular, caspase-1 is activated in different inflammasome complexes and several short peptide inhibitors have been developed 42 . However, due to the high similarity between caspase sequences, many peptide inhibitors designed to target caspase-1 also block the activity of other caspases, making data interpretation difficult. For example, the cell-permeable and irreversible caspase-1 inhibitor Y-VAD also inhibits caspase-4 and caspase-5. A small number of reversible specific inhibitors of caspase-1 (including VX-740 and VX-765) have entered clinical trials 42,43 : preclinical studies demonstrated that these inhibitors ameliorate disease severity in models of arthritis and inflammation, but some were withdrawn from clinical trials because of liver toxicity after long-term administration in animal studies. As such, no caspase inhibitors have yet reached Phase III.
Numerous research efforts have aimed to identify small molecules that can prevent inflammasome activation, including those that can inhibit the NLRP3 inflammasome 18,20,21,24 . However, the molecules identified to date are nonspecific and thus may interfere with several mechanisms that indirectly inhibit inflammasome activation. Peptide therapeutics is a growing area of interest as attractive alternatives to small-molecule inhibitors mainly due to their high potency and specificity 44 . Since the template sequence of the designed peptides is taken from the target molecule itself, in principle, they are less likely to initiate an immune response. We note that the effective stapled peptide dose used here (50-100 μM) are comparable to the average K D value for ASC PYD self-association (65 μM) measured by NMR titration 39 .
This work provides the first example of using peptide-based modulators to target components of the inflammasome pathway. The ability of the constrained helical peptides to inhibit the PYD interaction provides a good starting point to design similar peptide-based inhibitors for the other death domain families. The promising aspect will be addressed in future investigations aimed at understanding whether the ASC peptides designed in our study can block the activity of other PYD proteins. Studies addressing the bioavailability of these peptides would throw light on their potential as therapeutics.

Methods
Study approval. Ethical approval for all blood sources and processes used in this study was obtained from the National University of Singapore Institutional Review Board (license: NUS-IRB 12-044E). Subjects gave written informed consent in accordance with the Declaration of Helsinki. All experiments were carried out following the approved guidelines and regulations. Structure modeling. The ASC PYD -ASC PYD dimer (interacting through the type-I interface) obtained from the human ASC PYD filament structure solved by cryo-electron microscopy at atomic resolution (3J63.pdb, 3.8 Å) was used for the structural modeling. There the conformation of helix 2 peptide ( 16 TAEELKKFKLKLLSV 30 ) of the type-Ib monomer bound to the type-Ia interface of the other monomer was used as the initial model structure to design the stapled peptides (Fig. 1A). A hydrocarbon staple across positions 5 and 12 (I, I + 7) along the peptide was constructed. The peptide was mutated in silico at various positions (Table 1) and the resulting models subject to detailed optimization. This included carrying out extensive (3 independent simulations of 200 ns each per protein-peptide system) molecular dynamics simulations on the complexes in the aqueous phase followed by detailed structural and energetic analysis (described below). Docking simulations. All the designed peptides have positive electrostatic potential due to the presence of four/five positively charged residues. Conversely, the ASC PYD type-Ia interface is negatively charged. We used coarse-grained simulations (GO model type) to explore the binding of the designed peptide ms_ASC PYD #5 to ASC PYD , where the interaction between the protein and the peptide was modeled solely by electrostatic interactions between charged residues. Here coarse-grained simulations are advantageous over all-atom MD simulations in enabling access to timescales long enough to achieve the sampling of the binding process. Overall, we followed an approach to protein modeling similar to that described in previous studies 45,46 . Each protein residue was represented by two beads placed at the Cα-and Cβ-positions. Charged residues (K, R, H, D, and E) had a charge at the Cβ-position. The native topology of the protein and peptide was retained by applying a Lennard-Jones (L-J) potential during the simulation. The electrostatic interaction between charged residues was modeled by applying Debye-Hückel potential. As ASC PYD does not undergo significant conformational changes when interacting, all simulations were run at relatively low temperatures to allow the protein and peptide to fluctuate around their native states. The salt concentration was also kept low (10 mM) to reduce the charge screening effect. Since the distances between the charged beads in the coarse-grained model are longer than the corresponding distances in the all atom representations, the effective salt concentration is higher here, and 10 mM is equivalent to a salt concentration of ~30 mM. Each predictive simulation was started by placing the unbound peptide at one of 10 different positions at about ~35 Å (distance between the protein surface and the peptide center), ensuring that www.nature.com/scientificreports www.nature.com/scientificreports/ the peptides were distributed well around the protein for unbiased sampling. Each trajectory was simulated for 10 7 steps, and a total of 100 simulations (10 simulations for each of the 10 initial positions of the peptide around the protein) were run for adequate sampling. Coordinates were saved every 1000 steps and only the last 2000 conformations were collected during each simulation, resulting in a total of 2 × 10 5 analyzed conformations. As the number and positions of the positive residues are similar in all the peptides, all other peptides are expected to exhibit a similar binding behavior in docking simulations.
The simulated conformations were compared with the experimental type-I ASC PYD interface structure (Fig. 1A) to evaluate their binding accuracy. A structural similarity parameter (D) was calculated to quantify the degree of deviation of the simulated conformation from the experimental structure. We used a similar approach as described in a previous study 47 . Only Cβ beads of the interface residues of the protein (Asp6, Leu9, Asp10, Glu13, Asp48, Leu50, Asp51, and Asp54) and peptide (Lys6, Lys7, Lys9, Ile10, and Lys11) were considered for calculating D using the following equation: where, i and j stand for the i th Cβ interface bead of the protein and the j th Cβ interface bead of the peptide, respectively, and r ij and r ij 0 are the distances between those bead pairs in the simulated and the experimental structure, respectively. N protein and N peptide are the number of protein and peptide interface residue beads, respectively. D quantifies the overall conformational similarity between the predicted binding interface of the peptide:ASC PYD complex and the experimental type-I interface structure (only helix 2 from the positive surface). The direction of the peptide with respect to the ASC PYD interface is considered when calculating pair-wise distances. Finally, to obtain a detailed structural quantification, interface residues were divided into two spatial groups that cover the two halves of the peptide, and D was calculated for each (termed as D 1 and D 2 ). A low D 1 or D 2 value indicates high conformational similarity.
The same simulation and sampling strategies as outlined above were used to explore the interactions between ASC CARD (2KN6.pdb, residue:115-195) and ms_ASC PYD #5. The distances found in the simulated conformations suggests that the peptide does not bind to ASC CARD .

Molecular dynamics (MD) simulations. The bound conformation of helix 2 ( 16 TAEELKKFKLKLLSV 30 )
to ASC PYD (Fig. 1A) was used as the initial model structure for all-atom MD simulations of peptide-protein complexes ( Table 1). The ASC PYD type-I protein-protein complex structure was prepared by taking subunit A and B from the filament (3J63.pdb) for simulation. For different peptides, residues at different positions were mutated (Table 1) in the starting structure while keeping the binding posture of the peptide unaltered. The hydrocarbon linker (staple) in the various designed stapled peptides ((2R)-2(7-octenyl) alanine and (2S)-2(4-pentenyl) alanine) was modeled using the XLEAP module of AmberTools12. The RESP (Restrained Electrostatic Potential) based atomic charges of the hydrocarbon linker were derived from the R.E.D. server using the RESP-A1A (HF/6-31G*) charge model and Guassian_2009_C.01 quantum mechanics program 48 . Other force field parameters were derived from all-atom ff99SB 49 force field in Amber12 50 . The N-terminus of the protein and peptide was acetylated, the C-terminus of the protein was methylated, and the C-terminus of the peptide was amidated. The starting structures were placed in a cuboid water box such that the minimum distance from the edge of the box was at least 10 Å. The TIP3P water model was used for solvation 51 . The solvated systems were neutralized by adding chloride ions using the LEaP module of AmberTools12. A total of 10 peptide-protein complex systems (seven stapled and three unstapled) and one protein-protein complex system were prepared (Table 1), all of which were then used in the subsequent MD simulation steps.
Energy minimizations and MD simulations were carried out using the PMEMD module of AMBER12 employing the all-atom ff99SB force field parameters. Three independent explicit-solvent MD simulations using different initial atomic velocities were carried out for each of the eleven systems to improve the sampling. The solvated systems were initially relieved of any unfavorable interactions by subjecting them to 1,000 steps of energy minimization, which involved using 500 steps each of steepest descent, followed by conjugate gradient algorithms. The conjugate gradient algorithms were used as follows: i) first, Cartesian restraints were imposed on the solute while the solvent molecules were then allowed to relax around it, ii) next, the solvent was restrained and the solute was energy minimized, iii) finally, no restraints were imposed and the whole system was minimized. A force constant of 500 kcal mol -1 Å 2 was used during the restraining steps. The systems were then gradually heated to 300 K over a period of 30 ps using the NVT ensemble. Following this, each system was equilibrated under NPT conditions for 500 ps. They were then subjected to the production phase of MD simulation using the NPT ensemble for a period of 200 ns each. This procedure resulted in 33 (11 × 3) simulations with a combined simulation period of 6.6 μs. The simulation temperature of 300 K was set using Langevin dynamics 52 , with a collision frequency of 2.0 ps −1 . The pressure was maintained at 1 atm using weak-coupling 53 with a pressure relaxation time of 2 ps. Periodic boundary conditions in x, y and z directions were appropriately applied. The particle mesh Ewald (PME) method was used to treat electrostatic interactions 54 with a grid spacing of 1.0 Å. The SHAKE algorithm 55 was applied to constrain all bonds involving hydrogen atoms, allowing for a time step of 2 fs. A cutoff distance of 9 Å was implemented for nonbonded interactions. We note that all the peptides remained in bound conditions similar to the starting conformation throughout the simulations. Coordinates were saved every 10 ps.
Binding energy decomposition and salt-bridge distance analysis. The contribution of each peptide residue to the binding energy of complex formation was calculated by applying the free energy decomposition method 56 on 2,000 equally-spaced structures extracted from the MD simulation trajectories. Binding energies (2019) 9:4913 | https://doi.org/10.1038/s41598-019-41211-3 www.nature.com/scientificreports www.nature.com/scientificreports/ were computed by the MMPBSA.py script 57 in AMBERTools12 using the MM/GBSA (Molecular Mechanics/ Generalized Born Surface Area) method. Water molecules and Cl −1 ions were stripped from the extracted structures, and the solvent effect was represented using a Generalized Born Solvation Model 58 . A salt concentration of 150 mM was used. The nonpolar contribution to solvation free energy was estimated from the solvent accessible surface area (SASA) using the ICOSA method. Energy values are represented in Table 1 as an average of the values obtained from each of the three simulation runs. Distance calculations from the trajectories were performed using the PTRAJ module in AmberTools12.
Peptide synthesis. The linear and stapled peptides were synthesized by American Peptide Company, Inc.
(Sunnyvale, CA USA) by replacing the two respective Leu residues at positions I and I + 7 of the linear peptide with the olefin-bearing unnatural amino acids (R)-2-(7′octenyl) alanine and (S)-2-(4′-pentenyl) alanine, respectively, and stapled via olefin metathesis using the Grubbs catalyst. All peptides were purified using HPLC to >90% purity. They were amidated at their C-terminus and acetylated at their N-terminus. Peptides with FAM-labeled at the N-terminus were also synthesized by American Peptide Company, Inc. with the C-terminal amidated and was purified using HPLC to >90% purity.
Cell stimulation. Human monocytic THP-1 cells (1 × 10 7 cells) were plated in a 10-cm tissue culture dish and treated with PMA (200 nM) overnight to differentiate monocytic cells into macrophages. Cells were then washed and allowed to rest in fresh media for another day. For stimulation, differentiated THP-1 cells were trypsinized and seeded in 96-well plates (2 × 10 6 /well). Peripheral blood mononuclear cells were isolated by Ficoll-Hypaque density gradient centrifugation from buffy coats obtained from anonymous donors provided by the National University Hospital Blood Bank, Singapore. Isolation of monocytes was performed using negative selection with the Monocyte Isolation Kit II (Miltenyi Biotech), following the manufacturer's instructions. Monocytes were plated in tissue culture non-treated 96-well plates (2.5 × 10 5 /well) in 250 µl RPMI 1640-Glutamax media supplemented with 3% human AB serum (Gemini Bio Products).
Cells were incubated with FAM-NLRP3 PYD (12.5-25-50 μM), FAM-ASC PYD #5 or FAM-ASC PYD #6 (50-100 μM) peptides for 4, 8 and 18 h. The cells were then washed twice with PBS and fixed with 4% PFA for 10 min before permeabilization and incubation with an ASC antibody (4 µg/ml, Santa Cruz Biotechnology, 22514-R) in blocking/permeabilization buffer (10% goat serum, 1% FBS and 0.5% Triton-X in PBS) for 1 h. The cells were subsequently washed three times with blocking/permeabilization buffer prior to incubation with Alexa Fluor 568-labeled goat anti-rabbit antibody (1 µg/ml, Invitrogen, A11011) together with DAPI for 30 min, and imaged on an FV1000 Olympus confocal microscope using an x100 objective with an additional zoom of 2. Z-stack images were acquired and the maximum projections are shown.

Data Availability
All the data are available upon request.