HCK regulates NLRP12-mediated PANoptosis

Significance

The cytosolic innate immune sensor NLRP12 induces inflammatory cell death, PANoptosis, in response to homeostatic disruptions and infection. NLRP12-mediated PANoptosis has been implicated in hemolytic and inflammatory diseases, but the regulatory mechanisms controlling its activation remain unclear. Our study identifies tyrosine protein kinase hematopoietic cell kinase (HCK) as a regulator of NLRP12-mediated PANoptosis. Given the importance of NLRP12 in innate immunity and inflammation, our findings suggest strategies targeting NLRP12 or its regulator, HCK, for therapeutic development.

Abstract

NOD-like receptors (NLRs) are a highly conserved family of cytosolic pattern recognition receptors that drive innate immune responses against pathogens, pathogen-associated molecular patterns, damage-associated molecular patterns, and homeostatic disruptions. Within the NLR family, NLRP12 was recently identified as a key regulator of PANoptosis, which is an innate immune, lytic cell death pathway initiated by innate immune sensors and driven by caspases and RIPKs through PANoptosome complexes. While NLRP12 activation is critical for maintaining homeostasis, aberrant activation has been implicated in a broad range of disorders, including cancers and metabolic, infectious, autoinflammatory, and hemolytic diseases. However, the molecular mechanisms of NLRP12 activation remain poorly understood. Here, we identified hematopoietic cell kinase (HCK) as a regulator of NLRP12-mediated PANoptosis. HCK expression was significantly upregulated in response to NLRP12-PANoptosome triggers. Moreover, Hck knockdown inhibited NLRP12-mediated PANoptosis. Computational analyses identified residues in the putative interaction interface between NLRP12 and HCK, suggesting that HCK likely binds NLRP12 in the region between its NACHT domain and pyrin domain (PYD); removal of the NLRP12 PYD abrogated this interaction in vitro. Overall, our work identifies HCK as a regulator of NLRP12-mediated PANoptosis, suggesting that it may serve as a potential therapeutic target for mitigating inflammation and pathology.

The innate immune system acts as the first line of defense against invading pathogens and sterile insults (1, 2). Pattern recognition receptors (PRRs) are components of the innate immune system that sense specific pathogen-associated molecular patterns (PAMPs), endogenous damage-associated molecular patterns (DAMPs), and homeostatic alterations to activate inflammatory signaling and cell death pathways (1–3). Recent studies have shown that several members of the nucleotide-binding oligomerization domain (NOD)-like receptor (NLR) family of cytosolic PRRs can activate PANoptosis, an innate immune, inflammatory cell death pathway initiated by innate immune sensors and driven by caspases and receptor-interacting serine/threonine protein kinases (RIPKs) through PANoptosome complexes (4–10). Given their essential roles in inflammatory cell death and immune responses, NLRs have emerged as potential therapeutic targets (3, 5). Despite this interest, only certain members of the NLR family, such as NLRP3, have been extensively studied, and the broader roles, activation mechanisms, and functions of many other NLR proteins remain poorly understood.

Among these lesser-known NLRs is NLRP12. Composed of an N-terminal pyrin domain (PYD), a central NACHT domain, and a C-terminal leucine rich repeat region (LRR), NLRP12 is highly expressed in patients with hemolytic diseases, such as malaria and sickle cell disease (SCD), and in patients with infections associated with hemolysis (7, 11–16). Additionally, NLRP12 has been shown to drive PANoptosis in response to heme and PAMPs or TNF (7). NLRP12 has also been suggested to play a role in regulating inflammation through the NF-κB and MAPK signaling pathways (17, 18) and to modulate metabolic inflammation, reducing obesity and insulin resistance in murine models (19). Furthermore, given its critical functions in cell death and inflammation, dysregulation of NLRP12 is linked to various inflammatory and autoimmune disorders, including familial cold autoinflammatory syndrome 2 (FCAS2) and systemic lupus erythematosus (SLE) (20–22). In contrast, NLRP12 suppresses colorectal tumorigenesis, triple-negative breast cancer, and hepatocellular carcinoma in both in vitro and in vivo systems (17, 23–25). Collectively, these findings highlight the importance of appropriately regulating NLRP12 activation to modulate pathophysiology across the disease spectrum. However, the upstream regulatory mechanisms controlling NLRP12 activation and NLRP12-mediated PANoptosis remain poorly understood, creating a crucial gap in the ability to develop therapeutics targeting this pathway.

Therefore, in this study, we sought to identify key regulators of the NLRP12-mediated PANoptosis pathway to find potential therapeutic targets. By analyzing kinase gene expression in hemolytic diseases, we identified hematopoietic cell kinase (HCK), a nonreceptor tyrosine kinase in the Src family (26), as a highly upregulated gene. Knockdown of Hck blocked the biochemical activation and execution of cell death in response to the NLRP12-PANoptosome trigger heme plus TNF, suggesting that HCK regulates NLRP12-mediated PANoptosis. Furthermore, computational approaches identified the putative interaction interfaces between HCK and NLRP12, and we found that the NLRP12 PYD was critical for its interaction with HCK. Our analysis suggested key residues that consistently formed interactions to stabilize the NLRP12–HCK complex through complementary charge interactions. Overall, our study combines biochemical, genetic, and computational approaches to identify HCK as a regulator of NLRP12-mediated PANoptosis, suggesting that HCK may be a potential therapeutic target for reducing inflammation and immunopathology.

Results

HCK Mediates NLRP12-Dependent Inflammatory Cell Death, PANoptosis.

Members of the NLR family of cytosolic innate immune sensors are known to be regulated by posttranslational modifications to control the activation of their downstream pathways (5, 27). Phosphorylation is one of the most common posttranslational modifications, and NLRs are frequently regulated by kinases (5, 28, 29). However, the role of kinases in the regulation of NLRP12, a key NLR that drives inflammatory cell death, PANoptosis, in response to hemolytic triggers (7), remains unclear. We hypothesized that potential kinases regulating NLRP12 would be upregulated in conditions where NLRP12 was activated, such as in hemolytic diseases. We therefore analyzed publicly available gene expression datasets from hemolytic diseases, including whole blood (WB) from patients infected with the malarial parasite Plasmodium falciparum; bone marrow (BM) CD71⁺ cells from patients infected with Plasmodium vivax; and monocytes and BM CD34⁺ cells from patients with SCD (11–14). Among the kinases expressed in these datasets, we observed that HCK was the most highly upregulated (Fig. 1A). In vitro, we also observed that HCK mRNA and protein expression were upregulated in response to the hemolytic trigger heme plus TNF in BM–derived macrophages (BMDMs) (Fig. 1 B and C), suggesting that HCK is upregulated under conditions where NLRP12-mediated PANoptosis occurs. To determine whether HCK regulated NLRP12-mediated cell death, we used small interfering RNA (siRNA) to silence Hck in BMDMs, which effectively reduced HCK protein expression (Fig. 1C). Heme plus TNF stimulation induced cell death in BMDMs treated with control siRNA, while those treated with Hck- or Nlrp12-targeting siRNA exhibited significantly reduced cell death (Fig. 1 D and E). Additionally, heme plus TNF stimulation led to the release of DAMPs associated with lytic cell death in BMDMs, such as lactate dehydrogenase (LDH) and high mobility group box 1 (HMGB1), and the release of these DAMPs was decreased upon treatment with Hck- or Nlrp12-targeting siRNA (Fig. 1F). Furthermore, we confirmed that Hck siRNA treatment did not inhibit NLRP12 expression (Fig. 1G), suggesting that HCK acts directly on NLRP12 once it is expressed, rather than regulating further upstream. We also observed that the regulatory effect of HCK was specific to NLRP12-mediated cell death, as siRNA knockdown of Hck did not reduce cell death in response to other NLR triggers, including NLRP3 (LPS plus ATP), NLRC4 (Salmonella), and NLRP1b (ValboroPro) triggers (SI Appendix, Fig. S1 A–F). Together, these data indicate that HCK specifically regulates NLRP12-mediated cell death.

Fig. 1.

HCK mediates NLRP12-dependent inflammatory cell death, PANoptosis. (A) Heatmap depicting the expression profile of kinases in malaria WB, malaria BM CD71⁺ cells, SCD monocytes, and SCD BM CD34⁺ cells, compared with healthy controls. (B) Measurement of the relative expression of Hck mRNA normalized to β-actin expression in wild type (WT) BMDMs treated with media or heme plus TNF for 42 h. (C) HCK protein expression in BMDMs electroporated with control siRNA (siControl) or Hck siRNA (siHck) treated with media or heme plus TNF for 42 h. For loading control, β-actin is shown. (D–F) Representative images of cell death (D), quantification showing percentage of cell death (E), and immunoblot analysis of LDH and HMGB1 from the supernatant (F) of BMDMs electroporated with siControl, siHck, or Nlrp12 siRNA (siNlrp12) treated with heme plus TNF for 42 h. [Scale bar, 50 μm (D).] (G) Fold change mRNA expression of Nlrp12 normalized to 18S expression in WT BMDMs electroporated with siControl or siHck treated with media or heme plus TNF for 24 h. (H) Immunoblot analysis of pro- and activated caspase-1 (CASP1; P45 and P20, respectively); pro-, activated, and inactivated gasdermin D (GSDMD; P53, P30, and P20, respectively); pro- and activated gasdermin E (GSDME; P53 and P34, respectively); pro- and cleaved caspase-8 (CASP8; P55 and P44/P18, respectively); pro- and cleaved caspase-3 (CASP3; P35 and P19/P17, respectively); pro- and cleaved caspase-7 (CASP7; P35 and P20, respectively); and phosphorylated and total MLKL (pMLKL and tMLKL) in BMDMs electroporated with siControl, siHck, or siNlrp12 in media at the 0 h timepoint, or treated with heme plus TNF for 42 h. For loading control, β-actin is shown. Three or more independent experiments were performed, and the data shown are from a single representative experiment (B–H). Mean ± SEM is shown (B, E, and G). Statistical analysis was performed using the one-way ANOVA (E and G), or the unpaired t test (B). *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Since we observed that HCK regulated NLRP12-mediated cell death (Fig. 1 D and E), we next sought to determine the biochemical effect of HCK on PANoptosis by assessing the activation of PANoptosis effectors and executioners. We observed that the heme plus TNF-induced activation of caspase-1, GSDME, caspase-8, caspase-3, caspase-7, and pMLKL was reduced in cells treated with Hck- or Nlrp12-targeting siRNA when compared to control siRNA-treated cells (Fig. 1H). Additionally, the cleavage of GSDMD to its P20 fragment, which is produced downstream of caspase-3 activation (30, 31), was also reduced upon treatment with Hck- or Nlrp12-targeting siRNA (Fig. 1H). Collectively, these results suggest that HCK regulates the activation of NLRP12-dependent PANoptosis.

NLRP12 Interacts with HCK.

Based on the observation that loss of HCK prevented NLRP12-mediated PANoptosis, we next sought to mechanistically determine how HCK regulated this pathway. We first assessed whether HCK directly interacted with NLRP12 to regulate cell death. Co-immunoprecipitation (co-IP) using HCK as bait showed that NLRP12 could be pulled down with HCK (Fig. 2A), suggesting that NLRP12 and HCK likely interact. To understand this interaction at the molecular level, we next sought to model a putative NLRP12:HCK complex. An earlier report suggested that the C-terminal 30 amino acids of HCK are essential for NLRP12 binding (32). Moreover, NLRP3, a protein similar to NLRP12, is regulated via phosphorylation of its PYD-linker-NACHT region by BTK (29), a tyrosine kinase closely related to HCK. Therefore, we hypothesized that the PYD-linker-NACHT region of NLRP12 (Fig. 2B) could serve as a potential site for HCK’s activity. To identify potential HCK phosphorylation sites in this multidomain region of NLRP12, we used NetPhos v3.1 (33) to predict phosphorylation sites. Since there is no specific HCK phosphorylation site predictor, we used a broad Src tyrosine kinase predictor. Our analysis identified several potential phosphorylation sites within the NLRP12 sequence (Fig. 2C). Based on the hypothesis that the PYD-linker-NACHT region would be the most likely area for modifications in NLRP12, we examined possible HCK phosphorylation sites in this multidomain interface. We identified three tyrosine residues here (Y129, Y132, Y157), each with a phosphorylation potential >0.3 (Fig. 2 B–E). Pairwise sequence alignment of NLRP12 across species and with other NLRP family members, such as NLRP3, indicated that these three identified residues were evolutionarily conserved.

Fig. 2.

HCK interacts with NLRP12. (A) Co-IP of NLRP12 with HCK. (B) Domain architecture of HCK and NLRP12, with the HCK SH3, SH2, and HCK-KD and NLRP12 PYD (pyrin domain), NACHT, and LRR shown. (C) Plotted potential for phosphorylation at each residue of NLRP12 by Src family kinases. (D) Schematic representation of the hypothesized mode of NLRP12–HCK complex formation and phosphorylation. (E) Zoomed-in view of chosen NLRP12 putative phosphorylation sites in the FISNA region of the NACHT. Three independent experiments were performed, and the data shown are from a single representative experiment (A).

To further understand the potential interaction between NLRP12 and HCK, we used a molecular docking approach to construct the model of the NLRP12:HCK complex. Since the structure of human NLRP12 has not yet been determined, we employed AlphaFold2 (AF2) (34, 35) to generate a starting structure for NLRP12. Studies have shown that the kinase domain of Src family kinases, including the HCK kinase domain (HCK-KD), is sufficient for functional interactions with its targets (32, 36). Therefore, we used HCK-KD for subsequent experiments. Similar to NLRP12, the structure for the apo-form of HCK-KD has not been solved, so we used AF2 to generate a model of HCK-KD. The accuracy of the modeled residues over the structure was evaluated using the predicted local Distance Difference Test (plDDT) (SI Appendix, Fig. S2). The NLRP12 and HCK models were built with most regions modeled using a large number (>6,000) of database sequences (SI Appendix, Fig. S2A). Our analysis showed a plDDT score of >90 calculated for most of the domain regions of NLRP12 and HCK-KD, demonstrating accurate modeling (SI Appendix, Fig. S2B). However, the linker regions in NLRP12 connecting PYD and NACHT and connecting NACHT and LRR, as well as the loop region in the kinase domain core of HCK-KD, obtained plDDT scores in the range of 30 to 60 (SI Appendix, Fig. S2B), indicating regions of lower confidence.

After modeling the protein structures and identifying probable interaction sites, we utilized the protein–protein docking method HADDOCK v2.4 (37, 38) to model the interaction of NLRP12 with HCK-KD. In this process, we assessed whether the conserved tyrosine residues predicted as potential phosphorylation sites were exposed or buried within the modeled structures of NLRP12. We particularly focused on whether the tyrosine side chain (hydroxyphenyl group) was exposed, as this would be the location for phosphorylation by HCK when in close proximity. Although all predicted sites had the potential to be phosphorylated by HCK, we determined that Y129 of NLRP12 was unlikely to be the key phosphorylated residue, as it had a buried side tyrosine chain in the starting structure; Y129 was approximately 7 Å away from the succeeding tyrosine with surface-exposed side chains (NLRP12: Y132) (SI Appendix, Fig. S3 A and B). Therefore, we chose to build the NLRP12:HCK-KD complex using the surface-exposed residue Y132 as a residue constraint for NLRP12. An initial study showed that the C-terminal 5 residues of HCK are required for binding to NLRP12 (32); therefore, we chose the C-terminal 5 amino acid loop region (Y522-P526) as a constraint for HCK (SI Appendix, Fig. S3 A and B). We generated the ambiguous interaction restraint by defining active and passive residues. The phosphorylation site of NLRP12 (Y132), along with the C-terminal 5 amino acid loop of HCK, were selected as active residues in their respective molecules, while passive residues were defined by selecting the solvent-accessible neighbors of active residues. The HADDOCK method employs a series of energy minimizations (rigid body and semi-flexible), orientations and translations, and explicit solvent-based refinement to optimize the interaction defined in the constraint space of the two protein partners. It then clusters the generated docked conformations using a fraction of common contacts (cutoff 0.60) and sorts the clusters using the average HADDOCK score. We followed the recommendations of the HADDOCK program to select the optimal conformation of the NLRP12:HCK-KD complex for further analysis, including selecting the cluster with the highest HADDOCK score and examining multiple models within the chosen cluster. The best docking pose of the NLRP12:HCK-KD complex achieved a score of −71.9 ± 12.9, with only seven members in its cluster. However, upon visual analysis, the cluster (score: −64.3 ± 4.5) with the largest number of members (57 members) for NLRP12:HCK-KD dock poses exhibited clashes at the protein interface. We also modeled the third residue (NLRP12: Y157) and found no suitable docked conformations. These observations indicate that advanced techniques exploiting large conformational space, such as replica exchange and umbrella sampling, may be suitable for obtaining favorable NLRP12:HCK-KD heteromolecular interactions by selecting these constraints. Together, our modeling data suggest that NLRP12 and HCK can directly bind, which may be a key regulatory step to drive NLRP12-mediated PANoptosis.

Molecular Dynamics Simulations Suggest That HCK Forms a Complex with NLRP12.

Based on our modeling results, we continued all further analyses using the NLRP12 (Y132)-HCK docked complex (Fig. 3A). To understand the interaction preferences of HCK with NLRP12 and provide insights into the regulatory mechanisms, we performed molecular dynamics simulations for 500 ns on the top docked poses of the NLRP12:HCK-KD complexes using the DESMOND module (Schrödinger suite v2023.1). Restraints were applied between the chosen tyrosine residues in the NLRP12:HCK-KD interface during simulations, similar to the restraints used in docking. Comparing the docked pose to the simulated pose, we observed sustained proximal interactions between the Y522 residue of HCK and the probable phosphorylation site of NLRP12 (Y132) (Fig. 3B). The RMSD over the simulation time showed that the NLRP12:HCK-KD complex achieved stable associations after 100 ns and fluctuated with an RMSD of ~7 Å (Fig. 3C). RMS fluctuation (RMSF) measurements showed that a few amino acid residues of the PYD and linker regions connecting PYD with NACHT and NACHT with LRR underwent larger fluctuations throughout the simulation (Fig. 3D). In contrast, the HCK-KD experienced fewer fluctuations due to its stable interactions with NLRP12 (Fig. 3D). We compared the initial docked and final simulated poses of each NLRP12:HCK-KD complex and identified the interacting hydrogen bonds (Table 1). These results suggest that HCK and NLRP12 form a stable complex mediated by hydrogen bonding interactions.

Fig. 3.

Molecular dynamics simulations suggest that HCK forms a complex with NLRP12. (A) Ribbon view of the docked NLRP12–HCK model. Distinct domains are colored in different colors, with the HCK-KD and NLRP12 PYD, NACHT, and LRR shown. (B) Zoomed in view of the simulated NLRP12–HCK putative phosphorylation site at 0 ns and 500 ns. The location of the NLRP12 Y132 and HCK Y522 residue hydroxyl groups are noted in red. (C) RMSD of the NLRP12–HCK complex over the simulation time frame. (D) RMSF of residues by domain.

Table 1.

Consistent hydrogen bonding interactions observed between the HCK-KD and NLRP12

Region	Residue		% of H bonds throughout simulation
	NLRP12	HCK
PYD	MET001	GLU505	45.54
PYD	ARG003	ASP513	78.22
PYD	ARG007	ASP512	30.69
Linker	ASP097	ARG375	69.31
Linker	GLY101	THR311	76.24
NACHT	GLN257	GLN525	30.69
NACHT	THR260	PRO526	30.69
NACHT	SER270	GLN523	46.53
NACHT	GLU281	SER392	29.7
NACHT	GLU315	LYS352	30.69

Key residues in the PYD, linker, and NACHT regions of NLRP12 form consistent, stabilizing hydrogen bonding interactions with the HCK-KD.

The NLRP12:HCK-KD Complex Is Stabilized at the PYD–NACHT Linker Region.

After confirming that NLRP12:HCK-KD formed a heterodimer complex (Fig. 3 A–D), we next sought to identify the possible interface regions between NLRP12 and HCK within the NLRP12:HCK-KD complex. Three interacting interfaces were identified between NLRP12 and HCK-KD (Fig. 4A). The PYD and linker region of NLRP12 developed stable H bond interactions with C-terminal and core residues of HCK, which we defined as interfaces 1, 2, and 3, (Fig. 4A and Table 1). In all these NLRP12:HCK-KD complexes, the restrained residues did not develop H bonds between themselves. We observed that the constraints initially carried from the HADDOCK dock poses, which were subsequently defined as restraints in the simulation of the NLRP12:HCK-KD complex, maintained an average distance of 15.5 Å (SI Appendix, Fig. S4). This allowed different interface sites of the NLRP12:HCK-KD complex to form consistent H bonds during simulations. Furthermore, we employed the dynamic cross-correlation matrix (DCCM) method to detect correlative motions of amino acid residues involved in the NLRP12:HCK-KD complex during simulations. Certain key residues (phosphorylation sites and residues of interface sites) retained interactions and, therefore, participated in positive correlative motions (SI Appendix, Fig. S5). Since the C-terminus of the HCK-KD is mostly composed of loop elements, no and/or anticorrelations were expected in certain regions of the interface. The DCCM analysis suggests that HCK maintained a stable orientation at the PYD-linker-NACHT multidomain interface of NLRP12 throughout the simulations.

Fig. 4.

HCK interacts with NLRP12 at the intervening region of PYD and NACHT. (A) Model of the NLRP12–HCK complex after 500 ns of simulation. Distinct domains are colored in different colors, with the HCK-KD and NLRP12 PYD, NACHT, and LRR shown. Three critical interfaces are observed; interfaces 1 and 3 occur between the HCK-KD and the NLRP12 NACHT region. Interface 2 occurs between the HCK-KD and the NLRP12 PYD. (B) Analysis of the electrostatics of the HCK and NLRP12 proteins at 500 ns. (C) Co-IP of HCK with NLRP12–ΔFISNA and NLRP12–ΔPYD. Three independent experiments were performed, and the data shown are from a single representative experiment (C).

The NLRP12:HCK-KD Complex Is Electrostatically Complementary.

Tyrosine kinases tend to phosphorylate tyrosine residues that are surrounded by acidic amino acids, and they use electrostatic interactions to recognize and bind to these complementary sites on substrates (26, 39, 40). To determine whether this is the case for the NLRP12:HCK-KD complex, we used the DelPhi program to apply the linear Poisson–Boltzmann equation to calculate the electrostatic energies of the simulated complex. Subsequently, we visualized the electrostatic potentials on their accessible surface areas to examine electrostatic complementarity at the identified interface sites. Electrostatic potential analysis of the NLRP12:HCK-KD complex identified that interface 1 (constraint site) of the complex featured both shape and electrostatic complementarity, with a mix of neutral regions (Fig. 4B). Interface 2 also contained several charged residues, resulting in a high frequency of consistent heteromolecular H-bonding residues (Fig. 4B and Table 1). Interface 3 primarily consisted of neutral regions within the NACHT domain, interspersed with patches of suitable electrostatics, and a smaller proportion of consistent H bonds (Fig. 4B and Table 1). Furthermore, we assessed the protein–protein interaction preferences of the simulated NLRP12:HCK-KD complex using knowledge-based statistical potential via the PIZSA (Protein Interaction Z-Score Assessment) program. This evaluation determines whether the NLRP12:HCK-KD complex possesses the ability to interact, based on potentials derived from databases of protein–protein partners. All the initial and simulated systems achieved a Z-score of >1.0, indicating a stable association (SI Appendix, Fig. S6A). The binding affinities of these complexes, calculated using the PRODIGY program, showed that the initial dock poses (ΔG of ~−7.0 kcal/mol) were improved in the final simulated conformation (minimum ΔG of ~−12.0 kcal/mol) (SI Appendix, Fig. S6B). Taken together, these results suggest a stable association of the NLRP12:HCK-KD complex, with notable binding affinity and electrostatic complementarity.

To validate these results in vitro, we expressed full-length HCK, NLRP12 without the PYD (NLRP12–ΔPYD), and NLRP12 without the FISNA domain of NACHT (NLRP12–ΔFISNA). According to our model, which showed that the PYD and FISNA were critical for mediating the interaction of NLRP12 with HCK, we hypothesized that both constructs would show reduced binding with HCK. We overexpressed these proteins in 293T cells and performed co-IPs using HCK as the bait. Co-IP analyses identified that the NLRP12–ΔPYD construct showed almost no binding to HCK, while NLRP12–ΔFISNA retained a low level of interaction with HCK (Fig. 4C). Together, these results support the in-silico data, suggesting that HCK binds NLRP12 at the linker region of PYD and NACHT.

Discussion

The NLR family of innate immune sensors mediates inflammatory signaling and cell death pathways in response to pathogens, PAMPs, DAMPs, and homeostatic disruptions, and NLR dysregulation has been linked to a variety of inflammatory diseases (1, 5, 20, 41, 42). Therefore, understanding the molecular mechanisms of NLR activation is crucial for identifying therapeutic targets to modulate their functions in immunopathology. Mechanistically characterizing regulatory molecules in these pathways is of paramount importance, as targeting the NLRs directly has proven challenging (5).

NLRP12, a member of the NLR family, is known to be involved in autoinflammatory disorders, hemolytic diseases, metabolic conditions, and cancers (7, 17, 19–25, 43–45). Despite recent studies identifying NLRP12’s key role in innate immune cell death, PANoptosis, and its activating triggers (7), little more is known about regulators of NLRP12-mediated PANoptosis. Here, we found a regulatory role for HCK in mediating NLRP12-dependent PANoptosis in response to heme plus TNF treatment. HCK is likely activated through membrane-bound Toll-like receptor (TLR) signaling via the TLR adaptor protein MyD88. TLRs and MyD88 are known to play crucial roles in driving NLRP12-dependent inflammatory cell death in response to heme plus PAMPs or heme plus TNF (7), and MyD88 can regulate HCK expression (46). Additional studies have also linked TLRs, MyD88, and HCK activation (47, 48). Collectively, these studies suggest that TLR signaling through the adaptor MyD88 upregulates HCK to drive its activation and the subsequent NLRP12-mediated inflammatory cell death, PANoptosis, in response to heme plus TNF, though other regulatory factors may also be involved and require further study.

Given the importance of protein–protein interactions for NLRP12–PANoptosome assembly and PANoptosis activation (7, 8), we further investigated the potential for an NLRP12–HCK interaction. Our computational and biochemical analyses suggest that HCK binds to NLRP12 in the intervening region between the PYD and NACHT domains. These results are further supported by a previous yeast two hybrid system screen, where HCK binds specifically to NLRP12’s PYD+NACHT, with the C-terminal 30 amino acids of HCK being sufficient for this interaction (32). However, further biochemical characterization of this interaction and NLRP12 phosphorylation in endogenous settings is limited by several factors, including the transient nature of the interaction between kinases and their substrates, low stoichiometry of tyrosine phosphorylation, and lack of specific murine NLRP12 antibodies. Given these technical limitations, alternative biochemical approaches, such as advanced proteomics and microscopy, would be useful in the future to capture the precise interface and phosphorylation sites involved in HCK-mediated regulation of NLRP12.

Overall, our results suggest that HCK is an upstream regulator of NLRP12-mediated inflammatory cell death, PANoptosis. Given the implications of NLRP12-mediated PANoptosis in infectious, inflammatory, and hemolytic diseases and cancers, HCK may be an attractive therapeutic target for improving patient outcomes across a variety of diseases.

Materials and Methods

Sequence Retrieval and AlphaFold2 Modeling.

The protein sequences of human NLRP12 and HCK were retrieved from the UniProtKB database using the identifiers P59046 and P08631, respectively (49). Domain boundaries for HCK were obtained from published literature, while those for NLRP12 were identified by performing pairwise sequence alignment using the EMBOSS Needle program with human NLRP3 and subsequently cross-verified with UniProtKB domain annotations (50). Structure modeling of these protein sequences was performed using AlphaFold2 via ColabFold v1.5.5 (34, 35). The complete protein modeling tasks for each sequence were performed independently using the St. Jude High-Performance Computing (HPC) facilities and CNVRG AI/ML platform. Sequence similarity searches were conducted against the UniRef and environmental datasets using the MMseqs2 program, and multiple sequence alignments were generated in "unpaired" mode to obtain a large number of database sequences (51). The resulting protein models were then visually represented using PyMOL 2.8, with rendering based on the plDDT score provided in the temperature field, to identify regions of high and low confidence. The protein models were prepared using the Protein Preparation Wizard of Schrödinger suite v2023.1 with the following steps: bond angles were assigned, hydrogens were replaced and rebuilt, heterogeneous states at pH 7.4 were generated, and protonation states were assigned using the PROPKA method (52). Subsequently, the models were energy-minimized using the OPLS4 force field (53) with the steepest gradient technique for 1,000 iterative steps until convergence to heavy atoms at 0.30 Å RMSD.

Phosphorylation Site Prediction and Protein–Protein Docking.

Phosphorylation sites in the NLRP12 protein sequence were predicted using the NetPhos v3.1 program with the "kinase specific" predictor to identify sites with phosphorylation potential. NetPhos employs artificial neural networks to predict kinase-specific sites and score their potentials (54). The modeled kinase domain of HCK was docked with the modeled NLRP12 structure using the HADDOCK program (37, 38). Phosphorylation sites predicted in the PYD-linker-NACHT multidomain interface of NLRP12 and the C-terminal residues (Y522-P526) of HCK were selected as active site residues, with the surrounding residues designated as passive site residues to guide interface docking. The HADDOCK method employs three successive steps: randomization of the input molecules’ orientation and rigid-body minimization, semiflexible simulated annealing of the rigid-body docked poses in the torsion angle space, and final refinement of the optimized poses in explicit solvent (37, 38). All the resultant dock poses are then scored by the HADDOCK scoring function, which is a weighted sum of different energy components, including electrostatics, van der Waals, desolvation, restraints violation, and buried surface area. The poses are clustered based on RMSD measure and average HADDOCK score.

Dynamic Simulations of NLRP12–HCK Complexes.

The docked pose of the NLRP12–HCK complex was simulated for a time scale of 500 ns using the DESMOND module of the Schrödinger suite v2023.1 (academic license) (55). These docked poses were initially prepared following the standard procedures of the Protein Preparation Wizard, with atom types assigned using the OPLS4 force field (53). The systems were then solvated with the single point charge water model, enclosed in an orthorhombic box with periodic boundary conditions, and neutralized with counterions using the System Builder module. Following equilibration, the molecular dynamics simulations were conducted under NPT (constant number of atoms, pressure, and temperature) ensemble conditions. Position restraints were applied to the phosphorylation residues within the PYD-linker-NACHT multidomain interface of NLRP12 and the C-terminal Y522-P526 residues of the HCK-KD. System dynamics were maintained using the Nosé–Hoover thermostat at 300 K (relaxation time = 1 ps) and the isotropic Martyna–Tobias–Klein barostat at 1.01325 bar (relaxation time = 2 ps) (56, 57). The smooth particle mesh Ewald method, combined with the RESPA integrator, was employed to evaluate short- and long-range electrostatics with a cutoff of 9 Å (58, 59). Simulation frames were recorded at 500 ps intervals, resulting in a total of 1,000 frames. The dynamic simulations were performed using the St. Jude HPC.

Analysis of Simulation Frames and Intermolecular Interactions.

RMSD and RMSF were analyzed using the Simulation Interaction Diagram module. Simulation frames in PDB format were subsequently examined for intermolecular interactions via the PDBePISA webserver, with hydrogen bonds detected using HBPLUS v3.06 (60, 61). DCCM were computed for each NLRP12–HCK trajectory using the Bio3D R package, with frames initially converted to DCD trajectory files using the Visual Molecular Dynamics visualizer (62, 63). Electrostatic calculations for the simulated NLRP12–HCK complexes were performed using the DelPhi program (64), applying the AMBER force field at pH 7.0 (65). The linear Poisson–Boltzmann solver was utilized with a grid resolution of 1.0 Å, internal and external dielectric constants set at 4 and 80, respectively, and surface calculations using a probe radius of 1.40 Å. Interaction preferences and binding affinities of the NLRP12–HCK complexes were assessed using the PIZSA and PRODIGY webservers, respectively (66, 67). Visualization of protein structures and electrostatic potentials was conducted using ChimeraX v1.8 (68), and data plots were created using GraphPad Prism v10.3.0.

Co-Immunoprecipitation.

For immunoprecipitation in the overexpression system, HEK293T cells (CRL-3216, ATCC) were seeded into six-well plates and transfected with a combined total of 2 mg of the following plasmids (1:1, molar ratio): pGENlenti-GFP-hHCK-MYC (human HCK cloned in pGENlenti mammalian expression vector) and pGENlenti–GFP–hNLRP12–MYC (human NLRP12, NLRP12–ΔPYD, or NLRP12–ΔFISNA genes cloned in pGENlenti mammalian expression vector). Transfected cells were then incubated for 24 to 36 h. Cells were collected and lysed in ice-cold lysis buffer containing 20 mM Tris-HCl (pH 7.4), 150 mM NaCl, 1% Triton X-100, 10% glycerol, 1 mM Na₃VO₄, 2 mM PMSF, EDTA-free protease inhibitor cocktail (A32965, Thermo Fisher Scientific), phosphatase inhibitor cocktail (4906845001, Sigma), and 25 mM Z-VAD-FMK (S7023, Selleck-chem). After centrifugation at 16,000 × g for 10 min, the lysates were incubated with either IgG control antibody (#3900S, Cell Signaling Technology [CST]) or anti-HCK antibody (sc-101428, Santa Cruz) overnight at 4 °C. The immunoprecipitated proteins were then added to washed Protein A/G magnetic beads (88802, Thermo Fisher Scientific) and incubated for 2 h at 4 °C. Subsequently, the beads were washed 4 times with lysis buffer and boiled in sodium dodecyl sulfate (SDS) loading buffer at 100 °C for 5 min. Immunoprecipitants in sample buffer were subjected to immunoblotting analysis with 0.5% of input sample and 12% of pull-down sample per lane. After blocking nonspecific binding with 5% skim milk, membranes were incubated overnight with primary antibodies against Myc (#2276, CST, 1:1,000). Membranes were then washed and probed with the appropriate horseradish peroxidase (HRP)–conjugated secondary antibodies (anti-mouse [315-035-047], Jackson ImmunoResearch Laboratories). Immunoblot images were acquired on an Amersham imager using Immobilon Forte Western HRP Substrate (WBLUF0500, Millipore) or SuperSignal West Femto Maximum Sensitivity Substrate (34096, Thermo Fisher Scientific).

Differentiation of BMDMs and siRNA-Mediated Transfection.

Primary BMDMs from wild-type mice were grown for 6 d in BMDM growth medium, containing IMDM (12440053, Thermo Fisher Scientific) supplemented with 30% L929 conditioned media, 10% heat-inactivated fetal bovine serum (HI-FBS; S1620, Biowest), 1% penicillin and streptomycin (15070-063, Thermo Fisher Scientific), and 1% non-essential amino acids (11140-050, Thermo Fisher Scientific). On day 6, cells were harvested, washed with PBS two times, and resuspended in BMDM growth medium. BMDMs were then seeded at a density of 1 × 10⁶ cells/well in 12-well plates or 5 × 10⁵ cells per well in a 24-well plate in BMDM growth media and grown overnight.

To knockdown gene expression in BMDMs through siRNA, BMDMs were cultured until day 5, washed twice with PBS, and suspended in 10 mL PBS. They were then centrifuged at 300 × g for 5 min and suspended in PBS at a concentration of 10⁷ cells per mL. Then, 5 picomoles non-targeting siRNA (D-001206-14-20, Horizon Discovery), mouse-specific Hck siRNA (M-040986-00-0005, Horizon Discovery), or mouse-specific Nlrp12 siRNA (M-060234-01-0005, Horizon Discovery) per million cells was added. Transfection was achieved by electroporation (Neon Transfection System kit, MPK5000, Thermo Fisher Scientific). After transfection with siRNA, BMDMs were resuspended in BMDM growth medium and seeded at a density of 5 × 10⁵ cells per well in a 24-well plate. The medium was replaced with fresh media the next day. BMDMs were stimulated 48 h posttransfection.

Hemin Preparation.

The 100 mM Hemin (heme [ferriprotoporphyrin IX chloride]; H9039, Sigma or H651-9, Frontier Scientific) stock was prepared by dissolving heme with filter sterilized 0.1 M NaOH and neutralizing to a pH of 7.2 with 1 M HCl. New stocks were prepared before each experiment.

Bacteria Culture.

Salmonella enterica serovar Typhimurium (Salmonella) strain SL1344 was cultured in Luria–Bertani media (3002-031, MP Biomedicals) and incubated overnight under aerobic conditions at 37 °C. The culture was then diluted 1:10 in fresh LB media for 3 h at 37 °C to obtain log-phase bacteria. For infection experiments, S. Typhimurium (MOI 1) was used.

Cell Stimulation.

BMDMs were stimulated in DMEM (11995-065, Gibco) containing 10% HI-FBS and 1% penicillin and streptomycin, using combinations of 50 μM heme and 100 ng/mL TNF (315-01A, Peprotech), 100 ng/mL ultrapure LPS (tlrl-3pelps, InvivoGen), and 5 mM ATP (101275310001, Roche), and 10 μM ValboroPro (5.31465, Millipore Sigma). A total of 250 μL media containing the indicated ligands was used for stimulation in 24-well plates. For LPS plus ATP stimulation, BMDMs were primed with 100 ng/mL ultrapure LPS for 4 h, followed by stimulation with 5 mM ATP for 45 min. Additionally, 2 µM MCC950 (inh-mcc, Invivogen) was included in LPS plus ATP stimulation experiments where indicated. For ValboroPro stimulation, cells were treated with 10 μM ValboroPro for 16 h to evaluate cell death.

Real-Time Imaging for Cell Death.

Cell death kinetics were monitored with the IncuCyte S3 or SX5 (Sartorius) live-cell analysis system. BMDMs were seeded into 24-well plates at a density of 5 × 10⁵ cells/well and then treated with the indicated stimuli. Propidium iodide was used to measure cell death, following the manufacturer’s protocol. Every 1 h, the plate was scanned to acquire fluorescence and phase-contrast images. PI-positive cells were marked with a red mask and quantified using the software package supplied with the IncuCyte imager.

Immunoblot Analysis for PANoptosis Molecules and Endogenous HCK.

Following the appropriate treatments, cells were lysed along with culture supernatants in caspase lysis buffer [containing 10% NP-40, 25 mM DTT, and a mixture of protease and phosphatase inhibitors (11697498001 and 04906837001, respectively, Roche)] and SDS sample loading buffer (with 2-mercaptoethanol) for probing caspase activation. For immunoblot analysis of signaling activation, culture supernatants were removed, and cells were washed once with Dulbecco’s PBS (DPBS; 14190-250, Thermo Fisher Scientific) before being lysed in RIPA buffer and SDS sample loading buffer. For immunoblot analysis of LDH and HMGB1 in supernatant, the supernatant was collected and centrifuged at 8,000 × g for 2 min. After removing cell debris, the obtained supernatant was mixed with 4× sample loading buffer. Proteins were separated on 8 to 12% polyacrylamide gels and transferred onto PVDF membranes (IPVH00010, Millipore) using the Trans-Blot Turbo system. To block non-specific binding, membranes were incubated in 5% skim milk, followed by overnight incubation with primary antibodies against: caspase-1 (AG-20B-0044, AdipoGen), caspase-3 (#9662, CST), cleaved caspase-3 (#9661, CST), caspase-7 (#9492, CST), cleaved caspase-7 (#9491, CST), caspase-8 (AG-20 T-0138, AdipoGen), cleaved caspase-8 (#8592, CST), pMLKL (#37333, CST), tMLKL (AP14272B, Abgent), GSDMD (ab209845, Abcam), GSDME (ab215191, Abcam), LDHA (Proteintech, 19987-1-AP), HMGB1 (ab79823, Abcam), HCK (sc-101428, Santa Cruz), and β-actin (sc-47778 HRP, Santa Cruz). Membranes were then washed and incubated with the appropriate HRP-conjugated secondary antibodies (anti-mouse and anti-rabbit). Immunoblot images were acquired using an Amersham Imager with Immobilon Forte Western HRP Substrate or SuperSignal West Femto Maximum Sensitivity Substrate.

RT-PCR Analysis.

Total RNA was extracted at indicated time points using TRIzol (15596026, Thermo Fisher Scientific). cDNA was synthesized with 1 μg of extracted RNA using the High-Capacity cDNA Reverse Transcription Kit (4368814, Applied Biosystems). Real-time quantitative PCR was performed with SYBR Green (4368706, Applied Biosystems) as the fluorescent reporter on an Applied Biosystems 7500 real-time PCR instrument. The mouse primer sequences used were Hck- Forward primer: TGCTTGTCTGTTCGAGACTTTG, Reverse primer: TCTTGTAGTGGAGCACGAGTT; β-actin – Forward primer: GGCTGTATTCCCCTCCATCG, Reverse primer: CCAGTTGGTAACAATGCCATGT.

For Nlrp12 transcript analysis, total RNA was extracted and isolated at the indicated time points with the PureLink RNA Mini Kit (12183025, Invitrogen) according to the manufacturer’s instructions. RNA was quantified with a NanoDrop 2000 (Thermo Fisher Scientific). Then, 5 µg of RNA was reverse transcribed with the High-Capacity cDNA Reverse Transcription Kit (4374966, Thermo Fisher Scientific). The amplified cDNA was diluted to 100 ng/µL, and real-time quantitative PCR was performed with the TaqMan custom-generated probe for mouse Nlrp12 (assay ID: APRWP2W) and TaqMan Fast Advanced Master Mix (4444557, Applied Biosystems) on an Applied Biosystems 7500 real-time PCR instrument. The delta–delta cycle threshold method was used to normalize expression to that of the reference gene 18S (4319413E, Applied Biosystems).

Quantification and Statistical Analysis.

GraphPad Prism v10 software was used for in vitro data analysis. Data are shown as mean ± SEM. Statistical significance was determined using unpaired t tests (two-tailed) for two groups or one-way ANOVA (with Dunnett’s multiple comparison tests) for three or more groups. Statistical significance is represented as * for P < 0.05, ** for P < 0.01, *** for P < 0.001, and **** for P < 0.0001.

RNA-Seq Dataset Analysis.

Previously published gene expression datasets were reanalyzed to investigate the expression of Src family kinases. We analyzed four publicly available datasets: GSE34404 (11) (whole blood microarray of 155 West-African children), GSE136046 (12) (RNA-Seq of BM CD71⁺ cells from Brazilian patients with malaria at day 1 and 42 posttreatment), GSE102881 (13) (RNA-Seq of CD34⁺ cells from healthy donors and patients with SCD), and GSE168532 (14) (RNA-Seq of classical monocytes from healthy donors and patients with SCD). Each dataset was first processed through quality control measures, including normalization using the “normalize.quantiles” function from the preprocessCore v1.66.0 package, followed by log2 transformation of the read counts for further analysis. For differential gene expression analysis, we employed the limma v3.60.2 package (69). This involved using the “lmFit” function to construct linear models from expression matrices containing log-transformed counts. Models were designed to compare healthy versus disease conditions, with empirical Bayes moderation of SE applied using the “eBayes” function. Differentially expressed genes were identified based on a Benjamini–Hochberg adjusted P value threshold of <0.05 for the larger datasets (GSE34404 and GSE168532), and a raw P value threshold of <0.05 for the smaller datasets (GSE136046 and GSE102881) (70). The Src kinases were ranked by their mean fold-change (logFC) across these datasets and visualized using a heatmap generated with the Complex Heatmap v2.20.0 package (71). All calculations were performed using R v4.4.0.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

RNA sequencing datasets have been deposited in the Gene Expression Omnibus database [GSE34404 (72), GSE136046 (73), GSE102881 (74), and GSE168532 (75)]. All other data are included in the manuscript and/or SI Appendix.