Inherently Reduced Expression of ASC Restricts Caspase-1 Processing in Hepatocytes and Promotes Plasmodium Infection

Malaria, caused by Plasmodium parasites remains an unresolved global health burden that impacts more than half of the world’s human population. Upon inoculation into its mammalian hosts as sporozoites, Plasmodium undergoes obligatory replication and development in the hepatocytes, where the sporozoites transition into the merozoite stage that infects the RBCs. Infection of the RBCs would initiate the blood stage of malaria, which is responsible for almost all of the morbidity and mortality associated with this disease. The interventions that block the progression of Plasmodium to its blood stage are therefore expected to prevent clinical disease and transmission (1). However, our incomplete understanding of the fundamental biology of Plasmodium infection in the liver has prevented us from effectively targeting the parasites during this stage of its life cycle.

It is well established that innate immune responses generated in the liver can control Plasmodium, potentially limiting the incidence and severity of clinical malaria (2–6). We have shown that the inflammasome pathway is a key component in this process, wherein the receptor, absent in melanoma (AIM) 2, detects Plasmodium DNA and induces inflammasome-mediated caspase-1 activation and pyroptotic cell death in the infected hepatocytes (2). This process is instrumental in controlling Plasmodium infection in the liver and limiting its progression to the blood stage, as well as offering Plasmodium Ags to the Ag presentation machinery in the liver (2–4).

Inflammasomes are large macromolecular complexes generated in the host cell cytosol following the detection of pathogen-associated molecular patterns (PAMPs) such as DNA by intracellular pattern recognition receptors (PRRs). The inflammasomes are typically composed of such PRRs, the adaptor molecule ASC (apoptosis-associated specklike protein containing caspase recruitment domain [CARD]) when the PRRs do not possess their own CARD, and the zymogen pro–caspase-1. Pro–caspase-1 undergoes proximity-driven autoproteolysis at such inflammasome complexes to its constituent CARD, p20, and p10 subunits, following which the p20 and p10 subunits heterodimerize to generate enzymatically active caspase-1 (7). Caspase-1 formed in this manner is believed to proteolytically activate the membrane pore-forming molecule, gasdermin D (GSDMD), and the proinflammatory cytokines IL-1 and IL-18 (7–10).

It is noteworthy that the above paradigm was established via studies conducted primarily in cells of the myeloid lineage (7). The dynamics of pro–caspase-1 processing is quite distinct in the hepatocytes, wherein they undergo only partial proteolysis into a p32 caspase-1 species composed of unseparated p20 and p10 subunits (2). Although p32 activates GSDMD and induces pyroptotic cell death following exposure to PAMPs, it does not facilitate the maturation of IL-1 or IL-18 in the hepatocytes, much like when caspase-1p32 was artificially generated by mutating the pro–caspase-1 gene in bone marrow–derived macrophages (BMDMs) (2, 11). The molecular mechanism responsible for the natural restriction of pro–caspase-1 processing to the p32 form in hepatocytes has remained unknown. Filling this fundamental knowledge gap is critical in improving our current understanding of the innate immune responses in the liver.

Innate immune responses such as inflammasome-mediated caspase-1 activation are also key drivers of downstream adaptive immune responses (12). Caspase-1 function is necessary to recruit APCs that prime protective CD8 T cell responses against hepatotropic infections such as Plasmodium colonizing the liver following natural infections or vaccinations (4, 13–16). Determining why caspase-1 processing is limited in the hepatocytes would also pave the way to identifying potential new intervention strategies to enhance adaptive immune responses against pathogens infecting the liver, such as Plasmodium.

In this work, we show that pro–caspase-1 is terminally processed to the p32 caspase-1 species in hepatocytes because of inherently reduced expression of ASC. Transgenic overexpression of ASC resulted in the maturation of pro–caspase-1 into p20 and p10, generation of mature IL-1β and IL-18, and enhanced GSDMD-mediated pyroptotic cell death in the hepatocytes. This also resulted in significantly better innate immune control of liver-stage malaria. In addition to advancing our basic understanding of liver biology, our discovery is expected to have key implications in antimalarial drug and vaccine designs.

C57BL/6 (B6) mice were purchased from The Jackson Laboratory. All mice were housed with appropriate biosafety containment at the animal care units of the University of Georgia, and the animals were treated and handled in accordance with the guidelines established by it. Anopheles stephensi mosquitos infected with Plasmodium berghei ANKA (Pb) and P. berghei expressing luciferase (Pb-Luc) were reared at the SporoCore insectary facility at the University of Georgia.

Primary hepatocytes were isolated from mice as described in detail before (4). In short, the inferior vena cava of anesthetized (2,2,2-tribromoethanol, 300 mg/kg) mice were catheterized (Autoguard, 22-gauge, BD Biosciences) aseptically to perfuse the liver by draining through the hepatic portal vein. Steady-state perfusion of the liver was performed first with PBS (4 ml/min for 5 min), then with Liver Perfusion Medium (4 ml/min for 3 min, Life Technologies), and finally with the Liver Digest Medium (4 ml/min for 5 min, Life Technologies). The digested liver was excised, and a single-cell suspension was made and resuspended in a wash solution of 10% FCS (Sigma-Aldrich) in DMEM (Life Technologies). The hepatocyte fraction was recovered by centrifugation at 57 × g, from which the debris and the dead cells were removed by density gradient centrifugation with a 46% Percoll gradient (GE Healthcare). The remaining cells were counted and resuspended in DMEM with 10% FCS. Typically, 6 × 10⁵ of these cells, the hepatocytes, were plated on flat-bottomed, collagen-coated plates and incubated at 37°C, 5% CO₂. A total of 6–8 × 10⁵ primary mouse or human (obtained from BioIVT) hepatocyte cultures were infected with 2–4 × 10⁴ sporozoites in each well of a 6-well plate. Cultures were further incubated for the desired time to allow infection and liver-stage parasite development. Of note, prolonged incubation of primary hepatocyte cultures with Pb sporozoites is required to induce adequate infection and caspase-1 activation (2, 17). Therefore, the 16-h time point was used to assess caspase-1 processing in the Pb-infected hepatocytes. To serve as comparable controls, the hepatocytes were stimulated with LPS (100 ng/ml, InvivoGen) and ATP (5 mM, Sigma-Aldrich) for 16 h. LPS and ATP were resuspended in media. Disulfiram was used as a pretreatment in vitro in some experiments at 100 nM for 6 h before LPS and ATP addition. A total of 3 × 10⁶ BMDMs were treated with LPS (100 ng/ml, 3.5 h) followed by 5 mM ATP (0.5 h) as described in detail before (18). Culture supernatants and/or cell lysates were obtained at various time points postinfection (p.i.) or poststimulation as indicated.

Liver nonparenchymal cells (LNPCs) were isolated as previously described (19). In short, the livers were perfused and excised, and a single-cell suspension was made and obtained as described above. The liver homogenate was subjected to 50 × g centrifugation for 3 min to recover the total residual liver parenchymal cells in the supernatant, which was collected and centrifugated at 900 × g for 8 min (4°C) to collect the LNPCs. The collected LNPCs were resuspended 36% Percoll (GE Healthcare) solution made with FCS-supplemented DMEM and centrifugated at 859 × g for 20 min without braking. The LNPC fraction collected was used as is, or the Kupffer cells were isolated by cell adhesion to plastic for 2 h.

Distinct datasets were built for mice and humans. Mouse RNA-sequencing Illumina reads were retrieved from the National Center for Biotechnology Information Sequence Read Archive (https://www.ncbi.nlm.nih.gov/sra) under accession numbers SRR9961636, SRR9961637, SRR9961638, SRR9961644, SRR9961645, and SRR9961646 for hepatocytes and SRR6246979, SRR5714120, and SRR5714119 for macrophages. The human hepatocyte data used were from our published single-cell RNA-sequencing results (2). Data on monocytes were retrieved from the Sequence Read Archive under accession numbers SRR12539860, SRR12539861, and SRR12539862. The Illumina reads were aligned against the National Center for Biotechnology Information reference genome for mice (GRCm38) or humans (GRCh38) using HISAT2 version 2.1.0 (20). The alignment files were then sorted by SAMtools version 1.6 (21) and submitted to HTseq version 0.9.1 (22) to generate the count file. The count file was then used for the differential expression analysis under R using DEseq2 (23). The data were normalized and filtered for p values <0.01.

BMDMs were prepared as described in detail previously (18). In short, the bone marrow cells were grown in L-cell-conditioned IMDM (Thermo Fisher) supplemented with 10% FCS, 1% nonessential amino acids, and 1% penicillin-streptomycin for 5 d to differentiate into macrophages. On day 5, BMDMs were seeded in 6-well cell culture plates. The next day, the BMDMs were ready for use.

For sporozoite challenge experiments, salivary glands of parasitized A. stephensi mosquitoes were dissected, and the sporozoites were isolated, counted, and injected 2 × 10⁴ in 200 μl RPMI with 1% mouse serum (Innovative Research) retro-orbitally as described in detail before (24).

Cells obtained from the liver were stained with anti-CD45 (BioLegend, clone 30-F11), F4/80 (BioLegend, clone BM8), CD11c (Tonbo, clone N418), or CSF1R (BioLegend, clone AF598) to determine the frequencies and counts of the myeloid cells as presented in detail before (4). To determine CD8 T cell responses, blood was collected from mice through tail or retro-orbital bleeding, the RBCs were lysed with ammonium-chloride-potassium lysis buffer, and the leukocyte fraction was stained. The cells were stained for surface markers with appropriate Abs or the GAP50 tetramer (National Institutes of Health Tetramer Core Facility) in PBS for 30 min prior to washing and resuspending in PBS to analyze by flow cytometry as described in detail before (4, 25, 26). CSF1R⁺ APCs were designated as CD45⁺ F4/80⁺ CD11c⁺ CSF1R⁺ cells, and Plasmodium-specific CD8 T cells were designated as GAP50 tetramer⁺ CD8 T cells.

For T cell assays, staining was performed as previously described (27, 28). Briefly, spleens from mice were harvested, homogenized, and passed through a 70-μm mesh filter to obtain a single-cell suspension. The homogenate was treated with RBC lysis buffer (EasyStep, STEMCELL Technologies), washed, and resuspended in complete DMEM. A total of 2 × 10⁶ splenocytes were stimulated with 10 μg/ml of GAP50 peptide (SQLLNAKYL, GenScript) in the presence of 10 μg/ml brefeldin A (BioLegend) at 37°C for 5 h. Cells were then washed with FACS buffer (PBS supplemented with 1% FCS), resuspended, and stained with anti-CD8-PE (Tonbo, clone 53.67) and anti-CD44-BV605 (BD Biosciences, clone IM7) Abs for 20 min on ice. The cells were then washed twice with FACS buffer, fixed, and permeabilized by incubating for 30 min in Fix/Perm buffer (Tonbo) on ice. Cells were washed in Perm wash buffer (Tonbo) and stained with anti–IFN-γ allophycocyanin (BD Biosciences, clone XMG1.2) and anti–TNF-α-FITC (BioLegend, clone MP6-XT22) for 30 min on ice. Cells were then washed twice in Perm wash buffer and resuspended in FACS buffer for cytometric analysis. Data were acquired on a Quanteon flow cytometer (Agilent) and analyzed with FlowJo (FlowJo LLC).

Liver parasite burden was assessed by quantitative real-time RT-PCR for parasite 18s rRNA in the livers of mice challenged with sporozoites isolated from infected mosquitoes as described before (29). Total RNA was extracted from the liver at the indicated time points after Plasmodium infection using TRIzol (Sigma-Aldrich), followed by DNase digestion and cleanup with the RNA Clean and Concentrator kit (Zymo Research). Liver RNA (2 μg per sample) was used for qRT-PCR analysis for Plasmodium 18S rRNA using TaqMan Fast Virus 1-Step Master Mix (Applied Biosystems). Data were normalized for input to the GAPDH control for each sample and are presented as ratios of Plasmodium 18s rRNA to GAPDH RNA signals. The ratios depict relative parasite loads within an experiment and do not represent absolute values. Pb-Luc was used to assess the kinetics of replication and clearance of Plasmodium infection in the liver of mice. For bioluminescence detection, mice were injected with d-luciferin (150 mg/kg; SydLabs) i.p. 5 min prior to being anesthetized using 2% (v/v) gaseous isoflurane in oxygen and imaged using an IVIS 100 imager (Xenogen). The quantification of bioluminescence and data analysis were performed using Living Image version 4.3 software (Xenogen).

In the propidium iodide (PI) uptake assay for pyroptosis, PI staining distinguishes programmed cell death from whole-cell lysis (30). To determine the extent of pyroptotic cell death, hepatocytes transfected with designated plasmids were seeded at 2 × 10⁴ cells per well of an opaque-walled, clear-bottomed, 96-well plate (Corning) overnight. Culture media were replaced with DMEM without phenol red, and cells received the specified treatments or infections. All culture wells were supplemented with 6 mg/ml PI (Invitrogen) and were incubated at 37°C for 40 min. Subsequently, the cells were washed thrice with PBS, and the plate was sealed with a clear, adhesive optical plate seal (Applied Biosystems). The frequencies of PI-stained cells were estimated using a plate reader (Molecular Devices, SpectraMax iD3), and 100% cytolysis was calculated. The maximum signal was determined following treatment with 1% Triton X. Data were presented as experimental group PI signal/maximum PI signal as a percentage and were normalized using untreated sample levels. Untreated hepatocyte cultures overexpressing ASC or enhanced GFP (eGFP) exhibited no discernible differences in basal PI staining levels.

Hepatocytes (7 × 10⁵ cells) were transfected with various plasmids (at 6 μM or indicated concentrations) using the Mouse/Rat Hepatocyte Nucleofector kit (Lonza) following the manufacturer’s protocol, yielding ∼50% transfection efficiency in the primary cells. For overexpression of ASC, pcDNA3-N-Flag-ASC1 (Addgene, 75134) plasmid was used. For overexpression of pro–caspase-1, pcDNA3-N-Flag-Caspase-1 (Addgene, 75128) was used. The transfected cells were transferred to collagen-coated wells 16 h prior to infection or the treatments. For silencing, cells were transfected with ON-TARGETplus SMARTpool siRNAs (Dharmacon). Differentiated BMDMs were detached by treatment with trypsin and resuspended at a concentration of 10⁶ cells per 100 μl of nucleofector solution. These cells were transfected with ON-TARGETplus SMARTpool siRNAs (Dharmacon) using the Mouse Macrophage Nucleofector kit (Amaxa) following the manufacturer’s protocol. The transfected cells were transferred to plates and allowed to recover for 24–48 h prior to infections or treatments.

For imaging, hepatocytes were cultured overnight in collagen-coated chamber slides (ibidi), transfected with ASC-encoding plasmid, treated with LPS+ATP as described above, stained for fluorochrome-labeled inhibitors of caspases (FLICA) as per the manufacturer’s protocol using the FAM-FLICA-Caspase-1 (YVAD) assay kit (Immunochemistry), and visualized using fluorescence microscopy (Keyence). In short, the adherent cells were incubated for 1 h at 37°C, washed twice with 1× apoptosis buffer, and stained with Hoechst stain. The fluorescence signal intensity of the FLICA-stained inflammasome complexes, as well as the area, was determined using ImageJ software (National Institutes of Health) from the acquired images and presented as mean signal intensity, calculated as background-corrected mean fluorescence/area (31).

The following treatment regimen was used in this study: disulfiram (Sigma-Aldrich) 50 mg/kg i.p. in sesame oil.

Adeno-associated virus (AAV) vectors were generated by VectorBuilder by encoding the genes of interest under the control of the albumin promoter to limit target protein expression to the hepatocytes (32). The AAV-DJ strain was used as the background for efficient transduction of hepatocytes in vivo (33). Virus stocks were resuspended in PBS and inoculated i.v. at 1 × 10¹¹ genome copies/mouse. Experiments were performed 10–14 d after inoculation for efficient transduction and protein expression.

To quantify specific protein levels in culture supernatants of ex vivo cultured hepatocytes, ELISA was performed as described in detail before (34). In short, serial dilutions of supernatants or whole-cell lysates (normalized for total protein) were coated in triplicate in 96-well format in 0.1 M carbonate bicarbonate buffer (pH 9.6) overnight at 4°C, washed thrice with 0.05% Tween 20 in PBS (PBS-T), blocked for 60 min with 1% BSA/PBS, probed with murine anti-caspase-1 (AdipoGen, clone Casper-1), anti-ASC (Santa Cruz Biotechnology, F-9), anti–IL-1β (R&D Systems, polyclonal), or anti–IL-18 (MBL International, clone 125-25) as applicable for 60 min at 37°C, washed five times with PBS-T, probed with the corresponding HRP-conjugated secondary Abs in 1% BSA/PBS, and again washed three times with PBS-T before developing. All ELISAs were developed using a tetramethylbenzidine liquid substrate system (Sigma-Aldrich), stopped with 2 N sulfuric acid, and then read at 450 nm using an ELISA microplate reader (Molecular Devices, SpectraMax iD3).

Western blot analysis was performed as described before (2, 18). In short, the cells in culture were lysed along with the supernatant using radioimmunoprecipitation assay buffer or sample loading buffer containing DTT and NaDodSO₄ (SDS) at the indicated time points. The supernatant was included in the assays inducing caspase-1 activation because caspase-1 activation and cell death would release the cytosolic contents into the supernatant. The protein samples were run on 12% SDS polyacrylamide gel by electrophoresis and transferred to polyvinylidene difluoride (Millipore) or nitrocellulose (Bio-Rad Laboratories) membranes. After blocking with 5% skimmed milk or Odyssey blocking buffer (LI-COR Biosciences) for 1 h at room temperature, the membranes were probed with the following primary Abs: mCaspase-1p20 (AdipoGen, clone Casper-1), mCaspase-1p10 (AdipoGen, clone Casper-2), mIL-1β (Cell Signaling Technology, clone D3H1Z), mIL-18 (MBL International, clone 39-3F), mGSDMD (Abcam, clone EPR19828), or mASC (Santa Cruz Biotechnology, clone F-9) at 4°C overnight. Then, they were washed with Tris-buffered saline containing 0.1% Tween-20 (TBST) four times and incubated at room temperature for 45 min with polyclonal secondary anti-rabbit, anti-rat, or anti-mouse chemiluminescence (Jackson ImmunoResearch) or IRDye-conjugated (LI-COR Biosciences) Abs as applicable. After four washes with TBST, the proteins were visualized using a chemiluminescence detection reagent (Millipore) or directly by fluorescence (LI-COR Biosciences). Loading controls (LCs) reflect the total amount of protein in the specified lane and rely on the housekeeping protein β-tubulin when whole-cell lysates were analyzed. When culture supernatants also constituted the samples, an arbitrary protein band in the corresponding SDS-PAGE stained with Coomassie brilliant blue dye acted as the LC, as published before (2, 35, 36). This approach is known to provide a more precise comparison across samples (37). Relative band densities were not averaged across replicate blots, because the signal intensities representing the same amount of protein vary drastically with varying imaging conditions, such as the exposure time, the detection systems used, etc., and is known to be nonlinear when infrared or chemiluminescence detection methods are applied (38, 39).

Chemical stripping of the Western blot membranes when employed followed the manufacturer’s protocol (LI-COR Biosciences). In short, membranes were incubated for 7 min in fluorescent stripping buffer, washed three times in PBS, and reprobed with the primary Abs. Band density quantification reflects the relative protein amounts as a ratio of each protein band signal intensity to the lane’s LC, determined using ImageJ software (National Institutes of Health).

Data were analyzed using Prism version 7 software (GraphPad Software) and as indicated in the figure legends.

On the basis of studies in BMDMs, caspase-1 activation is believed to occur through autoproteolysis of pro–caspase-1 into its constituent CARD, p20, and p10 subunits, following which the p20 and p10 subunits heterodimerize to generate enzymatically functional caspase-1 (Supplemental Fig. 1A). In contrast, as we have shown previously in both human and mouse hepatocytes, pro–caspase-1 is terminally processed to a p32 caspase-1 species composed of unseparated p20 and p10 subunits (Supplemental Fig. 1A) (2). Caspase-1 p32 has been consistently observed in primary murine hepatocyte cultures infected with P. berghei (Pb) or treated with the standard inducers of the inflammasome pathway, LPS+ATP, and for up to 24 h after exposure (Supplemental Fig. 1B, 1C), indicating that it is a stable protein species (2). Why caspase-1 processing is naturally limited to its intermediate p32 form in the hepatocytes remains unknown.

We considered the possibility that the incomplete processing of caspase-1 in hepatocytes is due to alternate splicing of pro–caspase-1, making the interdomain linker (IDL) connecting the p20 and p10 domains not amenable to proteolytic cleavage. Nevertheless, bioinformatic analysis of our (4) or others’ (40–43) published transcriptomic data did not reveal any pro–caspase-1 splice variants in human or mouse hepatocytes. Furthermore, no detectable alternately spliced transcripts of pro–caspase-1 were detected in either naive, Plasmodium-infected, or LPS+ATP-stimulated primary murine hepatocytes (2).

The adaptor molecule ASC is critical for seeding the inflammasome platform (also called “ASC specks”) that assimilates pro–caspase-1 molecules locally in the host cell cytoplasm, imparting it a catalytically active quaternary structure and autoproteolytic activity (7, 31, 44, 45). Examination of our (4) and other publicly available transcriptomic data (40–43) indicated that both pro–caspase-1 and ASC transcripts are inherently less abundant in both mouse and human hepatocytes when compared with the cells of myeloid lineage from either species (Fig. 1A). This was corroborated by their baseline protein expression as well (Fig. 1B, 1C). A lower overall expression of pro–caspase-1 would result in fewer pro–caspase-1 molecules being recruited to the inflammasome complex, impeding its processing. A low expression of ASC would, on the other hand, result in its suboptimal oligomerization, leading to lower assimilation of pro–caspase-1 at the inflammasome complexes (46, 47). Therefore, the reduced expression of pro–caspase-1 or ASC was a potential explanation for the suboptimal processing of pro–caspase-1 in hepatocytes.

Stimulation with PAMPs such as LPS is known to increase the expression of intracellular PRRs and pro–caspase-1 in various cell types (48, 49). Therefore, to determine if the reduced expression of pro–caspase-1 and ASC remained intact in the context of PAMP stimulation in the hepatocytes, we treated both hepatocytes and BMDM controls with LPS+ATP. Although the difference in pro–caspase-1 expression between hepatocytes and BMDMs became less pronounced following PAMP stimulation, the expression of ASC continued to remain reduced in the hepatocytes despite LPS stimulation (Fig. 1D, 1E). Together, these findings indicated that the baseline expression of ASC in hepatocytes is naturally lower than in the myeloid cells.

The formation of dense ASC specks that bring pro–caspase-1 molecules in close proximity to each other in the cytoplasm is a critical determinant of efficient caspase-1 processing and function (31, 46). We predicted that transgenically enhancing the expression of pro–caspase-1 and/or ASC in hepatocytes would facilitate the formation of compact inflammasome complexes, leading to complete pro–caspase-1 processing, resulting in the generation of the caspase-1 p20 subunit. To test this hypothesis, we overexpressed pro–caspase-1 or ASC in primary murine hepatocytes (Fig. 2A, 2B). Overexpression of ASC, but not pro–caspase-1, in primary murine hepatocytes generated a discernible caspase-1 p20 cleavage product following the stimulation with LPS+ATP (Fig. 2C). eGFP served as the control for transgenesis. Overexpression of ASC also generated caspase-1 p20 following Pb infection (Fig. 2D). Notably, the overexpression of ASC also facilitated the generation of higher p32 levels in such hepatocytes, possibly due to the higher overall inflammasome-mediated caspase-1 activation in such cells.

Enhancing ASC expression also resulted in the generation of more condensed and catalytically active inflammasome complexes in the hepatocytes, as evidenced by ASC specks with relatively higher intensity of FLICA staining (Fig. 2E, 2F). It is known that the intensity of FLICA staining directly, as well as the size of ASC specks inversely, correlates with inflammasome and caspase-1 activity in cells (50). Taken together, the above findings suggested that the baseline expression of ASC in host cells may be a determinant of the extent of caspase-1 activation. This meant that tempering ASC expression in BMDMs would result in incomplete processing of caspase-1 and the generation of p32 in BMDMs following PAMP stimulation. To test this, we transfected BMDMs with increasing doses of siRNA to tune down ASC expression without completely eliminating it (Supplemental Fig. 2A). Although fully abrogating the expression of ASC prevented any caspase-1 activation in BMDMs following LPS+ATP stimulation (Supplemental Fig. 2B), tempering it resulted in the progressive appearance of p32, with concurrently diminishing p20 levels (Supplemental Fig. 2C). These observations further confirmed that the level of ASC expression in host cells is a key determinant of the extent of pro–caspase-1 processing in the context of ASC-mediated inflammasome activation.

Overexpression of ASC in primary murine hepatocytes also generated detectable levels of mature IL-1β and IL-18 (Fig. 2G–2J), as well as, higher levels of active GSDMD, in response to LPS+ATP stimulation (Fig. 2K). This phenotype was also accompanied by a significantly higher cell death response in ASC-transgenic hepatocytes following LPS+ATP stimulation or Plasmodium infection, compared with the control eGFP overexpressing hepatocytes (Fig. 2L). The frequency of eGFP-transgenic hepatocytes undergoing cell death following Pb infection is in the range of what has been observed in wild-type hepatocytes in previous studies (2). Therapeutically blocking GSDMD function using disulfiram negated the above phenotype when examined in the context of the standard LPS+ATP stimulation (Fig. 2M). These data indicated that the enhanced pyroptosis observed in ASC-transgenic hepatocytes was potentially driven by the enhanced GSDMD activation.

The data in Fig. 2C and Supplemental Fig. 2B indicated that higher ASC expression was also associated with slightly higher pro–caspase-1 levels in the hepatocytes and BMDMs. Enhancing ASC expression elevated the levels of pro– IL-1β and IL-18 as well (Fig. 2G, 2H). Although pro–caspase-1 is believed to be constitutively expressed in cells, proinflammatory cytokines such as IL-1 produced as a result of complete processing of caspase-1 can enhance pro–caspase-1 expression through autocrine or paracrine NF-κB–mediated signaling (49, 51). Similarly, activated IL-1 and IL-18 can further induce the expression of pro– IL-1 and IL-18 in a paracrine or autocrine fashion in the host cells (52). Nevertheless, the fact that reducing ASC expression in BMDMs resulted in partial processing of pro–caspase-1, as well as the absence of caspase-1 p20, following overexpression of pro–caspase-1 in hepatocytes pointed toward a direct role of ASC in regulating the extent of caspase-1 processing, plausibly independently of the altered pro–caspase-1 levels.

Taken together, the above data indicated that caspase-1 processing occurs incompletely in hepatocytes, possibly because of the lower expression of ASC, and that enhancing ASC induced complete processing of caspase-1 and production of mature IL-1β and IL-18 and improved GSDMD-mediated pyroptotic cell death when the inflammasome pathway was induced.

GSDMD-mediated pyroptotic cell death induced by the AIM2–caspase-1 axis is a key component of natural immunity to malaria in the liver (2, 3). Our ability to enhance pyroptosis in Plasmodium-infected hepatocytes by augmenting ASC expression suggested that the latter would result in better control of Plasmodium in the liver. To test this, we generated transgenic AAV on the AAV-DJ background where ASC (AAV-ASC) or control eGFP (AAV-eGFP) was placed under an albumin promoter. AAV-DJ variant was employed to ensure high transduction efficiency and protein expression in hepatocytes in vivo, and the albumin promoter was used to restrict the expression of the target proteins to the hepatocytes (33, 53–55). AAV-eGFP or AAV-ASC inoculated mice were challenged with luciferase-expressing Pb (Pb-Luc), and the kinetics of Plasmodium infection was determined (Fig. 3A). ASC overexpression was confirmed by Western blot analysis in primary hepatocytes isolated from the AAV-ASC inoculated mice (Fig. 3B). AAV-ASC inoculated mice showed comparable ASC expression in the total LNPCs and the Kupffer cells, implying that ASC overexpression was possibly restricted to just the hepatocytes in the liver as intended (Supplemental Fig. 3). Transgenically enhancing ASC expression in hepatocytes resulted in the rapid control of Plasmodium in mice (Fig. 3C, 3D). This phenotype was undone, however, by disulfiram treatment (Fig. 3E). These data suggested that increasing ASC expression in hepatocytes would result in improved innate immune control of Plasmodium in the liver, possibly through enhanced pyroptotic elimination of the infected hepatocytes. Of note, we have shown before that proinflammatory cytokines such as IL-1 do not have a direct role in the innate immune control of liver-stage malaria (2).

Intriguingly, enhancing ASC expression in hepatocytes also resulted in an increased influx of total CD11c⁺ APCs, as well as the specialized CSF1R⁺ CD11c⁺ APCs to the liver following Plasmodium infection (Fig. 4A–4C). This may be because of the increased pyroptotic cell death and the production of IL-1 and IL-18 associated with ASC overexpression in hepatocytes (56, 57). Nevertheless, neither IL-1β nor IL-18 was detected in the sera of mice inoculated with AAV-ASC following P. berghei infection (data not shown). Although not pertinent to the direct control of an ongoing infection, the monocyte-derived CSF1R⁺ CD11c⁺ APC subset is responsible for acquiring Plasmodium Ags from the infected hepatocytes undergoing cell death and priming protective CD8 T cell responses (4). Such CD8 T cells are the primary mediators of protection induced by live-attenuated vaccines such as the radiation or genetically attenuated sporozoite-based malaria vaccines (13, 58, 59). These findings therefore predicted that enhancing ASC expression in hepatocytes would lead to better vaccine-induced immunity to malaria.

To test this, we immunized mice inoculated with AAV-eGFP or AAV-ASC with radiation-attenuated sporozoites (RASs) before challenging them with virulent Pb sporozoites (Fig. 4D). Surprisingly, however, AAV-ASC inoculated mice induced significantly lower Plasmodium GAP50 epitope-specific CD8 T cell responses (Fig. 4E, Supplemental Fig. 4A), although the GAP50-specific CD8 T cells generated in both groups of mice were similarly functional (Supplemental Fig. 4B, 4C). The GAP50⁺ CD8 T cell responses act as a reliable surrogate for the overall protective CD8 T cell responses generated against liver-stage malaria and a key predictor of RAS vaccine-induced immunity in the mouse model (25). In support of the above observation, AAV-ASC inoculated mice exhibited suboptimal protection from Pb challenge following RAS immunization (Fig. 4F). Although seemingly counterintuitive, these results are in agreement with the published findings that stronger innate immune responses in the liver impede sporozoite-based vaccine-induced immunity to malaria (60) and is possibly because of the rapid elimination of RAS from the livers of AAV-ASC inoculated mice, precluding the generation of adequate quantity and breadth of protective epitopes following RAS vaccination.

These data show that enhancing ASC expression in hepatocytes induced significantly better control of Plasmodium infection in the liver, potentially by driving a more efficient GSDMD-mediated elimination of the infected hepatocytes. Although this enhanced the recruitment of CSF1R⁺ APCs to the liver, it impeded the generation of protective CD8 T cell responses and vaccine-induced immunity to malaria.

The existing model for the caspase-1 activation dynamics has been derived from studies conducted primarily in cells of myeloid lineage (2, 7, 11, 31). However, caspase-1 processing that occurs in hepatocytes following infections with Plasmodium or stimulation with the PAMPs such as LPS+ATP used to establish the model in myeloid cells does not fit this model (2). Despite its potential implications for basic research and therapeutics pertinent to the liver, the biological reasons behind such a deviation have remained unknown. In this study, we show that caspase-1 processing occurs unconventionally and incompletely in hepatocytes because of the relatively lower expression of ASC. By transgenically enhancing ASC expression in hepatocytes, we enabled complete proteolysis of pro–caspase-1, maturation of IL-1β and IL-18 that was otherwise not detectable, and enhanced GSDMD-mediated pyroptotic cell death. Reducing ASC expression in myeloid cells, on the other hand, resulted in incomplete processing of pro–caspase-1. These findings suggested that ASC expression is a key regulator of the extent of caspase-1 processing in host cells. Although increasing ASC in hepatocytes engendered better innate immune control of liver-stage malaria, it impeded the generation of Plasmodium-specific CD8 T cell responses and vaccine-induced protection against malaria.

The liver is an important barrier organ that mounts robust immune responses against a variety of pathogens, toxins, and allergens that access our body (61–63). The ability of the liver to maintain an immunotolerant environment is critical, however, for its functional integrity (61–65). Although home to a large collection of specialized immune cells such as the Kupffer cells, dendritic cells, NK cells, etc., over 90% of the volume of the liver is made up of its parenchymal cells: the hepatocytes (61, 66, 67). Therefore, how the hepatocytes respond to foreign Ags or infectious agents would greatly influence the overall immune portfolio of the liver (68). Hepatocytes limiting the extent of caspase-1 processing and proinflammatory cytokine responses while retaining the capacity to undergo pyroptosis may be a critical adaptation that maintains the immunotolerant nature of the liver without surrendering the ability to combat pathogens. Gaining a deeper mechanistic understanding of how various innate immune networks operate within the hepatocytes is necessary to harness the full immune potential of the liver and manage liver diseases such as those driven by inflammatory responses, such as liver fibrosis, cirrhosis, etc., and our work is a key step in this direction.

When caspase-1 processing was artificially limited to the p32 form by mutating the interdomain linker sequence of pro–caspase-1, BMDMs were unable to efficiently mature IL-1β in response to LPS+ATP stimulation or Salmonella typhimurium infection (11). Similarly, altering the ASC sequence through site-directed mutagenesis restricted the efficient formation of inflammasome complexes in BMDMs, impeding the maturation of IL-1β (46). Nevertheless, these alterations did not prevent GSDMD activation or the ability of the host cells to undergo pyroptotic cell death. Notably, although less efficient than the fully processed caspase-1, pro–caspase-1 is capable of activating GSDMD and inducing cell death to a certain degree (46, 69). This meant that the proinflammatory role of caspase-1 can be regulated in cells by limiting the extent of pro–caspase-1 processing and that various cell lineages may have ontogenically restricted or acquired the full range of functions offered by caspase-1 by altering the expression of inflammasome components such as the ASC adaptor. A variety of noncanonical caspase-1 cleavage products are observed in neurons (70, 71), epithelial cells (72), corneal stromal cells (73), or certain types of carcinomas (74–77). Various caspase-1 cleavage products are also visible in myeloid cells following infection or PAMP stimulation (78, 79). An intermediate product of pro–caspase-1 processing composed of unseparated CARD and p20 subunits, called p33, has been described in macrophages (31). In contrast to the long-lived p32 composed of p20 and p10 caspase-1 subunits, p33 in macrophages is composed of the CARD and p20 domains and is considerably short-lived (<30 min) (2, 31). Caspase-1 p32 is also catalytically functional, can be detected independently using caspase-1 p20- or p10-specific Abs, unlike p33, and immunoprecipitated with caspase-1 p20-specific Ab and detected with p10-specific Ab (2). These findings indicated that the caspase-1 p32 observed in hepatocytes is different from the caspase-1 p33 described in macrophages. Only future studies will help determine the origins and biological relevance of such “nonconformist” caspase-1 species observed in different cell types.

The convention of undertaking biochemical and functional characterizations of inflammasome pathways using immune cells belonging to the myeloid lineage has also created the notion that proinflammatory cytokine responses and pyroptotic cell death are inseparable consequences of caspase-1 activation (7, 80–82). Our observation that inflammasome-mediated caspase-1 activation in hepatocytes does not induce proinflammatory cytokines challenges this idea. We believe that the ability to process pro–caspase-1 into the p20 and p10 domains may have evolved with functional specialization in the immune cells of myeloid lineage. Increased availability of ASC in these cells would create a denser platform to bring pro–caspase-1 molecules in close proximity to each other and enable its complete proteolysis (46). Mutating ASC to impede their intermolecular interaction generates larger, less condensed ASC specks in macrophages, resulting in significantly reduced caspase-1–mediated functions (46). We showed that diminishing ASC expression in BMDMs reduced the efficiency of caspase-1 cleavage, leading to the appearance of caspase-1p32, presumably at the cost of caspase-1p20. The increased basal expression of ASC may have enabled the myeloid cells to mature proinflammatory cytokines such as IL-1 and IL-18, thereby fulfilling their signaling and recruitment functions. A lower expression of ASC, on the other hand, would limit caspase-1 processing in parenchymal cells such as the hepatocytes, which are not professional immune cells. Intriguingly, neutrophils also express ASC at lower levels than macrophages and exhibit unconventional caspase-1 processing dynamics and function following PAMP stimulation or pathogen encounter (2, 31, 73, 83). Certain pathogens, such as Sendai virus and paramyxoviruses, are known to regulate inflammasome activation and caspase-1 processing dynamics in host cells by targeting ASC oligomerization (84). Human adenovirus 5 can inhibit ASC phosphorylation and oligomerization to evade inflammasome-mediated caspase-1 activation (85). Therapeutically impeding ASC oligomerization has been shown to result in broad-spectrum inhibition of inflammasome activation and is a proposed treatment option for inflammasome-dependent inflammatory disorders (86). Targeting ASC expression and function may therefore offer a potential path to alter the extent and consequences of the induction of the inflammasome in the context of infectious or inflammatory diseases. We believe that the inherently reduced expression of ASC in hepatocytes may have played a significant role in the evolutionary choice of Plasmodium to undertake pre-erythrocytic development in these cells in mammals.

Although enhancing ASC expression resulted in the production of mature IL-1β and IL-18 in ex vivo cultured primary hepatocytes, this was not reflected in vivo in the AAV-ASC inoculated mice. Yet, there was enhanced recruitment of APCs to the liver following RAS vaccination in the AAV-ASC inoculated mice. This suggested that APC recruitment to the liver may be driven by the local cytokine signals in the liver. Only future studies will help decipher the mechanism by which CSF1R⁺ APCs are recruited to the liver following Plasmodium infection and how the inflammasome pathway in hepatocytes impacts it.

CSF1R⁺ CD11c⁺ APCs are instrumental in the generation of protective CD8 T cell responses following natural malaria infections and sporozoite-based antimalarial vaccination (4). Conventional wisdom would dictate that enhancing the recruitment of such CSF1R⁺ APCs to the liver would result in the generation of stronger immune responses and protection from future Plasmodium challenges. However, RAS vaccination in the context of ASC overexpression in hepatocytes resulted in diminished CD8 T cell responses and suboptimal protective immunity. Prior research has shown that prolonged survival of the hepatocytes harboring live-attenuated vaccine strains of Plasmodium is critical for the generation of robust immune responses against malaria (87). Adequate development and persistence of the attenuated parasites in hepatocytes is believed to make a larger quantity and broader range of protective epitopes available to the Ag presentation machinery in the liver (13, 87). Overexpression of ASC in hepatocytes likely results in rapid clearance of the RAS-infected hepatocytes. Of note, intact type I IFN signaling in hepatocytes, which facilitates such rapid clearance of Plasmodium from the liver, albeit via other potential mechanisms, also limits the immune responses and protection from malaria following sporozoite-based vaccination (2, 3, 60). Therefore, therapeutically limiting ASC oligomerization and function further in the hepatocytes, such as with the drug MM01, may be a viable strategy to enhance the efficacy of live-attenuated sporozoite-based vaccines against malaria (86). Our work, in addition to filling a fundamental knowledge gap in our understanding of the dynamics of the regulation of pro–caspase-1 processing in cells, offers a pathway to modify the extent and consequences of inflammasome processing in cells.

The authors have no financial conflicts of interest.

We thank the University of Georgia Center for Tropical and Emerging Global Diseases Flow Cytometry Core and the University of Georgia Center for Tropical and Emerging Global Diseases Sporocore staff for contributions, the National Institutes of Health Tetramer Core Facility for providing the GAP50 tetramer, and the members of the Kurup laboratory for their intellectual and technical support. We also acknowledge Dr. Kojo Mensa-Wilmot for giving access to fluorescence microscopy and Dr. Ronald Etheridge for valuable feedback on the manuscript.