Noncanonical Roles of Caspase-4 and Caspase-5 in Heme-Driven IL-1β Release and Cell Death Free

The interactions between inflammatory caspases and inflammasomes are critical for preventing uncontrolled inflammation and for mediating appropriate inflammation under infectious and sterile conditions. Inflammasomes are multiprotein complexes that provide the platform for recruitment and activation of inflammatory caspases and are essential for cellular inflammatory responses (1). The inflammatory caspases include human caspase-1, -4, and -5 and murine caspase-11 (1). This subset of the broader caspase protease family does not mediate apoptosis but specifically regulates inflammation by facilitating the activation and release of the proinflammatory cytokines IL-1β and IL-18 (2). Although the essential nature of inflammatory caspases in pathogen clearance is well established, their role in sterile inflammation (inflammation in the absence of infection) is less clear. Sterile inflammation occurs when nonpathogenic inflammatory stimuli activate inflammasomes. These stimuli are known as damage-associated molecular patterns (DAMPs) and are generally endogenous signals released by dying cells. This type of inflammation is important for wound healing and tissue regeneration but, if unchecked, can contribute to tissue damage associated with conditions such as ischemic stroke, myocardial infarction, and neurodegeneration (3). Despite the importance of DAMPs for triggering inflammation, the endogenous signaling molecules that trigger sterile inflammation are not fully resolved (4).

Heme has the features of a DAMP because it is released following RBC destruction and triggers an inflammatory response. Extracellular hemoglobin and heme are highly pro-oxidant molecules that are assiduously scavenged by haptoglobin and hemopexin, respectively. Excessive hemolysis can saturate and deplete the haptoglobin and hemopexin systems, resulting in free heme with strong proinflammatory capabilities (5). Heme has been shown to activate caspase-1 in mouse macrophages via assembly of the NOD-like receptor family pyrin domain containing (NLRP) 3 inflammasome (6). Heme has also been shown to activate TLR4 in murine endothelial cells to activate NF-κB (7). For IL-1β release to proceed, two signals are needed. Signal 1 activates NF-κB to induce pro–IL-1β expression and expression of additional inflammasome proteins, including NLRP3, in a process known as priming. Signal 2 provides an intracellular signal that induces inflammasome assembly and caspase-1 activation that cleaves pro–IL-1β to its mature form, which is released from the cell (8). Heme is naturally taken up and recycled by macrophages, providing a physiological intracellular signal 2 (9). Because of its ability to activate both caspase-1 and TLR4 and to be internalized by macrophages, extracellular heme has the properties of a DAMP that could potentially provide both signal 1 and signal 2 to initiate an effective inflammatory response.

Excessive release of heme from RBCs is a key feature of several pathological states, including sepsis, malaria, and sickle cell disease (SCD). SCD is the most prevalent inherited blood disorder, affecting ∼100,000 Americans and millions worldwide (10). The clinical manifestations of SCD arise from a complex pathophysiology, including chronic hemolytic anemia, increased susceptibility to infection, and vaso-occlusive events (11). Chronically elevated heme levels induce the inflammatory activation of monocytes, macrophages, and the endothelium (7, 12, 13). This unchecked inflammation can result in vaso-occlusion, acute chest syndrome, and organ damage (5, 7, 14). Heme-induced activation of monocytes and macrophages contributes to these severe complications through release of inflammatory cytokines, such as IL-1β, that trigger endothelial activation, upregulation of adhesion factors, and vaso-occlusion (15, 16). Indeed, in a study of children with SCD, patients having a vaso-occlusive pain crisis demonstrated elevated levels of proinflammatory cytokines IL-1β, IL-6, IL-10, TNF-α, and free heme (17). The role of the inflammatory caspases in heme-induced inflammation in the context of hemolytic disorders, such as SCD, has not been well studied.

Mice deficient in caspase-1 or the inflammasome proteins NLRP3 or apoptosis-associated speck-like protein containing a CARD (ASC) survive following hemolysis induced by a lethal dose of phenylhydrazine (6). Thus, caspase-1 is essential for an effective inflammatory response to hemolysis. However, the roles of caspase-4 and caspase-5 in this process are unknown. Caspase-4 and caspase-5 are the human orthologs of murine caspase-11. Caspase-4, -5, and -11 have been shown to be activated by intracellular LPS independent of inflammasomes, and they each have been shown to cleave the pore-forming protein gasdermin D (GSDMD) (18–20). Cleavage of GSDMD allows its N-terminal fragment to insert in the plasma membrane, forming a pore predicted to be 180 Å in diameter (21). This pore is of sufficient size to allow release of mature IL-1β but also permits influx of ions leading to cell swelling and a necrotic form of cell death called pyroptosis (22–24). Although caspase-1 can also cleave GSDMD, blocking caspase-1–dependent cleavage delays but does not inhibit pyroptosis (18). Therefore, the current thinking is that caspase-1 cleaves pro–IL-1β and pro–IL-18, and caspase-4, caspase-5, and caspase-11 cleave GSDMD to induce pyroptosis, allowing active cytokine release. It has been proposed that cytokine release and pyroptosis cannot be uncoupled (25). However, contradicting this theory, some reports show living cells releasing IL-1β (26, 27). In this study, we demonstrate that heme induces inflammasome-dependent caspase-1 activation and inflammasome-independent caspase-4 and caspase-5 activation. Furthermore, we show that both caspase-4 and caspase-5 are essential for IL-1β release, whereas only caspase-4 contributes to heme-induced cell death.

The following Abs were used: anti–caspase-1 (D7F10; Cell Signaling Technology, Danvers, MA), anti–caspase-4 (4450; Cell Signaling Technology), anti–caspase-5 (D3G4W; Cell Signaling Technology); anti-actin (C4; MP Biomedicals); anti-GSDMD (G7422; Sigma-Aldrich, St. Louis, MO), anti-GSDMD (N-terminal) (E7H9G; Cell Signaling Technology), and anti–IL-1β (MAB601; R&D Systems, Minneapolis, MN). All cell culture media reagents were purchased from Thermo Fisher Scientific (Waltham, MA). Ultrapure LPS (from E. coli O111:B4) was purchased from Invivogen (San Diego, CA). Unless otherwise indicated, all other reagents were purchased from Sigma-Aldrich.

The pBiFC.VC155 and pBiFC.VN173 plasmids encoding the prodomain of caspase-1 (C1-Pro), the caspase-4 prodomain (C4-Pro), or caspase-5 prodomain (C5-Pro) were described previously (28). Single mutations were introduced using QuikChange Site-Directed Mutagenesis Kit (Agilent Technologies, Santa Clara, CA). The bicistronic vector consists of C1-Pro VC and C1-Pro VN linked with a 2A peptide. Silent mutations were introduced into the second C1-Pro nucleotide sequence to prevent the sequence recombining out during cloning because of the presence of two identical C1-Pro sequences. The sequence was generated by Integrated DNA Technologies (Coralville, IA) and cloned into pRRL-MND-MCS-2A-mCherry-2A-Puro. pdsRedmito was purchased from Clontech Laboratories (Takara Bio, Mountain View, CA). pCMV-SPORT6.IPAF (NLRC4) was purchased from Thermo Fisher Scientific. Each construct was verified by sequencing.

Whole blood samples were obtained from patients with SCD who attend our hematology clinic as part of their routine care. The study was approved by the Baylor College of Medicine Institutional Review Board, and informed consent was obtained from all participants or their legal guardian (if the participants were minor). To isolate PBMCs, whole blood obtained from healthy blood donors or SCD patients was separated using the Ficoll-Paque (GE Healthcare, Pittsburgh, PA) gradient protocol (29). CD14⁺ monocytes were isolated from PBMCs using magnetic bead selection (Miltenyi Biotec, San Diego, CA). To differentiate cells into macrophages, monocytes were seeded at 5 × 10⁶–1 × 10⁷/10 cm dish in RPMI 1640 medium supplemented with FBS (10% [v/v]), GlutaMAX (2 mM), and penicillin/streptomycin (50 I.U./50 μg/ml) and GM-CSF (50 ng/ml). Cells were allowed to adhere overnight, and culture media were exchanged the following day for fresh GM-CSF–supplemented media. Media were exchanged every 2–3 d and cells were considered fully differentiated at 7 d, as determined by morphology. To differentiate into M2 macrophages, media were exchanged, and cells were incubated for additional 24 h in media supplemented with IL-4 (50 ng/ml). For the matched M1 macrophages, cells were incubated in fresh GM-CSF–supplemented media for an additional day.

The heme solution was prepared immediately before use by solubilizing 3.3 mg of porcine hemin (oxidized version of heme) in 100 μl of NaOH solution (0.1–1.0 M). The mixture was vortexed for 5 min in the dark. Nine hundred microliters of serum-free RPMI 1640 media were added to the resulting solution and vortexed for an additional 5 min in the dark, followed by filtration through a sterile 0.22-μm spin filter. The term “heme” is used generically to refer to both heme and hemin. Unless otherwise indicated, cells were treated with heme in the presence of 0.1% FBS in the culture media to prevent components present in FBS from sequestering and inhibiting heme. Heme is rapidly recycled by circulating macrophages; therefore, to mimic the transient exposure of cells to heme and to limit its toxicity, where indicated, cells were exposed to heme for 1 h followed by addition of an equal volume of culture media containing 10% FBS to inactivate the heme.

THP-1 cells were grown in RPMI medium containing FBS (10% [v/v]), supplemented with GlutaMAX (2 mM), penicillin/streptomycin (50 I.U./50 μg/ml), HEPES (10 mM), sodium pyruvate (1 mM), D-glucose (2.5 g/l), and 2-ME (50 pM). MCF-7 cells were grown in DMEM containing FBS (10% [v/v]), l-glutamine (2 mM), and penicillin/streptomycin (50 I.U./50 μg/ml). Caspase-1, caspase-4, and caspase-5 were deleted from THP-1 cells using an adaptation of the CRISPR/Cas9 protocol described in (30). Protospacer sequences for each target gene were identified using the CRISPRscan scoring algorithm [www.crisprscan.org (31)] or were purchased from Sigma-Aldrich for caspase-1. DNA templates for single guide RNAs (sgRNAs) were made by PCR using the pX459 plasmid containing the sgRNA scaffold sequence and using the following primers: negative control sgRNA (Ctrl) sequence, 5′-TTAATACGACTCACTATAGGCGCGATAGCGCGAATATATTgttttagagctagaaatagc-3′; ΔCASP1_53 sequence, 5′-TTAATACGACTCACTATAGGCGGTTTGTCCTTCAAACTTgttttagagctagaaatagc-3′; ΔCASP1_61 sequence, 5′-TTAATACGACTCACTATAGGTGTCTCATGGTATTCGGGAgttttagagctagaaatagc-3′; ΔCASP4_40 sequence, 5′-TTAATACGACTCACTATAGGGAAACAACCGCACACGCCgttttagagctagaaatagc-3′; ΔCASP5_46 sequence, 5′-TTAATACGACTCACTATAGGTCCTGGAGAGACCGCACAgttttagagctagaaatagc-3′; ΔCASP5_53 sequence, 5′-TTAATACGACTCACTATAGGTCAAGGTTGCTCGTTCTAgttttagagctagaaatagc-3′; and universal reverse primer, 5′-AGCACCGACTCGGTGCCACT-3′. The sgRNAs were generated by in vitro transcription using the HiScribe T7 High Yield RNA Synthesis Kit (New England Biolabs, Ipswich, MA). Purified sgRNA (0.5 μg) was incubated with Cas9 protein (1 μg; PNA Bio, Newbury Park, CA) for 10 min at room temperature. THP-1 cells were electroporated with the sgRNA/Cas9 complex using the Neon Transfection System (Thermo Fisher Scientific, Waltham, MA) at 1600 V, 10 ms, and three pulses. For the caspase-1– and caspase-5–deficient cell lines, two sgRNAs were selected at either end of the gene to delete the intervening region. Deletion of CASP1, CASP4, or CASP5 was confirmed by Western blot or by PCR. Single-cell clones were generated by single-cell plating of the parental cell line. Gene deletion in the single-cell clones was confirmed by sequencing.

For transfection of human monocyte-derived macrophages (hMDM), 1 × 10⁵ cells were transfected with the appropriate plasmid combinations or plasmid/small interfering RNA (siRNA) combinations using the Neon Transfection System (Thermo Fisher Scientific) and a 10-μl Neon Tip at 1000 V, 40 ms, and two pulses. Cells were transfected with amounts of the relevant expression plasmids, as described in the figure legends. A total of four wells were transfected for every plasmid and incubated in 200 μl of antibiotic-free media. After 1 h, 200 μl of complete growth media containing penicillin/streptomycin (50 I.U./50 μg/ml) was added. Expression was allowed for 24 h, and media were exchanged for fresh media prior to treatment. Control siRNAs were cyclophilin B siRNA (ON-TARGETplus SMARTpool; Dharmacon). ASC and NLRP3 siRNAs were ON-TARGETplus SMARTpool (Dharmacon, Lafayette, CO). A total of 1 × 10⁵ MCF-7 cells were transfected with appropriate plasmid combinations using Lipofectamine 2000 transfection reagent (Thermo Fisher Scientific) according to manufacturer’s instructions.

Cells were imaged using a spinning disk confocal microscope (Zeiss, Thornwood, NY), equipped with a CSU-X1A 5000 spinning disk unit (Yokogowa Electric, Tokyo, Japan) multilaser module with wavelengths of 458, 488, 514, and 561 nm, and an Axio Observer Z1 motorized inverted microscope equipped with a precision motorized XY stage (Carl Zeiss MicroImaging, Thornwood, NY). Images were acquired with a Zeiss Plan-Neofluar 40× 1.3 numerical aperture or 64× 1.4 numerical aperture objective on an Orca R2 charged-coupled device camera using ZEN 2012 software (Zeiss). Cells were plated on dishes containing coverslips (MatTek, Ashland, MA) coated with poly-d-lysine hydrobromide 24 h prior to treatment. For time-lapse experiments, media on the cells were supplemented with HEPES (20 mM) and 2-ME (55 μM). Cells were allowed to equilibrate to 37°C in 5% CO₂ prior to focusing on the cells.

Images were analyzed by drawing regions around individual cells and then computing average intensity of the pixels for each fluor using ZEN 2012 software (Zeiss). Data were scaled by the following formula: scaled point = (Max − x)/MaxDifference, where Max equals the maximum value in the series, x equals the point of interest, and MaxDifference equals the maximum minus the minimum value in the series.

THP-1 cells were plated at 1 × 10⁶ cells/ml and differentiated into macrophages by 24-h incubation in the presence of PMA (10 ng/ml) followed by 24-h incubation in RPMI medium. Cells were polarized into M1 macrophages by 20-h incubation with human IFN-γ (20 ng/ml; PeproTech) and Ultrapure LPS (10 pg/ml) (32). Cells were washed prior to treatment, as indicated in the figure legends. IL-1β concentration in harvested clarified supernatants was measured with the IL-1β DuoSet ELISA Kit (R&D Systems) according to the manufacturer’s instructions.

Cells were treated as indicated and collected by centrifugation. Cells were washed with PBS and resuspended in 100 μl of PBS supplemented with 5% FBS, 0.5% BSA, 2 mM EDTA, and 1 μl of 7-aminoactinomycin D (7-AAD; Thermo Fisher Scientific). The 7-AAD–positive cells were quantitated by flow cytometry and analyzed with FlowJo software (FlowJo, Ashland, OR).

Cells were treated as indicated. Cells were lysed in IP Lysis Buffer (50 mM Tris [pH 7.4], 150 mM NaCl, 0.1% SDS, and 1% NP-40) containing protease inhibitors (cOmplete Mini Protease Inhibitor mixture). Protein concentration was determined by BCA Protein Assay (Thermo Fisher Scientific). Fifty micrograms of total protein (lysate) or 50–60 μg culture media (supernatant) was resolved by SDS-PAGE and transferred onto a 0.45-μm nitrocellulose membrane (Thermo Fisher Scientific) and immunodetected using appropriate primary and peroxidase-coupled secondary Abs (Prometheus Protein Biology Products, San Diego, CA). Proteins were visualized by West Pico and West Dura chemiluminescence substrate (Thermo Fisher Scientific).

Statistical comparisons were performed using a two-tailed Student t test or one-way ANOVA and calculated using Prism 6.0 (Graph Pad) software.

It has been previously shown that heme can induce IL-1β release in LPS-primed murine bone marrow–derived macrophages and that heme can activate TLR4 to induce NF-κB activation in endothelial cells (6, 7). This suggests that heme can provide both the priming signal (signal 1) required to induce pro–IL-1β expression and the inflammasome activating signal (signal 2) required to induce inflammasome-dependent caspase-1 activation. To test this in human macrophages, we isolated CD14⁺ monocytes from peripheral blood from healthy donors and differentiated them into macrophages with GM-CSF. We exposed the macrophages to heme in the presence or absence of prior priming with LPS. Treatment with heme in nonprimed macrophages induced a modest increase in IL-1β. However, in order for heme to induce maximal IL-1β release, prior priming with LPS was required. Treatment with LPS alone resulted in similar IL-1β release to heme alone (Fig. 1A). These experiments were carried out in 0.1% FBS in the culture media. This is because components present in serum, including hemopexin and albumin, bind to heme with high affinity and prevent its uptake into cells (33–36). To show that the IL-1β release was specific to heme and nothing else present in the formulation, we increased the amount of serum in the culture media. As little as 1% FBS concentration in the cellular media was sufficient to block heme-induced IL-1β release (Fig. 1B), indicating that the IL-1β release from macrophages that we detected was due to the specific effects of heme.

Sickle cell macrophages are constantly exposed to high levels of extracellular heme due to persistent hemolysis of sickled RBCs (12). To determine the effect of this on IL-1β release, we compared hMDM from healthy donors with hMDM isolated from patients with SCD. We noted that heme-treated hMDM from patients with SCD were more sensitive to heme-induced IL-1β release relative to the control hMDM (Fig. 1C). However, priming with LPS was required for strong IL-1β release in both groups. Together, these results suggest that, whereas heme is able to activate caspase-1 to induce IL-1β release, it is not sufficient to prime the cells for IL-1β release.

It has been shown that macrophages from sickle mice express higher levels of M1 markers (12). To investigate IL-1β release by heme in the two extreme macrophage subtypes, we polarized the CD14⁺ monocytes toward M1 (proinflammatory) by treating them with GM-CSF for 7 d or toward M2 (anti-inflammatory) by treating them with IL-4 for a further day after the incubation with GM-CSF. LPS-primed M1 or M2 hMDM were stimulated with heme or with ATP. We showed lower levels of IL-1β release in M2 compared with M1 hMDM in response to both heme and ATP (Fig. 1D). In response to heme, M1 macrophages underwent cell death with a necrotic morphology. In contrast, M2 macrophages remained viable with an intact plasma membrane, but they appeared more flattened and granular than the unstimulated cells (Fig. 1E). Together, these results suggest that SCD macrophages have increased inflammatory caspase activity, resulting in increased IL-1β activation and that this may be due, in part, to the increased proportion of the M1 macrophage population in patients with SCD.

Given that heme induces IL-1β release in macrophages, we next investigated the ability of heme to activate caspase-1 and the other human inflammatory caspases, caspase-4 and caspase-5. Caspase-1 is activated upon proximity-induced dimerization following recruitment to inflammasomes (1, 37). The recruitment of caspase-1 to an inflammasome is mediated by interactions between specific protein interaction domains in the inflammasome proteins. The caspase recruitment domain (CARD) in ASC or in NLR family CARD containing 4 (NLRC4) binds to the CARD in the C1-Pro to recruit it to the inflammasome and facilitate induced proximity of the caspase (38, 39). To measure the induced proximity of each caspase, we used caspase bimolecular fluorescence complementation (BiFC) (40). BiFC uses nonfluorescent fragments of the yellow fluorescent protein, Venus (“split Venus”) that can associate to reform the fluorescent Venus complex when fused to interacting proteins (41). When the CARD-containing C1-Pro is fused to each half of split Venus, recruitment of caspase-1 to inflammasomes and the subsequent induced proximity results in enforced association of the two Venus halves (28). Thus, Venus fluorescence (BiFC) acts as a read-out for caspase induced proximity, the proximal step for activation. We transiently expressed C1-Pro (aa 1–102) fused to each of the split Venus fragments, Venus C (aa 155–239) and Venus N (aa 1–173) in hMDM (Fig. 2A). Following exposure to heme, we noted a significant increase in the proportion of cells that became Venus-positive. Surprisingly, and unlike the requirement for IL-1β release, this induction of caspase-1 BiFC did not require prior priming with LPS. When we transfected macrophages with the C4-Pro or C5-Pro BiFC pairs, we observed a similar pattern, in which caspase-4 and caspase-5 BiFC were induced by heme, independent of LPS priming (Fig. 2B, 2C). This suggests that caspase-1, caspase-4, and caspase-5 are each activated by heme in macrophages. When we analyzed the appearance and localization of the fluorescent complexes produced by heme in each case, we noticed some differences (Fig. 2D). The caspase-1 complex appeared as a single green punctum that is typical of ASC specks and of ASC-induced caspase-1 BiFC, which we previously reported (28, 42). Caspase-4 and caspase-5 BiFC appeared as a series of punctate spots located throughout the cytoplasm of the cell that did not colocalize with mitochondria. This indicates that there may be different mechanisms of action for caspase-1 compared with caspase-4 or caspase-5.

Next, we used the BiFC system to measure inflammatory caspase activation in M1 and M2 macrophages. We polarized macrophages into M1 and M2 as before and transfected donor-matched M1 and M2 cells with each of the inflammatory caspase BiFC pairs. Following exposure to heme, caspase-1, caspase-4, and caspase-5 BiFC were increased as before in M1 hMDM (Fig. 3A). Interestingly, whereas caspase-1 and caspase-4 BiFC were induced to the same extent in M1 and M2 macrophages, the M2 macrophages had impaired heme-induced caspase-5 BiFC. To explore this further, we measured caspase-1, caspase-4, and caspase-5 expression in M1 and M2 macrophages by immunoblot (Fig. 3B). Endogenous caspase-1 and caspase-4 were similarly expressed in both subtypes, and their expression was unchanged with the addition of heme. Caspase-5 expression was readily detected in M1 macrophages but was lowly expressed in unstimulated M2 macrophages. Heme did not induce caspase-5 expression in the M2 subgroup, but pretreatment with LPS restored the level of caspase-5 to the level detected in M1 cells. We hypothesized that the low level of endogenous caspase-5 in M2 hMDM was responsible for the inability of these cells to induce caspase-5 BiFC. To test this, we compared caspase-5 BiFC induced by heme in M1 and M2 hMDM with and without LPS priming. Priming of M2 macrophages with LPS restored their ability to induce the caspase-5 BiFC complex (Fig. 3C). This suggests that full-length endogenous caspase-5 is required to form the caspase-5 activation complex in response to heme.

Intracellular LPS has been shown to bind and induce oligomerization of caspase-4, caspase-5, or caspase-11 independently of inflammasomes (20). The oligomerization is mediated by CARD clustering, which leads to induced proximity and caspase dimerization. Because heme can promote caspase-4 and caspase-5 induced proximity and is naturally taken up and recycled by macrophages, we reasoned that heme may represent an intracellular trigger for noncanonical inflammatory caspase activation (activation independent of known inflammasomes). To test this, we investigated whether heme-induced caspase activation requires interactions with inflammasomes. Recruitment of caspases to inflammasomes is dependent on an intact CARD. The aspartate residue (D59) in caspase-1 is essential for the ASC–caspase-1 interaction, and mutation of this D59 residue blocks ASC-induced caspase-1 BiFC (28, 43). We previously showed that ASC overexpression does not induce caspase-4 or caspase-5 BiFC, but NLRC4 does (28). We modified the conserved CARD-binding residue found in caspase-4 (D59) and caspase-5 (D117) and showed that disruption of this residue also blocked caspase-4 and caspase-5 BiFC triggered by overexpressed NLRC4 (Supplemental Fig. 1A, 1B). Using the CARD-disrupting mutants in caspase-1 (D59R), caspase-4 (D59R), and caspase-5 (D117R), we tested whether heme-induced inflammatory caspase BiFC is independent of canonical inflammasome interactions. We transfected hMDM with the C1-Pro, C4-Pro, or C5-Pro BiFC pairs or with the corresponding CARD-disrupting mutant BiFC pairs. Following exposure to heme, caspase-1 (D59R) BiFC was significantly decreased compared with the wild-type (WT) caspase-1 reporter (Fig. 4A). This indicates that heme triggers recruitment of caspase-1 to the inflammasome for its subsequent activation. In contrast, heme-induced caspase-4 and capase-5 BiFC was not changed when the CARD mutant BiFC reporters were expressed (Fig. 4A). This suggests that heme activates caspase-4 and caspase-5 independently of canonical inflammasome interactions.

To confirm these results, we used siRNA to silence NLRP3, the inflammasome receptor that has been shown to be required for heme-induced caspase-1 activation (6). We similarly silenced the inflammasome adaptor protein ASC. ASC is essential for the assembly of most characterized inflammasomes, including the NLRP1, NLRP3, and AIM2 inflammasomes (1). It has been shown to be dispensable for the NLRC4 inflammasome, but ASC enhances NLRC4 inflammasome activity (44, 45). Therefore, if heme induces inflammatory caspase activation via the NLRP3 or alternative inflammasome complex, loss of ASC is predicted to block this activity. The siRNA-mediated silencing of both NLRP3 and ASC in hMDM reduced heme-induced caspase-1 BiFC to background levels (Fig. 4B, 4C, Supplemental Fig. 1C). In contrast, NLRP3 or ASC depletion had no effect on the levels of caspase-4 or caspase-5 BiFC induced by heme (Fig. 4B, 4C). Together, these results show that caspase-4 and caspase-5 are activated by heme independently of canonical inflammasomes.

Having identified caspase-4 and caspase-5 induced proximity following heme exposure, we set out to explore the functional outcomes of this mechanism. We used CRISPR/Cas9 to delete caspase-1, caspase-4, or caspase-5 from the monocytic THP-1 cell line. To most closely represent the M1 phenotype of primary macrophages, cells were first treated with PMA followed by incubation with IFN-γ and a low concentration of LPS (10 pg/ml) to polarize them to an M1-like state (32). The cells were then primed for 3 h with a higher concentration of LPS (100 pg/ml) followed by exposure to heme for a 1-h pulse. IL-1β release was induced by heme in LPS-primed M1 THP-1 cells to a similar level as we observed in M1 hMDM in parental THP-1 cells (WT) and in THP-1 cells edited with a control sgRNA (Fig. 5A). As expected, caspase-1 deficiency (shown as ΔC1) completely blocked the release (Fig. 5A). We next measured the impact of loss of caspase-4 or caspase-5. Deficiency of either caspase-4 or caspase-5 potently blocked heme-induced IL-1β release in LPS-primed M1 THP-1 cells (Fig. 5B). These results show that both caspase-4 and caspase-5 are required for IL-1β release induced by heme and that one caspase does not functionally replace for the action of the other. Caspase-4 loss in PMA-primed THP-1 cells had no effect on IL-1β release induced by the canonical stimulus nigericin, a bacterial potassium ionophore and potent inducer of the NLRP3 inflammasome (46, 47) but blocked release induced by high doses of the noncanonical stimulus LPS (Supplemental Fig. 2). Interestingly, IL-1β release induced by both of these stimuli was blocked by loss of caspase-5, indicating that caspase-5 is required for IL-1β release in response to both canonical and noncanonical stimuli.

Caspase-1, caspase-4, and caspase-5 have been shown to cleave GSDMD (19), the proposed mechanism by which inflammatory caspases induce both pyroptosis and release of mature cytokines. We therefore investigated GSDMD cleavage in M1-polarized THP-1 cells. We detected a 30-kDa band corresponding to cleaved GSDMD in control cells exposed to heme with and without LPS priming (Fig. 5C, lanes 3 and 5). In response to nigericin, the majority of the cleaved GSDMD was detected in the cellular supernatants, likely due to the elevated potency of nigericin compared with heme (Fig. 5C, lane 6). This is consistent with reports that the N terminus of GSDMD is released from cells when it is cleaved (22, 23). Deficiency in caspase-1 did not markedly reduce the amount of N-terminal GSDMD produced by heme treatment (Fig. 5C, lane 9), but did block GSDMD cleavage induced by nigericin (Fig. 5C, lane 12). In cells deficient in caspase-5, there was a mild reduction in heme-induced GSDMD cleavage but only in the absence of LPS (Fig. 5C, lane 21) and a slight reduction in response to nigericin. In contrast, loss of caspase-4 led to an almost complete impairment of GSDMD cleavage (Fig. 5C, lane 15) that was not observed following nigericin treatment (Fig. 5C, lane 18). The cleavage was also reduced in LPS-primed cells but not as effectively (Fig. 5C, lane 17). Together, these results suggest that caspase-4 is the major effector of GSDMD cleavage in response to heme, and this does not require priming. In the absence of caspase-1, we noted impaired induction of caspase-5 by M1 polarization. This effect was consistent across independently generated CRISPR/Cas9 caspase-1–deficient cell lines and in a cell line made with short hairpin RNA against caspase-1 (data not shown). This suggests that caspase-1 is required for the expression or stability of caspase-5.

We probed for IL-1β in the lysates and supernatants of the control and caspase-deficient cell lines. Heme treatment of LPS-primed THP-1 M1 control cells induced cleavage of pro–IL-1β, as measured by production of the p17 mature fragment (Fig. 5C, lane 5). Production of this fragment was completely blocked in caspase-1–deficient cells and partially blocked in caspase-5–deficient cells (Fig. 5C, lanes 11 and 23). This is consistent with the reduction of IL-1β release detected by ELISA in these cells in Fig. 5A, 5B. Heme-induced IL-1β cleavage in the absence of caspase-4 was increased in LPS-primed cells but, in contrast to nigericin-treated cells, the cleaved IL-1β was all detected in the cell lysates and not in the cellular supernatants (Fig. 5C, lanes 17 and 18). This suggests that, in response to heme, the release of mature IL-1β but not the cleavage is blocked in the absence of caspase-4.

When we probed the cells in Fig. 5C for pro–IL-1β, we noted a higher amount of pro–IL-1β in the lysates of caspase-4–deficient cells compared with WT, caspase-5–deficient, or caspase-1–deficient cells. In combination with the decreased GSDMD cleavage and increased mature IL-1β found in the lysates in these cells, we reasoned that this may be due to the cells being protected from cell death in the absence of caspase-4. To investigate the requirement of each caspase for heme-induced cell death, we first determined the level of death that was caspase dependent. We treated LPS-primed or unprimed THP-1 cells with heme in the presence or absence of the pan-caspase inhibitor qVD-OPH (Fig. 6A). Heme treatment of both primed and unprimed THP-1 cells induced substantial cell death as measured by 7-AAD uptake. Caspase inhibition significantly blocked this death, but notably, the reduction in cell death was not complete, and ∼30% of the cells were killed by heme in a caspase-independent manner. To determine the specific contribution of each inflammatory caspase to this death, we measured heme-induced cell death in caspase-1–, caspase-4–, and caspase-5–deficient THP-1 cells compared with WT THP-1 cells. We measured cell death at two time points: an early time point at 6 h and a later time point at 20 h. At 6 h, heme did not induce a significant amount of death above background, and there was no significant difference in the amount of death induced in the caspase-deficient cell lines (Fig. 6B). At the 20-h time point, loss of caspase-4 significantly decreased heme-induced cell death, whereas loss of caspase-1 or caspase-5 had no effect (Fig. 6C). This suggests that, among the inflammatory caspases, heme-induced caspase-4 activation is the primary effector of heme-induced cell death. Similar to the effect of total caspase inhibition, caspase-4 loss did not completely block the death, and ∼50% of the cells still died. This confirms that caspase-independent cell death mechanisms also contribute to heme-induced cell death.

To investigate the kinetics of heme-induced caspase-1 activation relative to cell death, we used time-lapse confocal microscopy. For this, we generated THP-1 cells stably expressing the C1-Pro BiFC components. We designed a bicistronic construct, in which the C1-Pro VC and C1-Pro VN are expressed in a single vector separated by the viral 2A self-cleaving peptide, similar to a caspase-2 reporter we previously described (48). This design ensures that the caspase-1 BiFC components are expressed at equal levels because they are translated from a single mRNA transcript. These cells also express a linked mCherry gene as a reporter for expression of the BiFC components. In addition, loss of mCherry fluorescence can be used to detect cell lysis. The caspase-1–BiFC complex was detected ∼18 h following the addition of heme. Its appearance was followed by lysis of the cell that was detected by loss of the mCherry protein upon cell rupture (Fig. 6D, Supplemental Video 1). Cell death occurred within 5 min of detection of the C1-Pro BiFC complex (Fig. 6E). Thus, recruitment of caspase-1 to inflammasomes in response to heme immediately precedes cell death. It is important to note that this death occurred in the presence of the pan-caspase inhibitor qVD-OPH. The inclusion of qVD-OPH in the time-lapse experiment was to prevent apoptotic cell death that causes the cells to lift off the coverslip and impairs imaging as the cells move out of the focal plane. This, together with the results from Fig. 6A, suggests that there is a substantial caspase-independent component to heme-induced cell death. However, the close timing of caspase-1 recruitment to inflammasomes and cell lysis indicates that these events are mechanistically linked, which could imply that caspases contribute to the death in a manner that does not require their catalytic activity.

Our data show that heme activates caspase-1, caspase-4, and caspase-5. However, there are significant differences in the upstream requirements for activation, the localization of the activation complexes, and the outcomes of activation of each caspase. We found that both caspase-4 and caspase-5 are required for heme-induced IL-1β release, whereas caspase-4 is the primary contributor to heme-induced cell death. Our results indicate that caspase-4 and caspase-5 have nonoverlapping functions in heme-induced inflammation and caspase-1 activation. Together, these data underscore the important functions of inflammatory caspases in heme-induced sterile inflammation.

Although SCD is primarily a disease characterized by anemia, many of its clinical complications are exacerbated by chronic inflammation. Sickle-shaped RBCs are more prone to hemolysis, releasing excess heme into the blood stream that overwhelms the body’s heme-scavenging systems. Consistent with this, we noted that heme-treated macrophages derived from patients with SCD released more IL-1β when compared with healthy controls. This suggests that SCD macrophages are more sensitive to heme-induced inflammation. Using PBMC transcriptome profiles of patients with SCD compared with healthy controls, van Beers et al. (49) showed that the SCD cohort had higher expression of many markers of innate immunity, including TLR4, NLRP3, NLRC4, CASP1, IL-1, and IL-18. They also showed a positive correlation between TLR4 expression and IL-6 expression in SCD cells. Lanaro et al. (50) showed increased expression of TNFα and IL-8 in SCD mononuclear cells. Because heme can activate TLR4 and, in turn, the transcription factor NF-κB that controls transcription of many of these cytokines and proteins, it is likely that a lifetime exposure to heme in patients with SCD, resulting from higher rates of hemolysis, contributes to these elevated levels. We originally suspected that the increased IL-1β release we detected from SCD macrophages after heme exposure would be due to this lifetime exposure to heme, which would prime the cells for inflammasome activation by inducing expression of pro–IL-1β and other inflammasome components. However, heme alone was insufficient to trigger IL-1β release in macrophages from patients with SCD or from healthy donors, indicating that an extra priming step is still required. Monocytes from patients with SCD have been shown to have increased levels of the macrophage markers CD14 and CD11b (15). In addition, liver macrophages from HbS sickle mice had increased surface expression of the M1 macrophage markers CD86, MHC class II, iNOS, and IL-6 (12). This suggests that sickle macrophages are likely to skew to the M1-like proinflammatory phenotype. Treatment of HbS sickle mice with the heme scavenger hemopexin reverted the M1 polarization, indicating that heme is the inducer of M1 polarization (12). Thus, we propose that the increased proportion of activated and proinflammatory M1 macrophages in patients with SCD is the reason why the cells, once primed by LPS, are more sensitive to heme-induced IL-1β release rather than the effect of heme priming. Interestingly, heme alone was sufficient to induce caspase-1, caspase-4, and caspase-5 activation. Caspase-1 activation by heme has been reported to be dependent on the NLRP3 inflammasome (6), and we confirmed that observation in this study. Unlike the other inflammasome receptors, NLRP3 requires a priming step for activation (51, 52). Our results may suggest that heme can prime for NLRP3 expression but does not provide enough of a priming signal to induce sufficient pro–IL-1β expression. Induction of NLRP3 expression has been shown to be dependent on reactive oxygen species (ROS) (8). Heme induces ROS, and blocking ROS has been shown to inhibit both caspase-1 cleavage and IL-1β processing (6). Therefore, heme-induced ROS generation may be a mechanism by which heme can prime for NLRP3 expression but not for pro–IL-1β expression.

Our observation that heme-induced caspase-4 and caspase-5 induced proximity does not require LPS priming could be explained in one of two ways. Similar to caspase-1, heme may be sufficient to prime the cells for caspase activation. However, we also show that, in contrast to caspase-1, caspase-4 and caspase-5 induced proximity is activated by heme independently of CARD-inflammasome interactions and of the common inflammasome proteins NLRP3 and ASC. The lack of a priming requirement for caspase-4 and caspase-5 induced proximity is consistent with caspase activation that does not require an additional upstream receptor. Caspase-5 is lowly expressed in many cell types, and its expression is induced by LPS and IFN (53). Despite this, primary M1-polarized macrophages express abundant caspase-5. M2 macrophages have reduced caspase-5 expression and display reduced caspase-5 induced proximity in response to heme. This observation may suggest that full-length, endogenous caspase-5 is required to promote assembly of the caspase-5 signaling platform. Indeed, when we reconstituted caspase-5 expression in M2 macrophages with LPS, we were able to restore the levels of caspase-5 induced proximity. These results are consistent with studies that have shown that caspase-4 and caspase-5 are direct intracellular receptors for LPS (20). LPS binds to the CARD in these caspases to trigger oligomerization of caspase-4 or caspase-5 in the absence of any known inflammasome protein. Our data indicate that heme acts in an analogous fashion to induce caspase-4 or caspase-5 oligomerization, either directly or through an as yet unidentified cytosolic mediator. Contrary to the report that LPS binding to caspase-4 or caspase-5 is CARD mediated (20), our results suggest that full-length caspase-5 is required for its scaffolding function because the BiFC construct contains only the CARD-containing prodomain, which is insufficient in the absence of endogenous caspase-5. Further supporting a mechanistic difference between heme-induced caspase-1 activation and heme-induced caspase-4 or caspase-5 activation, we observed differences in the size and localization of the caspase-4 and caspase-5 signaling complexes induced by heme compared with the single ASC-like speck observed for caspase-1. Heme is naturally taken up and recycled by macrophages (9), representing a physiological stimulus and, importantly, a trigger of sterile inflammation that directly engages caspase-4 and caspase-5. Thus, heme acts as a canonical stimulus by engaging inflammasome-dependent caspase-1 activation and also as a noncanonical stimulus by activating caspase-4 and caspase-5 independent of canonical inflammasomes.

Caspase-4 and caspase-5 are often assumed to be redundant proteins that directly phenocopy caspase-11 in mice. Studies using a transgenic mouse generated to express human caspase-4 indicated that caspase-4 does not completely phenocopy caspase-11, and a recent study showed a broader reactivity of caspase-4 to LPS compared with caspase-11 (54, 55). In addition, humans are more sensitive to endotoxemia than rodents (56). These studies highlight the important variations between the human and murine innate immune response and suggest that the presence of an additional inflammatory caspase in humans may contribute to these differences. We show that caspase-4 and caspase-5 are both required for heme-induced IL-1β release in M1 THP-1 cells, providing evidence of a cooperative regulation between these two caspases and caspase-1 rather than a redundant function, in which one can replace for the other. In contrast, loss of caspase-1 or caspase-5 on its own only marginally reduced GSDMD cleavage, whereas loss of caspase-4 was more effective at inhibiting GSDMD cleavage. Therefore, whereas the ability of caspase-4 to mediate IL-1β release can be explained by inhibition of GSDMD cleavage, caspase-5 appears to be able to inhibit IL-1β by additional means. Caspase-11 has been proposed to regulate caspase-1 (57), but the exact mechanism of this is unclear. Our work suggests that, in response to heme, caspase-5 functions primarily to regulate caspase-1, whereas caspase-4 mainly regulates GSDMD cleavage, and together, these proteins cooperate to regulate the production and release of mature IL-1β. This represents a bifurcation of the functions attributed to caspase-11 in humans.

Consistent with the different effects on GSDMD cleavage, caspase-5 did not impact cell death induced by heme in the same manner as caspase-4. This indicates that caspase-4 is the main effector of cell death induced by heme. However, there is still a considerable amount of caspase-independent death that is induced by heme. These results suggest that additional death mechanisms are engaged by heme. Heme has been shown to induce RIPK3-dependent necrosis (58), and this may therefore be a contributing mechanism to the death we observed. RIPK3-dependent necrosis has been shown to limit pathogenic inflammation and associated tissue damage without impairing IL-1β release (59). Given that cell death occurs immediately following detection of the caspase-1 activation platform, it is possible that a similar mechanism occurs in this study, in which the cell lysis is a result of concurrent activation of the RIPK3 pathway rather than caspase-dependent effects. Because K⁺ efflux is known to be necessary for NLRP3-induced caspase-1 activation (47), another explanation could be that the close timing of caspase-1 induced proximity and cell lysis indicates that caspase-1 is oligomerizing in response to early potassium efflux through a caspase-4/5–driven pore. The exact interplay between caspase-1 activation and caspase-4/5 activation that is triggered by exposure to heme and how this leads to cell death is something that requires further study.

The role of caspase-4 in heme-induced cell death may suggest that caspase-4 is a key contributor to tissue damage in humans. The association between caspase-11, excess pyroptosis, and inflammation-associated tissue damage is well established (57, 60). In a mouse model of SCD, the consequence of heme-induced inflammation is tissue damage that manifests as vaso-occlusion and death (7). Our results suggest that caspase-4 and not caspase-1 or caspase-5 is the primary contributor to heme-induced pyroptosis. Our model predicts that blocking caspase-4 would protect from tissue damage, whereas blocking caspase-1 or caspase-5 would protect from uncontrolled inflammation (fever, pain, etc.). Further exploration of the distinct roles of these caspases is required to fully understand how chronic inflammation can be controlled in hemolytic disorders.

We thank the members of L.B.-H.’s laboratory, past and present, for helpful discussions and careful reading of the manuscript. We thank Doug Green and the members of his laboratory for valuable suggestions. This project was supported by the Cytometry and Cell Sorting Core at Baylor College of Medicine with the assistance of Joel M. Sederstrom. The graphical abstract was drawn using BioRender software.

J.D.B. and G.M.V. have research funding from CSL Behring and Mitobridge-Astellas. B.A.R. is currently employed by Omniome Inc. The other authors have no financial conflicts of interest.