Liver-to-lung microembolic NETs promote gasdermin D–dependent inflammatory lung injury in sickle cell disease

Sickle cell disease (SCD) is a genetic disorder that affects millions of people worldwide.^1-4 Sickle cell anemia, the most common form of SCD, is caused by a homozygous mutation (SS) in the β-globin gene.^1,3 The mutant hemoglobin (HbS) polymerizes under hypoxic conditions, leading to erythrocyte sickling, vaso-occlusion, and hemolysis.^5-8 Vaso-occlusion contributes to the development of acute painful vaso-occlusive episodes, which are the primary reason for hospitalization of patients with SCD.¹ Clinical evidence suggests that acute chest syndrome, a type of acute lung injury and one of the leading causes of mortality among patients with SCD, is a sequela of an vaso-occlusive episode.^1,9,10 Histopathological findings in patients with SCD^11,12 and in vivo lung imaging in transgenic-humanized mice with SCD^13,14 have identified that occlusion of lung arterioles (pulmonary vaso-occlusion) by neutrophil-platelet–erythrocyte aggregates promotes the development of acute chest syndrome. This etiology also offers a therapeutic window to prevent the development of lung injury in SCD, provided new therapies are identified to treat pulmonary vaso-occlusion. Previously, we have shown that P-selectin blockade¹³ led to partial (∼50%) protection from pulmonary vaso-occlusion in mice with SCD. A major clinical trial also reported ∼50% reduction in frequency of hospitalization among patients with SCD receiving anti–P-selectin antibody (Ab) therapy.¹⁵ Altogether, these findings suggest that pulmonary vaso-occlusion in SCD may also be enabled by a yet unidentified P-selectin–independent mechanism, and new therapies targeting this unknown pathway would be necessary to prevent the development of acute chest syndrome in patients with SCD.

Hemolysis promotes sterile inflammation in SCD by releasing erythrocyte-derived damage-associated molecular-pattern molecules (DAMPs) such as hemoglobin and its oxidation-product heme that scavenge nitric oxide, generate reactive oxygen species (ROS), prime innate immune pathways, and promote the generation of neutrophil extracellular traps (NETs).^5,7,14,16-18 NETs are potent neutrophil-derived DAMPs, composed of externalized decondensed-chromatin decorated with citrullinated histones and neutrophil proteases.^19-22 Although recent findings suggest a role for NETs in promoting lung injury in mice with SCD¹⁸ and patients with SCD,²³ the innate immune signaling in neutrophils that promotes NETs generation and NETs-dependent lung injury in SCD remains largely unknown.^4,5 Emerging evidence supports a role for pore-forming protein gasdermin D (GSDMD)²⁴ in enabling NETs generation during bacterial infection^25,26; however, whether this pathway plays a role in the sterile inflammation of SCD remains unknown.

Here, we use intravital (in vivo) lung¹³ and liver²⁷ microscopy in mice with SCD and imaging flow cytometry of blood from patients with SCD in vitro to reveal for the first time that the sterile inflammatory milieu in SCD promotes caspase-4 (human) or caspase-11 (mouse)-dependent activation of neutrophil-GSDMD, which triggers shedding of NETs in the liver microcirculation. These NETs travel intravascularly (embolize) from the liver to the lung, to promote P-selectin–independent lung vaso-occlusion in SCD.

Refer to supplemental Data, available on the Blood Web site for a full description of reagents and methods.

Townes SS and AS mice have been used previously as the SCD and nonsickle control mice, respectively.^14,28-32 Townes mice with SCD (SS) and nonsickle control mice (AS),^14,28-32 P-selectin–deficient SCD (SCD-Selp^−/−) mice,^27,33 and GSDMD knock-out (Gsdmd^−/−) mice³⁴ were used.

Quantitative fluorescence intravital lung microscopy (qFILM) has been used widely for in vivo assessment of pulmonary vaso-occlusion in mice with SCD.^{13,14,28,33,35} In the current study, qFILM was used to assess pulmonary vaso-occlusion and detect NETs in the intact lung microcirculation of live mice following IV challenge with saline, oxy-hemoglobin (oxy-Hb), hemin, or lipopolysaccharide with or without pretreatment with N-terminus active domain (GSDMD-NT) inhibitors LDC7559,²⁶ NSA,³⁶ Disulfiram,³⁷ or the pan-caspase inhibitor Z-VAD-FMK,²⁶ haptoglobin,³¹ or TLR4-inhibitor (TAK242).³¹ qFILM was conducted with a Nikon multiphoton excitation fluorescence microscope and an APO LWD 25× water immersion objective with 1.1 NA. Before imaging, mice were anesthetized with an intraperitoneal injection of ketamine HCl and xylazine. A cannula was inserted into the right carotid artery, and tracheotomy was performed to facilitate mechanical ventilation with 95% O₂ and supply maintenance anesthesia (1% to 2% isoflurane). The left lung was surgically exposed, and a small portion of the lung was immobilized against a coverslip using a vacuum-enabled micromachined device as described elsewhere.^35,38 Next, fluorescent Abs and dyes were injected into the carotid artery catheter for visualization of the pulmonary microcirculation, extracellular DNA, and in vivo staining of neutrophil elastase (NE), citrullinated histones, neutrophils, and platelets, respectively. Pulmonary vaso-occlusions and NETs were quantified and compared between treatment groups using the strategy described in supplemental Methods and described previously.^{13,14,18,28,33,39}

The liver intravital microscopy experimental setup has been described previously in detail.^27,29,40 The same strains of mice and treatment groups as those used in qFILM studies were used. Mice were anesthetized with an intraperitoneal injection of ketamine HCl and xylazine. The right carotid artery was cannulated, and the right lobe of the liver was gently immobilized against a coverslip using a vacuum-enabled micromachined liver-imaging window described elsewhere.^27,29,40 Next, fluorescent dyes and Abs were injected into the carotid artery catheter, and intravital observations were conducted with a Nikon multiphoton excitation fluorescence microscope.

Venous blood was collected in BD Vacutainers (containing sodium citrate) from healthy race-matched control humans (control) and steady state (not in crisis) patients with SCD (SS or S/β⁰) in accordance with the guidelines set by the Institutional Review Board at the University of Pittsburgh and the Declaration of Helsinki.

Imaging flow cytometry was conducted using Image Stream equipment (Amnis, Seattle, WA), and data were analyzed using IDEAS application software version 6.2.187.0 (Amnis).

MojoSort mouse neutrophil negative-selection kit (Biolegend) was used to isolate neutrophils from murine blood using the protocol described elsewhere.⁴¹ Flow cytometry was used to confirm neutrophil purity (>98%) as described elsewhere.^13,42,43

Venous blood was incubated for 15 minutes with agonists or inhibitors. After the incubation, blood was transferred into a 15-mL BD Falcon tube containing 4.5 mL PolymorphPrep, and neutrophils were isolated as described elsewhere.⁴⁴ 

Supplemental Tables 1 and 2 show the list of primary and secondary Abs, respectively, and the dilutions used.

The primer sequences used in the quantitative reverse transcription polymerase chain reaction are listed in the supplemental Table 3.

Means were compared using the unpaired Student t test without or with Bonferroni correction or one-way analysis of variance with Games-Howell's multiple comparison test. Percentages were compared using the fourfold table analysis with χ² statistics.^13,14,28P < .05 was considered significant. Unpaired or paired 2-tailed Student t test was used to confirm the significance in western blot analyses.

Using an IV oxy-Hb triggered model of vaso-occlusive crisis in mice with SCD,³¹ we found that 10 μmol/kg IV oxy-Hb led to acute lung injury in mice with SCD but not control mice (Figure 1A; supplemental Figure 1). Intravital lung microscopy analysis (experimental scheme in Figure 1B) revealed that IV oxy-Hb–induced lung injury was associated with the development of lung vaso-occlusion in mice with SCD. Neutrophils (red) were observed trafficking through the pulmonary arteriole and into the pulmonary capillaries of control mice administered IV oxy-Hb (Figure 1C; supplemental Video 1), indicating the absence of in situ lung vaso-occlusion. In contrast, IV oxy-Hb led to occlusion of pulmonary arteriole “bottlenecks” (junction of a pulmonary arteriole with pulmonary capillaries) by large neutrophil-platelet aggregates (marked by dotted ellipses in Figure 1D; supplemental Video 2). Pulmonary vaso-occlusions were quantified using the strategy described in “Methods” and supplemental data.^13,14,33 Number of pulmonary vaso-occlusions per field of view (FOV; Figure 1E), percent FOVs with pulmonary vaso-occlusions (Figure 1F), and number of large pulmonary vaso-occlusions (area >1000 μm²) per FOV (Figure 1G) were significantly higher in mice with SCD administered IV oxy-Hb than control mice administered IV oxy-Hb or mice with SCD administered IV saline. Pulmonary vaso-occlusions were also defined as neutrophil-rich (neutrophil aggregates with few platelets) or platelet-rich (platelet aggregates with few neutrophils), and both groups (representative examples shown in Figure 1D) were found to be significantly higher in mice with SCD than in control mice administered IV oxy-Hb (Figure 1H). Intravital studies using the IV-hemin triggered model of lung injury in mice with SCD^30,31 led to similar findings (supplemental Figure 2; supplemental Videos 3-4). Intravital lung microscopy (experimental scheme in Figure 1I) revealed significant abundance of NETs (extracellular DNA colocalized with NE and/or citrullinated-histones^18,20,39) in the lung of mice with SCD administered IV oxy-Hb than mice with SCD administered IV saline (Figure 1J) or control mice administered IV oxy-Hb (Figure 1K). To our surprise, these NETs were sequestered primarily within the pulmonary arteriole bottlenecks (Figure 1L-M; supplemental Videos 5-6) and localized within the large-neutrophil aggregates occluding the pulmonary arterioles in mice with SCD (Figure 1N; supplemental Video 7). Intravital studies using the IV-lipopolysaccharide triggered model of lung vaso-occlusion in mice with SCD^13,14,33,45 led to similar findings (supplemental Figure 3). NETs components such as citrullinated histones may promote intravascular thrombosis, which contributes to the development of lung vaso-occlusion in mice with SCD.^45-47 Indeed, fibrin was present within and around the neutrophil-platelet aggregates occluding the pulmonary arterioles in mice with SCD administered IV oxy-Hb (supplemental Figure 4). Importantly, both pulmonary vaso-occlusions and NETs were significantly fewer in mice with SCD IV administered 10 µmol/kg oxy-Hb with 10 µmol/kg haptoglobin³¹ than mice with SCD IV administered 10 µmol/kg oxy-Hb alone (supplemental Figure 5).

Remarkably, intravital lung microscopy (experimental scheme in Figure 2A) revealed circulating NETs (cNETs) entering the lung microcirculation via the pulmonary arterioles in mice with SCD administered IV oxy-Hb (Figure 2B; supplemental Figure 6; supplemental Videos 8-9). Entrance of cNETs through the pulmonary arterioles suggested that cNETs could not be originating in the pulmonary circulation but were probably shed by neutrophils in a nonpulmonary vascular bed and then carried by the blood to the lung. Intravital microscopy images of 3 representative FOVs #1 (top row in Figure 2B; supplemental Video 8), #2 (bottom row in Figure 2B; supplemental Video 9), and #3 (supplemental Figure 6) at different time points show several cNETs (green; marked with dotted ellipses) entering the lung microcirculation (purple) via the pulmonary arterioles over few milliseconds in an SCD but not control mouse (supplemental Figure 7; supplemental Video 10) administered IV oxy-Hb. Time series of intravital microscopy images were quantified over several FOVs in the lung of 3 to 5 mice per group, to estimate the average number of cNETs entering per FOV of lung per minute, which was significantly higher in SCD than control mice administered IV saline (Figure 2C), mice with SCD administered IV oxy-Hb than IV saline (Figure 2D), and mice with SCD than control mice administered IV oxy-Hb (Figure 2E). Recently,³³ we generated SCD-Selp^−/− mice that genetically lack P-selectin in all tissues. Unexpectedly, the average number of cNETs entering per FOV of lung per minute was not significantly different between SCD and SCD-Selp^−/− mice (Figure 2F), suggesting that shedding of cNETs in mice with SCD is P-selectin–independent. Next, we used imaging flow cytometry (experimental scheme in Figure 2G) to confirm whether cNETs in SCD mice blood were indeed NETs shed by neutrophils rather than DNA released by other cells. A representative imaging flow cytometry image (Figure 2H) shows a cNET as an extracellular DNA fragment positive for both neutrophil-elastase and citrullinated-histones (NETs markers), confirming that cNETs were indeed neutrophil-derived NETs. Identical to the intravital findings in Figure 2F, imaging flow cytometry also revealed that cNETs were not only abundant in plasma of mice with SCD administered IV oxy-Hb but also not significantly different from SCD-Selp^−/− mice administered IV oxy-Hb (Figure 2I), thus again confirming the shedding of cNETs in mice with SCD to be P-selectin–independent. Identical to mice with SCD and consistent with a previous report,²³ imaging flow cytometry (experimental scheme in Figure 2J) revealed that cNETs were also present in steady-state SCD patient blood (representative image in Figure 2K) and significantly more abundant than in control human blood (Figure 2L). The clinical characterization of human subjects is shown in Table 1. Cell-free oxy-Hb released following intravascular hemolysis undergoes oxidation in the vasculature in vivo to release the major DAMPs heme (ferrous protoporphyrin IX) and its oxidized form hemin (ferric protoporphyrin IX)^4,5,17 that promote sterile inflammation and NETs generation.^4,5,18 Thus, to more closely mimic the in vivo pathology, human blood samples were incubated with hemin instead of oxy-Hb in all in vitro studies to compensate for the absence of vasculature. Imaging flow cytometry revealed that in vitro incubation of SCD patient blood with hemin led to further (significant) increase in cNETs concentration (Figure 2M), suggesting that the inflammatory milieu (DAMPs) in SCD promotes shedding of cNETs.

The liver microcirculation is one of the primary sites for NETs generation by neutrophils during systemic inflammation,^7,39,48,49 liver also serves as the sink for heme and hemoglobin released following intravascular hemolysis,^{3,17,29,48,50} both heme and hemoglobin are potent agonists of NETs generation,^5,7,16,18,48 and recent findings show that liver is chronically inflamed in mice with SCD.^27,29,51 Intravital liver microscopy⁴⁰ (experimental scheme in Figure 3A) revealed that NETs identified as neutrophils (red) releasing extracellular DNA (green) were abundant in the liver microcirculation (purple) of mice with SCD administered IV oxy-Hb (Figure 3B) but not IV saline (supplemental Figure 8). Intravital images (Figure 3B) and corresponding videos (supplemental Videos 11-13) of 3 representative FOVs show numerous neutrophils releasing NETs (marked by dotted ellipse) in the liver microcirculation of mice with SCD administered IV oxy-Hb. Figure 3C shows intravital images of 4 individual neutrophils releasing NETs in the liver microcirculation of mice with SCD administered IV oxy-Hb. These findings were further validated by confocal microscopy of freshly isolated (unfixed) liver slices (Figure 3D; supplemental Figure 9) to identify NETs based on colocalization of extracellular DNA (green) with neutrophil-elastase (red) and neutrophils (white). As shown in the representative confocal images (Figure 3D) and the corresponding image analysis (Figure 3F), NETs were significantly more abundant in the liver of SCD than control mice administered IV oxy-Hb. Remarkably, intravital liver microscopy revealed shedding of NETs by neutrophils in the liver microcirculation of mice with SCD administered IV oxy-Hb (Figure 3E; supplemental Videos 14-15). In Figure 3E (top row), a fragment of ex-DNA (green; marked with dotted ellipse) can be seen detaching from the neutrophil (red) at t = 0.1 seconds and then disappearing into the liver microcirculation (purple) by t = 0.3 seconds. In Figure 3E (bottom row), several fragments of ex-DNA (green; marked with dotted ellipse) can be seen detaching from the neutrophil (red) at t = 0.07 seconds, and then disappearing into the liver microcirculation (purple) by t = 0.3 seconds. Compared with the liver, NETs were rare in the kidney microcirculation of mice with SCD administered IV oxy-Hb (Figure 3G; supplemental Figures 10-11; supplemental Video 16).

We hypothesized that NETs shed by neutrophils in the liver embolize to the lung as cNETs. We assessed the effect of acutely interrupting the blood flow from the liver to the lung on the arrival of cNETs in the lung, by simultaneously ligating the hepatic artery and portal vein (strategy in supplemental Figure 12). As shown in supplemental Video 17, hepatic artery and portal vein ligation (liver clamping) led to blood stasis in the liver. Remarkably, following liver clamping, cNETs arriving in the lung of mice with SCD administered IV oxy-Hb were rare (Figure 3H; supplemental Figures 13A-C; supplemental Videos 18, 25, and 26) and significantly reduced (Figure 3I; supplemental Figure 13D).

Interferon-α and -β (type I IFNs) signal through the heterodimeric IFNAR1/2 receptor in an autocrine or paracrine manner to promote downstream type I IFN signaling, which leads to upregulation of several IFN-stimulated genes.^52,53 Recent evidence suggests that type I IFNs are significantly elevated at steady state (not in crisis) in the plasma of patients with SCD,^54,55 leading to activation of type I IFN signaling in SCD patient neutrophils.^54,56 We found a significant alteration in the transcript levels of several key components (marked by grey boxes in Figure 4B) of the type I IFN pathway in neutrophils isolated from the blood of SCD compared with control mice administered 10 µmol/kg IV oxy-Hb (experimental scheme in Figure 4A; purity in supplemental Figure 14). Compared with control mice neutrophils, SCD mice neutrophils manifested higher transcripts level of IFN-α (Ifna), IFN-β (Ifnb1), IFNAR1 (Ifnar1), and the major kinase Tyk2 (Tyk2) involved in the type 1 IFN signaling (Figure 4C-F). Transcript levels of caspase-11 (casp11), an IFN-stimulated gene known to be upregulated downstream of type 1 IFN signaling,^24,57-59 were also significantly elevated in neutrophils of SCD compared with control mice administered IV oxy-Hb (Figure 4B,G). Recently, neutrophil-caspase-11 was shown to be activated (cleaved) following cytosolic invasion by gram-negative bacteria and then cleave pore-forming protein GSDMD into the GSDMD-NT,²⁵ which triggered chromatin expansion, granule protein release, and extrusion of NETs following GSDMD-NT–mediated pore formation in nuclear, granule, and plasma membranes, respectively.^25,26 Interestingly, transcript levels of both GSDMD and its transcriptional regulator IRF2⁶⁰ were also significantly elevated in neutrophils of SCD compared with control mice administered IV oxy-Hb (Figure 4B,H). Surprisingly, caspase-11 protein expression was undetectable in neutrophils of control mice but present in both uncleaved (45 kDa) and active-cleaved (20 and 25 kDa) state in neutrophils of mice with SCD administered IV oxy-Hb (Figure 4I-K). Similar to caspase-11, its substrate GSDMD was also undetectable in neutrophils of control mice but present in both uncleaved (50 kDa) and active (cleaved) GSDMD-NT (30 kDa) state in neutrophils of mice with SCD administered IV oxy-Hb (Figure 4L-N). GSDMD-NT was undetectable in neutrophils of both control and mice with SCD administered IV saline (supplemental Figure 16).

Next, SCD or control human subjects’ blood without or with incubation with hemin (± inhibitors) was used for isolation of neutrophils (experimental scheme in Figure 5A; purity in supplemental Figure 17). Identical to mice with SCD (Figure 4L), protein expression of uncleaved GSDMD was significantly higher in neutrophils isolated from SCD than control human blood without (Figure 5B-C) or with hemin-incubation (Figure 5B,D). GSDMD-NT was also significantly higher in neutrophils isolated from SCD than control human blood without (Figure 5B,E) or with hemin-incubation (Figure 5B,F). Similar to the effect on cNETs concentration (Figure 2M), incubation of SCD patient blood with hemin also led to further (significant) increase in GSDMD-NT expression in neutrophils (Figure 5B,G). Recently, heme was shown to promote caspase-4 (human ortholog of murine caspase-11^24,57,58) -dependent cleavage of GSDMD in human macrophages.⁶¹ We found that the expression of caspase-4 was also significantly higher in neutrophils isolated from SCD than control human blood without (Figure 5H-I) or with hemin-incubation (Figure 5H,J). N-acetyl-l-cysteine (NAC) is an antioxidant,³¹ which was earlier shown to prevent heme-induced NETs generation by scavenging ROS.¹⁸ We found that caspase-4 was significantly reduced in neutrophils isolated from SCD patient blood incubated with both hemin and NAC compared with hemin alone (Figure 5H,K), suggesting a role for ROS in promoting caspase-4 expression in SCD patient neutrophils. Finally, GSDMD-NT expression was significantly reduced in neutrophils isolated from SCD patient blood incubated with hemin and caspase-4 inhibitor LEVD-CHO⁶² (Figure 5L-M) or hemin and NAC (Figure 5L,N) compared with hemin alone.

SCD patient blood incubated with 20 μM hemin (± inhibitors) was used in imaging flow cytometry (experimental scheme in Figure 6A) for detection of cNETs. In vitro incubation of SCD patient blood with hemin in the presence of a GSDMD-NT inhibitor LDC7559²⁶ (Figure 6B) or caspase-4 inhibitor LEVD-CHO (Figure 6C) or antioxidant NAC (Figure 6D) led to significant reduction in cNETs concentration compared with incubation with hemin alone. Also, imaging flow cytometry (experimental scheme in Figure 6E) revealed significantly reduced cNETs concentration in the blood of mice with SCD IV administered 10 µmol/kg oxy-Hb with 10 mg/kg GSDMD-NT inhibitor (LDC7559) than mice with SCD IV administered 10 µmol/kg oxy-Hb alone (Figure 6F). Remarkably, imaging flow cytometry also revealed significant reduction in the cNETs concentration in the blood of SCD-Selp^−/− mice IV administered 10 µmol/kg oxy-Hb with 10 mg/kg GSDMD-NT inhibitor (LDC7559) compared with SCD-Selp^−/− mice IV administered 10 µmol/kg oxy-Hb alone (Figure 6G).

We hypothesized that reduction in plasma concentration of cNETs following GSDMD inhibition is secondary to reduced shedding of cNETs in the liver microcirculation. Therefore, we assessed the effect of GSDMD-NT inhibition (using 3 inhibitors: LDC7559,²⁶ Necrosulfonamide,³⁶ and Disulfiram³⁷) or GSDMD deficiency (using Gsdmd^−/− mice³⁴) on NETs generation in the mice liver using intravital microscopy (experimental scheme in Figure 6E). As shown by the representative liver intravital images (Figure 6H), corresponding videos (supplemental Videos 19-20) and the image analysis (Figure 6I), NETs (marked by dotted ellipses) identified as neutrophils (red) releasing extracellular DNA (green), were significantly abundant in the liver microcirculation (purple) of mice with SCD administered 10 µmol/kg IV oxy-Hb but rare (only a single small NET visible in the Figure 6H, right panel) in the liver microcirculation (purple) of mice with SCD IV administered 10 µmol/kg oxy-Hb with 10 mg/kg GSDMD-NT inhibitor (LDC7559). IV administration of 10 µmol/kg oxy-Hb was innocuous to wild-type (WT) mice; however, twofold higher dose of IV oxy-Hb (20 µmol/kg) led to NETs generation in the liver microcirculation of WT mice as well, which was significantly attenuated in Gsdmd^−/− mice (Figure 6J; supplemental Figure 21). Next, we hypothesized that attenuation of NETs generation in the liver after GSDMD inhibition or deletion would translate to fewer cNETs arriving in the lung. Indeed, intravital lung microscopy analyses (experimental scheme in Figure 6E) revealed several-fold (significantly) fewer cNETs entering per FOV per minute in the lung of mice with SCD IV administered 10 µmol/kg oxy-Hb with 0.004 µmol/kg pan-caspase inhibitor Z-VAD-FMK²⁶ (Figure 6K) or 10 mg/kg GSDMD-NT inhibitor LDC7559 (Figure 6L) or 20 mg/kg GSDMD-NT inhibitor Necrosulfonamide³⁶ (Figure 6M) than mice with SCD IV administered 10 µmol/kg oxy-Hb. Concomitant to reduced NETs generation in the Gsdmd^−/− mice liver (Figure 6J), cNETs entering per FOV per minute in the lung were also significantly fewer in Gsdmd^−/− than WT mice administered 20 µmol/kg IV oxy-Hb (Figure 6N). Remarkably, cNETs entering per FOV per minute in the lung were also fourfold fewer in SCD-Selp^−/− mice IV administered 10 µmol/kg oxy-Hb with 10 mg/kg GSDMD-NT inhibitor (LDC7559 or disulfiram) than SCD-Selp^−/− mice IV administered 10 µmol/kg oxy-Hb alone (Figure 6O; supplemental Figure 22).

Next, intravital lung microscopy (experimental scheme in Figure 7A) was conducted in mice to assess whether preventing liver-to-lung embolization of cNETs by inhibiting or deleting GSDMD signaling leads to amelioration of pulmonary vaso-occlusion. Although large neutrophil-platelet aggregates (marked with dotted ellipses) were seen occluding the arteriolar bottlenecks, leading to impaired blood flow in the lung of mice with SCD IV administered 10 µmol/kg oxy-Hb (Figure 7B; supplemental Video 21), such aggregates were rare in the lung of mice with SCD IV administered 10 µmol/kg oxy-Hb with 0.004 µmol/kg pan-caspase inhibitor Z-VAD-FMK (Figure 7C; supplemental Video 22), or 10 mg/kg GSDMD-NT inhibitor LDC7559 (Figure 7D; supplemental Video 23), or 20 mg/kg GSDMD-NT inhibitor Necrosulfonamide (Figure 7E; supplemental Video 24). Erythrocytes (dark cells) were observed flowing unobstructed through the pulmonary arteriole and into the pulmonary capillaries, suggestive of the lack of pulmonary vaso-occlusion in mice with SCD IV administered oxy-Hb with Z-VAD-FMK (supplemental Video 22) or LDC7559 (supplemental Video 23) or Necrosulfonamide (supplemental Video 24). Concomitant to the reduction in number of cNETs arriving in the lung (Figure 6K-M), the number of pulmonary vaso-occlusions per FOV and the number of large pulmonary vaso-occlusions (area >1000 μm²) per FOV were also significantly reduced in mice with SCD IV administered oxy-Hb with Z-VAD-FMK (Figure 7F-G), LDC7559 (Figure 7H-I), or Necrosulfonamide (Figure 7J-K) than oxy-Hb alone. The twofold higher dose of 20 µmol/kg IV oxy-Hb also led to the development of lung vaso-occlusion in WT but not Gsdmd^−/− mice (Figure 7L-M; supplemental Figure 23). Both the number of pulmonary vaso-occlusions per FOV (Figure 7L) and the large pulmonary vaso-occlusions (area >1000 μm²) per FOV (Figure 7M) were significantly less in Gsdmd^−/− than WT mice administered 20 µmol/kg IV oxy-Hb. Identical to our previous report,³³ P-selectin deficiency led to ∼50% reduction in pulmonary vaso-occlusion in SCD-Selp^−/− mice administered 10 µmol/kg IV oxy-Hb (supplemental Figure 24); however, IV administration of oxy-Hb with the GSDMD-NT inhibitor (LDC7559 or disulfiram) led to a further threefold (significant) reduction in pulmonary vaso-occlusions per FOV (Figure 7N; supplemental Figure 25A; supplemental Figure 26) and absence of large pulmonary vaso-occlusions (area >1000 μm²) per FOV (Figure 7O; supplemental Figures 25B-26) in SCD-Selp^−/− mice.

Cell-free oxy-Hb released during intravascular hemolysis is scavenged by plasma haptoglobin, which chaperones it to the liver, spleen, and bone marrow for clearance.^1,3,17 However, SCD is associated with chronic depletion of haptoglobin, leading to impaired scavenging of cell-free oxy-Hb, which contributes to development of sterile inflammation in SCD.^4,5,16 Based on this, mice with SCD were systemically (IV) challenged with oxy-Hb to trigger vaso-occlusive crisis.^28,31 Intravital microscopy revealed that IV oxy-Hb promoted accumulation of NETs in the pulmonary arterioles and NETs-dependent lung vaso-occlusion by neutrophil-platelet aggregates, leading to development of lung injury in mice with SCD. To our surprise, these NETs were found to enter the lung via the pulmonary arterioles, suggesting that they originate in a nonpulmonary vascular bed. We found that these NETs were primarily shed by neutrophils in the liver microcirculation of mice with SCD and then transported by blood (as cNETs) to the lung, where they promoted lung vaso-occlusion. Indeed, cNETs were abundant in the peripheral blood of mice with SCD, and the levels were further elevated following challenge with IV oxy-Hb.

Interestingly, the type I IFN pathway was upregulated, and its downstream target caspase-11 and the caspase-11 substrate GSDMD were activated in neutrophils of mice with SCD given IV oxy-Hb. The cNETs were also abundant in SCD patient blood and further elevated following incubation with hemin, which was secondary to ROS and caspase-4–dependent activation of neutrophil-GSDMD. Inhibition of caspase-11 or GSDMD or absence of GSDMD prevented IV oxy-Hb–induced shedding of cNETs in the liver, their embolization to lung, and development of lung vaso-occlusion in SCD and Gsdmd^−/− mice. Importantly, shedding of cNETs and their embolization to lung were not affected by the absence of P-selectin, but significantly prevented following GSDMD inhibition in SCD-Selp^−/− mice given IV oxy-Hb. Finally, GSDMD inhibition completely abolished the remaining lung vaso-occlusion present in SCD-Selp^−/− mice given IV oxy-Hb. Taken together, our current findings (Figure 7P) suggest for the first time that the sterile inflammatory milieu (DAMPs) in SCD promotes caspase-4/11–dependent activation of neutrophil-GSDMD, which leads to P-selectin–independent NETs generation in the liver. These NETs detach from parent neutrophils in the liver and then arrive as cNETs in the lung to promote occlusion of pulmonary arterioles by neutrophil-platelet aggregates. Most importantly, GSDMD inhibition also abrogates P-selectin–independent lung vaso-occlusion in SCD.

The interpretation of our findings is associated with a few limitations that may inspire further investigation in future studies. First, although we did not observe a significant increase in NE transcript levels in SCD than control mice neutrophils, new evidence suggests that NE released into the cytosol following GSDMD-NT–mediated permeabilization of azurophilic granules may also contribute to GSDMD cleavage via a feed-forward loop.^26,63,64 Second, although our data demonstrate liver as a major source of cNETs arriving in the lung in SCD, contributions from other vascular beds including lung cannot be ruled out. Third, because of the technical limitations associated with intravital fluorescence microscopy, minor contribution of nonneutrophil DNA to cNETs cannot be ruled out. Fourth, GSDMD activation in cells other than neutrophils may also contribute to lung injury. Fifth, the effect of GSDMD inhibition on chronic organ injury in mice with SCD remains to be investigated in future studies. Sixth, the higher baseline GSDMD-NT expression in humans as compared with mice neutrophils is suggestive of the limitations associated with mice models.

Notwithstanding these limitations, our current findings introduce a novel paradigm that liver-to-lung translocation of DAMPs promotes lung injury in SCD and identify a new GSDMD-mediated, P-selectin–independent mechanism of lung vaso-occlusion in SCD. Despite recent advances in new therapies,^15,65 acute chest syndrome continues to be a major cause of morbidity among hospitalized patients with SCD,³ but a pharmacological therapy to prevent its clinical onset still remains elusive.⁶ The current study is the first to highlight the therapeutic potential of a multitarget therapy of blocking both P-selectin and GSDMD-dependent events, to prevent the development of acute chest syndrome in high-risk patients with SCD hospitalized with painful vaso-occlusive episodes. We hypothesize that such a combined anti-inflammatory approach could significantly control vaso-occlusive painful episodes and secondary acute chest syndrome in patients with SCD.

This work was supported by National Institutes of Health (NIH), National Heart, Lung, and Blood Institute grants R01HL128297 (P.S.) and R01HL141080 (P.S.), American Heart Association 18TPA34170588 (P.S.), funds from the Hemophilia Center of Western Pennsylvania and Vitalant (P.S.) and NIH, National Institute of Diabetes and Digestive and Kidney Diseases 1K01DK125617-01 (T.P.-S.) and funds from the Hemophilia Center of Western Pennsylvania (T.P.-S.). The Nikon multiphoton excitation microscopes were funded by NIH grants S10RR028478 and S10OD025041. R.V. was supported by American Heart Association predoctoral fellowship 19PRE34430188. T.W.K. was supported by American Heart Association postdoctoral fellowship AHA828786. T.B. was supported by American Society of Hematology Postdoctoral Scholar Award and the Research Restart Award. Anti-mouse fibrin Ab (clone 59D8) was a gift from Rafal Pawlinski at University of North Carolina–Chapel Hill.

The visual abstract was created using BioRender software.

Contribution: R.V. conducted intravital microscopy and imaging flow cytometry studies; T.W.K. conducted western blot and messenger RNA analyses; T.B. was involved in lung intravital microscopy studies; J.A.L. was involved in western blot analyses; E.T. and O.K. bred and genotyped all the transgenic mice used in this study; J.T. prepared oxy-Hb used in vivo mice studies; J.J. and E.M.N. provided blood samples from human subjects; T.P.-S. was involved in liver intravital and confocal microscopy studies; M.T.G. was involved in experimental design and manuscript writing; P.S. was responsible for experimental design, manuscript writing, and project supervision; and P.S., R.V., and T.W.K. wrote the manuscript with consultation and contribution from all coauthors.

Conflict-of-interest disclosure: P.S. receives funding (not relevant to the current study) as a part of sponsored research agreements with CSL Behring Inc, IHP Therapeutics, and Novartis Pharmaceuticals Corporation. P.S. is also the recipient of 2021 Bayer Hemophilia Award Program (not relevant to the current study). The remaining authors declare no competing financial interests.

Correspondence: Prithu Sundd, Division of Pulmonary, Allergy and Critical Care Medicine, Pittsburgh Heart, Lung and Blood Vascular Medicine Institute, University of Pittsburgh School of Medicine, BST E1255, 200 Lothrop St, Pittsburgh, PA 15261; e-mail: prs51@pitt.edu.