Nlrp3^A350VneoR knock-in mice of 10–14 weeks old, on C57BL/6 background, in which a loxP-flanked neomycin resistance cassette is inserted in reverse orientation into intron 2 of the Nlrp3 gene, were purchased from Jackson Lab (strain 017969). These mice have a missense mutation, A350V, in the exon 3 on the Nlrp3 gene, which resembles the mutation observed in human Muckle–Wells syndrome. Nlrp3^A350VneoR knock-in mice were crossed to Gp1ba-Cre^KI/+ mice (9) to generate mutation specific for MK/platelet lineage. The resulting Nlrp3 ^A350V/+/Gp1ba-Cre^KI/+ mice were used in the experiments with Nlrp3 ^A350V/+/Gp1ba-Cre^+/+ and Nlrp3 ^+/+/Gp1ba-Cre^KI/+ littermates serving as a control. As Nlrp3 and Gp1ba are both located close (~10 MBs) on chromosome 11, breeding efficacy of homozygous animals was very low and therefore we used heterozygous mice. Both male and female mice aged 8–14 weeks were used. All experiments were performed in accordance with UK law (Animal Scientific Procedures Act 1986), with approval of the local ethics committee and UK Home Office.

Blood was drawn from the inferior vena cava (IVC) of anesthetized mice to 10% anti-coagulant citrate dextrose (ACD). Blood counts were performed using an ABX Pentra 80 automated hematology analyzer (HORIBA). Platelets were isolated by centrifugation using Modified Tyrode’s Buffer (pH 7.4) and washed in the presence of prostacyclin (1 µg/ml). Platelet aggregation at 2 × 10⁸/ml was assessed by light transmission aggregometry (ChronoLog, USA) in glass vials at 37°C, stirring at 1,200 rpm for 6 min, in the absence or presence of an agonist.

Whole blood was perfused at arterial shear rate (1,000 s⁻¹) as described (10). In brief, whole blood was perfused over Horm collagen (coated at 100 µg/ml) through a µ-Slide VI 0.1. Blood was treated with 1 µM DiOC6 (ThermoFisher) and 40 µM PPACK. Images were acquired simultaneously of two fields with Z-stacks using an EVOS M5000 Imaging system (ThermoFisher) by fluorescence (20× lens). Platelet deposition was calculated as percentage of surface coverage using FIJI v2.1.

Deep vein thrombosis (DVT) was modeled by IVC stenosis as described (11). In brief, mouse was placed in the supine position and anesthetized with isoflurane–oxygen mixture, and laparotomy was performed through midline incision. Guts were exteriorized and IVC was gently separated from aorta. The IVC was closed over a spacer (30G needle) with a 7-0 prolene suture, after which the spacer was removed. This results in ~90% closure of the IVC lumen (12). All visible side branches were ligated. Mice were euthanized and thrombosis checked in 48 h.

The femur and tibia were harvested and cleaned from surrounding tissues. One side of the bone was cut using sterile scissors, before fixation in 4% formalin for 24 h. Bones were then decalcified in EDTA (500 mM) for 72 h, or until rigidity is lost, and embedded in paraffin. Tissue was sectioned to 8 µm, deparaffinized with xylene, and rehydrated before antigen retrieval in antigen unmasking solution (Vector Labs, UK) and blocking using REAL peroxidase blocking solution (DAKO, UK). Sections were stained using a Hematoxylin & Eosin (H&E) kit (abcam, UK). Tissue was dehydrated and mounted with VECTASHIELD anti-fade mounting medium (Vector Labs, UK). Tissue was imaged using a Zeiss AxioScan.Z1 slide scanner.

Spleens were harvested, weighed, and halved (coronally), before embedding in optical cutting temperature (OCT) and snap freezing in liquid nitrogen. Spleens were then cyrosectioned to 6 µm, mounted on slides, and stored at −80°C. Sections were fixed in ice-cold acetone before blocking in 3% BSA and 5% donkey serum in TBS for 1 h. Sections were incubated with primary antibodies Ter119 (Ter119-AF488, BioLegend), CD3e (145-2C11-Hamster, eBioScience), and CD19 (eBio1D3-Rat, eBioScience) at 4°C overnight. Sections were TBS-T-washed, before incubation with secondary antibodies for 1 h at room temperature, and then washed. Autofluorescence was quenched by ammonium chloride (50 mM), and tissue was flat mounted with VECTASHIELD anti-fade mounting medium (Vector Labs, UK). Organs were imaged using a Zeiss AxioScan.Z1 slide scanner, and images were compiled using FIJI v2.1.

Bone marrow (BM) was isolated from the femur and tibia of mice by centrifugation, as described (13), and filtered through a 70-µm filter unit. The spleen was homogenized through a Falcon 70-µm cell strainer using the flat end of a 1-ml syringe, before final filtration through a 70-µm filter unit. Whole blood was drawn into 10% ACD, and diluted 1:10 in Tyrode’s Buffer. Cells were incubated in the presence of conjugated antibodies for 20 min on ice: CD41 (MWReg30-APC, eBioScience), GPIbα (1C2-PE, BioLegend), GPVI (784808-AF647, R&D), CLEC-2 (17D9-PE, BioLegend), Ter119 (Ter119-AF488, BioLegend), CD62P (RMP-1-PE, BioLegend), and CD71 (C2F2-PE, BioLegend). For intracellular staining, cells were permeabilized using PBS–Triton X-100 (0.1%, Abcam) for 10 min, before staining for GATA-1, anti-TGF-β1-APC (TW7-16B4, BioLegend), and FAM-FLICA Caspase-1 assay kit (Immunochemistry Technologies). Cells were fixed in 4% formalin, before acquisition using an Accuri C6 flow cytometer (BD Bioscience, UK), and analyzed using FlowJo v10. In zymosan-induced peritonitis, peritoneal cells (all) and splenocytes (⅕) were aliquoted into 12 × 75 mm polystyrene FACS tubes and incubated 1:50 with mouse FcR block (130-092-575; Miltenyi Biotec) for 2 min at RT. Peritoneal cells and splenocytes were then stained with the following antibodies for 20 min at 4°C prior to washing and fixation with 2% PFA: anti-CD45.2 BV605 (clone 104; BioLegend), anti-CD11c PE-Cy7 (clone N418, ThermoFisher), anti-Siglec F TexasRed (clone E50-2440, BD), anti-Ly6C FITC (clone HK1.4, BioLegend), anti-Ly6G APC (clone 1A8, BD), and anti-F4/80 APC-eFluor780 (clone BM8, ThermoFisher). Splenocytes were also stained with anti-CD3 PECy7 (clone 145-2C11, ThermoFisher) and anti-CD19 APC (clone 1D3, ThermoFisher). Immediately prior to analysis, CountBright beads (Invitrogen) and Zombie Aqua (BioLegend) were added and samples were acquired using Fortessa-X20 and data analyzed offline using FlowJo (V-10.2.6).

Cells and CountBright beads were gated on using the forward (FSC) and side (SSC) scatter profiles. Doublets and dead cells were excluded using forward scatter area versus height (FSC-A and FSC-H), and positive zombie aqua signal, respectively. Leukocytes were identified based on positive expression of CD45, and subdivided into dendritic cells (CD11c⁺), eosinophils (SiglecF⁺), monocytes (Ly6C⁺), neutrophils (Ly6G⁺), macrophages (F4/80^hi), T cells (CD3⁺), and B cells (CD19⁺). The absolute numbers of each leukocyte population listed above were quantified using the CountBright beads (ThermoFisher) according to the manufacturer’s instructions, and the obtained value was multiplied by the denominator of the fraction of tissue stained to provide the total number of leukocytes within that tissue.

Washed mouse platelets were isolated at 4 × 10⁸/ml, mixed with 2× Laemmli Sample Buffer (Bio-Rad, USA), and boiled. Protein electrophoresis was performed in Tris-Glycine SDS running buffer and transferred to an Immuno-Blot PVDF Membrane (Bio-Rad, USA). Membranes were imaged for peroxidase post-ECL substrate incubation using a LI-COR Odyssey Fc Imager and quantified using ImageJ.

Peritonitis was induced by intraperitoneal (IP) injection of 0.1 mg zymosan A (Z4250-250mg; Merck, UK) from Saccharomyces cerevisiae. After 4 h, mice were sacrificed via CO₂ narcosis and cervical dislocation. The peritoneum was lavaged by injecting 5 ml of 5 mM EDTA (Sigma-Aldrich) diluted in PBS without Mg/Ca into the peritoneal cavity and massaging the peritoneal membrane to detach adherent cells. The peritoneal lavage fluid (PLF) was collected using a 19 G 2-inch needle to a 5-ml syringe and stored on ice. The spleen was collected into a 15- ml falcon tube containing PBS with Mg²⁺/Ca²⁺ and transported on ice. Tissues were processed immediately as described below.

PLF samples were centrifuged at 300 g for 5 min to pellet cells. PLF supernatants were collected and stored at −80°C for future analysis. Peritoneal cells were then resuspended in MACS buffer [0.1 mM ethylenediaminetetraacetic acid (EDTA) and 0.6% bovine serum albumin (BSA) in PBS without Mg²⁺/Ca²⁺; all Sigma-Aldrich] and stored on ice prior to analysis by flow cytometry. Spleens were weighed using Pioneer weighing scales (Ohaus, Switzerland) and placed onto 40-µm cell strainers (Grenier Bio-One, Austria). Using the plunger of a 5-ml syringe, the spleen was mashed and washed through the cell strainer into the collection tube with PBS with Mg/Ca. Cells were pelleted by centrifugation at 300 g for 5 min. To lyse red blood cells (RBCs), 3 ml of formulated RBC lysis buffer (155 mM NH₄Cl, 12 mM NaHCO₃, and 0.1 mM EDTA; all from Sigma-Aldrich) was added to spleen samples for 10 min at RT. Splenocytes were then washed twice by diluting cells with 10 ml of PBS with Mg²⁺/Ca²⁺ followed by centrifugation at 300 g for 5 min. Splenocytes were then resuspended in MACS buffer and stored on ice prior to analysis by flow cytometry.

Cytokines were measured in non-diluted PLF using Peprotech (UK) mini-ELISA kits according to the manufacturer’s instructions.

Blood smears were air-dried, fixed in methanol for 10 min, and stained with May–Grünwald Giemsa staining kit (Atom Scientific, UK) according to the manufacturer’s instructions.

Femoral bones were fixed in 4% paraformaldehyde, decalcified, and embedded in paraffin. Sections were stained with an Elastic Van Gieson Stain Kit (Verhoeffs) (Atom Scientific UK) according to the manufacturer’s instructions.

Data are presented as mean ± SD. For all peritonitis results, bar in dot plot represents median. The statistical difference between multiple groups was assessed by Kruskal–Wallis test or (for all peritonitis results) by one-way ANOVA or (for ELISA data and other comparison between two groups) by Mann–Whitney test using Prism 8 (GraphPad Software Inc, USA). Statistical significance is represented by asterisks: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

MK/platelet-specific NLRP3 gain-of-function mice were generated by crossing Nlrp3 ^A350V/+-floxed (fl) mice with the Gp1ba-Cre knockin (KI) mice. The Gp1ba-Cre transgene was used rather than Pf4-Cre because it has been shown to be more lineage-specific (9), eliminating confounding effects of hyperactivation of NLRP3 in immune cells. Nlrp3 ^A350V/+/Gp1ba-Cre^KI/+ mice developed normally ( Supplementary Figures 1A, B ) and did not demonstrate any visible abnormalities. First, we confirmed that the mutation results in increased activity of NLRP3 in cells of megakaryocytic origin. As activation of caspase-1 is one of the central outcomes of NLRP3 inflammasome function, we determined both pro-caspase-1 and the active protein in platelets using Western blotting. The intraplatelet level of both total and active caspase-1 was significantly higher in platelets from NLRP3 gain-of-function mice compared with controls ( Figures 1A–C ).

A variety of potential effects of NLRP3 hyperactivation could potentially be attributed to elevated plasma levels of active IL-1β secreted from platelets, especially given their abundance in the circulation. We therefore measured IL-1β concentration in plasma of both mutant and control mice. Surprisingly, the levels of the cytokine were similar in both control and mutant strains ( Figure 1D ). This could be based on the recently reported novel mechanism of IL-1β secretion from platelets, which is independent of inflammasome and caspases (14). Also, we used unchallenged mice and therefore these results cannot rule out potential participation of platelet-derived IL-1β in the presence of a challenge, for example, after platelet stimulation with LPS in vitro or following cecal ligation and puncture in vivo (15). Of note, the seeming difference between medians in the control groups was highly insignificant (p > 0.65 by ANOVA Kruskal–Wallis test).

Thus, any phenotypes observed in unchallenged Nlrp3 ^A350V/+/Gp1ba-Cre^KI/+ mice would be unrelated to systemic alterations in the IL-1β level.

Mutant mice had mildly decreased RBC counts (7,605 ± 1,340 vs. 9,214 ± 289 and 9,184 ± 430 × 10³/μl in controls, respectively) and hemoglobin (11.8 ± 1.9 g/dl vs. 13.7 ± 1.4 g/dl and 13.3 ± 0.9 g/dl in controls, respectively; Figures 1E, F ), whereas the numbers of white blood cells remained unchanged (8.1 ± 4.3 vs. 5.2 ± 2.6 × 10³/μl in control, p > 2.3). To explore the cause of the anemia, we assessed erythropoiesis in the BM of the mutant mice. Femoral bones of the mutant mice had visually reduced red color ( Figure 1G ) and lower numbers of nucleated cells ( Figure 1H ). Consistent with this finding, the proportion of cells positive for the erythroid marker Ter119 was dramatically decreased in the Nlrp3 ^A350V/+/Gp1ba-Cre^KI/+ mice ( Figure 1I ). In contrast, the number of MK did not differ between the mouse strains ( Supplementary Figure 1C ). These results imply that the numbers of the maturing erythroblasts are markedly reduced. No difference in BM fibrosis was observed ( Supplementary Figure 1D ). The percentage of reticulocytes (Ter119⁺CD71⁺ or Ter119⁺thiazole orange⁺, 10.9 ± 3.4 vs. 10.7 ± 0.7 and 11 ± 5.1 vs. 11.7 ± 1.8 in mutant and controls, respectively; p > 0.7) and young platelets (CD41⁺thiazole orange⁺, 10.8 ± 5.3 vs. 10.7 ± 0.8 in mutant and controls, respectively; p > 0.4) in the peripheral blood were similar in the mutant animals and controls. Peripheral blood smears of Nlrp3 ^A350V/+/Gp1ba-Cre^KI/+ mice did not visually differ from controls ( Supplementary Figure 1E ).

GATA1 is one of the major transcription factors supporting survival, proliferation, and differentiation of erythroid progenitors, downstream of the erythropoietin receptor (16). Thus, to investigate potential mechanisms of reduced RBC counts in Nlrp3 ^A350V/+/Gp1ba-Cre^KI/+ mice, we measured levels of GATA1 in cells of erythroid lineage (Ter119⁺) in the BM of the mutant mice; however, no differences compared with control mice were observed ( Figure 1J ). A role of MK-derived TGF-β1 in the transition of erythroid progenitors to the erythropoietin-sensitive stage has recently been reported (17). Reduced numbers of Ter119⁺ precursors in the BM suggested that this mechanism could be implicated in the observed phenotype. Indeed, TGF-β1 levels were increased in the BM MK, although compared with only one (Nlrp3 ^+/+/Gp1ba-Cre^KI/+) but not another (Nlrp3 ^A350V/+/Gp1ba-Cre^+/+) control ( Figure 1K ). It is also known that CD109, a TGF-β coreceptor, downregulates TGF-β signaling in hematopoietic progenitors (18), and its modified expression could therefore mediate reduced maturation of the early precursors even in the absence or minimal changes in the levels of TGF-β itself.

Different subtypes of MK have recently been described, including platelet-forming, immunoregulatory, niche-supporting, and endomitotic cells (19). As immunoregulatory MKs are enriched with inflammatory markers and genes implicated in cytokine synthesis and secretion, it is tempting to hypothesize that the erythropoiesis-modifying stimulus, regulated by NLRP3, originates from this MK subtype. The percentage of this MK subtype, recognized as CD41⁺ cells expressing S100A9, was similar in mutants and controls (7.7 ± 4.2 vs. 5.8 ± 2.1, respectively, p > 0.6), which, however, does not rule out the potential difference in their cytokine production activity.

These findings reveal a novel and unexpected role of MK/platelet NLRP3 in erythropoiesis. Understanding the molecular mechanism underlying this phenomenon, in particular, whether NLRP3 operates alone or as a part of inflammasome, requires further investigation.

The spleen was enlarged in the mutant mice, with its weight (normalized to the body weight) exceeding the weight of the spleen of control animals by 1.5-fold (6.7 ± 1.7 vs. 3.8 ± 0.6 and 3.9 ± 0.8 g in controls, respectively; Figure 1L ). Staining of spleen sections revealed vague borders between the red and white pulp, which could reflect progressive growth of spleen size, but no prominent visible differences in the content of Ter119⁺ erythroid progenitors, and T or B cells, between control and Nlrp3 ^A350V/+/Gp1ba-Cre^KI/+ mice ( Figure 1M ). This was further confirmed by flow cytometry, which showed similar proportion of Ter119⁺ cells and erythroblast subtypes, EryA, EryB, and EryC, as well as MKs in the spleen of mutant and control animals ( Supplementary Figures 1F–H ). Thus, the overall content of maturing erythroid precursors in the spleen is increased proportionally to its size, which likely reflects a compensatory effect for reduced BM erythropoiesis. Hyperactive NLRP3 in MK might create an inflammatory environment, which is known to suppress steady-state erythropoietin-induced BM erythropoiesis and simultaneously stimulate extramedullary stress erythropoiesis, in particular, in spleen (20). This mechanism could contribute to the observed phenotype.

As the NLRP3 activity in our mouse model is platelet-specific, we next assessed whether it affects key platelet functions. No difference in the platelet counts or expression of major platelet receptors, GPIbα, the integrin β3 subunit (CD41), CLEC-2, and GPVI was detected ( Supplementary Figures 2A–E ). The release of α-granules (as judged by surface P-selectin expression) was comparable in mutant vs. control platelets in response to either CRP or thrombin ( Supplementary Figures 2F, G ) as well as aggregation induced by CRP, thrombin, U46619 (a thromboxane A agonist), and H3 ( Supplementary Figures 2H–K ). It was previously reported that inhibition of the NLRP3 inflammasome through Btk kinase attenuated platelet activation, aggregation, and thrombus development in vitro (21). The discrepancy with our findings may be explained by the fact that genetic upregulation of NLRP3 activity in a lineage-specific manner may not necessarily produce an effect opposite to its inhibition. Moreover, potential additional downstream and off-target effects of Btk kinase inhibitors unrelated to the inflammasome cannot be ruled out.

No changes in platelet deposition on collagen surface at arterial shear rate was observed ( Supplementary Figures 2L–N ). Hyperactivity of NLRP3 in platelets also did not modify either thrombosis prevalence or thrombus size in the model of venous thrombosis induced by partial blood flow restriction in the IVC ( Supplementary Figures 2O–Q ). The latter has been shown to depend on platelets, as platelet depletion prevents thrombosis in the model (22).

To assess whether overexpression of Nlrp3 in the MK lineage plays a role in the setting of inflammation, we used peritonitis induced by intraperitoneal injection of zymosan. Nlrp3 ^A350V/+/Gp1ba-Cre^KI/+ mice had a robust influx of CD45⁺ leukocytes consisting of Ly6G⁺ neutrophils into the peritoneum, whereas the numbers of other cell types, such as CD11c⁺ dendritic cells, SiglecF⁺ eosinophils, Ly6B⁺ inflammatory monocytes, and F4/80⁺ macrophages, remained unchanged ( Figure 2 ). No difference in immune cell numbers in spleen was observed ( Supplementary Figure 3 ). The levels of a set of proinflammatory cytokines, including IL-1β, MCP-1, IL-6, and IL-2, were also elevated in PLF ( Figures 3A–D ). These findings imply that MK/platelet NLRP3 is one of the mediators of acute inflammation and corroborates the observation that IL-1β-deficient mice have reduced inflammatory response in zymosan-induced peritonitis (23). Interestingly, the mutant mice had also increased levels of a potent anti-inflammatory cytokine, IL-10 ( Figure 3E ), which inhibits Nlrp3 inflammasome and synthesis of IL-1β (24). The levels of G-CSF, a major factor stimulating production of neutrophils, were also increased ( Figure 3F ). This is in line with the reported ability of IL-1β to stimulate G-CSF production (25), whereas G-CSF, in turn, can upregulate IL-1β secretion (26). In addition, G-CSF blocks erythropoiesis in the BM (27), which could potentially worsen anemia in the context of MK/platelet NLRP3 hyperactivation. Surprisingly, keratinocyte chemoattractant (KC, CXCL1), a potent chemoattractant for neutrophils, was not increased ( Figure 3G ). Thus, the observed neutrophil egression is mediated by other chemokines. For example, MCP-1 can promote neutrophil recruitment (28). However, KC levels could have already returned to control levels at 4 h as its peak at 2 h has been previously reported (29). Levels of MIP-1α, MIP-1β, TNF-α, MIP-2, and RANTES also remained unchanged ( Figures 3H–L ). Thus, hyperactivity of NLRP3 in MK and platelets results in a robust increase in the inflammatory response in the peritonitis setting.