Nlrp3 activation causes spontaneous inflammation and fibrosis that mimics human NASH

Abstract

The NOD-like receptor protein 3 (NLRP3) inflammasome is a central contributor to human acute and chronic liver disease, yet the molecular and cellular mechanisms by which its activation precipitates injury remain incompletely understood. Here, we present single cell transcriptomic profiling of livers from a global transgenic Tamoxifen-inducible constitutively-activated Nlrp3^A350V mutant mouse, and we investigate the changes in parenchymal and non-parenchymal liver cell gene expression that accompany inflammation and fibrosis. Our results demonstrate that NLRP3 activation causes chronic extramedullary myelopoiesis marked by myeloid progenitors that differentiate into proinflammatory neutrophils, monocytes, and monocyte-derived macrophages. We observed prominent neutrophil infiltrates with increased Ly6g^HI and Ly6g^INT cells exhibiting transcriptomic signatures of granulopoiesis typically found in the bone marrow. This was accompanied by a marked increase in Ly6c^HI monocytes differentiating into monocyte-derived macrophages that express transcriptional programs similar to macrophages of non-alcoholic steatohepatitis (NASH) models. NLRP3 activation also downregulated metabolic pathways in hepatocytes and shifted hepatic stellate cells towards an activated pro-fibrotic state based on expression of collagen and extracellular matrix (ECM) regulatory genes. These results define the single cell transcriptomes underlying hepatic inflammation and fibrosis precipitated by NLRP3 activation. Clinically, our data support the notion that NLRP3-induced mechanisms should be explored as therapeutic target in NASH-like inflammation.

Introduction

Non-alcoholic steatohepatitis (NASH) is a highly prevalent inflammatory and fibrotic liver disease that represents one of the most common reasons for cirrhosis and liver transplantation (1). Despite intense drug development efforts, there are still no FDA approved therapeutics for NASH (2). This is in part due to a lack of consensus regarding which cellular and molecular targets are most pathophysiologically important and actionable. Diverse cell types have been implicated in NASH pathogenesis using bulk methods and surface immunophenotyping; however recently, single cell transcriptomics has emerged to offer a more unbiased characterization of cell-specific transcriptomes in normal and diseased liver (3). Single cell biology is now poised to reveal new secrets about the cellular and molecular mechanisms of liver inflammation and fibrosis (4).

Studies of myeloid cells during inflammation and fibrosis in NASH models have revealed dynamic heterogeneity (5, 6). Kupffer cells, the resident macrophages of the healthy adult liver, decline in number and are replaced by bone marrow-derived monocytes that differentiate into new liver macrophages. Bone marrow-derived liver macrophages have been found to exacerbate damage as well as accelerate recovery after pathologic insult (7-9). In the fibrotic liver, single cell analysis also enabled identification and characterization of a “scar-associated” macrophage derived from circulating monocytes(10).

NLRP3 activation can serve as a genetic non-diet-induced model of liver inflammation and fibrosis (11). The NLRP3 inflammasome has been implicated in a wide range of inflammatory conditions, such as cardiovascular disease, chronic kidney disease, inflammatory bowel disease, and numerous autoimmune diseases (12-15). Upon activation, the NLRP3 protein binds to the apoptosis-associated speck-like protein containing a C-terminal caspase recruitment domain (ASC) adaptor protein and caspase-1 forming the NLRP3 inflammasome complex, permitting the conversion of pro-interleukin 1β (pro-IL1 β) and pro-interleukin 18 (pro-IL18) into their active forms and inducing cell death in a process termed pyroptosis (16). A subset of humans with the systemic auto-inflammatory disease, cryopyrin associated periodic syndrome (CAPS), carry the Muckle-Wells activating A352V mutation in the NLRP3 protein (17). An analogous mutation was engineered into a mouse model known as Nlrp3^A350V, and its inducible variants are now commonly used as models of NLRP3-induced hyperinflammation (17, 18). In mice, inducible activated Nlrp3^A350V causes liver fibrosis, pyroptosis, and immune infiltration (11, 19). Small molecule blockade of NLRP3 in mouse models results in reduction of the NASH phenotype (20). However, little is known about the diversity of cellular phenotypes within the NLRP3-induced inflamed and fibrosing liver.

Here, we analyze the inflamed and fibrosed livers of a mouse carrying the global tamoxifen-inducible activated Nlrp3^A350V mutated gene that was previously demonstrated to exhibit inflammation and fibrosis in the absence of direct liver injury (11). We define single cell transcriptional changes that arise in immune and non-immune cells during NLRP3 inflammasome-activated liver injury. We show that NLRP3 activation results in extramedullary myelopoiesis, replete with myeloid progenitors, proinflammatory monocyte-derived macrophages, and neutrophils, that lead to hepatocyte metabolic dysfunction and hepatic stellate cell-mediated fibrosis.

Materials and Methods

Mouse strains

As previously described, Nlrp3^A350V/+CreT knock-in mice (NLRP3-KI) with an alanine 350 to valine (A350V) substitution and an intronic floxed neomycin resistance cassette, in which expression of the mutation does not occur unless the Nlrp3 mutants are first bred with mice expressing Cre recombinase, were used for this study (21). NLRP3-KI mice were bred to B6.Cg-Tg (Cre/Esr1)5Amc/J mice (obtained from the Jackson Laboratory) to allow for mutant Nlrp3 expression in adult models after administration of tamoxifen (22).

Temporal induction of mutant Nlrp3 expression

NLRP3-KI mice and control (Nlrp3^A350V/−CreT) mice were injected intraperitoneally with 50 mg/kg tamoxifen-free base (MP Biomedicals) in 90% sunflower seed oil from Helianthus annus (MilliporeSigma) and 10% ethanol daily for 4 days starting at 8 weeks of age as previously described (23).

Liver sample preparation

NLRP3-KI and control mice were sacrificed 4 weeks after initiation of induction of Nlrp3 expression. Livers were either prepared for single cell, single nuclei isolation or fluorescence activated cell sorting (FACS) or representative pieces of harvested liver tissue were either (a) fixed in 10% formalin for 24 hours, (b) embedded in OCT on n-butane nitrogen and then frozen at −80°C, (c) placed in 0.5 mL RNAlater Solution (Invitrogen), or (d) snap-frozen in liquid nitrogen and stored at −80°C. If livers were not prepared for cell isolation, then blood samples (~0.2 mL) were obtained by cardiac puncture.

Liver histology and immunostaining

Livers were sliced in 5-μm sections and were stained for hematoxylin and eosin (H&E). H&E staining was used to score the grade of liver inflammation while liver fibrosis was assessed with picrosirius red (PSR) staining. For PSR staining, liver sections were incubated for 2 hours at room temperature with an aqueous solution of saturated picric acid containing 0.1% Direct Red (MilliporeSigma). To study liver cell death, terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay was performed using manufacturer’s instructions (ApopTag Peroxidase In Situ Apoptosis Detection Kit, MilliporeSigma). Immunohistochemistry (IHC) staining for myeloperoxidase (MPO) (1:200, Thermo Fisher Scientific, catalog RB-373) was performed on formalin-fixed, paraffin-embedded livers according to manufacturer’s instruction and counterstained with Mayer’s Hematoxylin solution (Sigma-Aldrich).

RNA-seq

Bulk RNA sequencing (RNA-seq) was performed by Q2 Solutions – EA Genomics. Briefly, total RNA was isolated from fresh frozen tissue using the Qiagen miRNeasy Mini Kit. All samples had >100 ng of input RNA and an RNA integrity number (RIN) value ≥ 7.0. Sequencing libraries were created using the IlluminaTruSeq Stranded mRNA method, which preferentially selects for messenger RNA by taking advantage of the polyadenylated tail. Libraries were sequenced using the Illumina sequencing-by-synthesis platform, with a sequencing protocol of 50 bp paired-end sequencing and total read depth of 30 M reads per sample.

RNA-seq data generated in this study were deposited in the Gene Expression Omnibus (http://www.ncbi.nlm.nih. gov/geo) with accession number GSE140742.

Tissue Processing

Liver cell isolation has been described elsewhere (24). Briefly, mice were anesthetized by ketamine/xylazine injection and perfused in situ through the inferior vena cava with sequential Pronase E (0.4 mg/mL, MilliporeSigma) and Collagenase D (0.8 mg/mL, MilliporeSigma) solutions. The entire liver was removed and digested in vitro with Collagenase D (0.5 mg/mL), Pronase E (0.5 mg/mL) and DNAse I (0.02 mg/mL, MilliporeSigma). After 20 minutes, tissue was filtered through a 70-μm mesh. The cell solution was concentrated in preparation for flow cytometry using centrifugation.

Alternatively, mice were intubated and ventilated with 2% isoflurane. After exposing the heart via thoracotomy, livers were perfused with 10mL of ice-cold phosphate-buffered saline (PBS) via cardiac puncture of the left ventricle. Perfused livers were excised and enzymatically digested in 250 mg aliquots for 1 hour at 37°C under continuous agitation (1200 rpm) in 450 U/ml collagenase I, 125 U/ml collagenase XI, 60 U/ml DNase I, and 60 U/ml hyaluronidase (Sigma), and filtered through a 40 μm nylon mesh in FACS buffer (PBS with 2.5% bovine serum albumin).

Bone marrow cells were collected by flushing femurs with ice-cold PBS. The resulting solution was filtered through a 40 μm nylon mesh and treated with red blood cell (RBC) lysis (BioLegend). Blood was collected by cardiac puncture. The cellular fraction was collected into EDTA-containing tubes (Sigma), and erythrocytes were eliminated using RBC lysis.

Nuclei Isolation

Mice livers were weighed and minced before flash freezing with liquid nitrogen. Minced samples were resuspended in 0.5 ml nuclei lysis buffer (Millipore Sigma, Nuclei EZ prep, NUC101), 0.2 U/μl RNAse inhibitor (stock 40U/μl, Enzymatics Y9240L) and homogenized with a 2 ml dounce grinder for 5 strokes with A motor and at least 20 strokes with B motor (Sigma-Aldrich D8938). The lysates were filtered through 100 μm and 50 μm cell strainers (CellTrics filters 04-004-2328, 04-004-2327) after resuspending with another 1 mL of nuclei lysis buffer and 10 minutes of incubation. Then they were centrifuged at 5 00 x g for 5 min at 4°C to pellet nuclei. The nuclear pellet was subsequently resuspended in 1.5 mL nuclei lysis buffer with an incubation of 10 minutes followed by centrifugation at 500 x g for 5 min at 4°C to pellet nuclei. The pellet was subsequently washed once in 0.5 mL of nuclei wash buffer (freshly 2% bovine serum albumin (BSA) in 1xPBS, 0.2 U/μl RNAse inhibitor and 1mM of EDTA) and incubated for 5 minutes without disturbing the pellet. After incubation, 1 mL of nuclei wash buffer and resuspension buffer were added and nuclei were resuspended. The washed nuclei were centrifuged at 500 x g for 5 min at 4°C and resuspended in 1.5 mL nuclei wash buffer. The isolated nuclei were centrifuged at 500 x g for 5 min at 4°C and respectively resuspended with Nuclei wash buffer stained with 10 μg/ml 4′,6-diamidino-2-phenylindole (DAPI) before FACS (5mg/ml, Invitrogen D1306). After sorting using purity mode, DAPI+ nuclei were pelleted at 1000 x g for 15 min at 4°C, resuspended in 2% BSA and trypan blue stained nuclei suspension were quality controlled and counted using hemocytometer (Hausser Scientific 3110V).

Flow Cytometry

For single cell/nuclei barcoding, cellular or nuclei suspensions were stained with DAPI. We enriched live, single cells/nuclei by sorting FSC-W^LO, DAPI^LO cells or DAPI^HI nuclei using a Sony MA900. For FACS analysis of leukocyte subsets, cellular suspensions were stained at 4°C in the dark in FACS buffer (PBS with 2.5% bovine serum albumin) with DAPI to exclude dead cells, Ter119 (BioLegend, clone TER-119) to remove unlysed red blood cells, and mouse hematopoietic lineage markers directed against B220 (BioLegend, clone RA3-6B2), Cd49b (BioLegend, clone DX5), Cd90.2 (BioLegend, clone 53-2.1), NK1.1 (BioLegend, clone PK136). Secondary staining of leukocyte subsets was performed using Cd11b (BioLegend, clone M1/70), Ly6g (BioLegend, clone 1A8), F4/80 (Biolegend, clone BM8), Ly6c (BioLegend, clone HK1.4 or BD Bioscience, clone AL-21), and Tim4 (Biolegend, clone RMT4-54).

FACS analysis of hematopoietic progenitors was performed as previously described (24952646). Briefly, cellular suspensions isolated from bone marrow were stained at 4°C in the dark in FACS buffer with lineage markers including phycoerythrin (PE) anti–mouse antibodies directed against directed against B220 (BioLegend, clone RA3–6B2), Cd11b (BioLegend, clone M1/70), Cd11c (BioLegend, clone N418), NK1.1 (BioLegend, clone PK136), TER–119 (BioLegend, clone TER–119), Gr–1 (BioLegend, clone RB6–8C5), Cd8a (BioLegend, clone 53–6.7), Cd4 (BioLegend, clone GK1.5) and Il7rα (BioLegend, clone A7R34). Then cells were stained with antibodies directed against Ckit (BioLegend, clone 2B8), Sca1 (BioLegend, clone D7), SLAM markers Cd48 (BioLegend, clone HM48–1) and Cd150 (BioLegend, clone TC15–12F12.2), Cd34 (BioLegend, clone RAM34), Cd16/32 (BioLegend, clone 2.4G2).

Singe cell and single nuclei sequencing

Single cell/nuclei RNA sequencing (scRNA-seq or snRNA-seq) was performed via microfluidic droplet-based encapsulation, barcoding, and library preparation (10X Genomics). Paired-end sequencing was performed on an Illumina NovaSeq instrument. Low level analysis, including demultiplexing, mapping to a reference transcriptome and eliminating redundant unique molecular identifiers (UMIs), was conducted with 10X CellRanger pipeline. All subsequent scRNA-seq analyses were conducted using the Seurat R package (v3.1) as detailed in the Extended Materials and Methods.

Statistics

Statistical analysis was performed using GraphPad Prism software (Version 9). All data are represented as mean values +/− standard deviation unless indicated otherwise. A statistical method was not used to predetermine sample size. For comparisons FACS populations, a 2-tailed t-test was used to determine statistical significance. For comparison of single cell derived scores, we implemented a Wilcoxon Rank Sums Test to determine statistical significance. P values less than 0.05 were considered significant and are indicated by asterisks as follows: *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001.

Results

NLRP3 activation leads to hepatic fibrosis and cell death

NLRP3-KI and control mice (~8 weeks of age) were injected with tamoxifen daily for 4 days and then sacrificed 4 weeks later after collecting livers, blood, and bone marrow (Figure 1A). As previously demonstrated, severe hepatic fibrosis and hepatic inflammation ensues with a global inducible mutation of Nlrp3 (19, 25). IHC staining with H&E shows a large immune cell response and liver architectural distortions in the NLRP3-KI model compared to control (Figure 1B). PSR staining shows elevated fibrosis throughout the liver in the NLRP3-KI liver (1.8% of area) compared to control (0.3% of area, p = <.05) (Figure 1C). TUNEL staining shows a high degree of apoptosis in the NLRP3-KI liver (310 ± 170 apoptotic bodies/field of view (FOV)) compared to the control liver (64 ± 29 apoptotic bodies/FOV, p = < .0001) (Figure 1D).

Figure 1: Global activation of NLRP3 inflammasome leads to hepatic inflammation, fibrosis, and cell death.

(A) Experimental design. (B-D) H&E, PSR, and TUNEL stainings of mouse liver sections, respectively. (E) Volcano plot of up- and downregulated genes in livers of NLRP3-KI compared to control mice. (F) GO analysis of up- (green) and downregulated (purple) genes.

Bulk RNA-seq implicates innate immune system activation

To explore the mechanisms underlying NLRP3 inflammatory progression, we performed RNA-seq of the livers from NLRP3-KI and control mice. We detected 975 and 216 genes that exhibited more than 2-fold increase or decrease in mRNA expression, respectively (Figure 1E and Table S1). Gene ontology (GO) analysis revealed increased expression of genes associated with several innate immune pathways including chemokine and cytokine signaling, transendothelial leukocyte migration, tumor necrosis factor (TNF) signaling, nuclear factor Kappa B (NF-κB) signaling, phagocytosis, and complement cascades (Figure 1F, green). Meanwhile, genes associated with drug metabolism related to cytochrome P450, glutathione metabolism, amino acid metabolism, steroid synthesis, fatty acid degradation and retinol metabolism were downregulated.

Single cell RNA-Seq of the NLRP3-activated liver

To define the specific cell types contributing to NLRP3-mediated inflammation and its sequelae, we performed RNA sequencing on FACS sorted whole single cells and single nuclei collected from the livers of NLRP3-KI and control mice (Figure 2A). We integrated the resulting counts matrices to produce a harmonized object composed of control and NLRP3-KI samples (Figure 2B). This procedure facilitated downstream analyses by ensuring that comparisons were drawn between like cell types. Across two independent experiments, we obtained 45,862 transcriptomes, 25,537 of which were single cell transcriptomes (5,211 Control; 19,325 NLRP3-KI; n = 2) and 20,119 that were single nuclei transcriptomes (12,941 control; 7,178 NLRP3-KI; n = 2). The single cell data was composed predominantly of myeloid cells including neutrophils, monocytes, monocyte-derived macrophages (MdMs), and Kupffer cells (KCs). We also captured lymphocytes (B cells, T cells, NK cells), fibroblasts, and endothelial cells to a lesser extent. Clusters were annotated based on the expression of canonical markers revealed by unbiased clustering and DEG analysis (Figure 2C, D). Within the single nuclei data, we similarly observed clusters expressing marker genes associated with endothelial cells and immune cells, however the bulk of the data was represented by hepatocytes, hepatic stellate cells (HSCs), and cholangiocytes (Figure 2E, F). By quantifying cell cluster composition normalized by total cells per sample, we observed a marked increase in immune cells in the NLRP3-KI compared to control mice (Figure 2G, H). Since previous reports identified phenotypic heterogeneity amongst macrophages in NASH models, we first focused our attention on monocytes and macrophages(6, 26, 27).

Figure 2: Single cell and nuclei transcriptomics reveals immune, parenchymal, and nonparenchymal compartments in the liver of NLRP3-activated mice.

(A) Flow cytometry sorting panel for isolation of whole cell (top) and nuclei (bottom) for single cell/nuclei barcoding. (B) Integration (batch correction) strategy to classify similar cell types captured in control and NLRP3-activated mice. (C,E) UMAP plots of integrated control and NLRP3-activated cells from single cell (C) (5,211 Control; 19,325 NLRP3-KI; n = 2) and single nuclei samples (E) 12,941 control; 7,178 NLRP3-KI; n = 2). (D,F) Dot plots displaying top differentially expressed genes in single cell (D) and single nuclei (F) samples. Data shown is average expression of scaled gene expression by cluster. (G) Relative abundance of each cell type, as defined in (C-F), as a percentage of total single cell/nuclei transcriptomes captured. (H) Fraction of cluster belonging to control and NLRP3-activated (KI) mice. (Neu, neutrophil; ECs, endothelial cells; Mega, megakaryocytes; Hep, hepatocytes; CV, central vein; PN; peripheral node; HSCs, hepatic stellate cells; Chol, cholangiocytes)

NLRP3 activation shifts liver macrophage diversity

In mice, multipotent hematopoietic cells (Lineage⁻Sca1⁺Ckit⁺, LSK) give rise to common myeloid progenitors (CMPs, LS⁻K Cd16/32⁺Cd34⁺) which differentiate into myeloid restricted granulocyte monocyte progenitors (GMPs, LS⁻K Cd16/32⁻Cd34⁺) that mature into monocytes or neutrophils. KCs, the liver’s resident macrophage, derive from monocytes that seed the embryonal liver and are later able to self-replenish. They are phenotypically distinct and have the capacity to express proinflammatory and anti-inflammatory transcriptional programs (28, 29). The balance of proinflammatory KCs and anti-inflammatory KCs helps regulate liver inflammation (30). Recently, several studies have shown that during acute liver injury and NASH, macrophages derived from bone-marrow derived monocytes are critical mediators of pathogenesis (7, 9, 31, 32). These MdMs have been categorized into several phenotypic groups including proinflammatory, wound-healing, and immunosuppressive (33).

To examine NLRP3-induced heterogeneity in macrophages, we bioinformatically isolated and reclustered monocytes (as well as closely related progenitor cells), MdMs and KCs as defined during coarse clustering (Figure 3A, B; 2D). These cells spontaneously arranged into 11 clusters which, based on DEG analysis, we annotated as follows: monocyte progenitors (MPs, Ngp^HI, Camp^HI), preMono (Stmn1^HITop2a^HI), Ly6C^HI monocytes (Ly6c2^HICcr2^HI), interferon-stimulated gene (ISG) expressing monocytes (ISG^HI, Cxcl10^HIIsg15^HI), MdMs (Ccr2^HIH2-Aa^HI), lipid-associated macrophages (LAMs, Trem2^HISpp1^HICd9^HI), Cx3cr1-expressing macrophage (c-LAMs, Cx3cr1^HICcr2^HI), KCs (Timd4^HIClec4f^HI), angiotensin-converting enzyme (Ace) expressing cells (Ace^HIEar2^HI), dendr(34)itic cells (DCs, Cd209a^HI) and Ly6C^LO cells (Irf8^HIH2-Eb1^HI) (Figure 3B, D). The control mice, as expected, were predominately composed of KCs. In contrast, we observed a dramatic shift to monocytes, preMono and to a lesser extent MdMs and LAMs in NLRP3-KI mice (Figure 3C). Recent reports have implicated KC loss and gain of monocyte derived macrophages as a critical mediator of NASH and NAFLD (6, 35). Xiong et al characterized the emergence of a triggering receptor expressed on myeloid cells 2+ (Trem2+) NASH-associated macrophage (26) while Remmerie et al (27) described an osteopontin-expressing (Spp1) MdM with transcriptional similarity to the lipid-associated macrophages (LAM) found in adipose tissue (35) and scar-associated macrophages found in hepatic fibrosis (10). To explore if NLRP3 induction precipitates a similar monocyte-derived macrophage transcriptional phenotype, we performed in silico comparisons by creating an integrated single cell dataset derived from cells of diet-induced NASH models and NLRP3-KI/Control (Figure S1A). We quantified the relative abundance of monocyte, MdMs and KC subpopulations at control and pathological conditions (Figure S1B-D). While comparisons of the relative abundance of each subset are obscured by differences in cell isolation techniques and purification/exclusion strategies (for example, negative selection for polymorphonuclear cells in Xiong et. al.), we observe directionally similar induction of monocyte (Ly6c^HI, ISG^HI, H2-Aa^HI, Ly6c^LO) and macrophage subsets (LAMs, cLAMs) (Figure S1D). Likewise, the ‘LAM signature’ was significantly elevated in monocytes after NLRP3-induction as in NASH models (Figure S1E). These data show that NLRP3-activation is sufficient to induce monocyte-derived macrophage differentiation observed in murine NASH models and implicated in human cirrhosis.

Figure 3: NLRP3 inflammasome activation shifts liver macrophage diversity.

(A) UMAP highlighting monocytes, macrophages, and Kupffer cells and defining genes amongst all captured single cell transcriptomes. (B) UMAP plots of subset and reclustered monocytes split by condition. (C) Relative abundance of each subset normalized to control sample. (D) Dot plot of DEGs showing monocyte differentiation and specialization. (E) FACS analysis of monocytes, macrophages and KCs from livers of control and NLRP3-KI mice. (F) FACS analysis of myeloid progenitors from blood of control and NLRP3-KI mice. (G,H,I) Quantification of FACS populations (grey, n = 2) and KI mice (blue, n=3). * P value < .05, ** P value < .01, Mann Whitney Rank Sums test.

Given the continuum of monocyte-like clusters and the shift to progenitor, we hypothesized that NLRP3-activation induces extramedullary monopoiesis. To test this, we integrated single cell transcriptomes of myeloid cells isolated from two bone marrow (BM) samples (Cd11b+ and LS⁻K enriched) and blood (BLD, Cd11b⁺ enriched) of steady-state mice with the present liver transcriptomes (Figure S2A). As expected Ly6c^HI monocytes were abundant in the blood. UMAP analysis showed that MPs and preMonos states were enriched in LS-K sorted BM samples as well as NLRP3-KI liver samples, but not BLD, confirming our annotation of these cells as progenitors (Figure S2B, C). Further, trajectory analysis revealed similar gene modules in BM and NLRP3-KI derived cells from progenitor state (those present in solely BM samples) to mature states (those abundant in BLD) (Figure S2D). We validated several of these observations by FACS analysis of progenitors, monocyte, macrophages across multiple tissue compartments (Figure 3E, F). The quantity of F4/80^LOCd11b^HILy6c^HI cells and LS⁻K cells increased significantly in NLRP3-activated livers (Figure 3G, H). The quantity of hematopoietic cells (LSKs) was elevated in NLRP3-KI, though not to a statistically significant degree (p = .0678). To understand the source of progenitor cells, we analyzed leukocytes isolated from peripheral BM and BLD (Figure 3F). FACS analysis of BM progenitors a numerical increase in GMP production of NLRP3-KI mice relative to control (Figure S3). We did not observe changes in LSK quantities in the BM (Figure S3C) but did find numerical increases in the blood (p = .0558, Figure 3I). Taken together, we concluded that NLRP3-activation causes myeloid progenitors to infiltrate the liver and differentiate into monocyte-derived macrophages that transcriptionally phenocopy those of NASH models.

NLRP3 activation causes extramedullary granulopoiesis

We next focused our attention on neutrophils because of their previous association with NLRP3-induced liver injury(23), and because in livers of NLRP3-KI mice, we observed a 3-fold increase in relative abundance of neutrophils as well as increased GMP progenitors from which they derive. Acute liver inflammation typically involves the rapid recruitment of neutrophils, yet the specific functions that neutrophils play during liver injury and the signaling mechanisms for neutrophil recruitment are incompletely understood (36). One study found that liver fibrosis was neither worsened nor ameliorated by diminishing neutrophil recruitment (37), yet other studies have associated increased neutrophil activity with severity of hepatic disease (38). To better understand NLRP3-induced neutrophilia, we bioinformatically isolated and reclustered neutrophil transcriptomes from the Retnlg^HICxcr2^HIIl1β^HICsf3r^HI cluster and closely related myeloid progenitors (Figure 4A, 2D). Similar to monocytes, this resulted in a continuum of 5 neutrophil subtypes, which based on DEGs and subsequent analyses, we annotated as myeloid progenitors (MPs, Lgals1^HIPtma^HI), preNeu (Elane^HIFcnb^HI), intermediate neutrophils (Int, Camp^HIMmp8^HI), mature neutrophils (Mat, Il1b^HICcl6^HI), and ISG-expressing neutrophils (ISG^HI, Cxcl10^HI, Gbp5^HI) (Figure 4B, D). Given that monocytes and neutrophils share a common progenitor, we hypothesized that NLRP3 induction prompts extramedullary granulopoiesis, replete with progenitors and their progeny. Indeed, when PreNeu and Int neutrophil transcriptomes from NLRP3-KI were compared to integrated single cell transcriptomes from BM and blood of WT mice, analogous cells were only found in the BM compartment where progenitors reside (Figure S2B, C). Meanwhile, the relative abundance of PreNeu, Int, Mature and ISG^HI neutrophils was increased in NLRP3-KI compared to control Figure 4C). Trajectory analysis revealed a continuum of states with gene modules that mirrored granulopoiesis in the BM (Figure 4E, S2D). Furthermore, we found a strong quantitative similarity between BM- and BLD-specific transcriptomes and the continuum within the liver (Figure 4F). These observations were confirmed by flow cytometric analysis (Figure 4G). We found a significant increase in Cd11B^HI liver leukocytes as well as Cd11b^HILy6g^HI and Ly6g^INT neutrophils (Figure 4G,H). MPO staining confirmed a large infiltrate of neutrophils in the NLRP3-KI liver (480 ± 320 cells/FOV) compared to control (2 ± 2 cells/FOV, p = 0.0465) (Figure 4I, J).

Figure 4: NLRP3 inflammasome activation induces emergency granulopoiesis.

(A) UMAP plots highlighting MPs and neutrophils amongst all cells . (B) UMAP plot of subsetted and reclustered neutrophils split by condition. (C) Relative abundance of MP1, PreNeu, Intermediate, Mature, and ISG^HI neutrophils in control and NLRP3 induced mice, normalized to control sample. (D) Dot plot of DEGs demonstrating continuum of maturation. (E) Trajectory analysis showing pseudotime as a function of UMAP space. (F) Spearman rank coefficients (row normalized) comparing single cell transcriptomes of neutrophils shown in (B) to neutrophils isolated from bone marrow and peripheral blood of WT mice. (G) FACS analysis of myeloid cells and progenitors of livers and blood isolated from control and KI mice (n=3). (H) Quantification of Cd11b^HI, Ly6g^HI and Ly6g^INT cells by FACS analysis. (I) Representative image from KI mouse showing MPO staining. (J) Quantification of MPO staining (manually counted MPO+ cells per 20X FOV). (K) Neutrophil-specific NLRP3-induced genes as determined by Wilcoxon Rank Sums test. (L) Violin plots of persistent immaturity markers, amplified maturity markers, or entirely up-regulated genes in KI versus control.. * P value < .05, ** P value < .01, Mann Whitney Rank Sums test. (Blood, BLD; bone marrow, BM; Wild-type, WT. Myeloperoxidase, MPO.)

Having established that NLRP3 activation results in chronic myelopoiesis, we next asked if its induction perturbed the transcriptional state of mature neutrophils. We compared the transcriptomes of cells belonging to mature neutrophils of control and NLRP3-KI mice. Comparisons between other states were restricted due to the lack of immature cells in the control sample. We found up-regulation of roughly 50 genes including those which have been associated with hepatic inflammation and fibrosis (Figure 4K). Chitinase-like protein 3 (Chil3 or CHI3L1 in humans) has been shown to be expressed in hepatic macrophages of non-alcoholic fatty liver disease (NAFLD) patients (39) with serum levels of YKL-40 elevated in patients with hepatic fibrosis (40). Further, we categorized up-regulated genes into three groups: persistent, amplified, and elevated (Figure 4L). Persistent genes, such as Chil3, chitinase-like protein 1 (Chil1), neutrophilic granule protein (Ngp), and placenta associated 8 (Plac8) are gene markers for immature and intermediate neutrophils which typically turn off before reaching maturity but were found to be highly upregulated in the mature neutrophils in the NLRP3-KI sample. Amplified genes, such as resistin-like gamma (Retnlg), WAP four-disulfide core domain 17 and 21 (Wfdc17, Wfdc21, respectively), are typical gene markers for mature neutrophils, however, were expressed to a greater degree upon NLRP3 activation. Lipocalin-2 (Lcn2) was also found to be highly upregulated in the NLRP3-KI model, which has been shown in NASH models to be an important mediator of interactions between neutrophils and hepatic macrophages via CXC chemokine receptor 2 (CXCR2) signaling (41). Elevated genes, such as serum amyloid A1 and A3 (Saa1, Saa3, respectively), were notably upregulated compared to control neutrophils. Serum amyloid A (SAA) is an acute phase reactant that can activate the NLRP3 inflammasome cascade, induces synthesis of cytokines, and is a neutrophil chemotactic agent (42). We observed similar patterns in pseudotime analysis (Figure S2E). It is likely that NLRP3 activation has pluripotent effects on granulopoiesis, immune signaling, and neutrophil granule function.

Taken together, this data demonstrates that NLRP3 activation results in hepatic extramedullary granulopoiesis that begins with bone marrow release of myeloid progenitors, which seed the liver, replicate, and differentiate into mature neutrophils with heightened proinflammatory functions.

Hepatocytes change metabolic function

We next turned to hepatocytes to better understand the effects of NLRP3-driven pathology on the liver itself. Hepatocytes perform myriad functions and have been shown to perform these functions based on spatial location within the liver, such as cholesterol synthesis and urea cycle processing in the highly oxygenated periportal region and bile synthesis and drug metabolism in the chronic hypoxic pericentral region (43). Although liver injury commonly induces regenerative programs, in NASH, hepatocyte-enriched genes are generally downregulated without substantial commensurate proliferation (26, 44). Similar to the analysis above, we began with subsetting and reclustering of hepatocyte transcriptomes (Figure 5A, B). Subsets anchored to hepatocyte zonation markers as previously described (Figure S4A) (43). We performed DEG analysis and found that NLRP3 activation down-regulated ~2,000 genes while only up-regulating 116 genes, a pattern similar to previous reports in NASH livers (Figure 5C) (26). NLRP3-induced hepatocytes displayed dysfunctional metabolism as evidenced by the decrease in expression of genes associated with retinol metabolism, steroid hormone metabolism, glyoxylate metabolism as well as several other metabolic pathways (Figure 5D). We found little to no evidence of zonation-specific regulation (Figure S4). A satellite population emerged in the NLRP3-induced sample expressing genes associated with microtubule, tubulin, and cytoskeletal binding, an indication of cellular division and liver regeneration (cluster 5, Figure 5E). To test if NASH and NLRP3-induced hepatocytes converged on a similar transcriptional phenotype, we analyzed upregulated and downregulated genes in both models and found no correlation between NLRP3-induced genes and WD (Figure S5). This data demonstrates hepatocytes are changing functional priorities from metabolic pathways to cellular regeneration as compensation for apoptosis.

Figure 5: Dysfunctional liver metabolism in the context of chronic NLRP3 inflammasome induction.

(A) UMAP plot highlighting hepatocytes (snRNA-seq) defined in Figure 2E, F. (B) UMAP plots of subset and reclustered hepatocytes, split by condition. Clusters shown represent hepatocyte zonation from central vein to peripheral vein (see figure 3S for more detail). (C) Volcano plot of up- and downregulated genes in hepatocytes of NLRP3-KI compared to control mice. (D) GSEA of downregulated genes. (E) Relative abundance of cluster 7 in control and KI mice (top) and GSEA of cluster defining genes.

Hepatic stellate cell activation

HSCs have a wide array of functions in the normal liver such as vitamin A storage, vasoregulation, and ECM homeostasis. In response to injury, HSCs specialize into fibrogenic alpha smooth muscle actin (ɑ-SMA)-expressing myofibroblasts (45, 46). This activation is regulated by a complex interplay between HSCs and most cell types found in the liver, and these signaling pathways have been targeted for anti-fibrotic therapies (47). Subsetting and reclustering of HSCs resulted in two broad groups: a large group, which expressed high levels of prototypical HSC markers (Ntm, Ank3, Nrxm1) and subdivided into central vein (CV) and peripheral vein (PV) -associated HSCs based on their relative expression of Il34 and Adamtsl2, per previous reports. (Figure 6A, B, C). The second group expressed high levels of genes associated with ECM organization (Gas6, Iga8, Sulf, Smoc2, Eln) and collagen deposition (Col1a1, Col1a2, Col3a1, Col5a2) (Figure 6C, E). To determine if this cluster represented portal fibroblast or activated HSCs, we plotted prototypical HSC and fibroblast markers. We did not find significant expression of Gsn, Cd34, Clec3b or other fibroblast markers. Therefore, we conclude that the population is most consistent with activated HSCs (Figure 6D). While some activated HSCs were found in control mice, we observed a nearly two-fold increase in activated HSC numbers, from 6% to 10.5%, upon NLRP3-activation (Figure 6F). As an alternative to cluster membership, we constructed a collagen formation score (see methods). The collagen formation score was significantly higher in HSCs from NLRP3-KI mice compared to control mice, both in quiescent and activated states (Figure 6G,H). This was supported by the large amount of ɑ-SMA seen in NLRP3-induced livers (Figure 6I, J). To better understand immune involvement, we performed cell-to-cell communication inference analysis between myeloid and HSC subsets as defined above (Figure S6). This resulted in 43 significantly enriched receptor-ligand or cell-to-cell based communication networks, most of which were due to intra-myeloid signaling (Cxcl-, Il1-, Bst2- signaling networks, for example). As expected, activated HSCs were enriched for collagen and laminin signaling. Amongst the pathways involving directional signaling from immune cells to HSCs, we found enrichment of Visfatin (a pluripotent adipocytokine with intracellular and extracellular function, also known as nicotinamide phosphoribosyl transferase, NAMPT) signaling from ISG^HI neutrophils and monocytes, Spp1 signaling from LAMs, as well as complement and APP (amyloid precursors protein) signaling that was indiscriminately driven by most immune populations. These data reveal putative signaling cascades that may underly inflammation-driven fibrosis.

Figure 6: NLRP3 induction activates HSCs and promotes collagen deposition.

(A) UMAP plots highlighting HSCs for subsequent analyses (snRNA-seq) defined in Figure 2E, F. (B) UMAP plots of subset and reclustered HSCs, split by condition. (C) Heatmap of DEGs, Wilcoxon Rank Sums Test. (D). Violin plots of of prototypical HSC and fibroblast markers. (E) GSEA of cluster 2 defining genes. (F) Relative abundance of activated HSCs (cluster 2). (G) Feature plot showing Collagen Formation Score embedded onto UMAP plots, split by condition. (H) Violin plots of Collagen Formation Score split by cluster (left) and condition (right). (I,J) Representative images of ɑ-SMA staining from control and KI livers. **** P value < .0001, Wilcoxon Rank Sums Test (G) or unpaired t test (I).

Discussion

Using scRNA-seq, we show that constitutive activation of the NLRP3 inflammasome precipitates chronic extramedullary myelopoiesis, hepatocyte dysfunction, and profibrotic HSC activation in normal liver. This complements a recent report showing that selective inhibition of NLRP3 can reduce liver inflammation and fibrosis in obese diabetic mice (20). Here, we provide further evidence that NLRP3 activation is sufficient to cause NASH-like inflammation and fibrosis even in the absence of steatosis.

scRNA-seq and FACS analyses implicate bone marrow-derived neutrophils and monocytes as key drivers of NLRP3-induced liver pathology. Neutrophils play prominent roles in injury resulting from drug-induced liver injury, physicochemical injury, the response to cell death, and bacterial infection (48-51). They release elastase, MPO, reactive oxygen species (ROS), and secrete cytokines and chemokines, all which fuel tissue injury, promote inflammation, and recruit additional inflammatory cells (52). Interestingly, in some contexts, neutrophils are reported to play protective roles in resolution of inflammation and in liver repair (53). In the context of NASH, neutrophils have been relatively understudied compared to monocytes and macrophages, in part because neutrophil-preserving assays can be technically challenging. In contrast, monocytes, MdMs and KCs are more durable longer-lived cells and are consequently more amenable to in vitro experimentation. MdMs of NLRP3-activated mice share a remarkably similar transcriptional fingerprint to that of mice fed WD. Future studies are needed to define the relationship of WD and NLRP3 activation.

Our data provide a number of potential cell-specific driver genes of NLRP3-mediated liver fibrosis. Neutrophils in our NLRP3-activated mice exhibited elevated expression of Lcn2, which was recently implicated as a mediator of neutrophil-mediated liver inflammation and macrophage cross-talk in diet-induced NASH (41). Both NLRP3-activated neutrophils and MdMs highly expressed Chil3, similar to the human gene CHI3L1, which leads to production of chitinase-like protein YKL-40, a biomarker of hepatic fibrosis severity (39, 40). By perturbing chitinase-like proteins in neutrophils and MdMs, future studies can investigate cell-specific contributions and test whether they are essential for NLRP3-mediated liver fibrosis.

Our data raise many questions that will provide inspiration for future experiments. What molecular mechanisms regulate mobilization of myeloid progenitors from the bone marrow and why do they traffic to the liver? Does this occur in all tissues, or is spontaneous NLRP3-dependent inflammation and fibrosis unique to the liver? Does mobilization of bone marrow progenitors depend on NLRP3-dependent modulation of established retention factors such as Cxcl12/Cxcr4 or are other NLRP3-dependent mechanisms involved? What are the relative contributions of local versus systemic sources of NLRP3-activated myeloid cells? The observed increase in intrahepatic progenitors and neutrophils could result from enhanced production and mobilization from hematopoietic reservoirs or from local production by intrahepatic progenitors. Future studies using scRNA-Seq and flow cytometry from blood and bone marrow as well as chimeric models using bone marrow transplantation, adoptive transfer, or parabiosis could shed additional light on this question. Within the liver, how do NLRP3-activated myeloid cells cause HSC activation and hepatocyte dysfunction? Is NLRP3 essential for this pathogenesis or can other activating mutations drive myeloid infiltration to drive NASH-like inflammation and fibrosis in normal livers? What is clear is that single cell and nuclei transcriptomics will provide an information-rich lens through which to answer these and other mechanistic questions about liver pathobiology.

Conclusion

Using single cell transcriptomics, we defined and characterized the immune cell subsets within the inflamed liver microenvironment induced by NLRP3 inflammasome activation as well as changes in metabolic status in hepatocytes and HSC phenotypes. These results have important implications for dissecting the mechanisms of liver injury in NLRP3-driven pathologies.