Type I interferon signaling facilitates resolution of acute liver injury by priming macrophage polarization

Abstract

Due to their broad functional plasticity, myeloid cells contribute to both liver injury and recovery during acetaminophen overdose-induced acute liver injury (APAP-ALI). A comprehensive understanding of cellular diversity and intercellular crosstalk is essential to elucidate the mechanisms and to develop therapeutic strategies for APAP-ALI treatment. Here, we identified the function of IFN-I in the myeloid compartment during APAP-ALI. Utilizing single-cell RNA sequencing, we characterized the cellular atlas and dynamic progression of liver CD11b+ cells post APAP-ALI in WT and STAT2 T403A mice, which was further validated by immunofluorescence staining, bulk RNA-seq, and functional experiments in vitro and in vivo. We identified IFN-I-dependent transcriptional programs in a three-way communication pathway that involved IFN-I synthesis in intermediate restorative macrophages, leading to CSF-1 production in aging neutrophils that ultimately enabled Trem2+ restorative macrophage maturation, contributing to efficient liver repair. Overall, we uncovered the heterogeneity of hepatic myeloid cells in APAP-ALI at single-cell resolution and the therapeutic potential of IFN-I in the treatment of APAP-ALI.

Introduction

Acetaminophen (APAP) toxicity is one of the most commonly observed causes of acute liver injury (APAP-ALI), frequently impacting multiple organ systems with 30% mortality [1]. APAP overdose results in centrilobular hepatocyte necrosis and liver failure [2]. Treatment with N-acetylcysteine (NAC) can replenish GSH stores and prevent hepatic necrosis but must be given within 8 h post APAP administration [3].

Excessive APAP initiates sterile inflammation in the liver and is followed by severe systemic bacterial infection in approximately 35–40% of ALF patients [4]. Damage-associated molecular patterns (DAMPs) released by necrotic hepatocytes and pathogen-associated molecular patterns (PAMPs) translocated through portal circulation to the liver trigger acute inflammatory responses via pattern recognition receptors (PRRs), including Toll-like receptors (TLRs). Liver-resident Kupffer cells and infiltrating myeloid cells respond rapidly to PAMPs and DAMPs to manipulate hepatic inflammatory status [4] via the production of inflammatory cytokines, including type I interferons (IFN-I). IFN-I plays essential roles in the host response to viruses and other microbes, as well as in the resolution of myeloid-mediated inflammation [5, 6]. Hence, we propose that IFN-I might manipulate macrophage function to provoke tissue repair during APAP-ALI.

Myeloid cell infiltration is central to the pathogenesis of ALI, leading to necrosis of centrilobular areas [7]. Neutrophils are the earliest cell population to appear post injury, followed by monocytes, which develop into monocyte-derived macrophages (Mo-MFs). These two cell populations execute diverse functions during liver inflammation at both the initiation and resolution stages and exhibit numerous activation states regulated by various cytokines and chemokines [8, 9]. CCR2+Ly6C^hi monocytes are recognized to contribute to early damage, while restorative Mo-MFs expressing MerTK and Trem2 are thought to orchestrate the clearance of DAMPs or PAMPs by phagocytosis [10, 11]. The specific myeloid cell populations and their potential roles during APAP-ALI progression require further investigation.

In the current study, our initial observations using mice with genetic deficiencies revealed the importance of TLR/TRIF/IRF3-dependent IFN-I production and subsequent IFN-I signaling, including the recently identified phosphorylation of STAT2 at T403, in early tissue repair following APAP-ALI. Using single-cell RNA sequencing (scRNA-seq), we identified multiple subsets of liver nonparenchymal CD11b+ myeloid cells post APAP challenge and characterized their interrelationships and reciprocal cytokine-mediated dialog between different cell populations. IFN-I-producing Mo-MFs help to drive the maturation of restorative macrophages in part by promoting the production of CSF-1 by inflammatory neutrophils, which subsequently undergo apoptosis. CSF-1 directly promotes the acquisition of Mo-MF restorative properties. Using mice with a mutation disabling STAT2 T403 phosphorylation (T403A) [12], we demonstrated the role of IFN-I in driving these events, and its deficiency resulted in delayed recovery and increased mortality. Together, our results provided a comprehensive transcriptomic overview of liver myeloid cells in the early recovery phase post APAP-ALI and pointed to the essential roles of IFN-I in orchestrating the liver repair process, suggesting the potential therapeutic usage of IFN-I for APAP-ALI treatment.

Results

IFN-I production and signaling deficiency delayed tissue repair upon ALI

Using a well-established experimental recoverable APAP model (Fig. 1a), the resolution of hepatic damage was found to be impeded in mice deficient in TRIF (TRIF KO) and IFNAR1 (IFNAR KO) 48 h post APAP challenge, as evidenced by elevated liver necrotic area (Fig. 1b, c) and plasma levels of the hepatic enzyme ALT (Fig. 1d). This was also observed in mice deficient in IRF3, a critical signaling component downstream of TRIF in IFN-I production (Fig. 1e, f). To more precisely assess IFN-I signaling events, we further evaluated mice deficient in STAT2 T403 phosphorylation, which is dominant for IFN-I signaling activation (STAT2 T403A, MUT) [12]. At the peak injury stage (24 h), WT and MUT mice showed massive and comparable liver damage, while at the early tissue repair stage (48 h), MUT mice retained marked liver pathology with wide necrotic areas and higher ALT levels (Fig. 1g–i). Consistently, the overall survival of MUT mice after a half-lethal dose of APAP (450 mg/kg) was significantly lower (14%) than that of WT mice (50%) (Fig. 1j). Greater numbers of mouse deaths were also observed in IFNAR KO and TRIF KO mice (Fig. 1j). These data suggested that deficiency in IFN-I signaling or production contributed to reduced tissue repair rather than initial liver damage. To determine in which cell type(s) IFN-I signaling plays indispensable roles in regulating recovery, bone marrow chimeric mice were generated (Fig. 1k–m). Remarkably, MUT mice engrafted with WT bone marrow (WT to MUT) displayed better liver recovery than MUT control mice (MUT to MUT), whereas WT mice bearing MUT bone marrow (MUT to WT) showed less tissue repair than WT control mice (WT to WT). Consistently, liver recovery from APAP-ALI was also attenuated in IFNAR^flox/floxLysM^Cre mice, in which a myeloid-specific deletion of IFNAR was created (Fig. 1n–p). These data suggest that the expression of and response to IFN-I within distinct myeloid subsets are critical components of the liver repair process.

Fig. 1.

IFN-I production and activation are critical for the resolution of APAP-ALI. a Study design of recoverable APAP-ALI (300 mg/kg, i.p.). b Representative liver histological staining from WT, TRIF KO and IFNAR KO mice (n = 9 for each genotype) 48 h after APAP injection (i.p., 300 mg/kg in saline), necrosis area quantification (c) and plasma ALT levels (d) are shown. One-way ANOVA. e Representative liver histological staining of WT and IRF3 KO (n = 9 for each genotype) and f necrosis area quantification. Two-sided t-test. g Representative liver histological staining from WT (n = 45) and T403A (MUT) (n = 45) mice post APAP injection (i.p., 300 mg/kg) at the indicated time, necrosis area quantification (h) and plasma ALT levels (i) are shown. Two-way ANOVA. j Survival rates of WT (n = 8), MUT (n = 8), TRIF KO (n = 18) and IFNAR KO (n = 16) mice post-APAP injection at a half-lethal dose (i.p., 450 mg/kg). k–m Recoverable APAP-ALI experiment of bone marrow chimeric mice. APAP-ALI of recipient mice (n = 5 for each genotype) was assessed 48 h post APAP by representative liver histological staining (k), necrosis quantification (l) and plasma ALT levels (m). WT to MUT means donor bone marrow was from WT mice and recipient mice were MUT. So do other labels in (k). One-way ANOVA. n Representative liver histological staining of WT and IFNAR^flox/floxLysM^Cre (IFNAR cKO) (n = 8 for each genotype) 48 h post APAP injection (i.p., 300 mg/kg), necrosis area quantification (o) and plasma ALT levels (p) are shown. Two-sided t-test. Scale bar, 100 μm. *p < 0.05

scRNA-seq defined liver CD11b+ nonparenchymal cell diversity during APAP-ALI

Next, the liver CD11b+ nonparenchymal cells were enriched and processed through CellMicroArray (CMA) for the observation of cell morphology (Supplementary Fig. S1a). In both WT and MUT samples, the cell morphologies varied at 0, 24 and 48 h after APAP-ALI, with abundant granulocytes appearing at 24 h and transitioning to monocytic cells at 48 h. To explore the roles of IFN-I signaling in myeloid cells during APAP-ALI in detail, we generated scRNA-seq profiles of liver CD11b+ nonparenchymal cells from WT and MUT mice at 0, 24 and 48 h post APAP-ALI (Fig. 2a, 31,408 cells, CNP0001813). After rigorous quality control, 26,758 cells were retained (Supplementary Fig. S1b) and assigned to 15 distinct cell clusters by unbiased, graph-based clustering (UMAP) (Fig. 2b, Supplementary Fig. S1c and Supplementary Table S1), of which the classification accuracy was confirmed by marker gene expression (Fig. 2c and Supplementary Table S2). The proportion of each cell population varied greatly at different times post APAP challenge and by genotype (Fig. 2d and Supplementary Fig. S1c). Next, CD11b+ myeloid cell clusters were further selected for downstream analysis based on their CD11b (Itgam) expression levels (Fig. 2e). CD11b^-/ultra-low clusters, including contaminating endothelial cells (Ptprb+ and Clec4g+), B cells (Cd22+ and Cd19+), and NKT cells (Cd3d+ and Klrb1c+), were excluded (Supplementary Fig. S1d–f). Cluster 14 (plasma cells, Iglv1+ and Jchain+, Supplementary Fig. S1g) was also excluded to avoid overestimation since only 1 cell in Cluster 14 was profiled in the MUT 48 h sample. Therefore, a total of 25,721 CD11b+ cells divided into 11 clusters (Clusters 0–9 and 11) from three time points and two genotypes were analyzed. Using known signature genes, these 11 cell clusters were composed of 5 major CD11b+ populations, including 2 types of Kupffer cells (Timd4+Clec4f+), 5 types of macrophages (CCR2+), 2 types of neutrophils (CCR2-), 1 type of DC cell (CD11c+MHCII+) [13], and 1 type of dividing myeloid cell (Ki67+) (Fig. 2f). Correlation analysis indicated that these 5 major cell populations displayed relatively conserved transcriptomic signatures, verifying the above classification in Fig. 2b, f (Fig. 2g).

Fig. 2.

Comprehensive cellular overview of CD11b+ liver nonparenchymal cells post APAP-ALI. a Schematic diagram of the scRNA-seq workflow. b UMAP plots for the cell type identification of 26,758 high-quality single cells from WT and MUT littermates pooled from 3 mice per genotype per time point (0, 24 and 48 h post APAP-ALI). c Heatmap showing the top 10 DEGs in each cell type in (b). d Bar plots showing the proportion of cell types in each sample. e Violin plots showing CD11b (Itgam) gene expression in 15 distinct cell types. f Dot plot of the mean expression of canonical marker genes for the 11 major lineages as indicated. g Correlation heatmap and hierarchical clustering of the 11 major lineages

Neutrophilic population heterogeneity post APAP-ALI

Recent studies have shown that neutrophils, as the first recruited responders, exhibit both proinflammatory and pro-resolving characteristics during APAP-ALI [9, 14]. In our study, two neutrophil subsets (CSF1+ N1 and CSF1- N8) (Figs. 2f and 3a) were observed and defined as low-density activated neutrophils in gradient separation (Supplementary Fig. S2a) [15]. N1 neutrophils expressed more CXCR4 (1.9-fold) but less CD62L (0.73-fold) and Ly6G (0.46-fold) than N8 neutrophils, resembling the reported neutrophil aging process [16]. Maximum N1 cells were observed at the peak of liver injury (24 h) and remained through 48 h (Supplementary Fig. S2b). The number of N8 cells producing the proinflammatory mediators S100a8/9 and MMP8/9 peaked at 24 h and declined at 48 h (Supplementary Fig. S2b). Functional analysis was performed on the marker genes (Fig. 3b) and differentially expressed genes (DEGs) between N1 and N8 (Supplementary Fig. S2c–e) both confirmed the enriched earlier neutrophil activation responses such as degranulation, migration and cellular response to stimuli in N8 neutrophils and aging signaling such as translation, apoptosis and cell death signaling in N1. A similar conclusion was obtained by evaluating known pathway activity using gene set variation analysis (GSVA) [16] (Fig. 3c). In addition, N1-upregulated DEGs were mainly enriched in the previously reported more aged activated neutrophil subset G5c [17], while N8 was enriched in the less aged activated G5a subset (Fig. 3d). Pseudotime analysis also confirmed the trajectories from N8 to N1 (Fig. 3e) [18], in which early-stage neutrophils underwent actin reorganization and stress response, while apoptosis and cytokine production were enriched at the late stage (Supplementary Fig. S2f). These results suggested that neutrophils were heavily engaged in translation to produce cytokines post degranulation as they transitioned to apoptosis during the liver recovery process.

Fig. 3.

Transcriptomic heterogeneity of liver infiltrated myeloid cells post APAP-ALI. a Illustration of the identified liver infiltrated neutrophil populations (N1 and N8) and Mo-MFs (MΦ2, MΦ7, MΦ0 and MΦ3). b Functional analysis of marker genes for N8 by Metascape. c Heatmap showing different functions and pathways enriched in N1 and N8 by GSVA. d Enrichment of the top 30 DEGs of G5a, b, and c in DEGs of N1 and N8. e Monocle trajectories of N1 and N8 colored by pseudotime (left) or cluster identity (right). f Total cell counts of the 4 types of infiltrated Mo-MFs 0, 24 or 48 h post APAP injection in WT mice. g Heatmap showing different functions and pathways enriched in MΦ2, MΦ7, MΦ0 and MΦ3 by GSVA. h Monocle trajectories of MΦ2, MΦ7, MΦ0 and MΦ3 colored by pseudotime (left) or cluster identity (right). i Correlation matrix of bulk RNA-seq of sorted macrophages published (GSE55606) and identified 4 types of macrophages (MΦ2, MΦ7, MΦ0 and MΦ3)

Plasticity of infiltrated Mo-MFs post APAP-ALI

Liver Mo-MFs undergo a maturation process from proinflammatory status toward restorative status [19]. We identified four types of infiltrating Mo-MFs that accumulated gradually post APAP-ALI (MΦ2, MΦ7, MΦ0 and MΦ3) (Fig. 3a, f). MΦ3 macrophages peaked at 48 h, while the other three populations peaked at 24 h and declined at 48 h post-APAP. Marker genes related to phagosome activity (lysosome, proteasome and oxidative phosphorylation) were highly expressed in restorative MΦ3 macrophages, while the TNF and NFκB signaling pathways were enriched in proinflammatory MΦ7 macrophages (Supplementary Fig. S3a). The freshly infiltrating Ly6C^hi MΦ2 exhibited massive ribosome biogenesis and translation, and early restorative MΦ0 was enriched with wound healing activities (Supplementary Fig. S3a). GSVA also revealed strong enrichment of phagosome and lysosome activities in MΦ3 (Fig. 3g). IFN-I production and signaling were highly enriched in MΦ0. It is worth noting that a persistent IFNβ, but not IFNα, response was observed in late reparative macrophages (MΦ3), suggesting the possible differential functions of IFN-I subtypes in the recovery of APAP-ALI. Sprouting angiogenesis associated with macrophage dissemination was enriched in both MΦ7 and MΦ2. Functional enrichment analysis identified MΦ3 as a more mature restorative MΦ since phagosome and PPAR signaling pathways were enriched in MΦ3 upregulated DEGs versus MΦ0 (Supplementary Fig. S3b, c). Trajectory analysis revealed a continuum of cells branching into different states arising from a root populated largely by MΦ2 macrophages (Fig. 3h). At branch points, these cells split into proinflammatory macrophages MΦ7 and further early restorative macrophages MΦ0 and finally MΦ3. The corresponding functional analysis indicated that at the early stage, Mo-MFs underwent migration and cytokine response, followed by acute phase response and wound healing, while at the late stage, lysosome and cargo receptor activity were enriched (Supplementary Fig. S3d). In addition, correlation analysis with the previously reported three types of Mo-MFs sorted using flow cytometry (GSE55606) (Ly6C^hi, Ly6C^low and CX3CR1- recovered KCs) [19] showed that Ly6C^hi macrophages were highly correlated with proinflammatory MΦ7 and somewhat less correlated with freshly infiltrated MΦ2 and early restorative MΦ0, while Ly6C^low and recovered KCs were correlated largely with restorative MΦ3 and to a lesser extent with early restorative MΦ0 (Fig. 3i). These data suggested that freshly infiltrating MΦ2 underwent both proinflammatory changes to MΦ7 at the early stage and pro-restorative changes from MΦ0 through MΦ3 at later stages.

In addition, resident Kupffer cells (K4 and K9) and CD11b+ DCs declined dramatically and remained at low levels through 48 h post APAP-ALI (Fig. 2b, d). Resident capsular MΦ5 expressing high levels of CX3CR1 [13] and CD11b+Ki67+ cells decreased transiently but recovered at 48 h post APAP-ALI. The top KEGG/GO terms enriched in these populations are listed (Supplementary Fig. S3e–i).

Effects of the STAT2 T403A mutation on myeloid cell populations after APAP-ALI

To elucidate the essential role of IFN-I signaling in orchestrating these myeloid cell functions during APAP-ALI, 180 DEGs were identified in liver CD11b+ nonparenchymal cells of MUT versus WT mice (Fig. 4a). Genes downregulated in MUT CD11b+ cells were highly enriched for IFN-I signaling and restorative macrophage-related activities (Fig. 4b), while upregulated genes were mainly related to proinflammatory responses (Supplementary Fig. S4a). Hub gene analysis in the protein‒protein interaction network identified seven interferon-stimulated genes (ISGs) [20] out of the top 10 hub genes downregulated in MUT cells (Fig. 4c) and various proinflammatory genes in upregulated hub genes (Supplementary Fig. S4b) [20, 21], defining IFN-I activation as the core signaling pathway. Consistently, immunofluorescence staining results showed that tyrosine phosphorylation at STAT2 Y690 in response to IFN-I was mainly enriched in CD45+ immune cells (Supplementary Fig. S4c), especially CD11b+ myeloid cells (Fig. 4d, e), which were largely compromised in MUT mice at 24 h post APAP-ALI [12]. The expression of typical ISGs, such as IFIT1, in whole liver tissue during APAP-ALI was also hindered in MUT mice in vivo (Fig. 4f). Similarly, LPS-induced IFN-I-dependent tyrosine phosphorylation of STAT1 and STAT2 was also attenuated in M-CSF/CSF1-polarized BMDMs (M-BMDMs) from MUT mice and entirely absent in IFNAR KO mice in vitro (Fig. 4g), as was the expression of ISGs such as IFIT1 and CCL5 (Supplementary Fig. S4d, e). Consistently, anti-IFNAR neutralizing antibody could restrain LPS-induced STAT2 activation from 2 h in WT mice, pointing to IFN-I/IFNAR-mediated STAT2 activation post LPS challenge (Fig. 4h). Furthermore, mice receiving anti-IFNβ neutralizing antibody showed more severe liver pathology with larger necrotic areas (Fig. 4i, j) and higher plasma ALT and GDH levels (Fig. 4k, l). All these data suggested that IFN-I signaling enriched in CD11b+ cells participated in myeloid cell function regulation and contributed to liver recovery.

Fig. 4.

IFN-I signaling is attenuated in MUT mice post APAP-ALI. a Volcano plot showing DEGs of the purified whole CD11b+ nonparenchymal cells between WT and MUT mice. b Functional enrichment and c top 10 hub genes in the downregulated DEGs of MUT cells. d Immunofluorescence imaging of CD11b (red) and phospho-STAT2 (pY690) (green) in liver with DAPI (blue) 24 h post APAP. Scale bar, 100 μm. e Quantification of p-STAT2+ cells in CD11b+ cells in (d). Two-sided t-test. f Relative expression of IFIT1 in liver tissue of APAP-ALI. Two-way ANOVA. g Western blot of phospho-STAT1 (pY701), phospho-STAT2 (pY690), STAT1, and STAT2 expression in M-BMDMs from WT, MUT and IFNAR KO mice treated with LPS (100 ng/ml) at the indicated times. h M-BMDMs from WT mice were treated with vehicle or anti-IFNAR antibody (5 μg/ml) for 30 min, followed by LPS treatment for the indicated times. Phospho-STAT2 (pY690) and STAT2 expression were analyzed using western blotting. i–l WT mice (n = 10 for each group) were injected with anti-IFNβ antibody (1 μg/20 g b.w. in 200 μl sterile saline, i.p.) or isotype control post APAP-ALI at 12 h and/or 24 h before sacrifice at 36 h. Representative liver histological staining (i) and necrotic area quantification (j). Plasma ALT (k) and GDH (l) are shown. One-way ANOVA. m Heatmap showing representative ISG gene expression in 11 major clusters of WT and MUT mice. n Functional enrichment in the selected clusters of WT and MUT mice. *p < 0.05

At single-cell resolution, the expression levels of 100 different ISGs [20, 21] varied among the 11 major CD11b+ cell clusters (Fig. 4m) as well as within each cluster of 6 samples (Supplementary Fig. S5a). Despite this diversity, ISGs [20] represented a significant portion of the downregulated DEGs of MUT in each subpopulation (composites 4.6–25.2% of total DEGs) (Supplementary Fig. S5b and Supplementary Tables S3 and S4). Furthermore, an UpSet plot [22] was utilized to visualize the similarity of downregulated ISGs between these subpopulations in a matrix, with the rows of the matrix corresponding to the downregulated ISGs in each subpopulation of MUT mice and the columns corresponding to the shared gene numbers among the indicated cell types (Supplementary Fig. S6a). As shown in Supplementary Fig. S6a and Supplementary Table S5, classic ISGs such as Irf7, Ifi204 and Ifi27l2a were downregulated in MUT in all subpopulations. Functional analysis of each cluster revealed strong enrichment of IFN-I-related signaling and functions such as antiviral, antigen-presenting and interferon-signaling activities in WT mice (Fig. 4n and Supplementary Fig. S6b). For clusters such as MΦ0, MΦ3 and MΦ2, both “response to IFNα” and “response to IFNβ” functional terms were enriched with higher significance (indicated by q value) for IFNβ signaling. For clusters such as DC6, K9 and Div11, “response to IFNβ” but not “response to IFNα” was ranked. These data imply that IFNβ might play a more prominent role in STAT2 T403A-diminished IFN-I signaling in the recovery of APAP-ALI. Collectively, these data suggest that IFN-I signaling plays indispensable roles by differentially modulating gene expression in distinct myeloid cell populations.

IFN-I strengthens efferocytosis in restorative macrophages

Efferocytosis is an essential part of the resolution of ALI through the elimination of cell debris and apoptotic cells [10, 11]. Compared with other infiltrating Mo-MFs, MΦ3 cells exhibited the highest phagosome, lysosome and proteasome scores (Fig. 5a and Supplementary Fig. S7a), which were reduced in MUT Mo-MFs. When compared with our bulk RNA-seq data of bone marrow (BM) and M-BMDMs (GSE179637) [23], MΦ3 closely resembled CSF1-derived M-BMDMs (Supplementary Fig. S7b). Similarly, analyzing the bulk RNA-seq data of WT and MUT M-BMDMs (GSE179591), phagosome activities were functionally enriched in the downregulated DEGs in MUT (Supplementary Fig. S7c), and phagosome function-related gene expression was also reduced in mutant M-BMDMs (Supplementary Fig. S7d) and MΦ3 (Supplementary Fig. S7e). Consistent with these results, the uptake of apoptotic cells by WT M-BMDMs was markedly higher than that by their mutant counterparts both at rest and after IFNβ treatment (Fig. 5b, c).

Fig. 5.

Impaired efferocytosis due to STAT2 phosphorylation deficiency at T403. a Violin plot of lysosome and phagosome scores for liver Mo-MFs (MΦ2, MΦ7, MΦ0 and MΦ3) 0, 24 or 48 h post-APAP. ^# for the comparison of MΦ0, MΦ3 and MΦ7 vs. MΦ2, and * for comparison of WT and MUT samples of the indicated MΦ cells. Two-way ANOVA. b, c Schematic diagram of the macrophage efferocytotic procedure in vitro (b). M-BMDMs were incubated with apoptotic Jurkat cells for 6 h. The percentage of BMDMs engulfing Jurkat cells was analyzed by flow cytometry (c). Two-way ANOVA. d Hierarchical plot shows the inferred intercellular communication network for GAS signaling 48 h post APAP-ALI (upper panel). Relative contribution of each ligand‒receptor pair to the GAS signaling pathway (bottom panel). e Immunofluorescence imaging of Gas6, Trem2, MerTK and AXL in the liver with DAPI 48 h post APAP-ALI. Scale bar, 100 μm. f Dot plot of the mean expression of GAS signaling (Gas6, Axl and Mertk) for MΦ3 in WT and MUT mice. g Immunohistochemical staining of Gas6 in the livers of WT and MUT mice 48 h post APAP-ALI. h Quantification of Gas6+ cells in (g). Two-sided t-test. i MΦ3-related significant signaling pathways are ranked based on their differences in overall information flow within the inferred networks between WT and MUT mice 48 h post APAP-ALI. GALECTIN represents Lgals9 signaling, GDF represents Gdf15, SPP1 represents Spp1, and ANGPTL represents Angptl4, as shown in Supplementary Fig. 7. ^# and *indicate p < 0.05

Furthermore, CellChat [24] was utilized to quantitatively analyze the intercellular communication networks from scRNA-seq data through methods abstracted from graph theory, pattern recognition, and manifold learning. A hierarchical plot drawn using CellChat provides an informative visualization method to highlight autocrine and paracrine signaling communication between cell subtypes for the indicated signaling pathway. As shown in Fig. 5d, we identified GAS signaling as an efferocytotic mediator by CellChat. As the primary ligand source, MΦ3 acted in both autocrine and paracrine manners to modulate MΦ0, K4, MΦ5 and itself, mainly through Gas6-Axl and Gas6-Mertk signaling (Fig. 5d). Consistent with this analytic result, Gas6 was highly enriched in Trem2+ MΦ3 cells and colocalized with TAM receptors (MerTK and Axl) within livers 48 h post APAP-ALI (Fig. 5e). Moreover, Gas6 and Axl were IFN-inducible genes [25], and the expression levels of Gas6 in MΦ3 and Axl and Mertk in targeted cells (MΦ0, MΦ3, K4 and MΦ5) were lower in MUT mice (Fig. 5f and Supplementary Fig. S7f, g). The decreased Gas6 expression within liver sections of MUT mice 48 h post APAP-ALI was also verified using immunohistochemical staining (Fig. 5g, h). These data suggest that Gas6-mediated efferocytosis in Mo-MFs contributed dominantly during damage repair of APAP-ALI.

Next, we compared the information flow for the indicated signaling pathways between WT and MUT mice 48 h post APAP-ALI using CellChat analysis (Fig. 5i). The information flow for a given signaling pathway is calculated by the sum of the communication probability among all pairs of cell groups in the inferred network. Pathways such as IGF, CXCL and IL2 (in black font) maintain a similar flow between WT and MUT. In contrast, the flow strength of GAS signaling was higher in WT samples, suggesting reduced GAS signaling in MUT CD11b+ cells. Moreover, four MΦ3-related pro-restorative pathways (GDF, GALECTIN, ANGPTL and SPP1) (in red font) were decreased, and a pro-recruiting CCL pathway (in green font) was increased in MUT cells (Fig. 5i and Supplementary Fig. S7h–l). Collectively, these data indicated that STAT2 T403 phosphorylation deficiency (MUT) diminished macrophage efferocytotic activity and other pro-restorative signaling by restraining IFN-I pathway activation.

CSF1 production by neutrophils is diminished due to impaired IFN-I activities

As a known ISG, CSF1 provokes mononuclear phagocyte differentiation and tissue homeostasis [26, 27] and has been reported to be a prognostic marker for ALI [28]. Analysis of the CSF1-CSF1 receptor (CSF1R) signaling network using CellChat identified neutrophil N1 as the most prominent source for CSF1 and Mo-MFs MΦ2 and MΦ0 as the dominant receivers (Fig. 6a, b). Consistently, CSF1 was expressed predominantly in N1 neutrophils and to a lesser extent in MΦ7 macrophages, while CSF1R was ubiquitously expressed in all infiltrating macrophages and resident Kupffer cells (Fig. 6c, d). The expression levels of both CSF1 and CSF1R in N1 (Fig. 6e) and MΦ7 cells (Supplementary Fig. S8a) were decreased in MUT cells 24 h post APAP-ALI, indicating the essential roles of IFN-I in neutrophil-mediated CSF1 signaling. To experimentally verify this finding, we isolated infiltrated neutrophils and non-neutrophil cells using flow sorting and magnetic bead separation, and the expression levels of CSF1 were analyzed by RT‒PCR. As shown in Fig. 6f and Supplementary Fig. S8b, consistent with the scRNA-seq data (Fig. 6), neutrophils (Ly6G+), but not other cell types, expressed high levels of CSF1, and the T403A mutation diminished CSF1 expression 24 h post APAP-ALI. At the protein level, according to the results of CSF1 immunostaining of liver sections from both WT and MUT mice, the expression of CSF1 was attenuated in MUT mice post APAP-ALI (Fig. 6g, h). Furthermore, an anti-IFNβ neutralizing antibody was applied to block IFN/JAK/STAT signaling during APAP-ALI, and the expression of CSF1 in liver sections was inhibited 24 h post APAP-ALI, according to RT‒PCR results (Fig. 6i), which also indicates the relevance of IFN-I with regard to CSF1 production.

Fig. 6.

Neutrophil-derived CSF1 expression was inhibited in MUT mice post APAP-ALI. a Heatmap and hierarchical clustering showing the relative importance of each cell group based on the computed four network centrality measures of CSF signaling of WT mice 24 h post APAP-ALI. b Relative contribution of each ligand‒receptor pair to the CSF signaling pathway. c, d Feature plots of Csf1 and Csf1r gene expression 24 h post APAP-ALI. e Dot plot of the mean expression of Csf1 and Csf1r for N1. f Neutrophils (CD11b+Ly6G+), Mo-MFs (CD11b+Ly6G-), nonmyeloid immune cells (CD45+CD11b-Ly6G-) and nonimmune cells (CD45-) were sequentially isolated from cell suspensions of liver tissues from WT and MUT mice 24 h post APAP-ALI using magnetic bead separation. The relative mRNA levels of Csf1 were analyzed using RT‒PCR. Two-way ANOVA. g Immunohistochemical staining of CSF1 in the livers of WT and MUT mice 24 h post APAP-ALI. h Quantification of CSF1+ cells in (g). IOD: integral optical density. Two-sided t-test. i WT mice were injected with anti-IFNβ antibody (1 μg/20 g b.w. in 200 μl sterile saline, i.p.) or vehicle 6 h post APAP-ALI. Mice were sacrificed 24 h post-APAP. The relative mRNA levels of Csf1 in liver samples were analyzed using RT‒PCR. One-way ANOVA. j N1-related significant signaling pathways are ranked within the inferred networks 24 h post APAP-ALI. k Alluvial plot showing outgoing signaling patterns of secreting cells. Violin plot of neutrophil degranulation (l) and aging score (m) for infiltrated neutrophils. n Neutrophils were treated with IFNβ or vehicle, and the percentage of Annexin-V+ cells was calculated. Two-way ANOVA. *p < 0.05

Next, we compared the information flow for the indicated signaling pathways between WT and MUT mice 24 h post APAP-ALI using CellChat analysis (Fig. 6j, k). The flow strength of CSF signaling was lower in MUT samples, suggesting reduced CSF signaling in MUT CD11b+ cells. In addition, autocrine proinflammatory IL-1 and pro-recruiting CXCL signaling by neutrophils were classified (Fig. 6j, k and Supplementary Fig. S8c, d) and their expression (Supplementary Fig. S8e, f) and signaling strength (Fig. 6j) were increased in MUT cells, indicating that neutrophils promoted more proinflammatory responses in MUT mice. The tendency of neutrophil degranulation and aging scores (Fig. 6l, m) [17] along the sequential N8 to N1-cell lineage transition (Fig. 3e) was hindered, and infiltrating neutrophils sustained a more proinflammatory status in MUT mice. Consistently, the percentage of Annexin-V+ apoptotic cells was increased upon IFNβ treatment in elicited peritoneal neutrophils ex vivo, which was compromised in MUT cells (Fig. 6n). Moreover, proinflammatory scores, such as chemotaxis, migration, and activation scores, were higher in MUT neutrophils, indicating increased neutrophil infiltration and exacerbated liver injury (Supplementary Fig. S8g–i). All these data suggested that IFN-I signaling participated in elevated late-stage neutrophil CSF1 production, reduced proinflammatory cytokine production, and neutrophil clearance.

STAT2 T403A mutation inhibited IFN-I production by interfering with IRF3/IRF7 activities during APAP-ALI

In addition to the decreased IFN-I signaling activation (Fig. 4d–f), IFNβ expression in liver tissues increased over time and peaked at 24 h post APAP-ALI in WT mice but showed only limited expression in MUT mice (Fig. 7a). IFNβ was predominantly colocalized with CD45+ immune cells (Supplementary Fig. S9a, b), particularly CD11b+ myeloid cells in liver tissue, and was markedly deficient in the MUT counterpart (Fig. 7b). To further determine the major IFNβ producer, CD11b+ cells were further dissected into neutrophils and Mo-MFs. As shown in Fig. 7c, the expression levels of IFNβ were much higher in CD11b+ Mo-MFs. To verify whether IFN-I production deficiency post PAMP/DAMP induction in MUT macrophages is restricted to CSF1-primed macrophages as in APAP-ALI, two well-established differentially polarized BMDMs (GM-CSF- and M-CSF/CSF1-primed BMDMs) [29] were investigated in vitro. Interestingly, LPS-induced IFNβ expression was markedly reduced in MUT CSF1-primed M-BMDMs but not GM-CSF-primed BMDMs (GM-BMDMs) (Fig. 7d and Supplementary Fig. S9c, d). The discrepancy in IFNβ induction in differentially polarized macrophages between WT and T403A implied that IFN-I production was especially important and regulated in CSF1-primed restorative macrophages both in vivo and in vitro.

Fig. 7.

The STAT2 T403A mutation led to diminished IFNβ production by decreasing IRF3/7 activities. a IFNβ concentration was determined by ELISA in the supernatant of liver post APAP-ALI. Two-way ANOVA. b Immunofluorescence imaging of CD11b (red) and IFNβ (green) with DAPI (blue) in the livers of WT and MUT mice 24 h post APAP. Scale bar, 100 μm. Quantification of IFNβ+ cells in CD11b+ cells is shown in the right panel. Two-sided t-test. c The relative IFNβ mRNA levels of the isolated neutrophils (CD11b+Ly6G+), Mo-MFs (CD11b+Ly6G-), nonmyeloid immune cells (CD45+CD11b-Ly6G-) and nonimmune cells (CD45-) from cell suspensions of liver tissues 24 h post APAP-ALI using a magnetic bead separation approach. One-way ANOVA. d IFNβ concentration was determined by ELISA in the media of LPS-challenged M-BMDMs from WT and MUT mice. Two-way ANOVA. e Heatmap showing IFN-I-related functions and pathways enriched in MΦ0 in WT and MUT mice 24 h post APAP-ALI by GSVA analysis. f Heatmap of the t values representing regulon activity change of MΦ0 between WT and MUT at 0, 24, and 48 h post APAP-ALI by SCENIC analysis. g, h Bone marrow cells (BM) were isolated and differentiated with GM-CSF (GM) or M-CSF (CSF1) (M) for 7 days. Cells were then harvested and processed for RNA-seq analysis. g Dot plot of the response to IFNβ-related gene expression of DEGs between the indicated comparisons. h IPA upstream analysis comparison of DEGs between the indicated groups. i Bone marrow cells of WT, MUT and IFNAR KO mice were primed by M-CSF (CSF1) for 7 days, followed by LPS challenge (100 ng/ml) at the indicated times. IRFs expression in M-BMDMs was determined using western blotting. j–l WT mice were injected with mIFNβ (50 ng/20 g b.w. in 200 μl sterile saline, i.p.) or vehicle 12 h post APAP-ALI before sacrifice at 36 h. Representative liver histological staining (j), necrotic area quantification (k) and plasma ALT (l) are shown. Scale bar, 100 μm. Two-sided t-test. m Schematic illustration of intercellular crosstalk between the major infiltrated myeloid lineages during the resolution phase of APAP-ALI (left) and IFN-I production and activation signaling pathways (right). The protein or phosphorylation marked in red represents the authorized targets, which is essential for the resolution phase of APAP-ALI. *p < 0.05

MΦ0 was identified as the major IFN-I-producing population (Fig. 3g), and the STAT2 T403A mutation impaired IFN-I production 24 h post APAP-ALI (Fig. 7e), which was confirmed using canonical pathway analysis by Ingenuity Pathway analysis (IPA, QIAGEN) (Supplementary Fig. S9e). Single-cell regulatory network inference and clustering (SCENIC) analyses [30] revealed that the regulon activities of the known transcription factors driving IFN-I production, including IRF3, IRF7, Nfκb1 and Nfκb2, were upregulated in MΦ0 from WT but not MUT mice post APAP-ALI (Fig. 7f), consistent with the impaired liver recovery observed in IRF3 and TRIF KO mice (Fig. 1b–f). Upstream regulator analysis by IPA also identified IRFs and other transcription factors that might account for the diminished IFN-I secretion in MUT MΦ0 (Supplementary Table S6). These data indicated that STAT2 T403 phosphorylation not only attenuated IFN-I signaling but also diminished upstream TRIF/IRF activities during APAP-ALI.

The MΦ0 cells were identified as the precursor of the restorative MΦ3 cells (Fig. 3h). Therefore, we hypothesized that IFN-I production might participate in the regulation of CSF1-polarized macrophage maturation. According to our bulk RNA-seq data (GSE179637), IFN-I-related functions were enriched in the upregulated DEGs of M-BMDMs versus GM-BMDMs (Supplementary Fig. S9f, g). Gene sets for the response to IFNβ were highly expressed in M-BMDMs compared with GM-BMDMs, BM or MUT M-BMDMs, implying the essential role of IFNβ during M-BMDM maturation (Fig. 7g). Upstream regulator analysis also pointed to IRF3/IRF7 activities in controlling IFN-I signaling activation in M-BMDMs (Fig. 7h). As classical ISGs [31], the changes in IRF3/IRF7 activities might partially result from their altered expression. Indeed, STAT2 T403A and IFNAR depletion downregulated the expression of IRF3/7/9 and STAT2 in both the steady state and after LPS challenge (Figs. 4g and 7i and Supplementary Fig. S9h–j). However, no obvious expression changes in STAT2 and IRF3 were observed in MUT GM-BMDMs at a steady state (Supplementary Fig. S9k). The diminished STAT2 phosphorylation in MUT GM-BMDMs results from insufficient canonical IFN-I signaling activation (Supplementary Fig. S9l) but not IFNβ production changes (Supplementary Fig. S9d). IFNβ-induced STAT2 activation was also hindered in MUT M-BMDMs (Supplementary Fig. S9m). Taken together, STAT2 T403 phosphorylation deficiency hindered both IFN-I production and IFN-I activation in CSF1-polarized macrophages due to attenuated upstream IRF3/7 activities.

Since IFN-I is tightly involved in the regulation of myeloid cell functions, we reasoned that IFN-I might be a therapeutic agent to promote the resolution of ALI. Notably, IFNβ treatment [6] at 12 h post APAP-ALI resulted in decreased plasma ALT levels and accelerated tissue repair in the liver (Fig. 7j–l). Collectively, post APAP-ALI, neutrophils (N8) are first recruited into the damaged liver tissue, produce MMPs and S100s, and undergo degranulation and aging to late-stage neutrophils (N1). Sequentially, Ly6C^hi Mo-MFs (MΦ2) with the highest CCR2 expression levels infiltrated and then differentiated into both Ly6C+ proinflammatory MΦ7 macrophages and intermediate anti-inflammatory MΦ0 macrophages. IFN-I produced by MΦ0 functions on aged N1 to promote CSF1 secretion, neutrophil aging and apoptosis, which also similarly affect MΦ7. In addition, MΦ0-produced IFN-I can also act on MΦ2 and themselves to prime restorative macrophage maturation (MΦ3). CSF1 was prominently produced by N1 cells, accompanied by IFN-I signaling, polarizing MΦ2 to restorative Trem2+ MΦ3 through IFN-I-producing MΦ0 (Fig. 7m). Compromising these pathways by preventing STAT2 T403 phosphorylation results in limited IFN-I production and hinders the resolution of APAP-ALI (Fig. 7m).

Discussion

There is a general consensus that during APAP-ALI, infiltrating myeloid cells play versatile roles in orchestrating the hepatic immune response and tissue repair. However, knowledge of the dynamic changes and crosstalk among these cells during APAP-ALI remains incomplete. In this study, four distinct macrophage and two neutrophil subpopulations and their cytokine-driven communication pathways were identified. Of particular interest, the IFN-I signaling pathway was proven to play a pivotal role in the modulation of multiple myeloid cell functions. IFN-I production by Mo-MFs is critical for CSF1 production by inflammatory neutrophils and promotes neutrophil apoptosis. CSF1 subsequently promotes the acquisition of restorative macrophages that are essential for liver repair. We propose that IFN-I signaling mediates the bidirectional crosstalk between macrophages and neutrophils, thereby defining a novel resolution circuit for APAP-ALI.

Mo-MFs exhibit high plasticity and are involved in both necroinflammatory and resolution phases of ALI. As the initial sensor of APAP-ALI, Kupffer cells produce CCL2 to recruit CCR2+ monocytes and CXCL1, CXCL2 and CXCL8 to attract neutrophils [32]. Infiltrating CCR2+Ly6C^hi Mo-MFs produce a variety of cytokines and chemokines that exacerbate liver inflammation [33] and subsequently undergo differentiation into restorative Ly6C^low/- Mo-MFs to suppress inflammation and initiate the liver repair process [19]. Ly6C^low/- Mo-MFs exhibit an immunosuppressive and restorative phenotype characterized by the expression of growth factors (HGF and IGF) and phagocytosis-related genes (MerTK and Trem2) [10, 11, 32]. Despite multiple studies, the complete myeloid cell atlas and differentiation landscape remain incompletely characterized in APAP-ALI. Recently, Elinav et al. [34] identified two types of Mo-MFs at the injury phase of APAP-ALI, including 71 IFN+ monocytes. While these cells showed an increased response to IFN, no further description of their potential roles was explored. Here, we provide an analysis of numerous IFN-responding cell populations and identify 3861 IFN-producing cells (MΦ0) using both bioinformatic and experimental strategies to thoroughly characterize their transcriptional profile and potential roles in promoting neutrophil apoptosis and restorative macrophage maturation. IFN-I produced by MΦ0 promotes the transition of MΦ2 into the mature restorative phagocytic Trem2+ MΦ3 phenotype, consistent with the SAMΦ populations described in fibrotic livers [35] and TAM-like macrophages in the HCC microenvironment [36]. We also noticed that the IFN-I response in the identified macrophage subpopulations showed a certain degree of difference between IFNα and IFNβ, with the IFNβ response lasting in the restorative macrophages (Fig. 3g). Differential responses to IFN-I subtypes were also observed in most clusters when comparing WT and MUT samples (Fig. 4n and Supplementary Fig. S6b). Although IFNα and IFNβ signaling share high similarities in signaling transduction and downstream target genes, differences exist in receptor affinity and specific downstream gene induction. IFNβ exhibits a higher affinity for IFNAR1 and IFNAR2 than IFNα [37, 38]. Moreover, it was reported that IFNβ is the only IFN-I that can induce the expression of restorative/immunosuppressive molecules IL10 and PDL1 [39–41]. IFNβ but not IFNα could significantly upregulate IL10 expression to reprogram resolution-phase macrophages in bacterial infection [6]. Moreover, dsDNA released by necrotic hepatocytes could activate STING, which primarily stimulates IFNβ production. According to the data we have now, we only know that exogenous IFNβ can ameliorate APAP-ALI (Fig. 7j–l), but we could not rule out the possible contributions from IFNα. Further studies with IFNα- or IFNβ-specific KO mice should answer these questions in the future.

As previously reported [6, 42, 43], IFNβ plays essential roles in the resolution of bacterial inflammation and is produced by nonphagocytic macrophages, which are derived from phagocytic macrophages. In addition to the influence on macrophage efferocytosis, IFNβ-producing nonphagocytic macrophages also exhibit decreased functions such as “focal adhesion”, “collagen fibril organization”, and “glycolytic process”, increased “oxidative phosphorylation”, and downregulation of the expression of “26 upregulated genes in idiopathic pulmonary fibrosis (IPF) patients”. In the current study, IFNβ-producing macrophages (MΦ0) are recognized in the early stage of liver repair and are identified as the precursor of phagocytic macrophages. Then, we compared the functions and related gene expression of previously identified “IFNβ producing nonphagocytic macrophages” [6] with IFNβ-producing MΦ0 and phagocytic MΦ3 in this study. Twenty-five out of 26 IPF genes were detected in our scRNA-seq data, and among them, 15 genes were downregulated in MΦ0 compared with MΦ3. Next, we calculated the GSVA scores of the indicated GO functional terms. Consistent with nonphagocytic macrophage characteristics, MΦ0 also exhibited decreased “collagen fibril organization” compared with MΦ3. However, inconsistent with IFNβ-producing nonphagocytic macrophages, “glycolytic process” and “focal adhesion assembly” were increased and “oxidative phosphorylation” was decreased in MΦ0 compared with MΦ3. Therefore, we concluded that IFNβ-producing MΦ0 in this study also exhibited anti-fibrotic gene signatures and functions, similar to the reported nonphagocytic macrophages. However, these two IFNβ-producing macrophage populations were dissimilar in other functional aspects, indicating that IFNβ-producing macrophages could play different roles during the resolution of inflammation.

Upon ALI, infiltrating neutrophils facilitate hepatic inflammation partially by recruiting inflammatory monocytes [32] and induce Mo-MF transition toward a reparative phenotype through ROS-dependent AMPK activation during resolution [9]. Reparative Mo-MFs in turn eliminate apoptotic neutrophils, allowing liver homeostatic recovery. Neutrophils may also promote Mo-MF anti-inflammatory reprogramming via CSF1 [26] or by impairing NFκB activation [44]. Mo-MFs producing IFNβ have been reported to induce neutrophil apoptosis to help resolve bacterial inflammation [6]. All these studies indicate versatile and complicated communication between neutrophils and macrophages to orchestrate the inflammatory process. Our scRNA-seq data fully recapitulate these concepts. During APAP-ALI, CSF1 produced by aged neutrophils drove the maturation of restorative Mo-MFs to resolve hepatic damage, while IFNβ produced by early restorative Mo-MFs induced neutrophil aging and apoptosis to suppress inflammatory responses. Apoptotic neutrophils could be eliminated by CSF1-polarized restorative macrophages through efferocytosis. The reciprocal interactions between these two cell populations facilitated liver repair post APAP-ALI.

Efferocytosis by restorative macrophages is essential for the resolution of APAP-ALI. According to our scRNA-seq analysis and experimental data, Gas6-mediated efferocytosis in Mo-MFs, mainly through Gas6-Axl and Gas6-Mertk signaling, contributed dominantly during damage repair of APAP-ALI. Since Axl and Gas6 are ISGs, their gene expression and signaling strength were diminished in MUT mice, which was responsible for the impaired efferocytotic activities during APAP-ALI. As numerous ISGs were downregulated in MUT cell populations, we hypothesized that other ISGs might also regulate macrophage efferocytosis to provoke liver repair post APAP-ALI. For example, CCL5, a classical ISG, was reported to be sustained in resolving exudates during peritonitis and promoting macrophage reprogramming upon efferocytosis [45]. In addition to TAM receptor (MerTK and Axl) and ligand (Gas6), we also explored the gene expression patterns of the remaining TAM receptor (Tyro3) and ligand (Pros1) [46, 47]. They are also functionally relevant to efferocytosis (phagocytosis) but did not contribute to Gas6 signaling, as predicted by CellChat analysis in this study. The TAM receptor Tyro3 was merely detected in the cell populations we analyzed (data not shown). Protein s (Pros1) was mainly expressed in Kupffer cells (K4) and exhibited the highest expression at a steady state (0 h). As Kupffer cell populations were significantly decreased during APAP-ALI, the contribution of Pros1 in the Kupffer cell population is unlikely to be the major contributor to efferocytosis.

The role of IFN-I in host defense against viruses is well established, and macrophage-derived IFNβ was recently reported to help resolve bacterial inflammation [6]. In previous work, we demonstrated that STAT2 T403 phosphorylation drives a conformational switch of the STAT1/STAT2 heterodimer to engage an antiviral IFN-I response [12]. The current study further expands the role that STAT2 T403 phosphorylation plays not only as a critical feature of IFN-I signaling but also its production during nonviral inflammatory responses. Endogenously produced IFN-I was established to promote phagocytic activities in CSF1-polarized macrophages as early as 1985 [48]. Subsequently, Hamilton et al. suggested that endogenous IFN-I could prime CSF1-polarized macrophages, which displayed increased MyD88-independent IRF3/STAT1 activation after LPS stimulation in vitro [49]. Moreover, exogenous IFN-I could upregulate IRF7 and TRIF expression, IRF3 phosphorylation, and NFκB activation in macrophages upon TLR engagement [50]. In addition to positive feedback loop regulation, STAT2 has been shown to directly interact with certain upstream regulators, such as p65, to modulate its transcriptional activities [51, 52]. Our data suggested that the upstream regulon activities of IRF3, IRF7 and NFκB were downregulated in T403A macrophages. We hypothesize that STAT2 T403 phosphorylation might participate in the regulation of IFNβ production by both direct and indirect interactions with upstream regulators.

The present study thus provides a comprehensive transcriptomic landscape of liver CD11b+ cells post APAP-ALI at single-cell resolution and clarifies the essential role of IFN-I signaling during the resolution phase.

Materials and methods

In vivo mouse models

All animal experiments were performed under protocols approved by the Committee of Experimental Animals of School of Medicine and Pharmacy, Ocean University of China (OUC-SMP-20190201). All mice used were male. C57BL/6J TRIF, IRF3 and IFNAR knockout mice were sourced from the Jackson Laboratory, and C57BL/6J STAT2 T403A mice were sourced from Cyagen Biosciences. The C57BL/6J mice for IFNβ or anti-IFNβ antibody treatments and ex vivo primary culture were from Charles River Laboratories. All mice were provided water and chow ad libitum and maintained in a pathogen-free facility. Mice used in APAP-ALI experiments were 6–8 weeks of age.

Ex vivo primary culture

Mice used for peritoneal neutrophil isolation and BMDM polarization were male at 6–8 weeks of age.

APAP-induced liver injury (APAP-ALI)

Mice fasted overnight received an intraperitoneal (i.p.) injection of APAP (300 mg/kg, Sigma‒Aldrich) or saline. Mice were sacrificed at the indicated times post-APAP, and plasma ALT activities were evaluated by diagnostic kits (Nanjing Jiancheng Bioengineering Institute).

Preparation of liver CD11b+ nonparenchymal cells

Nonparenchymal cells were isolated from mouse livers as previously described with minor modifications [53]. Briefly, the liver was perfused with D-PBS solution via the inferior vena cava, digested with collagenase D (1 mg/ml) and minced with a Gentlemacs tissue dissociator (Miltenyi Biotec). The digests were filtered through a 100-μm cell strainer (BD Falcon) and washed with D-PBS several times, and the nonparenchymal cells were enriched by 30% Histodenz density gradient centrifugation. The collected nonparenchymal cells were immunostained with anti-CD11b microbeads (Miltenyi Biotec), and CD11b+ cells were collected by passing the microbead-conjugated cells through MACS separation columns in a magnetic field.

scRNA-seq library preparation and sequencing

Single-cell suspensions of freshly isolated CD11b+ nonparenchymal cells from MUT (T403A) and WT littermate controls were resuspended in PBS containing 0.04% ultrapure BSA. Single-cell transcriptomic amplification and library preparation were performed by BGI-Qingdao using single-cell 3’ V3 (10× Genomics) according to the manufacturer’s instructions. Bioinformatics methods for processing, demultiplexing, mapping and functional analysis are detailed in the supplementary information.

Statistics

Statistical significance of differences between the indicated samples was determined by unpaired Student’s t-test or one-way or two-way ANOVA using GraphPad Prism 9 software. ^#,*p < 0.05 is considered significant.

For details regarding the materials and methods used, please refer to the Supplementary information.

Author contributions

TH, CZ and JY directed all aspects of the project. QS, SD, XX, PP, YZhao, SL, ZZ, PH, PJ, YQ, WL, JZ, YX, JX, ZW, LW, MZ and YZhang performed the experiments. XL, QS and CY analyzed the seq data with significant contributions from XL in data visualization. XZ and CL discussed the clinical features of liver injury and reviewed the pathology sections. QS, CZ, JY and TH wrote the manuscript. YW, GS, YL and GRS read the manuscript and provided useful comments.

Funding

This work was supported by the Key R&D Program of Shandong Province (2020CXGC010503), Shandong Provincial Key Laboratory Platform Project (2021ZDSYS11), and Major Program of National Natural Science Foundation of China (81991525).

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41423-022-00966-y.