Targeting Hepatic Stellate Cell PD-L1 Alters Liver Inflammation and Fibrosis in CCl₄ Liver Injury Mouse Model

Abstract

Background & Aims

Programmed death-ligand 1 (PD-L1) on hepatic stellate cells (HSCs) is required for HSC activation and suppressing T and B lymphocytes. We tested whether targeting HSC PD-L1 influenced liver inflammation and fibrosis in a carbon tetrachloride (CCl₄) injury mouse model.

Methods

HSC-specific PD-L1 knockout (PD-L1^HSCKO) mice were created by crossing Cd274 floxed mice to Collagen1A1-Cre mice. CCl₄ was injected into PD-L1^HSCKO and PD-L1^HSCWT mice twice weekly for 6 weeks. Liver fibrosis was assessed by Trichrome and Picrosirius Red staining; HSC activation was determined by immunofluorescence and Western blot for HSC activation markers; liver inflammation was studied by multiplex immunofluorescence and cytokine profiling. Multiomics was leveraged to determine how targeting PD-L1 altered HSC producing collagens and cytokines/chemokines.

Results

Collagen deposition was reduced in CCl₄-injured PD-L1^HSCKO livers compared with CCl₄-injured PD-L1^HSCWT livers; myofibroblast density was lower in CCl₄-injured PD-L1^HSCKO livers compared with CCl₄-injured PD-L1^HSCWT livers. CCl₄-injured PD-L1^HSCKO livers had higher lymphocyte densities (GranzymeB+, CD8a+, CD20+) but lower Kupffer and myeloid cell densities (F4/80+ and CD11b+) compared with CCl₄-injured PD-L1^HSCWT livers. Serum aspartate aminotransferase and alanine aminotransferase, however, were similarly elevated by CCl₄ in both groups. Spatial and bulk-cell transcriptomics revealed a global transcriptomic change of HSCs induced by PD-L1 targeting. A targeted proteomics identified that HSC secretion of a group of cytokines/chemokines, including growth/differentiation factor 15, granulocyte-macrophage colony-stimulating factor, C-X-C motif and C-C motif chemokines, was altered upon PD-L1 targeting, highlighting the role of HSC PD-L1 in HSC/Kupffer and HSC/myeloid cell interactions during HSC activation and fibrosis development.

Conclusions

Targeting HSC PD-L1 altered HSC transcriptome and liver inflammation, and suppressed liver fibrosis, representing a potential therapeutic strategy for liver fibrosis.

Graphical abstract

Summary.

Targeting programmed death-ligand 1 of hepatic stellate cells (HSCs) by Cre/loxP recombination alters HSC transcriptome and HSC secretion of cytokines/chemokines so as to suppress Kupffer/macrophage accumulation in carbon tetrachloride-injured murine liver required for HSC activation and liver fibrosis.

Liver fibrosis is a reversible pathologic condition resulting from chronic liver diseases, such as alcoholic liver disease, metabolic dysfunction-associated steatotic liver disease (MASLD), iron overload, autoimmune hepatitis, biliary obstruction, and viral hepatitis B and C.1, 2, 3 Myofibroblasts activated from hepatic stellate cells (HSCs) are responsible for the excessive accumulation of extracellular matrix (ECM) and liver fibrosis. Liver fibrosis can progress to more severe, irreversible complications such as cirrhosis, liver failure, and portal hypertension if not intervened in time.^4,5 Current therapeutic strategies primarily rely on targeting the disease-triggering factors such as alcohol, obesity, and viral hepatitis, which are limited and cannot reverse the fibrotic process completely. For example, Rezdiffra is the first and only drug approved by the United States Food and Drug Administration (FDA) to treat metabolic dysfunction-associated hepatitis (MASH) with moderate to advanced fibrosis along with diet and exercise, but it does not treat liver cirrhosis. Understanding the mechanisms underlying HSC activation is therefore imperative for identifying new targets for liver fibrosis treatment.

Programmed death-ligand 1 (PD-L1), encoded by the CD274 gene, is a transmembrane protein expressed on tumor cells and tumor stromal cells, including dendritic cells, macrophages, fibroblasts, and T cells.6, 7, 8, 9 Its receptor, programmed death 1 (PD-1), is expressed on T cells and B cells. Upon binding to PD-1 receptor, PD-L1 effectively inhibits the proliferation of T cells and B cells and induces cell apoptosis. Upregulation of PD-L1 has been detected in various cancers, which is one of the mechanisms underlying the immune evasion of cancer.9, 10, 11 The role of PD-L1 in liver diseases such as MASLD and viral hepatitis has received increasing attention, as overexpression of PD-L1 may alter the immune microenvironment of the liver, potentially suppressing cytotoxic T cells and reducing liver injury and damage.^12,13

We recently reported that PD-L1 expression by HSCs is upregulated by transforming growth factor beta 1 (TGFβ1) stimulation, and that HSC PD-L1 is required for TGFβ1-stimulated myofibroblastic activation of HSCs via protecting TGFβ receptor I mRNA and TGFβ receptor II protein from degradation.¹⁴ To investigate whether HSC PD-L1 represents a therapeutic target of liver fibrosis, we injected carbon tetrachloride (CCl₄) into control and HSC-specific PD-L1 knockout (KO) mice. We found reduced liver fibrosis in HSC-specific PD-L1 KO mice compared with control mice. Multiplex immunofluorescence (IF) and imaging demonstrated that cell densities of lymphoid cells (GranzymeB+, CD8a+, and CD20+ cells) were increased, whereas those of myeloid cells (F4/80+ and CD11b+ cells) were reduced in HSC-specific PD-L1 KO livers compared with control livers. Single-cell transcriptomics detected PD-L1 transcripts from activated-HSCs/myofibroblasts of CCl₄-injured murine livers. Spatial transcriptomics supported that targeting HSC PD-L1 led to a global transcriptomic change affecting HSC activation and HSC-producing collagens and cytokines/chemokines. A targeted proteomics identified that targeting PD-L1 altered HSC secretion of a panel of cytokines/chemokines, which may account for the lower Kupffer or myeloid cell densities observed in the HSC-specific PD-L1 KO livers. Thus, HSC PD-L1 is critical for HSC/Kupffer or HSC/myeloid cell interaction required for initiation and amplification of signals for HSC activation and liver fibrosis.

Results

Targeting HSC-specific Cd274/PD-L1 by Cre/LoxP Recombination Suppresses Murine Liver Fibrosis Induced by CCl₄

To obtain mice with PD-L1 deficiency in activated-HSCs/myofibroblasts, we set up a crossing between a floxed Cd274 mutant mouse line¹⁵ and a Collagen1A1-Cre (Col1A1Cre) transgenic mouse line.^16,17 Polymerase chain reaction (PCR) was performed to genotype mice so we obtained Cd274F/FCol1A1Cre mice (herein referred to as PD-L1^HSCKO mice) and Cd274+/+Col1A1Cre mice (control; herein referred to as PD-L1^HSCWT mice). Because the Col1A1Cre transgene only transmitted to male mice, 11 male PD-L1^HSCKO mice and 8 age-matched male PD-L1^HSCWT mice were subjected to CCl₄ injection (1 μL per gram of body weight diluted in oil).¹⁸ Six weeks later, we found that the livers of mice that received CCl₄ injection were pale-colored compared with dark-red in mice that received oil only (vehicle) (Figure 1A left). CCl₄ injection elevated the liver weight of PD-L1^HSCWT mice (Figure 1A) (P < .01), which was not obvious in PD-L1^HSCKO mice (Figure 1A). Results of serum biochemistry, however, revealed similar elevation of aspartate aminotransferase (AST) and alanine aminotransferase (ALT) by CCl₄ injection in both genotypes, indicating comparable liver injuries in the 2 groups of mice (Figure 1B) (P > .05).

Figure 1.

Liver fibrosis induced by CCl₄ was reduced in PD-L1^HSCKOmice compared to PD-L1^WTmice. (A) Left, PDL1F/FCol1A1Cre (PD-L1^HSCKO) and matched PDL1+/+Col1A1Cre mice (control; PD-L1^HSCwt) were subjected to CCl₄ injection (1 μL CCl₄ per gram of body weight) or corn oil injection (vehicle) 2 times per week for 6 weeks. Mouse liver was isolated, weighed, and photographed at the end point. Representative pictures of the mouse liver are shown. Right, liver weight was elevated by CCl₄ injection compared with oil injection in control mice, and this effect was attenuated in PD-L1^HSCKO mice. ∗∗P < .01; ∗∗∗P < .001 by ANOVA, n = 3, 3, 8, and 11 mice. (B) Murine serum was collected at the end point and serum biochemistry revealed that AST and ALT were elevated by CCl₄ injection compared to oil injection, and the changes were comparable in the 2 groups. ∗∗P < .01; ∗∗∗P < .001 by ANOVA, n = 3, 3, 8, and 11 mice. (C and D) Representative pictures of Trichrome staining (C) and Sirius Red staining (D) of murine liver sections are shown on the left. Quantitative data revealed that collagen deposition induced by CCl₄ was suppressed in PD-L1^HSCKO livers compared with PD-L1^HSCwt livers. ∗∗∗P < .001; ∗∗∗∗P < .0001 by t-test, n = 8 and 11 mice. For quantification, 5 to 10 microscopic fields were randomly selected from a liver section for densitometry analysis with the ImageJ software so that the average density of blue or red was obtained for the mouse. Bars, 400 μm.

Trichrome staining and Picrosirius Red staining of liver sections revealed that CCl₄ injection promoted collagen deposition in the liver of PD-L1^HSCWT mice, and this effect of CCl₄ was attenuated in PD-L1^HSCKO mice (Figure 1C and D) (P < .001). Because activated HSCs/myofibroblasts are a major contributor to collagen synthesis and deposition in the liver, we performed IF staining for α-smooth muscle actin (αSMA) and platelet-derived growth receptor α (PDGFRα), 2 markers of activated-HSCs/myofibroblasts. We found that IF densities of αSMA and PDGFRα were lower in CCl₄-injured PD-L1^HSCKO livers compared with CCl₄-injured PD-L1^HSCWT livers (Figure 2A and B) (P < .0001). Western blot (WB) analysis showed that the protein levels of αSMA, desmin (another HSC marker), collagen 1, PDGFRα, and PD-L1 were all reduced in CCl₄-injured PD-L1^HSCKO livers compared with CCl₄-injured PD-L1^HSCWT livers (Figure 2C) (P < .05). WB confirmed that the Col1A1-Cre transgene was induced to express Cre recombinase in murine livers in response to CCl₄ injection (Figure 2D) (P < .0001). Because primary murine HSCs undergo spontaneous activation in culture,^19,20 we collected murine HSCs at day 5 culture for αSMA IF, which revealed that HSCs of PD-L1^HSCKO mice exhibited suppressed spontaneous activation compared with those of PD-L1^HSCWT mice (Figure 2E) (P < .01) (Data represent 3 independent repeats with similar results). Thus, targeting HSC PD-L1 by Cre/LoxP approach suppresses HSC activation and liver fibrosis in mice in response to CCl₄-mediated liver injury.

Figure 2.

Targeting HSC PD-L1 by Cre/LoxP suppresses myofibroblastic activation of HSCs in CCl₄-injured mice. (A and B) Representative pictures of IF staining for αSMA (A) and PDGFRα (B) on murine liver sections are shown on the top. Quantitative data below revealed that αSMA and PDGFRα IF densities were lower in PD-L1^HSCKO livers compared with PD-L1^HSCwt livers. For IF quantification, 5 to 10 microscopic fields were randomly selected from a liver section for IF density analysis with the ImageJ software so as to obtain the average IF density for the mouse. ∗∗∗∗P < .0001 by t-test, n = 8 and 11 mice. Bars, 400 μm. (C) WB analysis demonstrated that in control mice, CCl₄ injection enhanced liver expression of collagen 1, αSMA, desmin, PDGFRα, and PD-L1 compared with oil injection, and this effect of CCl₄ was suppressed in PD-L1^HSCKO mice. ∗P < .05; ∗∗P < .01 by t-test, n = 5 and 5. (D) WB detected cre protein expressed in the liver of control mice as the result of CCl₄ injection. ∗∗∗P < .001 by t-test, n = 3 and 3 mice. (E) HSCs isolated from PD-L1^HSCKO and PD-L1^HSCWT mice were cultured for 5 days and collected for αSMA IF. Confocal microscopy detected that activation of PD-L1^HSCKO HSCs was suppressed compared with PD-L1^HSCWT HSCs in vitro. ∗∗P < .01 by t-test, n = 10 and 15 microscopy fields. Data represent 3 independent repeats with similar results. Bar, 200 μm.

Targeting HSC PD-L1 Alters Liver Inflammation of CCl₄-injured Murine Liver

Activated-HSCs/myofibroblasts regulate liver inflammation, liver injury, and fibrosis by direct or indirect interaction with the immune cells in the liver. Additionally, PD-L1 is an immune check point molecule suppressing T and B lymphocytes. The information led us to perform multiplex IF and imaging to determine the influence of PD-L1 targeting on liver inflammation, which was completed by the Visikol Multiplex IF and Imaging platform. Three CCl₄-injured PD-L1^HSCWT livers and 3 PD-L1^HSCKO livers were subjected to multiplex IF with a panel of immune cell markers, such as CD8a for the detection of cytotoxic T lymphocytes, GranzymeB for cytotoxic T lymphocytes and natural killer cells, B20 for B lymphocytes, F4/80 for Kupffer/macrophages, and CD11b for myeloid-derived leukocytes. Hepatocytes were labeled by a pan cytokeratin antibody; activated HSCs/myofibroblasts were labeled by anti-αSMA; cell nuclei were labeled by DAPI.

Multiplex IF demonstrated that, in a CCl₄-injured PD-L1^HSCWT liver, GranzymeB+ cells (red, Figure 3Ac), CD8a+ T cells (orange, Figure 3Ad), and CD20+ B cells (white, Figure 3Ae) were recruited into the space where they were in proximity to the αSMA-positive fibrotic septa (purple, Figure 3Aa), and that F4/80+ Kupffer cells (light green, Figure 3Af) and CD11b+ myeloid cells (rose-red, Figure 3Ag) were in fact recruited onto the fibrotic septa. Because Kupffer/macrophages express CD11b and F4/80, coregistration and image overlapping enabled the differentiation of Kupffer/macrophages (arrowheads, CD11b+F4/80+, Figure 3Ai) from myeloid-derived leukocytes (arrows, CD11b+, Figure 3Ai). The most striking difference between the 2 groups of mice was that F4/80 IF signals and the organization of the F4/80+ cells into fibrotic septa were weak in CCl₄-injured PD-L1^HSCKO livers (Figure 3Ba), consistent with the data of αSMA IF (Figure 3Bb).

Figure 3.

Targeting HSC PD-L1 alters the immune cell landscape of CCl₄-injured murine liver. (A) Multiplex IF and imaging was performed on liver sections of mice that received CCl₄ injection. A set of IF pictures for αSMA (purple), pan-cytokeratin (brown), Granzyme B (red), CD8a (orange), CD20 (white), F4/80 (green), CD11b (rose-red) on PD-L1^HSCWT liver is shown in Aa-Ag. Cell nuclei were counterstained by DAPI (blue). An image containing 4 makers is shown in Ah and 2 markers in Ai. Bars, 100 μm. (B) Representative pictures of F4/80 IF (green, Ba) and αSMA IF (purple, Bb) revealing weak αSMA and F4/80 IF signals in CCl₄-injured PD-L1^HSCKO mouse liver. Cell nuclei were counterstained by DAPI (blue). Bars, 100 μm.

To analyze how the immune cell landscape was altered by targeting PD-L1 of HSCs in the murine liver, areas of interest (AOIs) were randomly selected from liver sections of 3 CCl₄-injured PD-L1^HSCWT livers and 3 CCl₄-injured PD-L1^HSCKO livers for immune cell quantification. Data of 3 mice each group were analyzed, and compiled data support that the cell densities of CD8a+, GranzymeB+, and CD20+ cells were significantly higher, whereas those of F4/80+ and CD11b+ cells were significantly lower in CCl₄-injured PD-L1^HSCKO livers compared with CCl₄-injured PD-L1^HSCWT livers (Figure 4A–E) (P < .05; n = 3 mice per group. Data of 6 individual mice are also shown). Thus, targeting HSC/myofibroblast PD-L1 led to opposite effects on the lymphoid and myeloid cells in CCl₄-injured murine liver.

Figure 4.

Targeting HSC PD-L1 alters the immune cell landscape of CCl₄-injured murine liver. (A–C) Multiplex IF images for CD8a (orange, A), GranzymeB (red, B), and CD20 (white, C) are shown on the left, with quantitative data on the right. Cell nuclei were stained by DAPI (blue). Targeting HSC PD-L1 led to higher densities of CD8a+ cells (A), GranzymeB+ cells (B), and CD20+ cells (C) in CCl₄-injured murine livers. ∗P < .05; ∗∗P < .01; ∗∗∗P < .001 by t-test, n=3 mice each group. Data of 6 individual mouse livers are also shown. Five to 7 AOIs, each containing more than 500 cells, were randomly selected from a mouse liver section for quantifying CD8a+ cells; 5 to 9 AOIs, each containing more than 1000 cells, were randomly selected from a mouse liver section for quantification of GranzymeB+ cells; and 6 to 9 AOIs, each containing more than 400 cells, were randomly selected from a mouse liver section for quantification of CD20+ cells. Bars, 100 μm. (D and E) Representative IF images of F4/80 (green, D) and CD11b (rose-red, E), and quantitative data are shown. Targeting HSC PD-L1 led to lower densities of F4/80+ cells (D) and CD11b+ cells (E) in CCl₄-injured murine livers. ∗P < .05 by t-test, n = 3 mice per group. Data of 6 individual mouse livers are also shown. Six to 9 AOIs, each containing more than 1600 cells, were randomly selected from a mouse liver section for quantification of F4/80+ cells, and 4 to 11 AOIs, each containing more than 350 cells, were randomly selected from a mouse liver section for quantification of CD11b+ cells. Bars, 100 μm.

Cd274/PD-L1 Transcripts Were Detected in Activated HSCs/myofibroblasts of CCl₄-injured Murine Liver by Single-cell Transcriptomics

PD-L1 transcripts were detected from primary human HSCs.¹⁴ To determine whether HSCs of CCl4-injured murine liver express PD-L1 transcripts, we leveraged publicly available single-cell RNA sequencing (scRNA seq) dataset GSM6507612 (series GSE212039 [RS045]), which was obtained by sequencing all cells of a liver of a mouse that received 8 times CCl₄ injection.^20,21 The majority of the cells were hepatocytes, and only 130 fibroblasts were identified by the R toolkit Seurat (Figure 5A). Additionally, 7 quiescent HSCs (Q-HSCs) and 68 activated HSCs/myofibroblasts (A-HSCs) were identified (their expression of quiescence markers and activation markers is shown in Figure 5B). Cd274 transcripts were detected from A-HSCs, but not from Q-HSCs (Figure 5C). Because only a small number of activated-HSCs had Cd274 transcripts in this dataset, we next analyzed GSM5257942 (Series GSE212047 [RS039]), which was obtained by scRNA seq of Lrat+ cells sorted from a liver of a mouse that received a profound fibrosis regimen of 19 times of CCl₄ injection.²¹ The cells were divided into 12 clusters, and none expressed Gfap transcripts (Figure 5D). As an example, we analyzed cells in clusters 2, 4, and 7, and found that the cells in the 3 clusters all expressed Cd274 transcripts (n = 29, 21, 17, respectively) (Figure 5E). Thus, Cd274/PD-L1 transcripts were detected in A-HSCs of CCl₄-injured murine liver at the single-cell resolution.

Figure 5.

Single-cell RNA sequencing (scRNA seq) detected Cd274/PD-L1 transcripts from activated-HSCs/myofibroblasts of CCl4-injured murine liver. (A) GSM6507612 dataset from the series GSE212039 (RS045) was downloaded and analyzed with an R toolkit Seurat. Hepatic cells were divided into 10 groups by the software. (B) Left, hepatic fibroblasts were further divided into 3 clusters. Right, expression of example marker genes by the 3 clusters is shown by a DotPlot. High levels of Gfap, Lrat, and Des transcripts mark Q-HSCs whereas high levels of Des, Col1a1, and Acta2 mark A-HSCs. (C) Two UMAPs revealing Cd274/PD-L1 transcripts in A-HSCs, but not in Q-HSCs. (D) GSM5257942 from the series GSE212047 (RS039) was analyzed by the R toolkit Seurat, and the murine Lrat+ fibroblasts were divided into 12 clusters with their expression of example marker genes shown by a dot-plot. (E) A violin plot revealing Cd274/PD-L1 transcripts in the cells of clusters 2, 4, 7 (n= 29, 21, and 17 cells).

Spatial Transcriptomics Revealed a Global Transcriptomic Change in Cd274-targeted HSCs/Myofibroblasts

Targeting PD-L1 led to alteration of the TGFβ transcriptome of human HSCs.¹⁴ To determine how PD-L1 targeting influenced the transcriptome of murine HSCs, spatial transcriptomics was performed with murine liver sections, which was completed with the NanoString GeoMx Digital Spatialomics Profiler (GeoMx DSP).^19,22 To that end, CCl₄-injured PD-L1^HSCKO and PD-L1^HSCWT liver sections were subjected to in situ hybridization to the GeoMx Mouse Whole Transcriptome Atlas Panel targeting 20,177 murine transcripts, followed by Desmin IF to label A-HSCs and CD45 IF to label immune cells.^19,22 In each AOI, Desmin+CD45- cells were designated as purple for transcriptomic profiling (Figure 6A top). Five AOIs from 3 PD-L1^HSCWT livers and 4 AOIs from 3 PD-L1^HSCKO livers were subjected to transcriptomic profiling, leading to 13,260 transcripts detected from the myofibroblasts of an AOI. A volcano plot revealed 6964 downregulated and 457 upregulated transcripts in PD-L1-deficient HSCs/myofibroblasts compared with control HSCs/myofibroblasts (Figure 6A lower) (the horizontal line indicates P = .05, and the vertical lines indicate a fold change of 2), supporting a global transcriptomic change induced by targeting PD-L1 of HSCs in mice.

Figure 6.

Targeting HSC PD-L1 induces a global transcriptomic change in the myofibroblasts of CCl₄-injured murine liver. (A) Upper, CCl₄-injured liver sections were subjected to spatial transcriptomics with the NanoString GeoMx Digital Spatial Profiler. A-HSCs were labeled by desmin IF (purple), and 2 representative AOIs selected for transcriptomic profiling are shown. Lower, a volcano plot showing 6964 downregulated and 457 upregulated transcripts as the result of targeting PD-L1 of HSCs in mice. The horizontal line indicates P < .05 and vertical lines show a fold change of 2. (B) GSEA with M2 pathways revealed the pathways of HSCs that were impacted by PD-L1 targeting based on NES >1 and P < .05. (C) A gene set related to collagen formation impacted by PD-L1 targeting is shown. The enrichment of gene transcripts is shown by an enrichment plot (left), and the levels of the transcripts are shown by a heatmap (right). P < .0001, n = 5 and 4. The bar represents the minimum (blue) to the maximum expression level (red).

As revealed by gene set enrichment analysis (GSEA) and the mouse molecular signatures database (M2 database), the transcriptomic changes induced by PD-L1 targeting were associated with numerous biological processes of the cells, such as ECM, inflammation, protein transport and modification, cell division and death, metabolism, intracellular signaling, and others (Figure 6B) (nominal P < .05; NES > 1). For example, 76 transcripts in COLLAGEN FORMATION (Figure 6C) and 83 transcripts in INFLAMMATORY RESPONSE and CHOLESTEROL UP (Figure 7A top) were affected by PD-L1 targeting. Twenty-six intracellular signaling pathways were also affected with the MCBRYAN PUBERTAL TGFB1 TARGETS UP and the JOHANSSON GLIMAGENESIS BY PDGFB UP on the top of the list followed by other pathways such as KRAS, hypoxia, interleukin (IL), hepatocyte growth factor receptor Met, Wingless/Int-1 (WNT), Hedgehog, fibroblast growth factor (FGF), G-protein coupled receptor (GPCR), and so on (Figure 6B and Figure 7A).

Figure 7.

Targeting HSC PD-L1 induces a global transcriptomic change in the myofibroblasts of CCl₄-injured murine liver and cultured human HSCs. (A) Three gene sets affected by targeting HSC PD-L1 are shown. The enrichment of transcripts is shown by the enrichment plots (left) and gene transcript levels are shown by heatmaps (right). P < .0001, n = 5 and 4. The bar represents the minimum (blue) to the maximum expression level (red). (B) A heatmap showing the transcripts encoding ECM in control and PD-L1 KO myofibroblasts as detected by spatial transcriptomics. P < .05 by ANOVA, n = 5 and 4. The bar represents the minimum (blue) to the maximum expression level (red). (C) A heatmap revealing the transcripts encoding ECM in control and PD-L1 knockdown human HSCs as detected by RNA sequencing (GSE167173). (D) A heatmap revealing the transcripts encoding receptors related to murine HSC activation in control and PD-L1 KO myofibroblasts as detected by spatial transcriptomics. P < .05 except Met by ANOVA, n = 5 and 4. (E) A heatmap revealing the transcripts encoding receptors related to HSC activation in control and PD-L1 knockdown human HSCs as detected by RNA sequencing (GSE167173).

Data Comparability in Spatial Transcriptomic and Bulk-cell RNA Sequencing Datasets

To validate the Spatial transcriptomic data, we manually extracted the transcripts of our interest from the dataset and organized them into 3 categories, transcripts of ECM, transcripts encoding plasma membrane receptors, and transcripts encoding cytokines/chemokines and adhesion molecules. Thirty different murine collagen transcripts were detected by spatial transcriptomics, and 19 of them were downregulated by PD-L1 targeting in HSCs/myofibroblasts (P < .05) (Figure 7B). We have previously obtained the TGFβ transcriptome of primary cultured human HSCs by RNA sequencing (GSE167173).¹⁴ Data comparison revealed that the counterparts of the 19 murine collagens were all detected from TGFβ1-activated human HSCs, with the majority of them downregulated by PD-L1 knockdown except those of COL4A5, COL6A1, COL6A2, COL13A, COL18A, COL24A (Figure 7C). Spatial transcriptomics also detected downregulation of transcripts of 18 different plasma membrane receptors related to HSC activation, including the receptor of FGFs, transforming growth factor β (TGFβ) (Tgfbr1 and Tgfbr2), platelet-derived growth factors (PDGFs), ILs, hepatocyte growth factor (HGF), leukemia inhibitory factor (LIF), or nerve growth factor (NGF), as well as Toll-like receptors (TLRs) (P < .05, Figure 7D). The transcripts of the receptors, except those of ILR13, were detected from activated human HSCs with those of FGFR1, FGFR2, FGFR3, TGFβR1, PDGFRB, IL6ST, MET, IL17RA, NGFR, TLR1, and TLR3 downregulated by PD-L1 knockdown (Figure 7E). As we previously reported, TGFβR1 transcripts were downregulated by PD-L1 targeting in cultured human HSCs,¹⁴ whereas TGFβR2 transcripts were not¹⁴ (Figure 7E). In addition, human HSCs expressed transcripts of IL11R, IL13RA, and IL21RA, which were all downregulated by PD-L1 knockdown (Figure 7E). Thus, the transcripts of collagens and receptors related to HSC activation were affected by PD-L1 targeting in both murine HSCs/myofibroblasts and TGFβ-activated human HSCs.

Cytokines/chemokines such as FGFs, PDGFs, TGFβ, granulocyte-macrophage colony-stimulating factor (GM-CSF), macrophage colony-stimulating factor (M-CSF), and vascular endothelial growth factors (VEGFs), and adhesion molecules such as intercellular adhesion molecules (ICAMs) and vascular cell adhesion molecule-1 (VCAM-1) are implicated in the pathogenesis of inflammatory diseases, fibrosis, cancer, and metabolic diseases.23, 24, 25, 26, 27, 28, 29 Some of them are particularly important, and they are required for monocyte proliferation, monocyte-to-macrophage differentiation, and monocyte/macrophage recruitment to sites of tissue injury.^23,30, 31, 32 Insulin-like growth factor-binding proteins (IGFBPs) are related to immune response and fibrosis development as well.33, 34, 35 Spatial transcriptomics detected the transcripts of 20 cytokines/chemokines expressed by murine HSCs/myofibroblasts, and 18 of them were downregulated by PD-L1 targeting (P < .05) (Figure 8A). Their counterparts were detected from TGFβ1-activated human HSCs except those of IGFBP1 (Figure 8B) and TGFβ1-activated human HSCs expressed IGFBP2 transcripts (Figure 8B). Both datasets reveal consistent downregulation of the transcripts of FGF2, FGF7, TGFBI, vascular endothelial growth factor A [VEGFA], IGF2, IGFBP3, IGFBP6, NGF, CSF1, and VCAM1 by PD-L1 targeting (Figure 8A and B).

Figure 8.

Targeting HSC PD-L1 alters cytokine/chemokine transcripts of HSCs and the cytokine repertoire of CCl₄-injured murine liver. (A) A heatmap revealing the transcripts encoding a panel of cytokines/chemokines in control and PD-L1 KO myofibroblasts as detected by spatial transcriptomics. P < .05 by ANOVA except Igfbp3 and Igfbp4, n = 5 and 4. (B) A heatmap revealing the transcripts encoding cytokines/chemokines in control and PD-L1 knockdown human HSCs as detected by RNA sequencing (GSE167173). (C) A heatmap revealing the transcripts encoding another panel of cytokines/chemokines in control and PD-L1 knockout myofibroblasts as detected by spatial transcriptomics. P < .05 by ANOVA except Ccl17, Ccl9, and Cxcl1, n = 5 and 4. (D) A heatmap revealing the transcripts encoding another panel of cytokines/chemokines in control and PD-L1 knockdown human HSCs as detected by RNA sequencing (GSE167173). (E) Liver lysates were subjected to cytokine profiling with a Proteome Profiler Mouse XL Cytokine Array kit (ARY028 R & D System). Forty-three prominent cytokines/chemokines detected with their names are shown on the bottom. (F) A bar graph revealing that 11 cytokines/chemokines were affected by targeting HSC PD-L1 in CCl₄-injured livers. ∗P < .05; ∗∗P < .01 by t-test. n = 4 and 4 mice.

Both CC and CXC chemokine ligands (CCLs and CXCLs) are chemoattractants inducing directional movement of leukocytes by interacting with their receptors on the target cells, crucial for recruiting monocytes/macrophages, neutrophils, eosinophils, and other effector cells from the blood to sites of infection.36, 37, 38 Proinflammatory ILs, such as IL1β and IL6, are critical for initiating the immune response by stimulating immune cell recruitment and activation, and increase vascular permeability.³⁹ Spatial transcriptomics detected 15 murine CC chemokine transcripts with 13 downregulated by PD-L1 targeting, 6 CXC chemokine with 5 downregulated, and 6 interleukin transcripts with all downregulated (P < .05) (Figure 8C). In TGFβ1-activated human HSCs, however, only 4 murine CC chemokine counterparts were detected with none downregulated by PD-L1 knockdown (Figure 8D). The counterparts of murine CXCL1, CXCL2, CXCL12, and CXCL16 transcripts were detected from TGFβ-activated human HSCs, and additionally, CXCL10 and CXCL11 transcripts were detected (Figure 8D). Of the 6 human CXC chemokine transcripts, 4 exhibited a trend of downregulation by PD-L1 knockdown (Figure 8D). The counterparts of the 6 murine IL transcripts were detected with 3 showing a trend of downregulation by PD-L1 knockdown in human HSCs (Figure 8D). Thus, except for CC chemokines, the majority of the cytokines/chemokines we analyzed were affected by PD-L1 targeting at the transcript level in both murine and human activated-HSCs.

Targeting HSC PD-L1 Alters the Cytokine Repertoire of CCl₄-injured Murine Liver

To determine how targeting HSC PD-L1 influenced the cytokine repertoire of murine liver, we used a targeted proteomics approach. To the end, tissue lysates of CCl₄-injured PD-L1^HSCKO and PD-L1^HSCWT livers were subjected to cytokine profiling with a Proteome Profiler Mouse XL Cytokine Array kit (R & D System, ARY028), which enabled simultaneous detection of 111 mouse cytokines/chemokines.²² We detected 92 cytokines/chemokines from each murine liver by the assay, and 43 of them presented at relatively high levels, including CCL16, CCL21, CXCL16, IL-15, IL-33, IL-7, IL-28A/B, ICAM-1, VCAM-1, P-selectin, IGFBP-1, IGFBP-2, IGFBP-6, matrix metalloproteinase-2 (MMP-2), MMP-9, angiopoietin-2, angiopoietin-like 3, HGF, adiponectin, pentraxin-2, C-reactive protein (CRP), chemerin, and so on (Figure 8E). The cytokines/chemokines detected are related to inflammation and fibrosis. For example, P-selectin is key for the initial recruitment of leukocytes to sites of tissue injury by binding to P-selectin glycoprotein ligand-1 (PSGL-1) expressed on hematopoietic cells, such as neutrophils, eosinophils, lymphocytes, and monocytes^40,41; angiopoietin-2 causes inflammation by promoting vascular leakage⁴²; MMPs contribute to tissue remodeling, inflammation, and fibrosis^43,44; Pentraxin 2 and CRP promote innate immune responses by binding to and influencing macrophages, monocytes, and neutrophils⁴⁵; HGF is an anti-inflammatory factor by inhibiting the nuclear factor kappa B (NFκB) signaling and pro-inflammatory cytokine expression.⁴⁶ Whereas adiponectin is known to increase insulin sensitivity and block inflammation, chemerin is another adipokine promoting inflammation and affecting insulin signaling, steroidogenesis, and thermogenesis.^47,48

Eleven liver cytokines/chemokines were identified as the targets of HSC PD-L1, as they were altered in CCl₄-injured PD-L1^HSCKO livers compared with CCl₄-injured PD-L1^HSCWT livers (red in Figure 8E and F). Adiponectin, chemerin, and Reg3G were upregulated in CCl₄-injured PD-L1^HSCKO livers compared with CCl₄-injured PD-L1^HSCWT livers (P < .05 or P = .05), whereas 8 others were downregulated (P < .05 or P = .05) (Figure 8F). The 11 cytokine/chemokine targets are associated with inflammation: pentraxin 2, CRP, and Reg3G are regulators of innate immunity; IGFBP1, IL-7, CRP, and proprotein convertase 9 (PCSK9) are proinflammatory, whereas FGF1, HGF, and adiponectin are antiinflammatory. Indeed, CRP and PCSK9 have been used in clinic as 2 indicators for accessing inflammation of patients. Thus, the cytokine repertoire of CCl₄-injured murine liver was altered by targeting PD-L1 of HSCs in mice.

Targeting PD-L1 Alters HSC Secretion of Cytokines/Chemokines in the Absence of TGFβ1 Stimulation

To determine the influence of PD-L1 targeting on HSC secretion of cytokines/chemokines, we pursued cytokine profiling of HSCs. As the effect of PD-L1 targeting on cytokine/chemokine transcripts was somewhat comparable in human and murine activated-HSCs, we performed cytokine profiling of primary human HSCs. To the end, conditioned medium (CM) of control HSCs and PD-L1 knockdown HSCs, with or without TGFβ1 stimulation, were collected for cytokine profiling by a Proteome Profiler Human XL Cytokine Array Kit (R & D Systems ARY022B), which enabled the detection of 105 human cytokines/chemokines. We detected 90 cytokines/chemokines from serum-starved human HSCs, and 34 of them presented at relatively high levels, including CCL2, CCL5, CCL7, CCL19, CXCL1, CXCL5, CXCL12, IL-6, IL-8, IL-17A, VEGFA, FGF-7, FGF-19, M-CSF, GM-CSF, macrophage migration inhibitory factor (MIF), angiogenin (stimulator of new blood vessels), angiopoietin-1, and so on (Figure 9A). MIF is a proinflammatory cytokine involved in various immune responses, a potential drug target for sepsis, rheumatoid arthritis, and cancer.^49,50

Figure 9.

Targeting PD-L1 alters cytokine/chemokine secretion of primary human HSCs. (A) CM of serum-starved HSCs were collected for cytokine profiling with a Proteome Profiler human XL Cytokine Array kit (ARY022B R & D System). Thirty-four prominent cytokines/chemokines detected and their names are shown on the bottom. (B) A bar graph revealing that 15 cytokines/chemokines were affected by targeting PD-L1 of HSCs. ∗P < .05; ∗∗P < .01; ∗∗∗P < .001; ∗∗∗∗P < .0001 by t-test. n = 4 and 4. (C) CM of TGFβ1-stimulated HSCs were collected for cytokine profiling; 34 prominent cytokines/chemokines were detected and their names are shown on the bottom. (D) A bar graph revealing that 14 cytokines/chemokines were altered by PD-L1 targeting. ∗P < .05; ∗∗P < .01; ∗∗∗P < .001; ∗∗∗∗P < .0001 by t-test. n = 4 and 4.

Fifteen cytokines/chemokines were identified as the targets of PD-L1 knockdown in serum-starved HSCs (red Figure 9A). CXCL1, CXCL5, CXCL12, CCL5, and CCL7 were downregulated by PD-L1 knockdown (Figure 9B) (P < .05). GM-CSF, affecting the function of both macrophages and eosinophils, and adhesin molecule VCAM-1 were also downregulated (Figure 9B) (P < .001). Growth/differentiation factor 15 (GDF-15), a cytokine known to protect mice from liver inflammation via metabolic reprogramming macrophages and skewing them towards anti-inflammatory M2-like phenotype,⁵¹ was the only one upregulated by PD-L1 targeting in serum-starved human HSCs (Figure 9B) (P < .05). In addition, HSC secretion of VEGFA, angiogenin, and angiopoietin-1 was suppressed by PD-L1 targeting (Figure 9B) (P < .05).

Targeting PD-L1 Alters TGFβ1-stimulated Cytokine/Chemokine Secretion From HSCs

TGFβ1 was enriched in CCl₄-injured murine liver. We next performed cytokine profiling of HSCs that were under TGFβ1 (5 ng/mL) stimulation for 24 hours, leading to the detection of 34 prominent cytokines/chemokines (Figure 9C). The cytokine/chemokine profile shared a very high degree similarity to that of unstimulated HSCs (Figure 9A and C). By comparing the cytokine profile of TGFβ1-stimulated control HSCs with that of TGFβ1-stimulated PD-L1 knockdown HSCs, 14 HSC cytokines/chemokines were identified as the targets of PD-L1 knockdown (red, Figure 9C and D) (P < .05). Consistent with the data of unstimulated HSCs, targeting PD-L1 suppressed secretion of CXCL5, CXCL12, and VCAM-1, and elevated secretion of GDF-15 by TGFβ1-stimulated HSCs (P < .05) (Figure 9C and D). In addition, it suppressed secretion of CXCL10 and elevated secretion of the following 9 cytokines/chemokines, IL-1, ICAM-1, osteopondin (OPN), chitinase-3-like protein 1 (CHI3L1), Dickkopf-related protein 1 (dkk-1), urokinase receptor (uPAR), CD147 (basigin or extracellular matrix metalloproteinase inducer), GM-CSF, and MMP-9 (P < .05) (Figure 9D). OPN is a chemoattractant for macrophages regulating their migration, survival, phagocytosis, and proinflammatory cytokine production⁵²; CHI3L1 is strongly upregulated in various inflammatory and immunological diseases, including cancers, Alzheimer’s disease, and atherosclerosis⁵³; dkk-1 protein inhibits the Wnt signaling⁵⁴; uPAR is a receptor of urokinase being a part of the plasminogen activation system; and soluble uPAR (suPAR) is a biomarker of inflammation.⁵⁵ Together with the data of unstimulated HSCs, we concluded that PD-L1 targeting-led elevation of GDF-15 secretion and suppression of CXCL5, CXCL12, VCAM-1 secretion may contribute in part to fewer F4/80+ cells and CD11b+ cells in CCl₄-injured PD-L1^HSCKO livers compared with CCl₄-injured PD-L1^HSCWT livers.

Targeting PD-L1 Alters HSC Secretion of Cytokines/Chemokines Via Transcriptional and Post-transcriptional Mechanisms

We have transcriptomic data for unstimulated HSCs and TGFβ1-stimulated HSCs (GSE167173),¹⁴ we therefore compared the transcripts of the cytokine/chemokine targets identified in this study (Figure 10A and B). For the 14 downregulated cytokines/chemokines identified from un-stimulated HSCs, 7 had downregulated transcripts in response to PD-L1 knockdown (CXCL1, CXCL12, VCAM1, VEGFA, angiopoietin-1 [ANGPT1], thrombospondin 1 [THBS1], and IL18BP) (P < .05) and 4 had unchanged transcripts (P > .05) (CXCL5, FGF-7, GM-CSF, and angiogenin), whereas 3 had upregulated transcripts, although the increases did not reach statistical significance (LIF, CCL5, and CCL7) (P > .05) (Figure 10A). The transcripts of GDF-15 were upregulated upon PD-L1 knockdown, consistent with the result of cytokine profiling (P < .05) (Figure 10A).

Figure 10.

Transcriptional and posttranscriptional mechanisms are involved in cytokine/chemokine secretion of PD-L1 knockdown HSCs. (A) Left, the transcripts of 15 cytokines/chemokines identified from unstimulated HSCs were extracted from GSE167173 with their expression levels shown by a heatmap. Right, 3 bar graphs revealing 7 downregulated, 1 upregulated, and 7 unchanged transcripts in serum-starved PD-L1 knockdown HSCs compared with control HSCs. ∗P < .05; ∗∗P < .01; ∗∗∗P < .001; ∗∗∗∗P < .0001 by t-test. n = 3 and 3. (B) Left, 13 transcripts of the 14 cytokines/chemokines identified from TGFβ1-stimulated HSCs were extracted from GSE167173 with their expression shown in a heatmap. Right, 3 bar graphs revealing 3 downregulated, 3 upregulated, and 7 unchanged transcripts in response to PD-L1 targeting in TGFβ1-stimulated human HSCs. ∗P < .05; ∗∗P < .01 by t-test. n = 3 and 3.

For the 14 cytokine/chemokine targets identified in TGFβ1-stimulated HSCs, their transcripts were successfully extracted, except those of uPAR. Among the 4 cytokines/chemokines with their secretion downregulated by PD-L1 knockdown, 3 had downregulated transcripts (CXCL10, CXCL12, and VCAM1) (P < .05) (Figure 10B), and 1 had unchanged transcripts (CXCL5) (P > .05) (Figure 10B). Among the 9 cytokines/chemokines with their secretion upregulated by PD-L1 knockdown, 3 had upregulated transcripts (SPP1/OPN, CHI3L1 and MMP9) (P < .05), 4 had unchanged transcripts (IL-11, ICAM1, DKK1, CD147) (P > .05), and the transcripts of GDF-15 and GM-CSF had a trend of upregulation, although the changes did not reach statistical significance (P > .05) (Figure 10B). Together, these data suggest that targeting HSC PD-L1 may influence the baseline cytokine/chemokine secretion and TGFβ1-stimulated cytokine/chemokine secretion of HSCs by transcriptional and post-transcriptional mechanisms.

HSC-derived Cytokines/Chemokines Regulate the Chemotaxis and Proliferation of Monocytes/Macrophages

To obtain evidence supporting the function of HSC-secreted cytokines/chemokines, we pursued in vitro studies to test whether: (1) HSC-derived THBS1, CXCL1, and CXCL5 were chemoattractants for monocytes/macrophages; and (2) HSC-derived GM-CSF was a mitogen for monocytes/macrophages. We studied the 4 cytokines/chemokines for the following reasons: (1) the anti-inflammatory role of GDF15 on macrophages has recently been demonstrated⁵¹; (2) THBS1 is involved in liver inflammation and fibrosis, particularly in the conditions like acute-on-chronic liver failure and fatty liver disease, but its mechanism of action has not been fully elucidated^56,57; (3) secretion of high level of CXCL1 and CXCL5 by HSCs was downregulated by PD-L1 knockdown in un-stimulated HSCs (at an intermediate activation state); and (4) secretion of GM-CSF by un-stimulated HSCs was drastically suppressed by PD-L1 knockdown. We collected CMs of HSCs and used them in the experiments, as shown below, to test our hypotheses.

Control human HSCs (expressing nontargeting [NT] short hairpin RNA [shRNA]), THBS-1 knockdown human HSCs (expressing THBS1 shRNA), CXCL1 knockdown human HSCs (expressing CXCL1 shRNA), CXCL5 knockdown human HSCs (expressing CXCL5 shRNA), and GM-CSF knockdown human HSCs (expressing CSF2 shRNA) were generated by lentiviral transduction. Gene knockdown by shRNA lentiviruses in HSCs was confirmed by WB or IF (Figure 11A). Because both monocyte-derived macrophages and Kupffer cells contributed to the F4/80+ cells accumulated in CCl₄-injured murine livers, the role of HSC CM on chemotaxis of human CD14+ blood monocytes was assessed by Boyden chamber assay. We found that the CM of control HSCs promoted chemotaxis of human monocytes compared with basal medium, and this effect was suppressed by knocking down THBS1, CXCL1, or CXCL5 of HSCs (Figure 11B) (P < .0001). Human CD14+ monocytes were next stimulated with M-CSF to induce monocyte-to-macrophage differentiation (50 ng/mL for 4 days and 100 ng/mL for additional 4 days) (Figure 11C). Boyden chamber assay for chemotaxis of monocyte-derived macrophages led to similar results (Figure 11D) (P < .001). Moreover, we found that the CM of control human HSCs promoted proliferation of human CD14+ monocytes in vitro and that the effect was suppressed by knocking down GM-CSF of HSCs (Figure 11E) (P < .0001). Thus, HSCs can regulate chemotaxis and proliferation of monocytes/macrophages via secreting cytokines/chemokines.

Figure 11.

HSCs promote migration and proliferation of monocytes/macrophages by secreting cytokines/chemokines. (A) Left, HSCs were transduced with shRNA lentiviruses and knockdown of THBS1 by shRNA lentiviruses was detected by WB. Right, IF revealed that CXCL1, CXCL5, or GM-CSF was effectively knocked down by shRNA lentiviruses in HSCs. Bars, 20 μm. (B) CD14+ human monocytes were subjected to Boyden chamber assay with CM of HSCs as a stimulant. CM of control HSCs promoted the chemotaxis of monocytes compared with basal medium, and this effect of CM was suppressed by knocking down THBS1, CXCL1, or CXCL5 of HSCs. ∗∗∗∗P < .0001 by ANOVA, n=6. Data represent 3 independent repeats with similar results. (C) A protocol inducing monocyte-to-macrophage differentiation by M-CSF in vitro is shown (4-day-incubation with 50 ng/mL M-CSF followed by another 4-day-incubation with 100 ng/mL M-CSF). Bars, 100 μm. (D) Boyden chamber assay was performed for the chemotaxis of monocyte-derived macrophages. Migrated cells were stained by DAPI, and representative pictures are shown on the top. Bar, 100 μm. The CM of control HSCs promoted the chemotaxis of macrophages, and this effect of CM was suppressed by knocking down THBS1, CXCL1, or CXCL5 of HSCs. ∗∗∗P < .001; ∗∗∗∗P < .0001 by ANOVA, n = 6. Data represent 3 independent repeats with similar results. (E) CD14+ human monocytes were subjected to MTS proliferation assay with the CM of HSCs as a stimulant. The CM of control HSCs promoted monocyte proliferation was compared with basal medium, and this effect of CM was suppressed by knocking down GM-CSF of HSCs. ∗∗∗∗P < .0001 by ANOVA, n = 6. Data represent 3 independent repeats with similar results.

Discussion

It has been proposed that the immune-privileged status of the liver is contributed in part by HSC PD-L1-mediated suppression of T cells and B cells of the liver.^58,59 Our prior data demonstrated an immune-independent function of PD-L1 whereby TGFβ1-stimulated PD-L1 expression of HSCs promotes the TGFβ signaling and activation of HSCs into tumor-promoting myofibroblasts in vitro and in an experimental liver metastasis mouse model.¹⁴ In parallel to HSC PD-L1, the fibroblasts of the lung also express PD-L1 critical for the TGFβ signaling, invasion of the fibroblast in vitro, and lung fibrosis in mice.^60,61 We therefore tested whether targeting HSC PD-L1 influenced liver fibrosis and inflammation in response to CCl₄-induced liver injury in this study. We found that targeting HSC PD-L1 indeed suppressed accumulation of Kupffer cells and myeloid cells in the liver and in liver fibrosis induced by CCl₄ in mice. A targeted proteomics revealed that HSC secretion of a panel of cytokines/chemokines was altered by PD-L1 targeting, which may explain, in part, why the Kupffer cell and myeloid cell densities were lower in HSC-specific PD-L1 KO livers compared to control livers. Multiomic data suggested an involvement of transcriptional and post-transcriptional mechanisms in the alteration of cytokine/chemokine secretion of PD-L1 knockdown HSCs. Together, our data support a critical role of HSC PD-L1 in liver inflammation and liver fibrosis in response to chronic liver injury.

Although the role of HSCs in liver fibrosis is well-studied and recognized, the role of HSCs in liver inflammation is incompletely understood, and whether HSCs are proinflammatory or antiinflammatory remains a matter of debate.⁶² Studies by Yu et al and Li et al supported that HSCs protect liver parenchyma from the infiltration of T cells and B cells via their expression of PD-L1.^58,59 In contrast, Reifat found that more than 25% of CD4+ T cells were colocalized with HSCs in murine liver following hepatic ischemia/reperfusion and that deactivation of HSCs with a pharmacologic reagent JWH-133 attenuated CD4+ T cell recruitment and reduced liver injury compared with control mice, suggesting that HSCs may facilitate CD4+ T cell infiltration and liver injury associated with liver ischemia/reperfusion.⁶³ Here, we show that targeting HSC-specific PD-L1 by a genetic approach led to higher densities of CD8+ T cells, GranzymeB+ T and NK cells, and CD20+ B cells in CCl₄-injured livers (Figure 4), supporting a suppressive role of HSC PD-L1 for lymphocyte activation in the liver.

Analyzing our bulk-cell RNA sequencing data revealed that the majority of the transcripts of plasma membrane receptors related to HSC activation were downregulated by PD-L1 knockdown in TGFβ1-activated human HSCs, similar to those in activated-HSCs/myofibroblasts of murine liver (Figure 7D and E). Interestingly, targeting PD-L1 induced downregulation of PDGFRα transcripts in murine HSCs/myofibroblasts, but not in activated human HSCs, which was in sharp contrast to PDGFRβ transcripts, which were downregulated in both (Figure 7D and E). The same discrepancy occurred for TGFβR1 and TGFβR2 transcripts. IF and WB confirmed the downregulation of PDGFRα in CCl₄-injured PD-L1^HSCKO livers compared with CCl₄-injured PD-L1^HSCWT murine livers (Figure 2), indicating that the biology of PDGFRα (and TGFβR2) of HSCs could be influenced by in vitro cell culture conditions. The HSCs in culture received TGFβ1 stimulation, whereas the murine HSCs in CCl₄-injured livers had activation in numerous signaling pathways, such as TGFβ1 signaling, FGF signaling, PDGF signaling, IL signaling, and TLR-mediated signaling, which acted together to form complex signaling networks to modulate gene transcription. Moreover, cultured human HSCs were plated on plastic substrate in contrast to murine HSCs in a soft liver, and the substrate stiffness is known to alter HSC biology and phenotype by a p300-dependent epigenetic mechanism.¹⁶

The immune checkpoint inhibitors have revolutionized cancer therapy; however, they can cause aberrant immune activation affecting the liver in 5% to 30% of patients, depending on the agent(s) used and underlying factors, termed as immune-mediated liver injury from checkpoint inhibitors.^64,65 Various lymphocyte populations, cytokines, and the secondary activation of the innate immune system are believed to be responsible for the liver injury.⁶⁴ Lobular or centrilobular (zone 2 or 3) hepatocellular injury pattern was the common histologic feature of immune checkpoint inhibitor hepatitis, although other liver injury patterns presented^66,67 (the cholangitic liver injury pattern was associated with competing causes or concomitant chemotherapy⁶⁷), similar to the hepatocellular damage caused by CCl₄.^68,69 However, its inflammation was attributed to the lobular accumulation of mononuclear cells with or without eosinophils,⁶⁷ different from the portal inflammation induced by CCl₄. As cytotoxic T lymphocytes also contributes to the hepatotoxicity,^64,70 we analyzed localization of CD8a+ T cells in the CCl₄-injured murine livers and categorized them into 2 groups: portal tract CD8 T cells (associated with the portal tract, white arrowheads in Figure 12, left), and lobular CD8 T cells (not associated with the portal tract, white arrows in Figure 12, left). We found that the infiltration of CD8 T cells into the lobules was actually impaired in CCl₄-injured PD-L1^HSCKO livers compared with CCl₄-injured PD-L1^HSCWT livers (Figure 12), despite the fact that there were more CD8 T cells in the CCl₄-injured PD-L1^HSCKO livers. Hepatocytes, sinusoidal endothelial cells, Kupffer cells, and dendritic cells are other liver resident cells that express PD-L1,71, 72, 73, 74 so targeting PD-L1 of HSCs could only result in a small PD-L1 protein reduction in the murine liver compared with the patients who received anti-PD-L1 or anti-PD1 immunotherapy. As cell/cell and cell/stroma interplays are major determinants of liver injury,^75,76 the altered liver cytokine repertoire in PD-L1^HSCKO livers also contributed to the phenotype of why liver injury was not aggravated by targeting HSC PD-L1 of the mice.

Figure 12.

Localization of CD8 T cells in CCl₄-injured murine livers.Left, representative CD8a IF pictures showing CD8 T cells associated with the portal tract (arrowheads; portal tract CD8 T cells), and cells not associated with the portal tract (arrows; lobular CD8 T cells). Trichrome staining pictures are shown on the bottom. Right, quantitative data revealing more portal tract CD8 T cells in CCl₄-injured PD-L1^HSCKO livers compared with CCl₄-injured PD-L1^HSCWT livers. ∗P < .05 by t-test, n = 3 and 3 mice. Data of 6 individual mice are also shown. Bars, 100 μm.

The bidirectional interactions between macrophages and HSCs contribute to fibrosis initiation and progression.^75,76 Kupffer cell activation is regarded as an initial event in response to liver injury, and Kupffer cell-derived cytokines, such as TGFβ1, act on HSCs to induce their activation. Seki et al showed that CM of HSCs promoted Kupffer cell migration in vitro, and HSC expression of mRNA of Cxcl1, Cxcl2, Cxcl10, Ccl2, Ccl3, and Ccl4 was enhanced by LPS.⁷⁷ In this study, we found that targeting HSC PD-L1 in mice suppressed CCl₄-induced recruitment of macrophages and leukocytes onto the liver fibrotic septa, and that HSC secretion of cytokines/chemokines was altered by PD-L1 targeting in cultured human HSCs. The elevation of HSC secretion of GDF-15 by PD-L1 targeting may contribute to the macrophage phenotype of the mice, as GDF-15 is known to suppress macrophage function and liver fibrosis.⁵¹ Our in vitro studies revealed that promotion of the chemotaxis of monocytes/macrophages by HSC CM was suppressed by knocking down CXCL1, CXCL5, or THBS1 of HSCs (Figure 11B and D), and that promotion of monocyte proliferation by HSC CM was suppressed by targeting GM-CSF of HSCs (Figure 11E), supporting that monocyte/macrophage behaviors are indeed regulated by HSC-derived cytokines/chemokines.

Because the F4/80+ cells accumulated in CCl₄-injured murine liver derived from both monocytes of the blood and Kupffer cells, we obtained primary human Kupffer cells commercially available for the assays. We did not get results because the viability of the cells was very poor, and the cells could not attach, migrate, and grow in culture. We did not obtain proliferation data of the monocyte-derived macrophages either possibly because they were at a terminally differentiated state. Our future studies could include induced pluripotent stem cell (iPSC)-derived human Kupffer cells as they express Kupffer cell specific markers and cytokines, representing a renewable cell source for human Kupffer cells, which may help us overcome the issues of using primary human Kupffer cells in mechanistic studies, such as the high cost and difficulty in culture.⁷⁸

In summary, we found that targeting HSC PD-L1 altered liver inflammation and suppressed HSC activation and liver fibrosis in a CCl₄ injury mouse model, whereas it did not exacerbate CCl₄-mediated liver injury. In addition to our previously identified mechanism that targeting PD-L1 of HSCs led to the defective TGFβ signaling hampering HSC activation,¹⁴ this study demonstrated that targeting PD-L1 altered HSC secretion of cytokines/chemokines so as to suppress HSC/Kupffer interactions required for the initiation and amplification of a signaling loop that promotes HSC activation and liver fibrosis. HSC PD-L1 therefore represents a target to inhibit HSC activation and liver fibrosis.

Materials and Methods

Mice and CCl₄ Injection

The Cd274/PD-L1 floxed mutant mouse line was originally created by Dr Arlene H. Sharpe’s laboratory.¹⁵ It was crossed to a Collagen1A1-Cre (Col1A1Cre) transgenic mouse line^16,17 in our laboratory to create HSC-specific PD-L1 knockout mice for liver metastasis study of colorectal cancer.¹⁴

The protocol of CCl₄ injection into mice was approved by the Institutional Animal Care and Use Committee of University of Minnesota. To generate HSC-specific PD-L1 knockout mice for CCl₄ injection, Cd274/PD-L1 floxed mutant mice were bred to Col1A1Cre transgenic mice, and littermate-matched PD-L1^HSCKOand PD-L1^HSCWTmice (control mice) were identified by PCR-based genotyping. Eight age-matched male PD-L1^HSCWT mice (control) and 11 male PD-L1^HSCKO mice were subjected to intraperitoneal injection of CCl₄ (Sigma-Aldrich #319961) twice a week for 6 weeks (1 μL/g of body weight diluted in oil). An additional 3 mice in each genotype were subjected to oil injection only as the controls. All mice were sacrificed 48 hours after the last CCl₄ injection, and their livers were weighted and isolated for WB analysis, cryosectioning for IF staining, and histopathology.

Facial Vein Blood Collection

Blood collection from mouse facial vein (FV) was approved by the Institutional Animal Care and Use Committee of the University of Minnesota, which was performed before mice were sacrificed. Blood collection was done by poking the rear end of the jaw bone of a mouse with a 4-mm sharp blade followed by collecting 0.4 mL of blood into an Eppendorf tube. Samples were allowed to clot at room temperature for 30 minutes followed by cooling at 4°C for 30 minutes. After centrifugating at 14,000 g for 15 minutes, 200 μL of serum were collected from each sample and sent for measurement of liver enzymes, such as AST and ALT, which was done by the Clinical Pathology Laboratory at the Veterinary Medical Center, University of Minnesota.

Hematoxylin and Eosin Staining, Picrosirius Red Staining, and Trichrome Staining

Formalin-fixed liver tissues were sent for embedding with paraffin so paraffin-embedded formalin-fixed sections were made for hematoxylin and eosin (H&E) staining, Picrosirius Red staining, and Trichrome staining, which were done by the service provided by the Mayo Clinic Laboratories. The images of Trichrome staining were obtained under a Zeiss AxioLab 5 microscope with a 20× lens and the Zen 2.3 lite software (Zeiss). The images of Picrosirius Red staining were obtained under a Zeiss Axio observer with Apotome with a 10× lens and the Zen 2.3 lite software.^19,20 The areas stained blue or red in the pictures were quantitated with the ImageJ Software (National Institutes of Health). To the end, 5 to 10 microscopic fields (AOIs) were randomly selected from a mouse liver section for densitometry analysis with the ImageJ software so as to obtain the average value of blue or red for the mouse. N = 8 mice in one group and 10 in another group were subjected to analysis and compiled data were used to make dot-plots and draw conclusions.

WB Analysis and Antibodies

The lysates of murine liver were made with a RIPA buffer, containing 1% Nonidet P-40, 1% sodium deoxycholate, and 0.1% SDS.⁷⁹ The buffer also contained PMSF, NaF, Na3VO4 and protease inhibitor cocktail (11873580001 Lot #45868600, Roche). After the tissues were homogenized with a Bio-Gen electronic PRO200 homogenizer (Pro-PK-01200S PRO Scientific), they were cleared by centrifugation at 15,000 rpm for 5 minutes followed by protein quantification with the DC Protein Assay kit (5000111 Lot# 64561846, Bio-Rad). Thirty-five μg of total protein of a liver lysate were loaded into a well of a gel for electrophoresis. After proteins were transferred onto a nitrocellulose membrane (GE10600079, Lot# A29548300, MilliporeSigma), the membrane was incubated with a primary antibody diluted in 5% nonfat milk followed by incubation at 4°C overnight. It was next incubated with a secondary antibody conjugated with horseradish peroxidase (HRP) (GENA931, or GENA934, MilliporeSigma) for 1 hour at room temperature on the following day. The signals were developed by incubating the blot with a chemiluminescence substrate (Immobilon Forte, WBLUF0500, Lot# 242155. MilliporeSigma) and obtained by a ChemiDoc MP Imaging System (Bio-Rad).

Primary antibodies used were: anti-PD-L1 (13684 Cell Signaling Technology); anti-type 1 collagen (1310-01 Lot# K0721-M222, SouthernBiotech), anti-αSMA (ab5694 Lot# GR283004-43, Abcam), anti-desmin (Y66) (ab32362 Lot# 1012085-33, Abcam), anti-PDGFRα (3164 Lot# 5, Cell Signaling Technology), anti-GAPDH (2118 Lot#16, Cell Signaling Technology), anti-Cre (D7L7L Cell Signaling Technology), and anti-THBS-1 (Cat# 37879, Lot# 1, Cell Signaling Technology).

IF Staining, Microscopy, and Antibodies

Frozen liver tissues were sectioned at 7 μm, followed by air-drying overnight at room temperature. For IF, the sections were first fixed with 4% paraformaldehyde (UN2213 Lot# Bo136490, MilliporeSigma) for 10 minutes and blocked with 10% goat serum to mask the nonspecific antibody binding sites. The sections were then incubated with a primary antibody at 4°C overnight, followed by incubation with Alexa Fluor-conjugated second antibody and DAPI on the following day. Lastly, the sections were mounted in anti-fading mounting medium (P36934 Lot# 1474820, SlowFade Gold Antifade Mountant, Thermo Fisher Scientific) for microscopy, which was done under a Zeiss Axio observer with a 20× lens and the Zen 2.3 lite software.^80,81 The αSMA antibody was from Abcam (5964, Lot# GR283004-43), and the PDGFRα antibody was from Cell Signaling Technology (3164, Lot# 5). For IF quantification, numerous AOIs were randomly selected from a liver section, and the ImageJ software was used to calculate “Integrated IF Density” so that the average IF density of αSMA or PDGFRα was obtain for the mouse. N = 8 mice in one group and 10 in another group were subjected to analysis, and combined data were used to make dot-plots and draw conclusions.

For αSMA IF on murine HSCs, murine HSCs were isolated from PD-L1^HSCKO and PD-L1^HSCWT mice for in vitro culture using the protocol as we described previously.^16,19 Murine HSCs in culture were fixed at day 5 with 4% paraformaldehyde for 10 minutes followed by permeabilization with 0.02% Triton X-100 for 3 minutes. The cells were then proceeded to blocking, antibody incubation, and mounting procedures as described above. αSMA antibody was from MilliporeSigma (A5228 Lot# 0000083803; 1:4,000 dilution), and cell nuclei were counterstained by DAPI. Confocal microscopy was done under a Zeiss LSM 900 with Airyscan 2 with a 40× lens and the Zen 2.3 lite software.^19,20,22 For data analysis, more than 10 microscopic fields were randomly selected from each cell group for quantitating αSMA+ HSCs so as to obtain the average rate of activated-HSCs/myofibroblasts for the 2 cell groups. Conclusion was drawn based on 3 independent repeats with similar results.

Other antibodies used were: anti-GM-CSF (sc-32753, Lot# H2324, Santa Cruz Biotechnology); anti-ENA-78 (CXCL5) (sc-377026, Lot# H0116, Santa Cruz Biotechnology); and anti-GROα (CXCL1) (sc-514065 Lot# C1923, Santa Cruz Biotechnology).

Multiplex IF Labeling and Imaging, and Antibodies

Multiplex IF labeling and imaging of liver sections were completed by the service provided by Visikol, Inc. Formalin-fixed liver sections of CCl₄-injured mice were subjected to deparaffinization (StatLab, 7400-1), rehydration, and antigen retrieval, which was done by boiling the sections in citrate buffer (Sigma, C9999) for 20 minutes in a pressure cooker. Blocking unspecific antigen binding sites was performed by incubating the sections with 6% donkey serum (Innovative Research, IGDNSER500ML) containing 0.3M glycine (Fisher Scientific, 212485) for 1 hour at room temperature. Incubation with a primary antibody panel was done at 4°C overnight (see Table 1 for panels and dilution details), followed by incubation with the secondary antibodies for one hour at room temperature. Cell nuclei were stained with DAPI (Invitrogen, D3571), and the sections were sealed with fluoro gel and Tris-buffer (Electron Microscopy Sciences, 17985-10) for imaging. Antibodies stripping was performed with Visikol’s EasyPlex solution at 56°C, followed by 3 washes, and the procedure was repeated after each panel staining. IF imaging was performed with a Zeiss Axioscan 7 with a 20× lens (0.345 μM/pixel), and the images were coregistered by nuclear staining. Intensity corrections were performed using a Gaussian blur (sigma = 1000) for background subtraction. To analyze cell density of immune cells, numerous AOIs were randomly selected from each liver section for quantifying CD8a+, CD20+, GranzymeB+, F4/80+, or CD11b+ cells, as well as the total number of cells based on DAPI staining so as to obtain the average cell density for each of immune cell types for the mouse. The liver sections of 3 individual mice per group were subjected to analysis, and combined data of 6 mice were used to make the dot-plots and draw conclusions. To analyze CD8 T cell localization, CD8a+ cells associated with the immune cells of the portal tract were defined as portal tract CD8 T cells, and in contrast, CD8a+ cells not associated with the portal tract were defined as lobular CD8 T cells. The liver sections of 3 individual mice per group were subjected to analysis, and combined data of 6 mice were used to make the dot-plots and draw conclusions.

Table 1.

Antibodies Used for Multiplex IF Labeling and Imaging

Panel	Primary antibody	Catalog #	Dilution	Secondary antibody	Catalog #	Dilution
1	CD4	GTX44529	1:100	Donkey αRat 488	A21208	1:1000
1	Granzyme B	AF1865	1:100	Donkey αGoat 594	A11058	1:1000
1	CD20	ab64088	1:100	Donkey αRabbit 647	A31573	1:1000
2	CD3	ab11089	1:100	Donkey αRat 488	A21208	1:500
2	CD8a	ab209775	1:250	Donkey αRabbit 647	A31573	1:500
3	α-SMA	NB300-978	1:100	Donkey αGoat 594	A11058	1:1000
3	CD14	1700-1-AP	1:200	Donkey αRabbit 647	A31573	1:1000
4	CD11b	NB110-89474AF594	1:100	N/A	N/A	N/A
4	CD27	AF574	1:100	Donkey αGoat 647	A21447	1:500
5	F4-80	#70076	1:500	Donkey αRabbit 594	A21207	1:500
5	Pan- cytokeratin	sc-8018	1:500	Donkey αMouse 647	A31571	1:500

IF, immunofluorescence.

Spatial Transcriptomics With the NanoString GeoMx Digital Spatial Profiler (GeoMx DSP)

Cryosections of murine liver were prepared according to the Slide Preparation User Manual (MAN-10150-01). After fixed with 10% buffered formalin phosphate (SF100-4 Fisher Chemicals) overnight, they were sent to the University of Minnesota Genomics Center for Spatial transcriptomics with GeoMx DSP.^19,22 Antigen retrieval was done by boiling the slides in 1x Tris-EDTA (pH 9.0) for 1 hour followed by Proteinase K digestion (1 μg/mL) at 37°C for 5 minutes to expose RNA targets. In situ hybridization was done by incubating the sections with the GeoMx Mouse Whole Transcriptome Atlas Panel (20176 target probes and 210 negative probes) overnight. On the following day, Desmin IF (ab32362 Abcam) was performed to label activated-HSCs , CD45 IF (NovusBio NBP1-44763) was performed to label immune cells, and SYTO 13 was used to stain cell nuclei. IF signals were acquired by the GeoMx instrument for AOI selection from 3 control mouse livers and 3 HSC-specific KO mouse livers. Ultraviolet light was applied to cleave tags of the probes bound to the cells so the tags were used for library preparation and sequencing, which was done by Illumina NextSeq 2000. FASTQ files were converted into probe counts for each gene transcript, and the data were analyzed by the GeoMx DSP Analysis Suite software.^19,22 Lastly, the data were exported as an Excel file for further bioinformatic analysis. The data are in the Gene Expression Omnibus (GEO) (GSE279640).

Analysis of Omic Datasets With Bioinformatics

Spatial transcriptomic dataset (GSE279640) and RNA sequencing dataset (GSE167173) were analyzed with the GSEA online platform with M2 or C2 curated gene sets.^19,20,22 Enriched pathways were identified based on normalized enrichment scores (NES) from the GSEA results, with significance determined by P value < .05. Enrichment plots were generated to visualize the findings, and the genes significantly enriched in the selected pathways were chosen for heatmap visualization. All heatmaps were created using the online platform Morpheus (https://software.broadinstitute.org/morpheus/). Pathway bar graphs were created with Microsoft Excel (v16.82), and volcano plots were generated with GraphPad Prism (version 10) based on fold change and false discovery rate (FDR) from the RNA sequencing data comparison.^19,20,22

Single-cell RNA (scRNA) transcriptomic datasets GSM6507612 (series GSE212039 [RS045]) and GSM5257942 (Series GSE212047 [RS039])²¹ were downloaded from GEO. They were analyzed with an R toolkit Seurat package (V5) with a series of steps of normalization, identification of variable genes, and scaling. Principal component analysis (PCA) was followed to extract key features, and the cells were separated with the nearest neighbor and Louvain clustering algorithms,^20,82 and uniform manifold approximation and projection (UMAP) was create to visualize cell clusters. We next annotated cells with a focus on fibroblast subpopulations using the SingleR package and the MouseRNAseq database from Celldex, and the data were integrated into Seurat object’s metadata.^20,83 Expression of the transcripts of key genes such as Lrat, Des, Gfap, Acta2, Col1a1, and Cd274 in the fibroblast subgroups were further characterized to study their differential expression patterns.^20,83,84

Cell Culture and Gene Knockdown

Primary human HSCs were obtained from ScienCell Research Laboratories (#5300, ScienCell Research Laboratories) and cultured in Dulbecco's Modified Eagle Medium (DMEM) (SH30081; Cytiva) containing 10% fetal bovine serum (FBS; 35-011-CV Lot# 18120001, Corning), penicillin, and streptomycin. Cells with passage <9 were used for experiments. Human CD14+ blood monocytes were purchased from Lifeline Cell Technology (LL-0080 Lot# 08920) and cultured with RPMI 1640 medium (LM-0025, Lot# 11825, Lifeline Cell Technology). Human peripheral blood mononuclear cells (PBMCs) were obtained from American Tissue and Cell Culture (ATCC, PCS-800-011) and cultured with RPMI 1640 medium supplemented with 10% FBS as well. To induce monocyte-to-macrophage differentiation, CD14+ human monocytes or PBMCs were incubated with M-CSF (216-MCC, Lot# PUJ1123101, R&D Systems) for 8 days (50 ng/mL for 4 days and 100 ng/mL for an additional 4 days).^85,86

The following shRNA lentiviral constructs were purchased from MilliporeSigma: CD274 shRNA construct (TRCN0000056915);¹⁴ THBS1 shRNA constructs (TRCN0000226402 and TRCN0000226403); CXCL5 shRNA constructs (TRCN0000312612 and TRCN0000057935); CXCL1 shRNA constructs (TRCN0000057941 and TRCN0000057939); CSF2 shRNA constructs (TRCN0000371974 and TRCN0000058431). A construct encoding NT shRNA was used as the control (SHC202 MilliporeSigma). Lentiviral packaging was done as we did previously.^17,87,88 In brief, 2 μg of a shRNA lentiviral construct were mixed with 2 plasmids encoding the viral packaging elements, p8.91 (1.5 μg) and pMD.G (0.5 μg), followed by cotransfection into HEK293T cells in a 100-mm dish (80% confluence) by Effectene Transfection Reagent (#301425 Lot# 178012326, Qiagen). Cell medium containing viruses was collected 48 hours and 72 hours later and cleared by filtration with a 45-μm syringe filter. After aliquoted, viruses were stored at −80°C. To transduce HSCs, 50% of a virus-containing medium (1:1 dilution with complete DMEM medium containing 8 μg/mL polybrene) was added to HSCs, followed by overnight incubation at 37°C. HSCs were collected 72 hours later for further analyses.

Cytokine/Chemokine Profiling of Cultured HSCs and Murine Liver

HSCs were transduced with NT shRNA or PD-L1 shRNA lentiviruses, and HSC CM was collected for cytokine/chemokine profiling with a Proteome Profiler human XL Cytokine Array kit (ARY022B R & D Systems), according to the manufacturer-recommended protocol.²² Briefly, 1 mL of CM was mixed with 15 μL of a cocktail of biotinylated antibodies recognizing 105 human cytokines/chemokines followed by incubation for 2 hours at troom temperature. The antibody/CM mixture was then added onto a nitrocellulose membrane spotted with 105 different human cytokine/chemokine antibodies followed by incubation at 4°C overnight. After washing off unbound antibodies and proteins, streptavidin–HRP and Chemi substrate were applied to detect chemiluminescence signals, which were done by a ChemiDoc MP Imaging System (Bio-Rad).²² Signals were quantitated by the ImageJ software.

To study the cytokine repertoire of murine liver, liver tissues were lysed with the RIPA buffer containing 1% Nonidet P-40, 1% sodium deoxycholate, and 0.1% SDS. The buffer also contained PMSF, NaF, Na₃VO₄ and a protease inhibitor cocktail. After homogenization with a Bio-Gen electronic PRO200 homogenizer, the lysates were cleared by centrifugation at 15,000 rpm for 5 minutes and protein quantification was done with the DC Protein Assay kit. Five hundred μg of total protein per sample, diluted to 1 mL with a buffer provided by the kit, were subjected to cytokine profiling according to the procedure as described above, which was completed with a Proteome Profiler Mouse XL Cytokine Array kit (ARY028 R & D Systems).

Boyden Chamber Migration Assay

HSCs were transduced with lentiviruses to generate control HSCs (expressing NT shRNA) and HSCs with knockdown of THBS1, CXCL1, CXCL5, or GM-CSF. After confirming knockdown by WB or IF, CM of HSCs was collected and used as a stimulant for Boyden chamber migration assay. To study monocyte migration, 30 μL of HSC CM were added into a bottom well of a Boyden chamber (2p48783 NeuroProbe), and 50 μL of monocyte single-cell suspension were added into a top well (15,000 cells per well). The 2 wells were separated by a polycarbonate filter containing 8-μm pores (PFB8, Lot# 7109415, NeuroProbe). Six replicates were set up for each CM group. After incubation at 37°C for 4 hours, monocytes migrated into the lower wells were counted under a microscope. To study migration of monocyte-derived macrophages, 30 μL of HSC CM were added into a bottom well of a Boyden chamber, and 50 μL of single-cell suspension of macrophages were added into a top well (3000 cells per well). After incubation at 37°C for 4 hours, cells migrated into the lower well and attached to the lower side of the filter were fixed with 4% paraformaldehyde (UN2213 Lot# Bo136490, MilliporeSigma), stained by DAPI, and counted under a fluorescence microscope.^19,89 The experiment was repeated 3 times independently.

Cell Proliferation Assay

To assess the role of HSC CM for monocytes proliferation, an MTS-based assay was performed with the CellTiter 96 AQueous One Solution Cell Proliferation Assay kit (G3582, Promega). Five thousand CD14+ monocytes suspended in 100 μL of HSC CM were seeded into each well of a 96-well-microplate, and 6 replicates were set up for each CM group. Cell densities were determined at time points 0, 24, 48, and 72 hours by adding 20 μL of CellTiter 96 AQueous One Solution Reagent into each well, followed by incubation at 37°C for 4 hours and recording the absorbance at 490 nm with an Epoch microplate reader (11-120-570 Thermo Fisher Scientific).^19,22 The experiment was repeated 3 times independently.

Statistical Analysis

Data are expressed as mean ± standard error of the mean (SEM). Two-tailed Student’s t-test was used to analyze data of two groups. Analysis of variance (ANOVA) followed by post hoc test was used to analyze data of more than 2 groups, which was completed by using the Graph-Pad Prism 6 Software (Graphpad Software, Inc). P < .05 is considered statistically different.