Congestion Enriches Intra-hepatic Macrophages Through Reverse Zonation of CXCL9 in Liver Sinusoidal Endothelial Cells

Abstract

Background & aims: Congestion alters the microenvironment of the liver sinusoid along the portal-central axis. We studied spatial changes in immune cells in the sinusoid that contribute to congestive fibrosis and portal hypertension (PHTN).

Methods: To visualize the distribution of immune cells in congestive hepatopathy (CH), we performed imaging mass cytometry (IMC) on liver tissue from patients with CH, Fontan-associated liver disease (FALD), and controls. We performed partial ligation of the inferior vena cava (pIVCL) to simulate CH in mice and isolated primary liver cells for single-cell RNA-sequencing (scRNA-seq) to study zonation of liver sinusoidal endothelial cells (LSECs). After pIVCL, mice were treated with intraperitoneal injections of AMG487, an inhibitor of the CXCL9 receptor, or a neutralizing antibody to CXCL9.

Results: Intra-hepatic macrophages are enriched in CH and FALD. Given the role of CXCL9 in macrophage patterning in the liver, we performed RNA in situ hybridization (RNAish) in CH and determined that CXCL9 was highly expressed in LSECs in FALD, suggesting that LSECs recruit macrophages in CH. After pIVCL, treatment with AMG487 or an antibody to CXCL9 attenuated portal pressures, fibrosis, and intra-hepatic macrophages. To study changes in LSECs that promote macrophage chemotaxis, we performed scRNA-seq after pIVCL and sham procedures. Analysis revealed 3 LSEC subpopulations according to sinusoidal location. RNAish identified peri-central LSECs as the predominant source of CXCL9 in FALD. In vitro analyses revealed that β-catenin and hypoxia inducible factor-1 α regulate CXCL9 transcription in peri-central LSECs.

Conclusions: CXCL9 derived from peri-central LSECs enriches intra-hepatic macrophages in CH and FALD, contributing to congestive fibrosis and PHTN. Strategies to target LSEC-derived CXCL9 may prevent the progression of CH and FALD.

Keywords: Fibrosis; Fontan-associated Liver Disease; Hypoxia; Imaging Mass Cytometry; Portal Hypertension.

Figure 1

Imaging mass cytometry identifies intra-hepatic immune cell populations in CH and FALD. (A) Schema of the experimental platform. Five-mm FFPE sections of human liver tissues were labeled simultaneously with 27 metal conjugated antibodies. Between 5 and 6 1000 mm × 1000 mm ROIs were selected for laser ablation per sample. Plumes of particles are carried over to CyTOF for signal quantification. Antibody labeling patterns were reconstructed and output as 32-bit images. Cell segmentation was performed to enable downstream image analyses. (B) Spearman correlation of per-cell average protein expressions, post cell segmentation and normalization. Distinct clusters of co-expressed structural and immune markers could be identified. (C) Higher protein expression along tissue edge distributions. Across ROIs, cells distributed along tissue edges had notably higher proportion of protein expression outliers compared with internal cells. (D) Definition of tissue edge and internal cells. Tissue edge/internal cells are defined by thresholding at 30 pixels (microns) away from the tissue edges. Outlier cells are defined as the 5% cells with the highest total protein expressions (ie, summing over all channels). Cells distributed along the tissue edges had a much higher proportion of such outliers, suggesting strong technical artifacts compared with internal cells. (E) Reduced density of other immune cell populations relative to intra-hepatic macrophages in CH and FALD. Most immune cell groups had reduced densities in diseased tissues when compared with healthy controls. Monocytes/macrophages had slightly elevated density in the FALD group, and proliferating cell densities were also elevated in both FALD and CH tissues.

Figure 2

Visualizing structural and immune markers in congested livers by IMC. (A) Identification of immune cell populations through phenotype-specific average protein channel expression. For each of the phenotyped immune cell subpopulations, we calculated phenotype-specific mean protein channel expressions, averaged across cells of the phenotype from all samples and ROIs. These were further Z-score transformed for each protein channel (subtracted by its mean across the phenotypes and divided by standard deviation) to facilitate visualization in the heatmap, which reveals clear correspondence between the immune cell populations and the IMC protein channels that targeted them. (B) tSNE visualization of immune cell populations identified through unsupervised cell clustering. (C) Enrichment of intra-hepatic macrophages relative to other cell populations in CH and FALD. Selected channel overlays for healthy controls, subjects with CH, and subjects with FALD demonstrate enrichment of intrahepatic macrophage populations in CH and FALD. Displayed channels are NA+/K+ ATPase (red), SMA (green), CD68 (blue), and CD16 (teal). The zoomed-in insets show macrophages as identified by expression of CD16 and CD68. Scale bars represent 100 microns. White asterixes are used to label the central vein, whereas white arrows demarcate the portal vein. (D) Reduced percentages of other immune cell populations relative to intra-hepatic macrophages in CH and FALD. Most immune cell groups had reduced percentages in diseased tissues when compared with HCs. Percentages of monocytes/macrophages are significantly elevated in the FALD group, and percentage of proliferating cells were also elevated in both FALD and CH tissues. A Wilcoxon test was completed to compare per-ROI cell percentages between conditions followed by multiple testing with the BH procedure. (∗q < 0.05; ∗∗q < 0.01; ∗∗∗q < 0.001). t-SNE, t-distributed stochastic neighbor embedding.

Figure 3

Immune cell subsets identified through unsupervised clustering analysis of IMC dataset display well-differentiated, cell-type-specific protein expression patterns. (A) Mean expression of each protein channel in different cell populations, across cells of the phenotype from samples and regions of different subject groups (normal, CH, and FALD). Overexpression of protein channels in the specific cell phenotype that they target is consistent between the 3 subject groups, supporting the accurate and consistent identification of cell phenotypes across different groups of patients. (B) Boxplots of immune marker expressions are compared across different immune cell subsets, with the per-ROI expressions identified according to the patient group they represent. Most protein markers displayed good specificity for the targeted immune cell group, such as CD11b and CD14 for macrophages/monocytes, CD19/CD20 for B-cells, and CD3/CD4 for T-cells.

Figure 4

Macrophage enrichment by CXCL9 enhances fibrosis and portal hypertension after pIVCL. (A) RNAish reveals increased expression of CXCL9 in LSECs 6 weeks after pIVCL (right). The zoomed-in inset shows CXCL9 expression in a LSEC after pIVCL. In contrast, CXCL9 is absent in LSECs 6 weeks after the sham operation (left). Intensity of CXCL9 fluorescence was quantified by ImageJ (National Institutes of Health) and displayed in the panel to the right of the images. CXCL9 is shown in red, the endothelial marker Lyve-1 is shown in green, and blue represents DAPI nuclear counterstain. The scale bar represents 20 microns. A t-test was used to obtain the P value; n = 6–7 (P < .001). (B) Circulating CXCL9 levels are significantly increased in mice 6 weeks after pIVCL compared with 6 weeks after sham surgeries (P < .05). A paired t-test was used to obtain the P value; n = 7. (C) Mice treated with AMG487 have significantly decreased portal pressures 6 weeks after pIVCL and sham surgeries when compared with DMSO-treated mice (n = 5–7 per group; ANOVA P < .05). (D) Sirius red staining and quantification of liver tissue after sham and IVC ligation procedures. Images are 10× magnification. Scale bar represents 200 microns. Quantification was performed with ImageJ (National Institutes of Health) and displayed in the adjacent graph (n = 5–7; ∗P ≤ .05 for all panels). (E) IHC staining for CD68 of liver tissue after sham and IVC ligation procedures. Images are 20× magnification. Scale bar represents 50 microns. Quantification was performed with ImageJ (National Institutes of Health) and displayed in the adjacent graph (n = 5–7; ∗P ≤ .05 for all panels).

Figure 5

Hepatic congestion leads to increased CXCL9 expression in peri-central LSECs. (A) RNAish was performed on liver biopsies obtained from patients with FALD (right) and HCs (left). LSECs express CXCL9 in patients with FALD but not in HCs. Intensity of CXCL9 fluorescence was quantified by ImageJ (National Institutes of Health) and displayed in the panel to the right of the images. CXCL9 is shown in red, the endothelial marker Lyve-1 is shown in green, and blue represents DAPI nuclear counterstain. The scale bar represents 50 microns. A t-test was used to obtain the P value; n = 7. (B) Circulating CXCL9 levels are significantly increased in patients with FALD compared with age- and gender-matched healthy controls (P < .05). A paired t-test was used to obtain the P value; n = 7.

Figure 6

Flow cytometry analysis of intra-hepatic leukocytes identifies a significant increase in CD45+/CD11b+/CXCR3+ macrophages after pIVCL. Quantitative assessment of intra-hepatic leukocytes after sham and pIVCL surgeries. (A) Schema of gating strategy used on isolated intra-hepatic leukocytes. (B) Quantification of intra-hepatic leukocyte populations after sham and pIVCL surgeries. Cd11c+ and CD45+/CD11b+/CXCR3+ immune cells were significantly increased after pIVCL compared with sham. A t-test was used to compare cell populations (n = 3; ∗P ≤ .05 for all panels). FSC-A, forward scatter area; FSC-H, forward scatter height; SSC-A, side scatter area; SSC-H, side scatter height.

Figure 7

Flow cytometry analysis of intra-hepatic leukocytes identifies CXCR3+ myeloid cells. (A) Flow cytometry dot plot of CD45+/CD11b+/CXCR3+ intra-hepatic leukocytes isolated after sham and pIVCL surgeries. (B) Staining of intra-hepatic leukocytes with CXCR3 antibody and isotype control reveals negligible non-specific staining (n = 3).

Figure 8

CXCR3 expression by immune cells in the liver. (A) Violin plot illustrating CXCR3 protein expression across various cell types using publicly available CITE-sequencing data from 39,996 cells derived from mouse livers. (B) Box plot showing the percentage of CXCR3+ cells across various cell types using publicly available scRNA-seq data from 81,872 cells derived from mouse livers. This data is accessible in the Sequence Read Archive (PRJNA1066342).

Figure 9

CXCL9-recruited macrophages contribute to fibrosis and portal hypertension after pIVCL. (A) Mice treated with a neutralizing antibody to CXCL9 have significantly decreased portal pressures 6 weeks after pIVCL and sham surgeries when compared with control-treated mice (n = 5–8 per group; ANOVA P < .05). (B) qPCR from whole-liver mRNA shows lower mRNA levels of α-SMA after pIVCL in mice treated with a neutralizing antibody to CXCL9 compared with mice treated with control (ANOVA P < .05). (C) Sirius red staining and quantification of liver tissue after sham and IVC ligation procedures. Images are 10× magnification. Scale bar represents 200 microns. Quantification was performed with ImageJ (National Institutes of Health) and displayed in the adjacent graph (n = 5–7; ∗P ≤ .05 for all panels). (D) IHC staining for CD68 of liver tissue after sham and IVC ligation procedures. Images are 20× magnification. Scale bar represents 50 microns. Quantification was performed with ImageJ (National Institutes of Health) and displayed in the adjacent graph (n = 5–7; ∗P ≤ .05 for all panels).

Figure 10

Profiles of transcriptomes after IVC ligation and sham surgeries. (A) t-SNE visualization of single-cell transcriptomes of primary cell populations from mice after sham operations and pIVCL based on condition (left panel) and conserved genes (right panel). (B) Primary cell type clusters identified based on marker gene expression profiles. Cluster 1 was identified as LSECs based on expression of key landmark genes. (C) Feature plot of expression of the LSEC markers Aqp1, PECAM1, f8, and Kdr. t-SNE, t-distributed stochastic neighbor embedding.

Figure 11

Single-cell sequencing and clustering. Schema of isolation process for scRNA-seq.

Figure 12

Hepatic congestion creates LSEC subpopulations according to location within the liver sinusoid and proximity to the central vein. (A) LSEC subpopulations and their spatial coordinates as identified through a spatial reconstruction algorithm. (B) Heat maps comparing expression of landmark genes in each of these 3 lobules after sham and IVC ligation procedures. (C) Violin plots demonstrating greater differential expression of peri-central landmark genes in pericentral LSECs (lobule 1) compared with middle LSECs (lobule 2) and peri-portal LSECs (lobule 3) after pIVCL and sham surgeries. (D) Violin plots demonstrating differential expression of landmark peri-portal genes in all 3 LSEC subpopulations after pIVCL and sham surgeries.

Figure 13

Characterization of LSEC subpopulations according to location in the liver sinusoid. (A) tSNE visualization of LSEC subpopulations 1 (peri-central), 2 (middle), and 3 (peri-portal). (B) Expression of the peri-central LSEC marker Wnt2. (C) Expression of the peri-central LSEC marker Cd117. t-SNE, t-distributed stochastic neighbor embedding.

Figure 14

Pathways regulating the inflammatory response are upregulated in peri-central LSECs after pIVCL. (A) Enrichr analysis reveals upregulation of cells and molecules involved in the local acute inflammatory response in peri-central LSECs. Pathways are sorted by P-value ranking. (B) Expression of CXCL9 is highest in peri-central LSECs compared with other LSEC subpopulations in our scRNA sequencing dataset. (C) Feature plot demonstrating that LSECs are the primary source of CXCL9 within the liver. (D) RNAish reveals increased CXCL9 in biopsies collected from patients with FALD (bottom row) compared with HCs (top row). CXCL9 colocalizes with Cd117, which is a marker of peri-central LSECs. CXCL9 also colocalizes with β-catenin. Intensity of CXCL9 fluorescence was quantified by ImageJ (National Institutes of Health) and displayed in the bottom left panel. Colocalization of CXCL9 and Cd117 fluorescence was quantified by ImageJ and compared via Pearson’s coefficient (bottom right panel). Cd117 is shown in green, CXCL9 is shown in red, and β-catenin is shown in teal. A DAPI nuclear counterstain was performed. The scale bar represents 50 microns. A paired t-test was used to obtain the P value; n = 6–7; ∗P < .05.

Figure 15

Pathway analysis reveals unique functions in peri-central LSECs compared with peri-portal LSECs. Pathway analysis reveals unique functions in peri-central LSECs compared with peri-portal LSECs. In contrast to the pathway analysis for peri-central LSECs (lobule 1), pathway analysis by Enrichr of peri-portal LSECs (lobule 3) reveals upregulation of pathways related to spinal cord injury and the nuclear receptors meta-pathway. Pathways are sorted by P-value ranking.

Figure 16

Expression of CXCL9 and CXCR3 after pIVCL. (A) Violin plot demonstrating differential expression of CXCL9 in LSECs (clusters 1 and 6) after pIVCL. (B) Violin plot demonstrating expression of CXCR3 after pIVCL. CXCR3 expression is mainly localized to clusters 5 (T cells) and 17 (dendritic cells) in this scRNA-seq dataset. These data corroborate other publicly available datasets, which demonstrates robust expression of CXCR3 protein with low CXCR3 mRNA expression in macrophages. This is likely due to low mRNA expression of surface marker proteins.

Figure 17

HIF-1$\mathbf{\alpha }$and β-catenin drive LSEC zonation in CH. (A) Ingenuity Pathway Analysis of upstream transcriptional regulators of CXCL9 with an activation Z-score >2.0 suggests that β-catenin and HIF-1 $\mathrm{\alpha }$ regulate CXCL9 transcription. (B) Treatment of LSECs with the hypoxia-mimetic agent DMOG increases CXCL9 mRNA levels. A t-test was used to obtain the P value; n = 3. (C) Transfection of LSECs with the pcDNA3-S33Y β-catenin mutant plasmid (Addgene plasmid #19286) leads to upregulation of CXCL9 mRNA levels at varying concentrations (ANOVA P < .05). (D) pcDNA3-S33Y-transfected LSECs that were treated with DMOG demonstrated further increases in CXCL9 mRNA expression compared with pcDNA-S33Y-transfected LSECs treated with a vehicle agent (Veh) (ANOVA P < .05). n = 3–4.

Figure 18

HIF1-$\mathbf{\alpha }$regulates CXCL9 transcription. Transfection of LSECs with siRNA to HIF1-$\mathrm{\alpha }$ decreases CXCL9 mRNA levels after 48 hours. A t-test was used to obtain the P value; n = 3; ∗P < .05.