Spatial proteogenomics reveals distinct and evolutionarily conserved hepatic macrophage niches

Abstract

The liver is the largest solid organ in the body, yet it remains incompletely characterized. Here we present a spatial proteogenomic atlas of the healthy and obese human and murine liver combining single-cell CITE-seq, single-nuclei sequencing, spatial transcriptomics, and spatial proteomics. By integrating these multi-omic datasets, we provide validated strategies to reliably discriminate and localize all hepatic cells, including a population of lipid-associated macrophages (LAMs) at the bile ducts. We then align this atlas across seven species, revealing the conserved program of bona fide Kupffer cells and LAMs. We also uncover the respective spatially resolved cellular niches of these macrophages and the microenvironmental circuits driving their unique transcriptomic identities. We demonstrate that LAMs are induced by local lipid exposure, leading to their induction in steatotic regions of the murine and human liver, while Kupffer cell development crucially depends on their cross-talk with hepatic stellate cells via the evolutionarily conserved ALK1-BMP9/10 axis.

Keywords: CITE-seq; Kupffer cell; NAFLD; across species; atlas; lipid-associated macrophage; liver; multi-omic; proteogenomic; spatial transcriptomics.

Graphical abstract

Figure S1

Cell types identified in transcriptomic studies depend upon cell/nuclei isolation technique used, related to Figure 1 Cells were isolated from livers of healthy C57B/l6 mice by either ex vivo or in vivo enzymatic digestion. Alternatively, livers were snap frozen and nuclei subsequently isolated following tissue homogenization by a sucrose gradient (3 mice per isolation method). Live cells/intact nuclei were identified and purified using flow cytometry. For the cells, either live CD45⁺, live CD45⁻ or live hepatocytes were sorted. 1 ex vivo digested sample and 1 in vivo digested sample were also stained with a panel of 107 (ex vivo cells) or 161 (in vivo cells) oligo-conjugated antibodies for CITE-seq analysis. FACS-purified cells/nuclei were loaded onto the 10× chromium platform and scRNA-seq, CITE-seq, or snRNA-seq performed. Following clean up and QC, cells from the same mice were pooled together in the same ratios (CD45⁺:CD45⁻:Heps) as found in the tissue as a whole before sorting, different mice were then pooled together and the data were analyzed using scVI. (A–C) UMAPs showing annotations of cell types and proportions of each cell type as a % of total cells in the UMAP isolated using (A) ex vivo digestion; 13,144 cells, (B) in vivo digestion; 24,014 cells and (C) nuclei; 8,583 nuclei. (D) Average number of genes/cell in the annotated mac, B cell, hepatocyte, endothelial, and stromal cell populations following each isolation method. ^∗p < 0.05 one-way ANOVA with Bonferroni post-test per cell type. (E) Correlation plot showing genes captured within the mac population when the liver is digested with the in vitro versus the in vivo digestion protocol. (F) Correlation plots showing genes captured within the mac, endothelial cell, and hepatocyte populations when cells are isolated using the in vivo digestion protocol or nuclei are isolated. (G–L) Confocal microscopy images to determine true abundance of (G) stromal cells and cholangiocytes (H) endothelial cells, (I) macs, (J) dendritic cells, (K) B cells, and (L) T cells in vivo. Scale bars, 200 μm. (M) The percentage of each population was calculated based on the percentage of a given population divided by the total number of nuclei. A threshold was applied to the DAPI channel (picture 1) in ImageJ (picture 2) and nuclei were automatically counted based on the ImageJ “analyze particles” plugin (size ). Due to the density of some liver zones, some nuclei were not automatically counted (arrow, picture 3). Those were then manually counted and added to the total number of nuclei. For the populations of interest, cells were counted manually based on specific markers (for example, CD3 for T cells, picture 4). Counting was performed blinded prior to analysis of the sequencing results. (N) Proportion of indicated cell types as a % of total cells identified in confocal microscopy images. Data are from 3–7 images per cell type taken from 2–4 mice.

Figure 1

A proteogenomic atlas of the healthy murine liver (A) Hepatic cells were isolated from healthy C57B/l6 mice by ex vivo (5 mice, 15 samples) or in vivo (5 mice, 19 samples) enzymatic digestion. Alternatively, nuclei were isolated by tissue homogenization (4 mice, 12 samples). Live cells/intact nuclei were FACS-purified. For cells, total live, live CD45⁺, live CD45⁻, live hepatocytes, or myeloid cells (live CD45⁺, CD3⁻, CD19⁻, B220⁻, NK1.1⁻) were sorted. 18 samples (7 ex vivo, 11 in vivo) were also stained with a panel of 107–161 barcode-labeled antibodies for CITE-seq analysis. All datasets were pooled together and after QC 185,894 cells/nuclei were clustered using TotalVI. (B) UMAP of sc/snRNA-seq data. (C) Tissue and capsule images from Visium analysis with clusters overlaid. (D) UMAP of zonation of Visium spots (left) and origin of the cells (right). (E) Zonation pattern mapped onto tissue slice. (F and G) Indicated cell signatures from sc/snRNA-seq mapped onto the Visium zonation data. (H) mRNA zonation pattern in Visium highly multiplexed protein analysis and VSIG4-ADT expression pattern (left) and zonated expression patterns of indicated antibodies (right). (I) MICS analysis of indicated proteins. (J) Molecular Cartography of indicated genes and cell types. (K) mRNA (Xcr1, Flt3l, Mafb, and Clec10a) and protein (MHCII and F4/80) expression in the same tissue slice. Scale bars, 50 μm. PV, portal vein; CV, central vein. Arrows indicate specific cell types, colors correspond to cell type/markers. Images are representative of 2–4 mice. See also Figure S2 and Tables S1, S2, S3, and S5.

Figure S2

Combination of CITE-seq, scRNA-seq, snRNA-seq, and spatial analyses enables identification of all hepatic cell types including bona fide cell doublets, related to Figure 1 (A and B) Top DEGs (A) and DEPs (B) for cell types from Figure 1B. (C) Distinct profiles of cells or nuclei within the UMAP depending on isolation protocols; 71,162 cells from ex vivo digestions, 96,066 cells from in vivo digestions, and 18,666 nuclei. Numbers on plots represent numbers of cells/nuclei per population. (D) Correlation plots showing genes captured within the KC, B cell and neutrophil populations with and without addition of CITE-seq antibodies. (E) Expression of VSIG4, CD206, and ESAM (protein, top) and Vsig4, Mrc1, and Esam (mRNA, bottom). (F) UMAP showing clusters of cells when only minimal QC for gene number and % mitochondrial genes is performed; 17,669 cells pooled from 3 samples. Expression of Cd5l, Cd19, and Kdr by the clusters facilitating identification of cell types per annotation. (G) CITE-seq data from (F) in Flow-Jo showing expression of CD206 and ESAM in total KCs (left) and total B cells (middle). Numbers represent % of entire KC or B cell population. Identified populations were then mapped back onto the original UMAP (right). (H) Expression of CD31, CD26, and CD38 by indicated populations. (I) Heatmaps showing expression of top DEGs between KC1s and LSECs (left), KC2s and KC1s + LSECs (middle) and B cell2s and B cell1s + LSECs (right). (J) 3D reconstruction of murine liver following perfusion with antigen fix to inflate endothelial cells and staining with antibodies against CD31, CD206, and F4/80. (K) UMAP showing clusters generated from Visium analysis of liver tissue (4 samples) and liver capsule (1 sample). (L) Top unbiased genes defining zonation trajectory from portal to central vein in Visium. (M) Expression of Glul and Epcam by confocal microscopy (left), annotation of portal, periportal, mid, and central regions on same tissue section (middle) and overlay of both datasets (right). (N) Identification of cholangiocyte (left) and cDC (right) signatures on zonated Visium spots. (P) Molecular Cartography showing expression of indicated zonated hepatocyte mRNAs in liver tissue. Data are representative of 2 mice. (O) Expression of Itgae (encoding CD103) in the UMAP of the total liver (left) and flow cytometric analysis of total cDC1s for CD103 and MHCII expression in the healthy murine liver (right).

Figure 2

A population of macrophages reside around the bile duct in the healthy murine liver (A) UMAP of murine myeloid cells (71,261 cells/nuclei) isolated from Figure 1B and re-clustered with TotalVI. (B and C) Top DEGs (B) and DEPs (C) between cell types. (D) Expression of Gpnmb and Cd207. (E) Expression of VSIG4 and F4/80 (left) or MHCII, CD11c, and DAPI (right) by confocal microscopy. Capsule macs identified by white arrows. Scale bars, 50 μm. (F) Molecular Cartography of indicated genes at liver capsule. (G) Expression of VSIG4, F4/80, GLUL, and DAPI (left) or F4/80 or CCR2 (right, inset) by confocal microscopy. Scale bars, 100 μm. (H) Molecular Cartography of indicated genes at portal triad. PV, portal vein; CV, central vein; HA, hepatic artery; BD, bile duct. Arrows indicate specific cell types, where color corresponds to markers. Images are representative of 2–4 mice. (I and J) Top GO terms for KCs (I) and bile-duct LAMs (J). (K) Representative image showing expression of VSIG4 (red) CD19 (yellow) and CD3E (magenta) by MICS analysis (left) and % of B or T cells found with/without a KC per field of view (right). Data are pooled from multiple fields of view in 2 mice. ^∗∗∗p < 0.001 Student’s t test. (L) Mice (29-week-old) were treated with 3.5 mg/kg LPS or PBS and 2 h later, livers were harvested without the capsule. KCs and LAMs were FACS-purified and expression of Il1b, Tnf, IL10, and Il18 was examined by qPCR, compared with b-actin. ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001, ^∗∗∗∗p < 0.0001, one-way ANOVA with Bonferroni post-test. See also Figure S3 and Table S2.

Figure S3

Validated flow cytometry gating strategy for murine myeloid cells, related to Figure 2 (A) CITE-seq data from the murine myeloid cells in Figure 2A were exported as an FCS file and an in silico gating strategy identified in FlowJo software. (B) Application of the in silico gating strategy with a 21-color flow cytometry panel. Myeloid cells were pre-gated as live CD45⁺ lineage cells (Ly6G⁻CD19⁻NK1.1⁻B220⁻CD3⁻). Data are representative of 3 experiments with 3–6 mice per experiment. (C) cDC1s, cDC2s, migratory cDCs (Mig. cDCs), peritoneal macs (Peri. Macs), KCs, and non-KC macs (non-KCs) were FACS-purified using gating strategy in (B), mRNA was isolated and qPCR performed to examine expression of indicated genes defining each population to validate their identity. Data are representative of 2 experiments with n = 3–6. (D) Putative peritoneal macs were FACS-purified using gating strategy in (B) and expression of Gata6 was examined by qPCR compared with other hepatic myeloid populations. Data are from a single experiment with n = 6. (E) Peritoneal macs as a % of total macs recovered from the liver using different digestion techniques (in vivo, ex vivo, or capsule) or in supernatants in which livers were washed following removal from the mouse but prior to digestion (wash). Data are from a single experiment with n = 4. ^∗p < 0.05, ^∗∗p < 0.01 one-way ANOVA with Bonferroni post-test compared with wash data. (F) Expression of CD14 and CD207 within the non-KC mac population from (B) (left) and % of CD207⁺ and CD207⁻ populations among total macs in livers digested using the ex vivo or in vivo protocols or in dissected and digested liver capsule (right). Data are representative of two experiments with n = 4–5 mice per experiment. ^∗∗∗∗p < 0.0001 mixed effects analysis with Tukey’s multiple comparison test. (G) Expression of VSIG4, F4/80, GLUL, and DAPI by confocal microscopy. Insets represent zones featured in Figures 2E, 2G, and S3I. (H) Molecular Cartography of indicated genes and cell types. Insets represent zones featured in Figures 2F, 2H, and S3J. (I) Expression of VSIG4, F4/80, GLUL, and DAPI by confocal microscopy at the central vein. Scale bars, 50 μm. (J) Molecular Cartography of indicated genes and cell types at central vein. (K) Expression of F4/80, EPCAM, CCR2, GPNMB, and DAPI by confocal microscopy at a portal vein (top) or F4/80 or GPNMB alone (bottom). Scale bars, 25 μm. (L) Quantification of % of Gpnmb & Trem2 counts over Adgre1 counts in indicated regions of tissue as assessed using Molecular Cartography data. Each dot represents an individual region. ^∗p < 0.05, ^∗∗∗∗p > 0.0001 one-way ANOVA with Bonferroni post-test. (M) Expression of DESMIN and F4/80 at the liver capsule and underlying parenchyma (left) or EPCAM, DESMIN and F4/80 at the bile duct by confocal microscopy. PV, portal vein; CV, central vein; HA, hepatic artery; BD, bile duct. Arrows indicate specific cell types, where color corresponds to cell type/markers. All images are representative of 2–6 mice.

Figure S4

Protein markers of murine CD45⁻ cell subsets, related to Figure 3 (A) CITE-seq data from the murine CD45⁻ cells in Figure 3A were exported as an FCS file and an in silico gating strategy identified in FlowJo. (B) Gated cell overlay of populations identified using strategy in (A). (C) Expression of CD90, CD204, CD73, and CD29 markers by indicated cell types. (D) Expression of indicated protein markers in 60-plex MICS analysis in endothelial cells. (E) Expression of DESMIN, EPCAM, LYVE1, and CD31 at a portal triad (left) with inset (right). (F) Expression of indicated protein markers in 60-plex MICS analysis in stromal cells. (G) Molecular Cartography of indicated genes and cell types at portal vein. PV, portal vein; CV, central vein. Arrows indicate specific cell types, where color corresponds to cell type/markers. All images are representative of 2–6 mice.

Figure 3

Hepatic macrophage populations reside in distinct niches (A) UMAP of murine CD45⁻ cells (83,410 cells/nuclei) isolated from Figure 1B and re-clustered with TotalVI. (B and C) Top DEGs (B) and DEPs (C) between cell types. (D) Indicated cell signatures from sc/snRNA-seq mapped onto the Visium zonation data. (E) Molecular Cartography of indicated genes at central vein (left) and 2 different portal triads (center and right). (F) UMAP of murine stromal cells (5,430 cells/nuclei) isolated from the UMAP in Figure 3A and re-clustered with scVI. (G) Top DEGs between different cell types identified. (H) Identification of mesothelial cell (top) and VSMC (bottom) signatures on zonated Visium data. (I and J) Molecular Cartography of indicated genes at the liver capsule (I) or the central vein (J; left) and portal triad (J; right). PV, portal vein; CV, central vein; HA, hepatic artery; BD, bile duct. Arrows indicate specific cell types, where color corresponds to markers. Images are representative of 2–4 mice. See also Figure S4 and Tables S3 and S4.

Figure S5

Combination of CITE-seq, scRNA-seq, snRNA-seq, and spatial analyses enables generation of a human liver atlas, related to Figure 4 (A) Murine lymphoid cells (B cells, T cells, NK cells, ILC1s, pDCs; 27,398 cells) were isolated from Figure 1B and re-clustered with TotalVI. (B and C) Top DEGs (B) and DEPs (C) for the cell types from Figure S5A. (D) CITE-seq data from Figure S5A were exported as an FCS file and an in silico gating strategy identified in FlowJo. (E) Human lymphoid cells (B cells, T cells, NK cells, ILC1s, pDCs; 105,790 cells) were isolated from Figure 4B and re-clustered with TotalVI. (F and G) Top DEGs (F) and DEPs (G) for the cell types from Figure S5E. (H) CITE-seq data from Figure S5E were exported as an FCS file and an in silico gating strategy identified in FlowJo. (I) Proportion of indicated cell types arising from patients with <10% (purple) or >10% steatosis (yellow). (J) Hepatic cells were isolated from 22 C57B/l6 mice fed either a standard diet (SD) or a western diet (WD) for 24 or 36 weeks to induce NAFLD and NASH by ex vivo (10 samples) or in vivo (12 samples) enzymatic digestion. Alternatively, livers were snap frozen and nuclei isolated by tissue homogenization (14 samples). Live cells/intact nuclei were purified using FACS. For cells, total live, live CD45⁺, live CD45⁻, live hepatocytes or myeloid cells (live CD45⁺, CD3⁻, CD19⁻, B220⁻, NK1.1⁻) were sorted. 10 samples were also stained with a panel of 107–161 barcode-labeled antibodies for CITE-seq analysis. All datasets were pooled together and after QC 121,980 cells/nuclei were clustered using TotalVI. (K) Murine lymphoid cells (B cells, T cells, NK cells, ILC1s, pDCs; 21,322 cells) from mice fed the SD or WD for 24 or 36 weeks were isolated from Figure S5J and re-clustered with TotalVI. (L) Proportion of indicated cell types arising from mice fed the SD (purple) or WD (yellow). (M) Top DEGs between CTLs isolated from mice fed the SD (purple) or WD (yellow).

Figure 4

Identification of bona fide Kupffer cells across species (A) Cells/nuclei were isolated from liver biopsies (∼1–2 mm³; 14 cells, 5 nuclei) from patients undergoing either liver resection, cholecystectomy or gastric bypass. Live cells/intact nuclei were FACS-purified. Either total live, live CD45⁺, and live CD45⁻ or live CD45⁺, CD3⁻, and CD19⁻ cells were sorted. 7 cell samples were stained with a panel of 198 barcode-labeled antibodies for CITE-seq analysis. All datasets were pooled together and after QC, 167,598 cells/nuclei were analyzed using TotalVI. (B) UMAP of sc/snRNA-seq data. (C) UMAP of Visium data from 4 patient biopsy samples. (D) Split of Visium spots based on % steatosis. (E) Healthy and steatotic Visium liver tissue with clusters overlaid and H+E staining to identify steatotic zones. (F) Zonation of Visium data (top) with zonation pattern mapped onto liver tissue (bottom). (G) Indicated cell signatures from sc/snRNA-seq mapped onto Visium zonation trajectory, healthy (top), steatotic (bottom). (H) Myeloid cells (40,821 cells) were isolated from Figure 4B and re-clustered with TotalVI. (I) Expression of VSIG4 protein (top) and CD5L mRNA (bottom). (J) Expression of VSIG4, F4/80, FOLRB, and GLUL combined with Cd5l/CD5L on murine (left) and human (H25; right) livers. Scale bars, 50 μm. Inset in bottom panels. Scale bars, 20 μm. Images are representative of 2–4 livers. (K) Livers (2/species) were isolated from healthy macaque, pig, chicken, hamster, and zebrafish. Cells were isolated by ex vivo digestion for CITE-seq (pig; 198 human antibodies) or scRNA-seq (hamster, chicken, and zebrafish), or nuclei were isolated for snRNA-seq (macaque). Total live cells (hamster, chicken, and pig), DsRed⁺GFP⁺ cells (zebrafish) or nuclei (macaque) were FACS-purified. Following QC, 8,483 nuclei (macaque) or 21,907 (pig), 5,965 (hamster), 7,457 (chicken), and 4,957 (zebrafish) cells were analyzed using TotalVI (pig) or scVI (macaque, hamster, chicken, and zebrafish) (top). KCs were identified using the human-murine KC signature and the signature finder algorithm (Pont et al., 2019) (bottom). See also Figures S6 and S7 and Tables S1, S2,S3, S5, S6,S8, and S9.

Figure S6

Combination of CITE-seq, scRNA-seq, snRNA-seq, and spatial analyses enables generation of a human liver atlas and identification of bona fide human KCs, related to Figure 4 (A and B) Top DEGs (A) and DEPs (B) for the cell types from Figure 4B. (C) Distinct profiles of cells or nuclei within the UMAP depending on isolation protocol used; 152,535 cells from ex vivo digestions and 15,063 nuclei. (D) Proportion of each cell type per patient profiled. (E) Proportion of indicated cell types as a % of total CD45⁺ cells calculated from ex vivo digested samples per surgery type. Ch; cholecystectomy, Re; resection, GB; gastric bypass. ^∗p < 0.05; one-way ANOVA with Bonferroni post-test. (F) Mapping of Visium UMAP zonation patterns onto tissue sections from patient H35 and H37. (G) Expression of indicated zonation genes in patients H35–H38 assessed by Molecular Cartography. (H and I) Expression of indicated proteins by MICS 100-plex protein analysis in the healthy (H) and steatotic (I) human liver. (J) Murine myeloid cells (cDC1s, cDC2s, Mig. cDCs, Macs, monocytes, and monocyte-derived cells; 42,922 cells) from mice fed the SD or WD for 24 or 36 weeks were isolated from Figure S5J and re-clustered with TotalVI. (K) Distribution of cells in UMAP originating from SD- (purple) or WD- (yellow) fed mice. (L) Proportion of indicated cell types arising from mice fed the SD (purple) or WD (yellow). (M and N) Flow cytometry analysis of indicated cell populations in SD and WD-fed mice (24 weeks). Representative gating strategies (M) and absolute number of indicated populations (N). ^∗p < 0.05, ^∗∗p < 0.01 Student’s t test. Data are from 2 independent experiments with n = 5–6 per diet. (O and P) Top DEGs (O) and DEPs (P) for cell types from Figure 4H. (Q) Top 25 Murine KC genes as expressed by the human myeloid cell clusters. (R) Mapping of KC signature onto Visium trajectory for healthy (purple) and steatotic (orange) livers. (S) Expression of VSIG4 mRNA within human myeloid cells. (T) Expression of VSIG4 (red) and CD163 (gray, top) or CD169 (gray, bottom) by MICS analysis in healthy human liver. (U) Representative images showing KC location (red) as assessed by MICS analysis in the healthy (left) and steatotic (right) human liver. PV, portal vein; CV, central Vein, dashed line indicates zones of steatosis. (V) Representative image of CD68 and CD163 staining in 10–15-year-old human liver paraffin sections. Image is representative of 6 different patients. (W) In silico gating strategy to isolate distinct myeloid cell populations identified from CITE-seq data. (X) Expression of VSIG4 and FOLR2 by live CD45⁺ cells also expressing CD14 in indicated human liver biopsies by flow cytometry. Data are representative of 21 biopsy samples analyzed.

Figure S7

Conserved and unique features of KCs across species, related to Figure 4 (A and B) Expression of human-murine KC signature genes across cell types in mouse (A) and human (B). (C) Unbiased identification of KCs in mouse and human using the human-murine KC signature and the signature finder algorithm (Pont et al., 2019). (D–H) Annotated UMAPs from indicated species and expression of top KC-specific genes compared with other cells per species. (I) Expression of previously identified core murine transcription factors (Bonnardel et al., 2019) by KCs across species. (J) Venn diagram showing convergence and divergence of expression of top 50 KC genes per species across species, see Table S9 for genes lists per species. (K) Top DEPs (identified with cross reactive human antibodies) in the pig CITE-seq data. (L) Expression of VSIG4 in the porcine liver by confocal microscopy. (M) Expression of VSIG4, CD68 (protein), and CD5L (mRNA) in macaque liver. PV, portal vein; HA, hepatic artery; BD, bile duct. All images are representative of 2 livers. (N and O) Conserved expression of indicated genes across CD45⁻ (N) and CD45⁺ (O) cell types and species.

Figure 5

LAMs are found at the bile duct in the healthy liver, but at zones of steatosis in the obese liver (A) Expression of indicated markers at the capsule of the healthy human liver (H14) by confocal microscopy. Scale bars, 20 μm. Arrowheads indicate capsule macs. (B and C) Representative images showing expression of indicated markers at the portal triad by confocal microscopy. Insets on right. Scale bars, 100 μm. Insets scale bars, 50 μm (B). Scale bars, 150 μm. Insets scale bars, 75 μm (C). (D) Human bile-duct LAMs identified using the murine bile-duct LAM gene signature and the signature finder algorithm (Pont et al., 2019). (E) Human LAM signature from scRNA-seq mapped onto the Visium zonation data, healthy (purple), steatotic (yellow). (F) Expression of indicated markers by confocal microscopy. Insets on right. Scale bars, 150 μm. Insets scale bars, 75 μm. (G and H) Expression of indicated genes in healthy (G) and steatotic (H) human liver. Insets on right. Images are representative of 2 patients per condition. (I) Mice were fed a western diet (WD) or standard diet (SD) for 36 weeks to induce NAFLD and Visium analysis was performed. Analysis is pooled from 1 liver slice from the SD condition and 3 liver slices from the WD condition. Zonation pattern and H&E staining (left) and LAM and KC location (right). (J) Heatmap showing DEGs between LAMs from SD (purple) and WD (yellow) fed mice (24 + 36 weeks pooled). (K) LAMs were FACS-purified from the liver of mice fed the SD or WD for 24 weeks (with removal of capsule prior to digestion), RNA was isolated and expression of indicated genes was assessed by qPCR relative to β-actin. ^∗p < 0.05, ^∗∗∗∗p < 0.0001, Student’s t test. Data are pooled from 2 independent experiments with n = 7–10 mice per group. Images from (A–C and F) are representative of 4–5 patients per condition. PV, portal vein; CV, central vein. See also Figure S8 and Table S2.

Figure S8

Evolutionarily conserved signals regulate LAM and KC development, related to Figures 5 and 7 (A) Confocal microscopy of healthy human liver showing expression of indicated markers. Scale bars, 200 μm. (B and C) Expression of conserved human-murine bile-duct LAM signature in human (B) and mouse (C) hepatic myeloid cells. (D) Proportion of indicated myeloid cell populations as a % of total myeloid cells in human liver biopsies profiled by scRNA-seq when divided based on presence of steatosis. (E) Mice were fed a western diet (WD) or standard diet (SD) for 36 weeks to induce NAFLD and Visium analysis was performed. Analysis is pooled from 1 liver slice from the SD condition and 3 liver slices from the WD condition. Shown are cluster and sample annotations. (F) Zonation of all cell types from Figure S5J in murine NAFLD map (SD&WD). (G) Differential NicheNet highlighting prioritized conserved (human-mouse) ligand-receptor (LR) pairs between indicated macs and their niche cells. LR pairs are grouped according to the niche cell type with highest ligand expression. (H) Expression of ALK1 (ACVRL1), BMP9 (GDF2), and BMP10 in human, mouse, and macaque livers where both KCs and stellate cells were profiled. (I) Livers were harvested from Clec4f-CrexAcvrl1^fl/fl mice or Acvrl1^fl/fl controls and KCs examined (top) and quantified (middle) using VSIG4 expression. Expression of indicated KC markers by mac populations in Clec4f-CrexAcvrl1^fl/fl or Acvrl1^fl/fl control mice (bottom). Data are pooled from 2 independent experiments with n = 14 per group. Student’s t test. ^∗∗∗∗p < 0.0001. (J) Expression of CD31 (ECs), DESMIN (stromal cells), F4/80 (Macs), and EPCAM (cholangiocytes) by confocal microscopy in Fcgr1-CrexAcvrl1^fl/fl mice and Acvrl1^+/+ controls. PV, portal vein; CV, central vein. Images are representative of 2 mice per group.

Figure 6

Macrophage niche cells may regulate macrophage locations in healthy versus steatotic liver (A) UMAP of human CD45⁻ cells (15,481 cells/nuclei) isolated from Figure 4B and re-clustered with scVI. (B) Top DEGs between cell types identified. (C) Indicated cell signatures from sc/snRNA-seq mapped onto the Visium zonation data, healthy (top), steatotic (bottom). (D) Molecular Cartography localizing distinct CD45⁻ cells. (E) Circos plot showing NicheNet predicted ligand-receptor pairs between KCs and LSECs, HSCs and hepatocytes uniquely expressed in the human liver (left) and UMAPs showing normalized expression of indicated chemokines in the human liver (right) and CCR1/Ccr1 in the human and murine NAFLD liver (bottom). (F) Zonation of indicated genes in the healthy and steatotic human liver. (G) UMAP of murine NAFLD (SD and WD) stromal cells (4,025 cells/nuclei) isolated from Figure S5J and re-clustered with scVI (left) and proportions of each cell type in SD- and WD-fed mice (right). (H) Top DEGs between cell types. (I) Mice were fed a western diet (WD) or standard diet (SD) for 36 weeks to induce NAFLD, and Visium analysis was performed. Analysis is pooled from 1 liver slice from the SD condition and 3 liver slices from the WD condition. Zonation of Ccl2⁺ fibroblasts and fibroblasts in SD- and WD-fed mice. See also Tables S3 and S10.

Figure 7

ALK1-BMP9/10 axis regulates KC development (A) Mouse BM monocytes were cultured in the presence of CSF1 and indicated concentrations of human ac-LDL, prior to analyzed for expression of indicated genes by qPCR. Data are pooled from 2 experiments. One-way ANOVA with Bonferroni post-test compared with 0 ng/mL. (B) NicheNet circos plot highlighting conserved ligand-receptor pairs and induced target genes between KCs and indicated niche cells in human and mouse. (C) Feature plots showing expression of ALK1 (Acvrl1) in human myeloid cells (left) and GDF2/BMP10 in CD45⁻ cells (right). (D) Livers were harvested from Fcgr1-CrexAcvrl1^fl/fl mice or Acvrl1^fl/fl or Acvrl1^+/+ controls and KCs examined (left) and quantified (right) using VSIG4 expression. (E) Expression of indicated KC markers by mac populations in Fcgr1-CrexAcvrl1^fl/fl or Acvrl^fl/fl control mice. Data are pooled from 3 independent experiments with n = 9 per group. Student’s t test. (F) Expression of indicated markers in livers of Fcgr1-CrexAcvrl1^fl/fl or Acvrl1^fl/fl or Acvrl1^+/+ control mice by confocal microscopy. Scale bars, 50 μm. Images are representative of 2 mice per group. (G) Schematic of chimera experiment setup. (H) % chimerism normalized to levels in blood Ly6C^hi monocytes in Clec4f-Dtr mice 7 or 13 days after DT administration following partial irradiation and receiving either Acvrl^fl/fl or Fcgr1-CrexAcvrl^fl/fl BM 4 weeks earlier. Data are pooled from 2 independent experiments with n = 10–12 mice per group. (I) Schematic of Fc trap experiment setup. (J) Representative FACS plots showing VSIG4 and CLEC2 expression by total macs. Numbers represent % of total mac population in the indicated gate (left) and % of VSIG4⁺ and VSIG4⁻ macs among total CD45⁺ cells in the different treatment conditions. One-way ANOVA with Bonferroni post-test. ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001, ^∗∗∗∗p < 0.0001. See also Figure S8.