Stellate Cells, Hepatocytes, and Endothelial Cells Imprint the Kupffer Cell Identity on Monocytes Colonizing the Liver Macrophage Niche

Abstract

Macrophages are strongly adapted to their tissue of residence. Yet, little is known about the cell-cell interactions that imprint the tissue-specific identities of macrophages in their respective niches. Using conditional depletion of liver Kupffer cells, we traced the developmental stages of monocytes differentiating into Kupffer cells and mapped the cellular interactions imprinting the Kupffer cell identity. Kupffer cell loss induced tumor necrosis factor (TNF)- and interleukin-1 (IL-1) receptor-dependent activation of stellate cells and endothelial cells, resulting in the transient production of chemokines and adhesion molecules orchestrating monocyte engraftment. Engrafted circulating monocytes transmigrated into the perisinusoidal space and acquired the liver-associated transcription factors inhibitor of DNA 3 (ID3) and liver X receptor-α (LXR-α). Coordinated interactions with hepatocytes induced ID3 expression, whereas endothelial cells and stellate cells induced LXR-α via a synergistic NOTCH-BMP pathway. This study shows that the Kupffer cell niche is composed of stellate cells, hepatocytes, and endothelial cells that together imprint the liver-specific macrophage identity.

Keywords: Bmp9; Id3; Kupffer cell; LXRa; Notch; Nr1h3; endothelial cell; fibroblast; liver; macrophage; monocyte; niche; stellate cell.

Graphical abstract

Figure 1

Replenishment of KC Pool by Ly6C^hi Monocytes (A and B) Expression of GFP, Ly6C, and F4/80 of monocytes (green gate), em-KCs (black gate), and mo-KCs (red gate) after DT injection in (A) Clec4f^DTR/+-Ccr2^GFP/+ mice or (B) Clec4f^DTR/+-Ccr2^GFP/GFP mice. Flow-cytometry plots are pre-gated on live CD45⁺CD11b⁺Lyve-1⁻SiglecF⁻Ly6G⁻ single cells. Data are representative of 2–3 experiments. (C) Proportion of Ly6C^hi monocytes (green lines), em-KCs (black lines), and mo-KCs (red lines) in the liver of Clec4f^DTR/+-Ccr2^GFP/+ mice (solid lines) or Clec4f^DTR/+-Ccr2^GFP/GFP mice (dashed lines) as a percentage of live CD45⁺ cells after DT injection. Pooled data are from 2–3 experiments; n = 5 (0,5d), 6 (PBS, 1d; 1,5d; 2d; 5d; 6d), and 8 mice (3d; 4d). (D) Heatmap showing the top 30 of upregulated genes in mo-KCs 3 days compared with 1 and 7 days after DT injection. (E) Expression of Ki-67 and EdU incorporation by em-KCs and mo-KCs in (top) Clec4f^DTR/+-Ccr2^GFP/+ mice or (bottom) in Clec4f^DTR/+-Ccr2^GFP/GFP mice after DT injection. Flow-cytometry plots are pre-gated as in (A). Data are representative of 2–3 experiments. (F) Percentage of EdU⁺ cells in indicated populations during the differentiation kinetic of mo-KCs in Clec4f^DTR/+-Ccr2^GFP/+ mice (solid line) or Clec4f^DTR/+-Ccr2^GFP/GFP (dash line). Pooled data are from 2–3 experiments; n = 5 (0,5d), 6 (PBS, 1d, 1,5d, 2d, 5d, 6d), and 8 mice (3d, 4d). Two-way ANOVA with Tukey post-test. ^∗p < 0.05, ^∗∗∗p < 0.001, ^∗∗∗∗p < 0.0001. (G) Percentage of EdU⁺ mo-KCs 3 days after DT injection in liver of mice treated with PLX3397 (α-CSF1R) or vehicle. Pooled data are from 2 experiments, n = 7. Mann-Whitney t test. ^∗∗∗p < 0.001. Related to Figure S1.

Figure 2

Recruited Monocytes Engraft in the Liver and Acquire KC-Specific Identity within 24 h after KC Depletion (A) F4/80 and Clec4F expression during embryogenesis. Gates indicate the different sorted populations used for the micro-array. (B–D) Heatmap showing expression by the indicated cell populations of (B) monocyte- and macrophage-related genes, (C) KC-associated transcription factors compared with other tissue-resident macrophages, and (D) KC-cores genes. (E) Shown in the top row is an in vivo two-photon microscopy analysis of livers from Clec4f^DTR/Cre-Ccr2^GFP/+-Rosa26^TdT/+ mice injected with PBS (left, corresponding to Video S2) or with DT for 10 h (middle, corresponding to Video S3) or 24 h (right, corresponding to Video S4) is illustrated. Shown in the middle row is automated tracking of GFP⁺ monocytes in the liver during 1 h. On the bottom are the average speed and confinement index (maximum displacement/path length) of individual monocyte tracks of the above conditions. Pooled data from 2–3 experiments. Scale bar, 20μm. (F) On top is the maximal intensity projection (MIP) of Clec4f^DTR/+-Ccr2^GFP/+ mouse liver sections stained for GFP (monocytes), MHCII, and DAPI in PBS control or 24 h after DT injection. MHCII⁻ monocytes (asterisks) are round and MHCII⁺ monocytes (arrowheads) are elongated. Scale bar, 50μm. On the bottom is a quantification of monocyte size 24 h after PBS or DT injection (left graph) or, depending on their MHCII expression, 24 h after DT injection (right graph). Pooled data from 2 experiments, n = 4. Student’s t test. ^∗∗∗∗p < 0.0001. Dots represent individual monocytes. (G) CD11c and MHC-II expression during mo-KC differentiation. Flow-cytometry plots are pre-gated as in Figure 1A and further gated on Ly6C^hiF4/80⁻ cells (Mono PBS; 0,5d and 1d) or F4/80⁺ cells (2d to em-KC PBS). Data are representative of 2–3 experiments. (H) Relative expression of KC-associated transcription factor mRNA in indicated populations and conditions. Pooled data from 3 experiments, n = 13. One way ANOVA with Bonferroni post-test. ^∗∗p < 0.01, ^∗∗∗p < 0.001, ^∗∗∗∗p < 0.0001.

Figure 3

Recruited Monocytes and Steady-State Kupffer Cells Are in Close Contact with Hepatic Stellate Cells in the Space of Disse (A) Still images of a time-series performed by intravital two-photon imaging of liver from Clec4f^DTR/Cre-Ccr2^GFP/+-Rosa26^TdT/+ mice showing engraftment of a monocyte and the extension of a dendrite 10–16 h after DT injection. (B) In 1 is the MIP of Clec4f^DTR/+-Ccr2^GFP/+ mouse liver sections, with CD31 (blue) and GFP (red). In 2 is the overlay of confocal microscopy and EM, allowing the identification of the monocyte and to delimit the blood vessels. In 3 is 2D EM slice showing the monocyte (red) passing through the endothelium (blue) and interacting with a droplet-rich cell (green) located in between hepatocytes (yellow). Shown in 4, 3D reconstruction of the EM. Data are representative of 2 experiments. Scale bar, 5 μm. (C) MIP of Clec4f^DTR/+-Ccr2^GFP/+ mouse liver sections stained for GFP (red), Desmin (green), CD31 (blue), and DAPI (gray) 24 h after DT injection, showing multiple HSC-monocyte interactions outside of blood vessels (arrows). Data are representative of 3 experiments. Scale bar, 10μm. (D) 3D reconstruction of confocal microscopy from Clec4f^DTR/+-Ccr2^GFP/+ mouse liver sections stained for GFP (red), Desmin (green), CD31 (blue), and DAPI (gray) 24 h after DT injection. Scale bar, 5μm. (E) MIP of PBS control Clec4f^{DTR /+} mouse liver sections stained for Clec4F (red), Desmin (green), CD31 (blue), and DAPI (gray). Data are representative of 5 experiments. Scale bar, 20μm. (F) Shown in 1 is the MIP of Clec4f^Cre/+-Rosa26^TdT/+ mouse liver sections with CD31 (blue) and TdTomato (red). In 2 is the overlay of confocal microscopy and EM allowing identification of the KC and delimiting of the blood vessels. In 3 is a 2D EM slice showing the KC (red) interacting with an HSC (green) outside of blood vessels (blue) and located in between hepatocytes (yellow). In 4 is the 3D reconstruction of the EM. Scale bar, 10μm. Related to Figure S2.

Figure 4

HSCs and LSECs Induce Monocyte Recruitment and Engraftment in the Liver after Kupffer Depletion (A) Principal component analysis of HSCs, LSECs, and hepatocytes at steady state, 12 h or 36 h after DT injection. (B) Heatmap of genes involved in granulocyte and agranulocyte adhesion and diapedesis upregulated by HSCs and/or LSECs after KC depletion according to ingenuity pathway analysis. (C) Analysis by ELISA of CCL2 concentration in the serum of mice after DT injection. Control mice consist of a mix of C57BL/6 DT-injected and Clec4f^DTR/+ PBS-injected mice (no differences observed). Pooled data are from 3 experiments. Control mice n = 5 (4h), 6 (10h, 12h, 16h), 9 (8h), and 10 mice (0h). KC-depleted mice n = 5 (0h); 11 (4h, 16h); 12 (12h); 15 (10h) and 16 mice (8h). Two-way ANOVA with Tukey post-test. ^∗p < 0.05; ^∗∗∗∗p < 0.0001. (D) On the left is the MIP of Clec4f^DTR/+-Ccr2^GFP/+ mouse liver sections stained for CCL2 (white), Desmin (green), GFP (orange), F4/80 (red), and CD31 (blue) 8 h after DT injection. CCL2 was mainly co-localized with HSCs (arrows), although a small amount could be found in monocytes (arrowheads) or in LSECs (asterisks). Data are representative of 2 experiments. Scale bar, 20 μm. (E) Quantification of CCL2⁺ cells at the indicated time points and normalized per mm² of tissue. Dots represent individual pictures. Pooled data are from 2 experiments; n = 4. Two-way ANOVA with Tukey post-test. ^∗∗∗∗p < 0.0001. (F) Quantification of VCAM-1 and selectin E expression by HSCs, LSECs, and hepatocytes after DT injection. Given that VCAM-1 was already expressed on HSCs at steady state, data are represented as median fluorescence intensity (MFI) fold increase as compared with PBS control. PBS controls were included in each individual experiment. Selectin E was represented as percentage of positive cells in each cell population. Pooled data are from 2–4 experiments; n = 7 (4h), 8 (10h, 16h, 24h, 48h), 9 (12h), 15 (PBS) and 17 mice (8h). One-way ANOVA with Bonferroni post-test. ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001, ^∗∗∗∗p < 0.0001. Related to Figure S3 and S4.

Figure 5

HSCs and LSECs Activation, Monocytes Recruitment, and Engraftment in the Liver Are Mediated by Both TNF and IL-1 (A) Schematic representation of the NicheNet analysis of upstream ligand-receptor pairs inducing the DE genes of LSECs and HSCs. Shown in (1) are potential upstream ligands based on HSCs and LSECs DE genes; in (2) (left) are potential receptors expressed by HSCs and LSECs associated with each potential ligands and (right) their expression in LSECs and HSCs; in (3) (top) are potential target genes of the top 3 of potential ligands and (bottom) their relative expression in PBS control or during KC depletion (12 h and 36 h). (B) MIP of Clec4f^DTR/+ mouse liver sections stained for CCL2 (white), Desmin (green), and F4/80 (red) 8 h after DT injection + isotype (top) or Anakinra + anti-TNF (bottom). Data are representative of 2 experiments. Scale bar, 50 μm. (C) Quantification of VCAM-1 and selectin E expression by HSCs and LSECs 10 h after PBS or DT injection administrated together with either isotype control antibody or with anti-TNF + Anakinra. VCAM-1 is represented as MFI fold increase as compared with PBS control. PBS controls were included in each individual experiment. Selectin E is represented as percentage of positive cells in each cell population. Pooled data from 3 experiments. n = 12. Two-way ANOVA with Tukey post-test. ^∗∗∗p < 0.001, ^∗∗∗∗p < 0.0001. (D) Ly6C and F4/80 expression in PBS-injected mice (PBS) or 24 h after (1) DT + isotype control, (2) DT + Anti-TNF, (3) DT + Anakinra, or (4) DT + anti-TNF + Anakinra. Flow-cytometry plots are pre-gated as in Figure 1A. Data are representative of 3 experiments. (E) Number of Ly6C^hi monocytes in the liver of Clec4f^DTR/+ mice 24 h after PBS or DT injection together with different combinations of isotype, anti-TNF, and Anakinra. Pooled data are from 3 experiments; n = 10 (PBS, anti-TNF) and 11 (Isotype, Anakinra, Anti-TNF + Anakinra). One-way ANOVA with Bonferroni post-test. ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001, ^∗∗∗∗p < 0.0001. (F) CD11c and MHCII expression on Ly6C^hi monocytes in PBS-injected mice (PBS) or 24 h after (1) DT + isotype control, (2) DT + Anti-TNF, (3) DT + Anakinra, or (4) DT + anti-TNF + Anakinra. Flow-cytometry plots are pre-gated as in Figure 1A. Data are representative of 3 experiments. (G) Number of Ly6C^hi monocyte subsets according to CD11c and MHCII expression in the liver of Clec4f^DTR/+ mice 24 h after PBS or DT injection together with different combinations of isotype control, anti-TNF and Anakinra. Pooled data are from 3 experiments; n = 10 (PBS, anti-TNF) or 11 (Isotype; Anakinra; Anti-TNF + Anakinra). Two-way ANOVA with Tukey post-test. ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001, ^∗∗∗∗p < 0.0001. (H) Number of Ly6C^hi monocytes in the liver of Clec4f^+/DTR-Tnf^flox/flox (n = 16) or Clec4f^Cre/DTR-Tnf^flox/flox (n = 12) mice 24 h after DT injection. Pooled data are from 3 experiments. t test. ^∗p < 0.05. (I) Number of the different Ly6C^hi monocyte subsets according to CD11c and MHCII expression in the liver of Clec4f^+/DTR-Tnf^flox/flox (n = 16) or Clec4f^Cre/DTR-Tnf^flox/flox (n = 12) mice 24 h after DT injection. Pooled data are from 3 experiments. Two-way ANOVA with Tukey post-test. ^∗∗∗∗p < 0.0001. Related to Figure S5.

Figure 6

CSF1 and BMP- and Notch-Signaling Pathway Serve as Potential Upstream Signals Inducing Monocyte-to-KC Differentiation (A) Schematic representation of the NicheNet analysis of upstream ligand-receptor pairs inducing the KC-specific identity. In 1, are the potential upstream ligands from HSCs, LSECs, or hepatocytes based on KC-associated transcription factors and on DE genes found between mo-KC 3 days after KC depletion and BM monocytes; in 2 (top) are potential receptors expressed by monocytes associated with each potential ligands and (bottom) their expression in BM mono and mo-KC during KC repopulation. (B) Circle plot showing links between (1) predicted ligands from HSCs (green), LSECs (blue), or hepatocytes (yellow) with (2) their associated receptors found on monocytes and (3) KC-associated transcription factors or DE genes (mo-KC 3d versus BM mono) potentially targeted by the ligand-receptors pairs. (C) On the left are genes potentially targeted by CSF1, BMP molecules, and/or Notch-signaling pathway and on the right their relative expression in BM mono or mo-KCs during KC repopulation.

Figure 7

HSCs, LSECs and Hepatocytes Imprint the KC Identity (A) MIP of Clec4f^DTR/+ mouse liver sections stained for (left) CSF1 (red), Desmin (green), CD31 (blue), and DAPI (gray) or (right) IL-34 (gray), Desmin (green), F4/80 (red), and CD31 (blue). At steady state, CSF1 is mainly produced by HSCs (arrows) with a small contribution of LSECs (arrowheads), whereas IL-34 is exclusively produced by HSCs. Data are representative of 2 experiments. Scale bars, 20 μm (left) and 50 μm (right). (B) Relative expression of KC-associated transcription factor mRNA in BM monocytes cultured 12 h alone or together with either HSCs, LSECs, or hepatocytes. Pooled data are from 2 experiments; n = 6. One way ANOVA with Bonferroni post-test. ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001, ^∗∗∗∗p < 0.0001. (C and D) Relative expression of KC-associated transcription factor mRNA (Nr1h3, Spic) or KC-core genes (Cd5l, Cdh5, Cd38, Cd207, Clec4f) in BM monocytes cultured during (C) 12 h or (D) 6 days on a feeder layer of OP9-DL4 or OP9-GFP with or without recombinant BMP2 or BMP9. Pooled data are from 3 experiments; n = 12 per group. Two-way ANOVA with Tukey post-test. ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001, ^∗∗∗∗p < 0.0001. (E) Representative histograms of CD38 or CD207 expression by BM monocytes cultured 6 days on a feeder layer of OP9-DL4 or control OP9-GFP with or without recombinant BMP2 or BMP9. Data are representative of 3 experiments. (F) Fold change of relative expression of Nr1h3 and Spic in Cd11c^hi MHCII^hi monocytes 24 h after DT injection from mice pretreated 24 h before DT injection with either isotype antibodies or a combination of anti-DLL1 and DLL4 antibodies. Pooled data are from 3 experiments; n = 12. t test ^∗p < 0.05, ^∗∗p < 0.01. Related to Figure S6 and S7.