Cluster and spatial lobular location of single-LSEC was identified in MCD-induced NASH mice

Elevated NASH activity scores (including steatosis, ballooning, lobular inflammatory cell infiltration, Additional file 1: Fig. S1A–E), more severe collagen accumulation (Additional file 1: Fig. S1B–F) and lipid deposition (Additional file 1: Fig. S1C–G), accompanied by more F4/80 staining (Additional file 1: Fig. S1D–H), were observed in MCD-fed mice than control mice. These findings demonstrated that MCD-fed mice developed to NASH. Liver NPCs from control or MCD mice were processed for scRNA-seq analysis (Fig. 1A). Clusters were annotated based on the gene expression of cell type-specific markers. A total of 21 single-cell clusters were revealed by a t-SNE plot (Fig. 1B). We focused on clusters of LSECs (cluster 0, 1, 5, 12, 17; Fig. 1B, C) characteristically expressing Cd31 and Vegfr (vascular endothelial growth factor)-3 (the EC marker genes; Fig. 1D). The cluster 0, 1 and 5 of LSECs were reclassified (cluster 12 and 17 with too few cells were omitted). Seven new clusters were identified between 2 groups (Fig. 1E). Additionally, LSECs of cluster 3, 4, 5 and 6, which had limited numbers, were omitted. A heatmap of the top 10 representative DEGs of all 7 clusters was shown (Fig. 1F). The spatial distribution of each cluster was determined based on the expression of well-known landmark genes using t-SNE and violin plots. Lyve1 (lymphatic vessel endothelial receptor 1) and Stab 2 (stabilin 2), known as LSEC markers, were expressed in most cells of cluster 0 and 1 (Fig. 1G). Interestingly, cells of cluster 1 also expressed periportal landmark such as Efnb2 (recombinant ephrin b2, Fig. 1G). As a VEC marker, Vwf (von Willebrand factor) was uniquely expressed in cells of cluster 2 and 3 (Fig. 1G). Rspo3 (recombinant R-spondin 3), Wnt9b and Wnt2 are markers of central and pericentral VECs. Rspo3 and Wnt9b were specifically expressed in cells of cluster 2 (Fig. 1G), whereas Wnt2 was primarily expressed in cells of cluster 2 and 0 (Fig. 2A). The scRNA-seq technology differentiated 3 special clusters of LSECs in NASH disease. Collectively, cluster 0 and 1 were defined as LSECs, and cluster 2 was defined as VECs.

Fig. 1

LSEC scRNA-seq analysis. A Liver single-cell isolation, detection and analysis workflow in MCD-induced NASH and control mice (each group n = 3). B t-SNE visualization of clusters based on the single-cell transcriptome. A total of 21 single-cell clusters (0–20) and 5 single-LSEC clusters (0, 1, 5, 12, 17) was shown. Each dot represented a single-cell, and each color represented a cluster. C t-SNE plots showed 4 clusters of single-LSEC population (cluster 0, 1, 5, 12) in control and NASH mice. D Paired t-SNE and violin plots showed the expression of marker genes of LSECs: Cd31 and Vegfr-3. E t-SNE plots showed 7 new clusters of single-LSEC population (cluster 0–6) in control and NASH mice. F Heatmap of the top 10 representative DEGs of LSEC clusters (cluster 0–6). G Paired t-SNE and violin plots showed the expression of landmark DEGs: Lyvel, Stab 2, Efnb2, Vwf, Rspo3 and Wnt9b

Fig. 2

Transcriptomic scRNA-seq and PCR analysis of clusters 0 and 1. In cluster 0: A Paired t-SNE and violin plots showed the expression of C-Kit, Cntfr, Gmpr and Wnt2. B, C GO and KEGG analysis. D The mRNA levels of C-Kit, Cntfr, Gmpr and Wnt2 in cluster 0 were examined by qPCR in pLSEC-Con and pLSEC-MCD group. In cluster 1: E Paired t-SNE and violin plots showed the expression of Msr1, Efnb1, Efnb2 and Il1a. F, G GO and KEGG analysis. H The mRNA levels of Msr1, Efnb1, Efnb2 and IL1a in cluster 1 by qPCR in pLSEC-Con and pLSEC-MCD group. In D and H, p-value indicated statistical significance compared to pLSEC-Con group

A cluster of C-Kit +-LSECs was identified in NASH

Cluster 0 was appraised based on the special expression of the top 10 representative DEGs (Table S4), including C-Kit, Cntfr (ciliary neurotrophic factor receptor), Gmpr (guanosine monophosphate reductase), Wnt2 (Fig. 2A) and 6 other genes (Akr1b8, Gm10600, Slc9a9, Id4, Mmd, and Rasgrp2, Additional file 1: Fig. S2A–F and Additional file 1: Table S4). C-Kit was the most representative marker of cluster 0 since nearly 67% LSECs in cluster 0 were C-Kit+ [percentage fold change (pct-FC) was the 1st, Table S4]. Analysis of GO and KEGG, DEGs of these LSECs were associated to regulation of hyaluronic acid (HA), ERK1/2, VEGFR, mesenchymal cell proliferation and PI3K-AKT pathway (Fig. 2B, C). The pLSECs were isolated from control and MCD mice (pLSEC-Con and pLSEC-MCD group). Then, the top 4 representative DEGs of cluster 0 were examined by qPCR. Compared with pLSEC-Con group, Cntfr, Gmpr and Wnt2 mRNA were upregulated, while C-Kit mRNA was downregulated in pLSEC-MCD group (p < 0.05, Fig. 2D). Therefore, a subgroup of C-Kit+-LSECs belonging to cluster 0 was identified, and they might participate in HA, ERK1/2, VEGFR and PI3K-AKT signaling transduction in NASH.

Another cluster of Msr1 +-LSECs was found in NASH

The top 10 representative DEGs of cluster 1 were Msr1 (macrophage scavenger receptor 1), Efnb1 (ephrin b1), Efnb2, Il1a (interleukin 1a) (Fig. 2E and Additional file 1: Table S4) and 6 other genes (Serpina3f, Lama4, Myzap, Dll4, Galnt15, and Chst2, Additional file 1: Fig. S3A–F and Additional file 1: Table S4). Msr1 was the most representative marker of cluster 1 since nearly 73% LSECs in cluster 1 were Msr1+ (pct-FC was the 1st, Additional file 1: Table S4). DEGs of these LSECs mainly regulated EC migration, vasculature development and angiogenesis (Fig. 2F, G). Compared with pLSEC-Con group, Msr1 and Efnb1/2 mRNA were upregulated, while Il1a was downregulated in pLSEC-MCD group (p < 0.05, Fig. 2H). Finally, another subgroup of Msr1+-LSECs classified under cluster 1 was found, and they appeared to participate in the regulation of endothelial functions in NASH.

The third cluster of Bmp4 +Selp +-VECs was revealed in NASH

The top 10 representative DEGs of cluster 2 included Tgfb2, Fmo2, Prss23, Samd5 (sterile alpha motif domain 5), Bmp4 (bone morphogenetic protein 4), Col6a3 (collagen 6α3), Gpm6a (glycoprotein m6a), Fstl1, Selp (selectin P, also LECAM3, CD62) and Rbms3 (Fig. 3A and Additional file 1: Fig. S4A–E and Table S4). In cluster 2, 70–75% VECs were Bmp4+Selp+ (pct-FC was the 5th and 9th, Additional file 1: Table S4). DEGs of these VECs mainly regulated collagen, extracellular matrix (ECM), and atherosclerosis pathways (Fig. 3B, C). In cluster 2, all of the top 10 representative DEGs were examined by qPCR. Compared with pLSEC-Con group, Samd5, Col6a3 and Gpm6a mRNA were upregulated, while Bmp4 and Selp mRNA were downregulated in pLSEC-MCD group (p < 0.05, Fig. 3D). Therefore, a subgroup of hepatic Bmp4+Selp+-VECs from cluster 2 was revealed, and they were probably involved in the regulation of fibrosis and atherosclerosis in NASH.

Fig. 3

Transcriptomic scRNA-seq and PCR analysis of cluster 2. A Paired t-SNE and violin plots showing the expression of Samd5, Col6a3, Gpm6a, Bmp4 and Selp. B, C GO and KEGG analysis. D The mRNA levels of Samd5, Col6a, Gpm6a, Bmp4 and Selp in cluster 2 by qPCR in pLSEC-Con and pLSEC-MCD group. In D, p-value indicated statistical significance compared to pLSEC-Con group

C-Kit +-LSECs could improve NASH and mitophagy in vitro

Among 3 clusters of LSECs differentiated by scRNA-seq, the cell number of Bmp4+Selp+-VECs was too limited and the mechanism of Msr1 in NASH was already intensively clarified [7]. Then we focused on the C-Kit+-LSECs whose distinct roles in the pathogenesis of NASH should be fully elucidated.

Flow cytometric analysis revealed an obvious decreased percentage of CD31+C-Kit+-pLSECs derived from MCD mice compared to control mice (pLSEC-Con vs. -MCD group: 41.9% vs. 31.0%, p < 0.05, Fig. 4A, B). To explore the influence of C-Kit+-LSECs on peripheral cells, including HCs and HSCs, in a steatotic environment, we cocultured pHCs or pHSCs with C-Kit+- or C-Kit−-pLSECs in PA treatment. Significantly decreased lipid droplets were observed in pHCs cocultured with C-Kit+-pLSECs in comparison with C-Kit−-pLSECs (C-Kit+- vs. C-Kit−-pLSECs group: 0.65-fold, p < 0.05, Fig. 4C, D). TNF-α proteins (green IF) in pHCs and α-SMA proteins (red IF) in pHSCs were obviously reduced when cocultured with C-Kit+-pLSECs than C-Kit−-pLSECs (C-Kit+- vs. C-Kit−-pLSEC group: TNF-α was 0.23-fold, p < 0.05, Fig. 4C, E; α-SMA was 0.40-fold, p < 0.05, Fig. 4C, F). Also, mRNA of TNF-α and α-SMA were downregulated in cells cocultured with C-Kit+-pLSECs versus C-Kit−-pLSECs (Fig. 4G). Costaining of LC3B (autophagy proteins, red IF) and COX4 (mitochondrial proteins, green IF) shown orange IF. Interestingly, the manifestation of orange pHCs cocultured with C-Kit+-pLSECs was 3.36-fold higher than those with C-Kit−-pLSECs (p < 0.05, Fig. 5A, B). The mitochondrial ROS products (red IF of mtSOX, Fig. 5A, C) or damaged mitochondria (red IF of mtKeima, Fig. 5A, D) were 0.40-fold and 0.46-fold lower in pHCs cocultured with C-Kit+-pLSECs than with C-Kit−-pLSECs, respectively (p < 0.05).

Fig. 4

C-Kit+-LSECs alleviate NASH in vitro. A, B CD31+C-Kit+-pLSECs isolated from Con and MCD mice were detected by flow cytometry. C–G pHCs or pHSCs were cocultured with PA-treated C-Kit+- or C-Kit−-pLSECs. ORO staining C and calculation D of lipid droplets in pHCs (× 400). IF staining C and calculation E of TNF-α (green IF) in PHCs (× 400). IF staining C and calculation F of α-SMA (red IF) in pHSCs (× 400). DAPI (blue) was used for nuclear staining. (G) The mRNA of TNF-α (in pHCs) and α-SMA (in pHSCs) was detected by qPCR. (H-M) HepG2 or LX2 cells were cocultured with 6 groups of TMNK-1 cells (BSA, PA, sh-NC + PA, sh-C-Kit + PA, ov-NC + PA, ov-C-Kit + PA). ORO staining H and calculation I in HepG2 cells (× 1000). IF staining () and calculation J of TNF-α (red IF) in HepG2 cells (× 500). IF staining H and calculation K of α-SMA (red IF) in LX2 cells (× 200). DAPI (blue) was used for nuclear staining. The mRNA of L lipid metabolism genes (APDN, FXR, PPAR-α, and LXR) and M inflammation and fibrosis genes (TNF-α, IL-6, Col1a, and α-SMA) was examined by qPCR. The p-value indicated statistical significance compared to the pLSEC-Con, C-Kit−-pLSEC, BSA, sh-NC + PA or ov-NC + PA group

Fig. 5

C-Kit+-LSECs improve the mitochondrial function of HCs in vitro. A–D pHCs were cocultured with PA-treated C-Kit+- or C-Kit−-pLSECs. IF staining and calculation of LC3B (red IF) and COX4 (green IF, A/B), mtSOX (red IF, A/C), and mtKeima (red IF, A/D) in pHCs (× 400). DAPI (blue) was used for nuclear staining. E–G HepG2 cells were cocultured with 6 groups of TMNK-1 cells. E, F IF staining and calculation of LC3B (red IF) and COX4 (green IF) in HepG2 cells (× 500). DAPI (blue) was used for nuclear staining. G The mRNA of mitophagy-related genes (PINK1, Parkin, LC3B) was examined by qPCR. The p-value indicated statistical significance compared to the C-Kit−-pLSEC, BSA, sh-NC + PA or ov-NC + PA group

Then, we cocultured HepG2/LX2 cells with 6 groups of TMNK-1 cells for additional validation of the effect of C-Kit. HepG2, cocultured with PA treated TMNK-1 cells, showed more lipid accumulation (Fig. 4H, I), downregulation of pro-lipolysis genes (ADPN: adiponectin, FXR: farnyl derivative X receptor, PPAR-ɑ: peroxisome proliferator-activated receptor-α, Fig. 4L) and upregulation of pro-lipogenesis genes (LXR: liver X receptor, Fig. 4L) compared to those cocultured with BSA treated TMNK-1 cells. Meanwhile, HepG2/LX2 cells cocultured with TMNK-1 cells of PA group displayed more TNF-α/α-SMA proteins (red IF, Fig. 4H, J and K), and upregulation of pro-inflammation and pro-fibrosis genes (TNF-α/IL-6, α-SMA/Col1a, Fig. 4M), compared to those cocultured with TMNK-1 cells of BSA group. Coculturing with TMNK-1 cells of C-Kit deficiency (sh-C-Kit) could aggravate the above lipotoxic damage to HepG2/LX2 cells, while coculturing with TMNK-1 cells of overexpressing C-Kit (ov-C-Kit) could reverse the above lipotoxic injury, compared to those with control cell groups (p < 0.05, Fig. 4H, M). Next, Pink1 (PETN-induced putative kinase 1)-mediated mitophagy pathway was detected. After incubation with PA-treated TMNK-1 cells, HepG2 cells revealed significantly decreased LC3B/COX4 costaining (Fig. 5E, F) and lower mRNA levels of Pink1, Parkin, and LC3B (Fig. 5G) than with BSA-treated cells, suggesting that Pink1-mediated mitophagy in HCs was inhibited. Additionally, incubation with sh-C-Kit TMNK-1 cells could repress Pink1-related mitophagy pathway in HepG2 cells to a greater extent than those with sh-NC cells (p < 0.05, Fig. 5E, G). Conversely, incubation with ov-C-Kit TMNK-1 cells might significantly improve Pink1-related mitophagy in HepG2 cells compared to those with ov-NC cells (p < 0.05, Fig. 5E, G). Therefore, C-Kit+-LSECs would alleviate NASH by improving hepatic steatosis, inflammation, fibrosis and mitophagy in vitro.

C-Kit +-LSECs could alleviated NASH and recovery mitophagy in vivo

Lower percentage of C-Kit+CD31+ cell (showed orange IF staining) was seen in hepatic sinusoids of MCD mice than control mice (MCD vs. control group: 0.37-fold, p < 0.05, Fig. 6A, B). To determine the state of C-Kit+-LSECs in real-world NASH, we also checked the percentage of C-Kit+CD31+ cells in severe NASH and AIH patients (as control). The livers of AIH patients contained abundant C-Kit+CD31+ cells, but the livers of severe NASH patients showed rare C-Kit+CD31+ cells in hepatic sinusoids (NASH vs. AIH group: 0.31-fold, p < 0.05, Fig. 6C, D). To determine the remedy function of C-Kit+-LSECs in NASH in vivo, we transplanted C-Kit+- or C-Kit−-BMCs into MCD-induced NASH mice (representing MCD_C-Kit+-BMC and MCD_C-Kit−-BMC group). Relative to MCD_C-Kit−-BMC mice, hepatic steatosis (Fig. 6E, F), lobular inflammation (Fig. 6E, G) and fibrosis (Fig. 6E, H) were significantly alleviated in MCD_C-Kit+-BMC mice (p < 0.05). The mRNA and protein levels of C-Kit (Fig. 6I, G and K), PPAR-α and FXR (Fig. 7A, D and E) were consistently higher, while TNF-α (Fig. 7B, D and E) and α-SMA (Fig. 7C, D and E) were accordantly lower in MCD_C-Kit+-BMC mice than in MCD_C-Kit−-BMC mice (p < 0.05). Then, the transition of Pink1-mediated mitophagy was examined in vivo. In MCD_C-Kit−-BMC mice, the IF value of hepatic costaining of LC3B/COX4 was increased by 2.89-fold compared with that in MCD_C-Kit−-BMC mice (p < 0.05, Fig. 7F, G). Compared to those in MCD_C-Kit−-BMC mice, the mRNA and protein levels of Pink1, Parkin and LC3B were significantly increased, and those of p62 were obviously decreased in MCD_C-Kit+-BMC mice (Fig. 7H, I and J). These results suggested that BMT of C-Kit+-BMCs could ameliorate Pink1-mediated mitophagy and MCD-induced NASH in vivo.

Fig. 6

C-Kit+-LSECs were lack in NASH and BMT of C-Kit+-BMCs could protect against NASH in vivo. A, B Representative images and calculation of IF staining for C-KIT (green IF) and CD31 (red IF) in liver tissues of control and MCD mice (× 200). DAPI (blue) was used for nuclear staining. Arrows (yellow) indicated C-Kit+CD31+-LSECs. C, D Representative images and calculation of IF staining for C-KIT (green IF) and CD31 (red IF) in liver tissues of NASH and AIH patients (× 400). DAPI (blue) was used for nuclear staining. C–K We transplanted C-Kit+- or C-Kit−-BMCs into MCD mice. Images and calculation of ORO (E/F), H&E (E/G) and Masson (E/H) staining of liver tissues in 2 groups. I Hepatic C-Kit mRNA was examined by qPCR in 2 groups. J, K Hepatic protein levels of C-KIT were examined by western blot in 2 groups. The p-value indicated statistical significance compared to control mice, AIH patients or MCD_C-Kit−-BMC mice

Fig. 7

BMT of C-Kit+-BMCs might improve Pink1-mediated mitophagy and NASH in vivo. We transplanted C-Kit+- or C-Kit−-BMCs into MCD mice. A–C Hepatic mRNA was examined by qPCR in 2 groups: A APDN, PPAR-α, FXR, LXR and SREBP-1c; B TNF-α and IL-6; C Col1a and α-SMA. D, E Hepatic protein levels of PPAR-α, FXR, α-SMA, and TNF-α were examined by western blot in 2 groups. F, G IF staining and calculation of COX-4 (green IF) and LC3B (red IF) in liver tissues of 2 groups (× 400). DAPI (blue) was used for nuclear staining. The liver mRNA H and protein I, J levels of Pink1-mediated mitophagy pathway were examined by qPCR and western blot in 2 groups. The p-value indicated statistical significance compared to MCD_C-Kit−-BMC mice