Age-related liver endothelial zonation triggers steatohepatitis by inactivating pericentral endothelium-derived C-kit

Aging leads to systemic metabolic disorders, including steatosis. Here we show that liver sinusoidal endothelial cell (LSEC) senescence accelerates liver sinusoid capillarization and promotes steatosis by reprogramming liver endothelial zonation and inactivating pericentral endothelium-derived C-kit, which is a type III receptor tyrosine kinase. Specifically, inhibition of endothelial C-kit triggers cellular senescence, perturbing LSEC homeostasis in male mice. During diet-induced nonalcoholic steatohepatitis (NASH) development, Kit deletion worsens hepatic steatosis and exacerbates NASH-associated fibrosis and inflammation. Mechanistically, C-kit transcriptionally inhibits chemokine (C–X–C motif) receptor (CXCR)4 via CCAAT enhancer-binding protein α (CEBPA). Blocking CXCR4 signaling abolishes LSEC–macrophage–neutrophil cross-talk and leads to the recovery of C-kit-deficient mice with NASH. Of therapeutic relevance, infusing C-kit-expressing LSECs into aged mice or mice with diet-induced NASH counteracts age-associated senescence and steatosis and improves the symptoms of diet-induced NASH by restoring metabolic homeostasis of the pericentral liver endothelium. Our work provides an alternative approach that could be useful for treating aging- and diet-induced NASH. The authors show that liver sinusoidal endothelial cell (LSEC) senescence promotes steatosis by reprogramming liver endothelial zonation and inhibiting C-kit, a type III receptor tyrosine kinase. Infusing C-kit-expressing LSECs in aged or diet-induced NASH mice counteracts senescence and steatosis.

Aging leads to systemic metabolic disorders, including steatosis. Here we show that liver sinusoidal endothelial cell (LSEC) senescence accelerates liver sinusoid capillarization and promotes steatosis by reprogramming liver endothelial zonation and inactivating pericentral endothelium-derived C-kit, which is a type III receptor tyrosine kinase. Specifically, inhibition of endothelial C-kit triggers cellular senescence, perturbing LSEC homeostasis in male mice. During diet-induced nonalcoholic steatohepatitis (NASH) development, Kit deletion worsens hepatic steatosis and exacerbates NASH-associated fibrosis and inflammation. Mechanistically, C-kit transcriptionally inhibits chemokine (C-X-C motif) receptor (CXCR)4 via CCAAT enhancer-binding protein α (CEBPA). Blocking CXCR4 signaling abolishes LSEC-macrophage-neutrophil cross-talk and leads to the recovery of C-kit-deficient mice with NASH. Of therapeutic relevance, infusing C-kit-expressing LSECs into aged mice or mice with diet-induced NASH counteracts age-associated senescence and steatosis and improves the symptoms of diet-induced NASH by restoring metabolic homeostasis of the pericentral liver endothelium. Our work provides an alternative approach that could be useful for treating aging-and diet-induced NASH.
Age-associated systemic metabolic disorders are recognized as major lifespan determinants [1][2][3] . Aging leads to deterioration of organ function, likely by accelerating vascular aging [4][5][6] . The fundamental changes in liver vasculature that occur with age and their role in the regulation of metabolic disorders, particularly NASH, are poorly understood.
Aging-induced metabolic disorders include steatosis, which evolves from excessive hepatic fat deposition 7,8 . Thus far, no Food and Drug Administration-approved therapeutic drugs have been developed for the treatment of NASH 9 . Historically, research has focused on unveiling the underlying mechanisms of aging or dietassociated NASH pathogenesis by studying hepatocytes 10,11 . However, growing evidence suggests that LSECs play pivotal roles in the regulation of NASH development [12][13][14] . Under pathological conditions, the number of LSEC fenestrae is reduced and the endocytosis of LSECs is impaired, which likely disrupts lipid exchange between hepatocytes and the blood, leading to hepatic lipid accumulation 15,16 . Furuta et al. reported that LSEC-derived vascular cell adhesion protein 1 (VCAM-1) promoted NASH pathogenesis 17 . Zhang et al. showed that targeting the maladapted endothelial niche alleviated fibrosis in NASH 18 . Recently, our group also found that Notch signalingmediated disruption of LSEC homeostasis aggravated diet-induced NASH in mice 14 . Article https://doi.org/10.1038/s43587-022-00348-z LSECs, which have gatekeeper roles in maintaining liver metabolic zonation 20 , can be roughly classified into three subpopulations: periportal (zone 1), sinusoidal (zone 2) and pericentral (zone 3) LSECs [21][22][23] .
The liver endothelium exhibits spatial heterogeneity along the axis of the liver lobule 19 . Single-cell transcriptomics has revolutionized the understanding of the spatial architecture of the liver endothelium 19 .   Although the spatial and molecular heterogeneity of LSECs is well recognized, the functional heterogeneity of the liver endothelium in manipulating hepatic pathologic challenges (including liver regeneration, fibrosis, steatosis and cancer) is understudied 24,25 . Whether LSECs originating from different hepatic zones have distinct roles in the pathogenesis of liver metabolic disease should be fully elucidated.
C-kit, also known as CD117, is a type III receptor tyrosine kinase 26,27 . In the liver, the distribution of C-kit shows spatial differences similar to those of some Wnt ligands, such as Wnt2 and Wnt9b 28 , which are regarded as markers of pericentral LSECs. Thus, C-kit has been used to label LSECs originating from the pericentral region of the liver [29][30][31] . We recently found that implantation of C-kit + LSECs into mice ameliorates toxin-induced liver injury by promoting hepatocyte regeneration through angiocrine signaling 31 . Thus, we were interested in investigating whether the pericentral liver endothelium protects against NASH.
In this study, we found that the normal expression and distribution of phenotypic and metabolic markers of the pericentral liver endothelium was disrupted in mice with aging-or diet-induced NASH. Inhibition of endothelium-derived C-kit aggravated hepatic steatosis and fibrosis by facilitating macrophage and neutrophil recruitment via CXCR4 signaling. Ultimately, implantation of C-kit + LSECs into mice improved their diet-or aging-induced NASH by restoring the metabolic homeostasis of the pericentral liver endothelium.

Aging is associated with hepatic steatosis in mice and humans
Models of aging were established using 24-month-old mice. With the exception of a reduction in the liver-to-body weight ratio, no significant changes in liver function or histology were detected in aged mice (Extended Data Fig. 1a-c). Senescence-associated β-galactosidase (SA-βgal), a hallmark of senescence, was used to identify senescent cells in the livers of aged mice. Compared to livers of younger (8-week-old) control mice, those of aged mice had more senescent cells (Fig. 1a). Oil Red O, Sirius Red and anti-F4/80 antibody staining showed signs of fat deposition deterioration, fibrosis and inflammation in the livers of aged mice (Fig. 1b), indicating that aging is associated with steatosis. Hepatic triglyceride (TG) content was higher in the livers of aged mice than that in control animals (Fig. 1c). Gene set enrichment analysis (GSEA) of RNA-sequencing (RNA-seq) data revealed that pathways associated with fibrosis and inflammation were activated in the LSECs of aged mice (Fig. 1d,e). Conversely, genes related to fatty acid oxidation and degradation were inactivated in the LSECs of aged mice (Fig. 1f,g), which proved that aging increased abnormal fat accumulation in the liver.
Additionally, clinical liver samples collected from patients (Supplementary Table 3) showed no difference in body mass index (BMI) between young and old groups (Fig. 1i). However, increased Oil Red O staining (Fig. 1h), elevated hepatic TG and total cholesterol (TC) (Fig.  1j), upregulated mRNA levels of senescence markers such as p16 and p21 and decreased expression of the oxidation-related gene PPARA (Fig. 1k) were observed in the livers of older patients, compared with controls. This suggests that older individuals exhibited more severe steatosis. To confirm the above findings, a public database was interrogated. Liver histology analysis of older individuals showed evidence of slight hepatocyte ballooning (Extended Data Fig. 1d). Furthermore, GSEA proved that fatty acid biosynthesis was enhanced in older mice and people alike (Extended Data Fig. 1e,f). These data collectively confirmed the existence of aging-associated steatosis.

Aging aggravates high-fat diet-induced metabolic disorders
After discovering the existence of aging-associated hepatic steatosis, we wished to clarify whether aged mice were more susceptible to dietinduced metabolic disorders. Thus, 24-month-old aged mice and their controls (8 weeks old) were subjected to a choline-deficient l-aminodefined (CDAA) diet for 10 weeks. We found that aging increased the liver-to-body weight ratio (Extended Data Fig. 2a), hepatic TG content (Extended Data Fig. 2b), hepatic senescence, hepatocyte ballooning and Oil Red O, Sirius Red and α-smooth muscle actin (αSMA) staining in mice fed a CDAA diet (Extended Data Fig. 2c). This indicates that aging aggravated diet-induced metabolic disorders.
Aging disrupts sinusoidal homeostasis and reshapes liver endothelial zonation by downregulating pericentral endothelial Kit Fig. 1a shows that SA-β-gal + cells distribute along the liver sinusoids. To clarify the origin of senescent cells, hepatocytes and liver nonparenchymal cells (NPCs) were isolated and characterized by quantitative PCR with reverse transcription (qRT-PCR). We found that LSECs highly expressed transcripts for senescence markers such as interleukin (IL)-1β, IL-1α, GATA4, p16, p53, C-C motif chemokine ligand 2 (CCL2), p21 and p27 (Fig. 2a). To confirm this finding, flow cytometry Fig. 1 | Aging-induced hepatic steatosis. a, SA-β-gal staining of liver samples collected from control (CT; 8-week-old) and aged (24-month-old) mice fed with a normal diet. A representative image of the aged liver is enlarged to better illustrate liver morphology. SA-β-gal + cells were quantified in control and aged groups of mice (n = 6, P = 0.0004). HPF, high power field. b, Oil Red O (n = 5, P = 0.0448), Sirius Red (n = 4, P = 0.0174) and F4/80 IF staining (n = 6, P = 0.0691) of control and aged liver samples. Positively staining areas were quantitatively compared. c, Hepatic TG content of control and aged mice (n = 6, P = 0.0161). d, Heatmaps showing expression of fibrosis-and inflammation-associated genes in LSECs of control and aged mice. e, GSEA of the RNA-seq data shown in d. ES, enrichment score. f, A heatmap showing expression of fatty acid-oxidation-related genes in LSECs of control and aged mice. g, GSEA of genes associated with fatty acid oxidation and degradation in the LSECs of control and aged mice. h, Oil Red O staining of liver samples collected from younger and older patients. Positively staining areas were quantitatively compared (n = 6, P = 0.0043). i, Age and BMI were compared between the younger and older groups (n = 6). j, A comparison of hepatic TG (n = 6) and TC (n = 8) contents of younger and older patients. k, The relative mRNA levels of P16 (Cdkn2a) (P = 0.0431), P21 (Cdkn1a) (P = 0.0409) and PPARA (P = 0.0437) were determined by qRT-PCR. The transcript for β-actin was used as an internal control (n = 5). Two-tailed Student's t-test was used for comparisons of two groups. Bars represent mean ± s.d.; NS, not significant; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Fig. 2 | Aging reshaped liver endothelial zonation and inactivated pericentral
endothelium-derived C-kit. Aged mice and their controls were fed with a normal diet. a, The relative mRNA levels of Il1b, Il1a, Gata4, p16, Trp53, Ccl2, p21 and p27 (Cdkn1b) were compared among hepatocytes (HCs), hepatic stellated cells (HSCs), LSECs and Kupffer cells (MΦ) isolated from aged mice (n = 3). Differences among multiple groups were compared using one-way ANOVA. b, Flow cytometry analysis of liver NPCs isolated from aged mice. VEGFR2 + cells were identified by first gating on SA-β-gal + cells. c, IF staining of aged liver samples for VE-cadherin (red), SA-β-gal (green) and Hoechst (blue). d, SEM images of control and aged liver samples. e, VE-cadherin (P = 0.0387) and LYVE1 (P = 0.0002) IF staining and SEM (P = 0.0162) images of control and aged livers. Positive areas of staining (n = 6) or the number of fenestrae (n = 5) were quantitatively compared. f, GSEA of expression of the top 50 genes (associated with liver EC-specific phenotypes and liver arterial, venous and capillary EC markers) in LSECs of control or aged mice. g, GSEA of pericentral and periportal EC marker genes in LSECs of control and aged mice. h, Comparison for the relative Kit mRNA level in LSECs isolated from control or aged mice (n = 3, P = 0.0059). i, C-kit IF staining of liver samples from control or aged mice. Positive areas of C-kit staining were quantitatively compared (n = 5, P = 0.0004). Two-tailed Student's t-test was used for comparisons of two groups. Bars represent mean ± s.d.; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.  analysis of liver NPCs from aged mice was performed. We found that 69.2% of SA-β-gal + cells were VEGFR2 + (Fig. 2b). Immunofluorescence (IF) staining further proved that senescent cells labeled using SA-βgal, p27 or the senescence staining kit were colocalized with vascular endothelial (VE)-cadherin + LSECs ( Fig. 2c and Extended Data Fig. 3a,b), suggesting that LSECs were the major contributors to the induction of aging-associated senescence. We therefore next investigated changes in the liver sinusoids of aged mice. Interestingly, scanning electronic microscopy (SEM) revealed that the livers of aged mice had a unique filiform appearance (Fig. 2d). Moreover, IF showed that expression of VE-cadherin and LYVE1, which are markers of liver endothelial cells (ECs), was lower in the livers of aged mice than that in young controls (Fig. 2e). Using SEM, we observed that the number of fenestrae, which are associated with LSEC differentiation, was reduced in the livers of aged mice (Fig. 2e), suggesting that aging accelerated the capillarization of liver sinusoids. Next, we aimed to investigate changes in endothelial gene expression associated with liver aging. Thus, LSECs were isolated from aged and control mice. As previously demonstrated 32 , liver ECs can be classified into distinct subpopulations. According to the GSEA of RNAseq data, the top 50 marker genes for liver ECs, arterial ECs, capillary ECs and venous ECs were significantly downregulated in aged mice ( Fig. 2f), implying that expression patterns of liver endothelial specification were disrupted by aging. We next compared expression of pericentral and periportal liver endothelial markers in LSECs derived from aged and control mice. Expression of pericentral and periportal liver EC-associated landmark genes was identified using sequencing data provided by Halpern et al. 33 . Interestingly, pericentral but not periportal EC markers were significantly downregulated on aging (Fig. 2g). Previously, we proved that C-kit was a marker of pericentral liver ECs, similar to Wnt9b, Wnt2a and R-spondin 3 (RSPO3) 20,31,34 . In the present study, we found that mRNA levels of Kit were lower in LSECs of aged mice than those of the control group (Fig. 2h). IF staining of C-kit proved that aging reduced hepatic C-kit expression (Fig. 2i). In mice maintained on a CDAA diet, aging also decreased the expression of C-kit and VE-cadherin (Extended Data Fig. 2d). These data collectively imply that the downregulation of endothelium-derived C-kit was likely associated with aging-related hepatic steatosis.

NASH inactivates pericentral liver endothelial markers, including C-kit
Next, we tried to clarify whether inactivation of endothelial C-kit was responsible for diet-induced NASH. According to mRNA-seq data, pericentral marker genes were extensively downregulated in NASH models; however, expression of periportal endothelial marker genes was not significantly altered (Extended Data Fig. 4a). Data from the public database confirmed that expression of Kit was downregulated in different NASH models (Extended Data Fig. 4b). Additionally, we evaluated the mRNA and protein levels of C-kit in LSECs isolated from mice with methionine-choline-deficient (MCD) diet-induced NASH. Western blot and qRT-PCR analyses confirmed that, in LSECs, C-kit was inactivated during the progression of NASH (Extended Data Fig. 4c,d). IF staining also showed evidence of reduced hepatic C-kit expression in MCD and CDAA diet-induced mouse models (Extended Data Fig. 4e,f). Collectively, these findings suggested that the inactivation of endothelial C-kit might be associated with the pathogenesis of dietinduced NASH as well.

Disruption of endothelial C-kit expression alters endothelial and metabolic gene profiles of LSECs
To precisely illustrate the role of endothelial C-kit in NASH, we established endothelial Kit-knockout (KO) mice by crossing Cdh5-Cre/ERT mice with Kit fl mice, as shown in Extended Data Fig. 5a. Western blot and qRT-PCR analyses confirmed successful depletion of Kit in the LSECs of these mice (Extended Data Fig. 5b,c). We observed no differences in serum parameters or morphology between endothelial Kit-KO and control mice, except for a significant increase in Oil Red O staining and increased liver-to-body weight ratios in Kit-KO mice (Extended Data Fig. 5d-g).
Liver endothelial gene expression profiles were markedly altered following Kit deletion. By reanalyzing previously published sequencing data 32 , we found that most of the top 50 liver EC marker genes were downregulated in LSECs of Kit-KO mice (Extended Data Fig. 6a), as were the top 50 metabolic marker genes (Extended Data Fig. 6b), suggesting that endothelium-derived C-kit played a critical role in maintaining LSEC homeostasis. Next, expression profiles of the top 50 arterial, venous and capillary EC marker genes were analyzed in LSECs of Kit-KO mice. Interestingly, we found that only capillary ECs, the majority of which are LSECs, lost their characteristic genotypes (Extended Data Fig. 6a). However, expression of arterial and venous EC marker genes fluctuated in Kit-KO LSECs (Extended Data Fig. 6a). GSEA confirmed the above findings (Extended Data Fig. 6c,d).

Endothelial C-kit depletion promotes LSEC senescence and aggravates hepatic steatosis
Next, we observed the impact of Kit deletion on cellular senescence. We used SA-β-gal staining to show increased accumulation of senescent cells in the livers of Kit-KO mice (Fig. 3a); most of these cells were LSECs (Extended Data Fig. 3c). Heatmap analysis and GSEA showed that aging-and senescence-associated pathways were activated in LSECs following C-kit depletion (Extended Data Fig. 6e and Fig. 3b).
We then pondered the impact of Kit deletion on hepatic lipid metabolism. Metabolic disorders were established in mice by feeding them an MCD or a CDAA diet for 6 or 10 weeks, respectively (Fig. 3c). Hematoxylin and eosin (H&E) and Oil Red O staining showed that endothelial C-kit depletion markedly aggravated hepatic steatosis and increased hepatic lipid contents of both MCD (Fig. 3d,f) and CDAA (Fig. 3e,g) diet-fed mice. In MCD diet models, expression of Acc1 (Acaca) and Fasn was upregulated in Kit-KO LSECs, while Acadvl was downregulated ( Fig. 3h). In mice fed the CDAA diet, expression of Fasn was upregulated, while that of Ppara, Acadvl and Mgll was downregulated (Fig. 3i), indicating that depletion of C-kit facilitated lipid synthesis and attenuated lipid oxidation and lipolysis. Besides, serum glucose levels were higher in Kit-KO mice fed a CDAA diet (Fig. 3j). All these findings prove that hepatic lipid metabolism was disturbed following C-kit depletion. Next, we evaluated the enrichment of hallmark genes in C-kit-deficient LSECs. Pathways associated with glycolysis, adipogenesis and oxidative phosphorylation were activated following the     Inflammation pathways Article https://doi.org/10.1038/s43587-022-00348-z loss of C-kit. In addition, pathways related to cholesterol homeostasis and fatty acid metabolism were also affected (Extended Data Fig. 6f). Moreover, heatmap analysis and GSEA confirmed that the process of hepatic lipid biosynthesis was greatly enhanced by depletion of endothelial C-kit (Fig. 3k,l).

Endothelial C-kit blockade provokes NASH-associated fibrosis and inflammation
Subsequently, we inspected whether NASH-associated liver fibrosis and inflammation were affected by the loss of C-kit. In mice with MCD diet-induced NASH, endothelial Kit KO increased the number of hepatic macrophages, basement membrane and liver fibrosis and impeded hepatocyte proliferation, which were identified by F4/80, laminin, Sirius Red and Ki67 staining, respectively (Fig. 4a). In accordance, mice fed a CDAA diet also exhibited aggravated fibrosis, inflammation and reduced hepatocyte-proliferation capacity (Fig. 4b). Next, molecules associated with fibrosis and inflammation were evaluated by qRT-PCR. In C-kit-depleted LSECs of mice fed the MCD diet, expression of Timp1, Col1agen-1 (Col1a1), Pdgfb, Il6 and Il1b was increased, compared with controls ( Fig. 4c,d). Likewise, mRNA levels of Sm22 (Tagln), αSMA (Acta2), Pdgfb, Tgfb (Tgfb1), Il1b, Tnfa (Tnf) and Il6 were upregulated after Kit KO in mice fed a CDAA diet (Fig. 4e,f), indicating that loss of endothelial C-kit exacerbated liver fibrosis and inflammation. The sequencing data further proved that hepatic inflammatory responses were largely promoted by Kit deletion (Fig. 4g,h).

Endothelial C-kit inhibits CXCR4 transcription via CEBPA
Because NASH often involves chemokine-mediated recruitment of inflammatory cells, we next investigated the alteration of chemokine pathways in LSECs from mice with different genotypes. As expected, heatmap analysis and GSEA confirmed extensive activation of chemokine signaling following C-kit depletion (Fig. 5a,b). Strikingly, CXCR4, one of the canonical chemokine receptors, was upregulated significantly at the mRNA level in LSECs from both control mice and mice fed the MCD diet following C-kit blockade (Fig. 5c). Depletion of endothelial C-kit also increased CXCR4 protein expression (Fig. 5d). Moreover, IF staining confirmed the increase in CXCR4 and stromal cell-derived factor 1 (SDF-1) expression in LSECs induced by C-kit deficiency in vitro (Fig. 5e). In addition, enzyme-linked immunosorbent assay (ELISA) was used to show that C-kit-deficient LSECs produced more SDF-1 in vitro (Extended Data Fig. 7a). We also evaluated CXCR4 expression in the livers of mice with NASH and found that, during the progression of CDAA diet-induced NASH, the mRNA level of Cxcr4 gradually increased (Extended Data Fig. 7b). Immunohistochemistry staining showed that CXCR4 expression was negatively correlated with C-kit expression in the livers of patients with NASH (Extended Data Fig. 7c). Collectively, these data imply that depletion of C-kit led to NASH development by upregulating CXCR4 expression.
To unravel the regulatory effect of C-kit on CXCR4, 109 genes that encode transcription factors and were significantly differentially expressed after Kit KO and 58 potential regulators of CXCR4 predicted from the Toolkit for Cistrome Data Browser (http://dbtoolkit.cistrome. org/) were analyzed. On comparing these lists of genes, we found nine that overlapped (Fig. 5f). We evaluated the top 20 potential regulators of CXCR4 according to their regulatory potential scores (Fig. 5g) and the top ten differentially expressed genes following C-kit disruption (Fig. 5h). Eventually, we decided to focus on Cebpa. After dissecting the promoter region of Cxcr4, we found that there were several putative binding sites for CEBPA, as predicted in previously published murine liver chromatin immunoprecipitation (ChIP)-sequencing data deposited in Cistrome Data Browser (http://cistrome.org/db/#/) (Fig. 5i). ChIP followed by quantitative PCR of isolated LSECs further showed that CEBPA bound strongly to the Cxcr4 promoter. Moreover, this binding could be markedly inhibited by C-kit inactivation (Fig. 5j,k), suggesting that C-kit inhibited CXCR4 by inhibiting the regulatory role of CEBPA.

Blocking CXCR4 abrogates macrophage and neutrophil recruitment in C-kit-deficient mice
As chemokine signaling modulates the recruitment of immune cells such as macrophages and neutrophils following injury, we explored pathways associated with macrophage and neutrophil biology. GSEA revealed that activation, migration and chemotaxis of macrophages and neutrophils were greatly enhanced after C-kit depletion (Fig. 6a,b). To clarify whether deletion of Kit promoted CXCR4-mediated macrophage and neutrophil recruitment, AMD3100, an inhibitor of CXCR4, was applied. Flow cytometry proved that the increase in CD11b + , F4/80 + , CD11b + F4/80 + and CD11b − F4/80 + liver macrophage numbers caused by C-kit depletion was not observed following AMD3100 administration in mice fed a CDAA diet (Fig. 6c,d).
To investigate cross-talk between LSECs and inflammatory cells, primary LSECs, macrophages and neutrophils were isolated and cultured in vitro (Fig. 6e). The transwell cell-migration assay  j,k, ChIP results. LSECs collected from wild-type, control or Kit-KO mice were subjected to ChIP using anti-CEBPA antibody, with preimmune immunoglobulin (Ig)G as a control. Precipitated chromatin fragments were further analyzed by qRT-PCR using primers spanning the most distal putative CEBPA-binding sites (n = 3). Two-tailed Student's t-test was used for comparisons of two groups. Differences among multiple groups were compared using one-way ANOVA. Bars represent mean ± s.d.; *P < 0.05, **P < 0.01, ***P < 0.001. Article https://doi.org/10.1038/s43587-022-00348-z demonstrated that enhanced migration of either macrophages or neutrophils to Kit-KO LSECs was abolished by AMD3100 (Fig. 6f,g). To avoid the effect of AMD3100 on myeloid subsets, short hairpin RNA (shRNA)-mediated Cxcr4 knockdown was adopted. As shown in Extended Data Fig. 7d,e, knockdown of Cxcr4 in cultured LSECs inhibited migration of macrophages and neutrophils, which indicated that cross-talk between liver ECs and myeloid cells relied on LSEC-specific CXCR4 signaling.

Loss of endothelial C-kit exacerbates diet-induced NASH by upregulating CXCR4
Subsequently, we investigated the role of CXCR4 in C-kit-regulated NASH progression. The strategy of AMD3100 administration is illustrated in Fig. 7a. H&E, Oil Red O, Sirius Red, myeloperoxidase (MPO) and F4/80 staining showed that blocking CXCR4 signaling successfully lowered the severity of hepatic steatosis, fibrosis and inflammation caused by C-kit deficiency in mice with either MCD (Fig. 7b,d)   Article https://doi.org/10.1038/s43587-022-00348-z (Fig. 7c,e) diet-induced NASH. Besides, elevated hepatic TC and TG content (Fig. 7f) and increased expression of Acc1 and Fasn (Fig. 7g) were counteracted by AMD3100 treatment in Kit-KO mice fed the CDAA diet.

Article
https://doi.org/10.1038/s43587-022-00348-z NASH. Serum and morphological analyses confirmed that mice with CDAA diet-induced liver malfunction and NASH could be treated with CXCR4 blockade (Extended Data Fig. 8a-c). In sum, these data prove that endothelial C-kit impeded NASH progression by counteracting the effect of CXCR4.

Splenic implantation of C-kit + LSECs into mice alleviates dietinduced NASH
As loss of endothelial C-kit accelerated NASH progression, we next explored whether regaining C-kit alleviated NASH. C-kit + LSECs were sorted using anti-CD117 antibody-coated magnetic beads. As shown by flow cytometry, the purity of C-kit + (CD117 + ) cells after isolation reached 97.77% (Extended Data Fig. 9a). C-kit + LSECs were infused into the spleens of mice fed the MCD diet (Fig. 8a); infusion of C-kit − LSECs was used as a negative control. One minute after infusion, spectral imaging showed the presence of fluorescently labeled C-kit + LSECs in the liver but not in the spleen, kidney, heart or lung (Extended Data Fig. 9b). Microscopy also confirmed the presence of CM-Dil Dye-labeled infused cells in the liver (Extended Data Fig. 9c). Even 12 or 24 h after cell infusion, a significant increase in hepatic fluorescence could be observed by spectral imaging (Extended Data Fig. 9d,e), implying that the infused cells could persist in the liver. Gene profiles were then compared between C-kit + and C-kit − LSECs by RNA-seq. The expression of chemokine receptors, pathways associated with immune cell recruitment and inflammation-related genes were all inactivated in C-kit + LSECs compared with C-kit − LSECs (Extended Data Fig. 10a-d). However, pericentral marker genes and the top 50 liver-specific EC and metabolic marker genes were upregulated in C-kit + LSECs (Extended Data Fig. 10d).
In the rescue experiment, implantation of C-kit + LSECs attenuated diet-induced hepatic steatosis, fibrogenesis and macrophage accumulation, while promoting hepatocyte proliferation by restoring hepatic C-kit levels (Fig. 8b) and inhibiting chemokine signaling, particularly CXCR4 (Fig. 8c). To confirm these molecular changes, RNA-seq was performed on LSECs isolated from mice with metabolic disorder following infusion of C-kit + LSECs. Heatmap analysis and GSEA confirmed the inactivation of chemokine-, inflammation-and fatty acid-biosynthesis-associated genes and the activation of fatty acid-oxidation-related genes (Extended Data Fig. 10g,h). These data suggest that mice with diet-induced NASH were rescued by infusion of C-kit + LSECs.

C-kit + LSEC infusion rescues mice from aging-associated senescence and steatosis
Ultimately, we performed rescue experiments with aged mice. RNAseq analysis showed that genes associated with aging and senescence were enriched in C-kit − LSECs (Extended Data Fig. 10e,f), implying that C-kit + LSECs had anti-aging attributes. Five days after cell infusion (Fig. 8d), morphological analyses showed that aging-associated senescence, steatosis, inflammation, C-kit inactivation and endothelial disruption were all reversed by the implantation of C-kit + LSECs. This was determined by SA-β-gal, Oil Red O, F4/80, C-kit and VE-cadherin staining, respectively (Fig. 8e). The decreased hepatic TG content (Fig. 8f) and downregulated mRNA levels of transcripts encoding senescence markers (including p16, p53, p27 and CCL2) (Fig. 8g) further confirmed the reversal of aging-associated senescence and steatosis. In the end, rescue with C-kit + LSECs restored pericentral endothelial markers and reduced expression of aging-and senescence-related genes (Extended Data Fig. 10i), implying that endothelial C-kit limited senescence by maintaining homeostasis of the pericentral liver endothelium.

Discussion
NASH accounts for a large proportion of aging-associated chronic liver diseases and is responsible for a sharp rise in morbidity and mortality. It is therefore important to understand the pathogenesis of aging-or dietinduced NASH and explore effective therapeutic strategies. Although hepatic lipotoxicity is the leading cause of steatohepatitis, growing numbers of studies suggest that LSECs play important roles in the development of NASH. In our recent investigation, homeostasis of the liver endothelium was disrupted in mice with diet-induced NASH 14 . In the present study, we demonstrated that aging perturbed the homeostasis of pericentral but not periportal ECs. Coincidentally, NASH-triggered LSEC maladaptation was also only observed in pericentral liver ECs 14 . These findings not only reveal the functional heterogeneity of LSECs but also imply that aging-and diet-induced forms of steatohepatitis probably develop via a shared mechanism.
C-kit is predominantly expressed in the pericentral (zone 3) liver endothelium. Aging-or diet-induced NASH downregulated pericentral liver endothelial landmark genes, including Kit. Once endotheliumderived C-kit was depleted, liver endothelial and metabolic homeostasis was disrupted, resulting in aggravated hepatic steatosis and NASH-associated fibrosis. Afterward, implantation of C-kit + LSECs counteracted diet-induced NASH and aging-associated steatosis. These findings imply that the C-kit + endothelium potentially protects hepatocytes against lipotoxicity. Previously, we reported that EC-derived Notch signaling aggravated diet-induced NASH 14 . Notch receptors and ligands are predominantly expressed by periportal (zone 1) liver ECs. Thus, pericentral endothelial C-kit decelerates NASH progression, while periportal EC-derived Notch has the opposite effect. Therefore, LSECs from different liver zones contribute distinctly to NASH pathogenesis.
As a chemokine receptor, CXCR4 modulates mobilization and recruitment of various immune cells [35][36][37][38] . In this study, we found that expression of CXCR4 was increased in LSECs lacking C-kit. Blocking CXCR4 signaling with AMD3100 diminished recruitment of macrophages and neutrophils into the liver of C-kit-deficient mice, decelerating NASH progression. Additionally, AMD3100 also reversed the increase in cellular senescence associated with the loss of C-kit, implying that C-kit-CXCR4-regulated cellular senescence accounts for the pathogenesis of NASH. Although our study indicated that upregulation of the transcription factor CEBPA following C-kit disruption contributes to increased expression of CXCR4, further experimental and mechanistic evidence is needed to elaborate this regulatory relationship.
Previously, Qing et al. reported that myeloid-derived yes-associated protein (YAP) stimulated macrophage-EC cross-talk in NASH 39 . In the current study, cross-talk between LSECs and myeloid cells was proven to be the determinant of NASH development. Inhibition of CXCR4 abrogated LSEC-induced migration of macrophages and neutrophils. Of note, flow cytometry confirmed that both CD11b + and CD11b − macrophages were affected by AMD3100 administration, indicating that both liver-resident and monocyte-derived macrophages were recruited via the C-kit-CXCR4 axis during NASH progression. However, more studies are needed to precisely characterize macrophages recruited in mice with NASH. Neutrophil extracellular traps have been reported to impede NASH 40,41 . In the present study, neutrophil recruitment was also modulated by the C-kit-CXCR4 axis. Whether In this study, we performed C-kit + LSEC implantation to rescue the loss of C-kit in aged mice or mice with NASH. To achieve this, we first performed tail vein injection; however, most of the infused cells were recruited to the lung and not the liver. To ensure that the implanted cells efficiently homed to the liver, splenic injection was performed. Spectral imaging confirmed successful homing of implanted cells to the liver and their persistence. Following infusion of C-kit + LSECs, the health of mice with liver metabolic disorders largely improved. Interestingly, chemokine and inflammatory signaling pathways were inactivated in the liver following C-kit + LSEC implantation. This finding not only shows that C-kit + LSECs improved the condition of mice with NASH by limiting the chemokine-mediated inflammatory response but also implies that expression patterns and therapeutic effects associated with infusion of these cells might persist over time. Additionally, in our recent study, we found that infusion of C-kit + LSECs alleviated toxin-induced liver injury by promoting hepatocyte proliferation via angiocrine signaling. Whether C-kit + LSECs protect hepatocytes from lipotoxicity through angiocrine signaling should be investigated in the future.
Aging-associated metabolic disorders greatly shorten the lifespan. In this study, we showed that aging led to loss of endothelial C-kit and that infusion of C-kit + LSECs attenuated aging-associated senescence and steatosis. Furthermore, C-kit + LSECs restored expression of pericentral endothelial markers and inhibited senescent genotypes, supporting our hypothesis that endothelial C-kit counteracts senescence by protecting the pericentral liver endothelium. Although the underlying mechanism is still unclear, C-kit + LSEC therapy provides an attractive rationale for treating aging-and diet-induced NASH in the future.

Clinical sample collection
Liver tissues used to evaluate aging-associated steatosis were obtained from patients who underwent a liver biopsy during cholecystectomy (the surgical removal of the gallbladder) at Xijing Hospital. Detailed patient information, including age, sex and BMI, are provided in Supplementary Table 1. NASH liver tissues were collected from (1) patients with biopsy-confirmed NASH who underwent a liver biopsy during bariatric (weight-loss) surgery and (2) non-NASH (control) patients who underwent cholecystectomy. None of these patients has been diagnosed with diabetes mellitus. All procedures were approved by the Human Ethics Committees of Xijing Hospital. All participants provided informed written consent.
Reagents AMD3100 (S8030), an antagonist of the chemokine receptor CXCR4, was purchased from Selleck Chemicals. Tamoxifen (T5648-5G) was purchased from Sigma-Aldrich. Antibodies used in this study are listed in Supplementary Table 2.

Animal studies
Male C57BL/6J mice (8 weeks old) were purchased from Charles River Laboratories. Mice were maintained in a specific pathogen-free facility on a C57BL/6J background. We used CRISPR-Cas9 homologous recombination technology to create Kit-floxed mice. Models of aging were established using 24-month-old mice, and some of the 24-month-old aged mice were purchased from a commercial company (Charles River). Mice were fed an MCD diet (A02082002BR, Research Diets) for 6 weeks or a CDAA diet (A06071309, Research Diets) for 10 weeks to build the NASH model. AMD3100 (5 mg per kg) was intraperitoneally administered to NASH model mice every other day for 2 weeks. All animals were supplied with sufficient water and food, and all animal experiments were performed in accordance with the guide for the Care and Use of Laboratory Animals prepared by the National Academy of Science. All animal work was approved by the Experiment Administration Committee of Airforce Medical University (Xi'an, China).

PCR assay with reverse transcription
Total RNA was extracted from cells and snap-frozen liver tissues using the TRIzol reagent (Invitrogen) and reverse transcribed into cDNA using the Evo M-MLV RT Premix Kit (Accurate Biology). Quantitative PCR was performed using the SYBR Green PCR Master Mix (Accurate Biology). Expression data were normalized to mRNA expression of the β-actin protein. The primer sequences are listed in Supplementary Table 3.

Western blotting
Western blotting was carried out using total protein extracts from LSECs and liver tissues. Cell and liver tissue lysates were homogenized in RIPA lysis and extraction buffer containing protease inhibitors (Millipore). Total protein was quantified using the BCA Protein Assay Kit (Beyotime Biotechnology), and equal amounts of protein were separated by SDS-PAGE and transferred to PVDF membranes. Membranes were blocked with skimmed milk and incubated with primary and secondary antibodies (Supplementary Table 2). Membranes were developed using chemiluminescence reagents (Millipore), and the proteins were visualized on the ChemiDoc MP Imaging System (Bio-Rad).

Senescence-associated β-galactosidase staining
Frozen liver tissue was embedded in optimum cutting temperature compound and sliced into 8-μm thick sections. The sections were then stained with the Senescence-Associated β-Galactosidase Staining Kit (Cell Signaling, 9860) according to the manufacturer's instructions.

Cell-migration assay
Liver macrophages and LSECs from Kit control and Kit-KO mice were collected. Liver macrophages were incubated with DiI dye at 37 °C for 30 min. A total of 1.5 × 10 5 liver macrophages (in DMEM supplemented with 10% FBS) were applied to the upper chamber, and 3 × 10 5 LSECs (in EC medium from ScienCell, supplemented with 10% FBS) were placed in the lower chamber of a transwell and incubated for 24 h. Three experimental groups were set up: Kit control, Kit KO and Kit KO with AMD3100. After 24 h, the upper chamber was removed, and the number of fluorescent liver macrophages was counted using a fluorescence microscope.

Cell isolation
Liver cells were isolated as described previously 14,42 . Briefly, mice were anesthetized and perfused with Hank's buffer containing 0.2 mg ml −1 collagenase IV (Sigma-Aldrich). The liver was removed and minced with a tissue homogenizer in buffer containing 100 μg ml −1 DNase I (Roche) and 0.2 mg ml −1 collagenase IV. Hepatocytes were collected by low-speed centrifugation at 50g for 3 min and resuspended in DMEM, followed by three washes with DMEM to remove cell debris. Supernatants were centrifuged at 400g for 6 min to collect NPCs. Liver NPCs were resuspended in 4 ml 17.6% iodixanol (Axis-Shield), and 4 ml 11.5% iodixanol and 2 ml DMEM were loaded on the top of the cell suspension. The cell suspension was centrifuged at 1,400g for 21 min in a swingout rotor without brake. HSCs were recovered from the interface between the DMEM and 11.5% iodixanol solutions, and LSECs and Kupffer cells were recovered from the interface between the 11.5% and 17.6% iodixanol solutions, mixed with an equal volume of DMEM and centrifuged at 400g for 8 min. Next, HSCs were washed three times with DMEM to remove cell debris. LSECs and Kupffer cells were resuspended and incubated with anti-CD146 magnetic beads (Miltenyi Biotec) for 15 min at 4 °C in the dark in the refrigerator. By magnetic separation with MS columns, CD146 + cells were collected by pushing the plunger into the MS column; CD146 + cells were LSECs. The eluted cells were Kupffer cells. The eluted fraction can be enriched and washed with DMEM three times to remove cell debris.

Cell transfection
Murine LSECs were isolated as previously described 42,43 . LSECs were cultured in EC medium containing 5% FBS and EC growth solution, 100 μg ml −1 streptomycin and 100 U ml −1 penicillin at 37 °C in a humidified atmosphere with 5% CO 2 . LSECs were transfected with NC-or CXCR4-shRNA-encoding lentiviral vector at a multiplicity of infection of 30 based on the recommended protocol. Forty-eight hours later, the culture medium was replaced with normal growth medium. Seventytwo hours after transfection, cells were collected or stained.

Chromatin immunoprecipitation
ChIP was performed using the SimpleChIP Enzymatic Chromatin IP Kit (magnetic beads) (Cell Signaling Technology, 9003), following the manufacturer's instructions 43 . Briefly, cells were collected in a 15-ml centrifuge tube and cross-linked by adding 1% (vol/vol) formaldehyde at room temperature for 10 min. The reaction was terminated by the addition of glycine for 5 min at room temperature, and cells were washed twice with cold PBS. Cross-linked chromatin was digested with micrococcal nuclease for 20 min at 37 °C, and 500-μl aliquots were sonicated on ice. ChIP was performed using the following antibodies: mouse anti-CEBPA (Proteintech, 18311-1-AP), mouse anti-histone H3 (a technical positive control) (Cell Signaling Technologies, 14269) and normal mouse IgG (a negative control) (Millipore, 12-371B). After overnight incubation at 4 °C, protein G magnetic beads were added, followed by further incubation for 2 h at 4 °C. Protein G magnetic beads were washed extensively with low-and high-salt ChIP buffer. After reverse cross-linking by incubation at 65 °C for 2 h, chromatin DNA was eluted with DNA-elution buffer, and the precipitated DNA fragments were extracted and analyzed by quantitative PCR. The primer sequences used for ChIP-PCR were designed according to the neighboring sequence of the CEBPA-binding site to Cxcr4 promoter regions as follows: Site1-F, CTTGTCTCAAGTTCTCCACCCGTAA; Site1-R, TGGCCGGATAATGTTCTCAGAGTTG; Site2-F, ACTTGCACGCTGTTTG-CAAACGT; Site2-R, CGCGTGCCGACTAAAGTCTCTAA.

Histology
Liver tissues were fixed in 4% paraformaldehyde for 3 h. H&E and Sirius Red staining, IF and immunohistochemistry were performed as previously described 42 . Oil Red O staining was conducted following instructions provided by Servicebio. Images were subsequently processed by CaseViewer.

Flow cytometry
For analysis of surface F4/80 and CD11b expression, mouse NPCs were isolated from the mouse liver as previously described and then stained with anti-mouse F4/80-PE (123109, BioLegend) and CD11b-APC (101211, BioLegend) antibodies. Cell viability analysis was performed using the 7-AAD Viability Staining Solution (420403, BioLegend). The 7-AAD + cells were then quantified by flow cytometry (MA900, Sony) and analyzed with Cell Sorter software (Sony). A total of 50,000 events were acquired within the FSC-SSC gate; cell debris (BSC-A versus FSC-A) and doublets (FSC-H versus FSC-A) were excluded.

C-kit + LSEC infusion
C-kit + cells were isolated from purified LSECs using anti-CD117 magnetic beads (Miltenyi Biotec). GFP-labeled C-kit + LSECs were isolated from GFP mice and injected into the spleen of C57BL/6 male mice. Livers were then collected at different time points after cell infusion and observed with the Caliper IVIS Spectrum imaging system. Besides, some of the C-kit + cells were incubated with CellTracker CM-Dil Dye (Invitrogen) for 30 min at 4 °C in the dark in the refrigerator, washed with PBS two times and then injected into the spleen of MCD mice; livers were collected at different time points after cell reinfusion and observed with the Caliper IVIS Spectrum imaging system.

Statistics and reproducibility
No statistical methods were used to predetermine sample sizes, but our sample sizes are similar to those reported in previous publications. No specific methods were used to randomly allocate samples to groups. Experiments were carried out in an unblinded fashion unless otherwise stated. No data were excluded from analysis. All data were analyzed using GraphPad software (version 8.30) and expressed as mean ± s.d. Data distribution was assumed to be normal but this was not formally tested. Statistical tests for each experiment are mentioned in the corresponding figure legends. Two-tailed Student's t-test was used for comparisons of two groups. Differences among multiple groups were compared using one-way ANOVA followed by Bonferroni's post hoc test. P values <0.05 were considered as a measure of statistical significance (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001).

Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability
The original RNA-seq datasets reported in this study have been deposited in the Gene Expression Omnibus database with the accession Article https://doi.org/10.1038/s43587-022-00348-z numbers GSE216592 and GSE216426. In addition, we also analyzed some published datasets including young and old human or mouse liver transcriptome-wide analysis datasets SE183915 and GSE167665, healthy and diet-induced NASH mouse RNA-seq data (GSE140994 and GSE119340) and ChIP-seq data (GSM1037658 and GSM1816817) (http:// cistrome.org/db/#/). All statistical data associated with this study are contained in the Supplementary Data. All other data are available from the corresponding author upon reasonable request. Source data are provided with this paper.