Human retinal pericytes cell line (HRMVP) and Human Umbilical Vein Endothelial Cell (HUVEC) were purchased from aoyinbio Co., Ltd (Shanghai, China). Cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM) medium supplemented with 10% fetal bovine serum (FBS) (#A3161002C, Gibco) and 1% penicillin–streptomycin (#15070063, Gibco). ERK1/2 inhibitors (#HY-10256, MCE) were purchased. The concentration of ERK1/2 inhibitors (SB203580) was used by 10 μM in vitro.

Every protein sample was loaded on 8% SDS- PAGE and then transferred into a PVDF membrane (ThermoFisher Scientific). And then membrane incubated with primary antibodies overnight at 4°C: anti-EN (#ab48185, Abcam), anti-Cyr61 (#ab230947, Abcam), anti-GAPDH (#10494-1-AP, Proteintech), anti-p-MAPK (#4370T, CST), and anti-MAPK (#4695T, CST). Cells were harvested and total RNA was isolated with kit (#R6834010000J28U100, omegabio). Reverse transcription was conducted using PrimeScript™ RT Master Mix (TaKaRa, Japan). Then quantitative PCR was performed using an SYRB GreenII kit (#DRR041A; TaKaRa, Japan). The following primers were used: human-Endosialin-forward (F): 5′- CTCCACACATTCGTGTTCGC- 3′ and reverse (R): 5′- CTGCTACGCTCTCTTCCCAC- 3′; human-CYR61-F: 5′- CAGGTGGGTGGGATGTGAA- 3′ and R: 5′- GGGAAACGCTGCTTCATTGG- 3′; and human-GAPDH- F: 5′- GCAACTAGGATGGTGTGGCT- 3′ and R: 5′- TCCCATTCCCCAGCTCTCATA- 3′.

All animal experiments of this study were approved by the Guidelines for the Care and Use of Laboratory Animals of Air Force Medical University. C57BL/6-Endosialinem1Smoc mice (systemic knockout) were purchased from Shanghai model organism company(#NM-KO-200094) and maintained in a 12h light/12h dark cycle with free access to food and water. And we choose littermate wild-type mice as control group. Each mouse was male and whether gender influences tumor promotion by Endosialin is unknown. Each mouse was inoculated with 2×10⁶ tumor cells (B16F10) in abdomen waiting for lung metastasis. The mice were sacrificed after 14 days.

The primary antibodies used for IF and IHC staining were anti-CD31 (#77699, CST), anti-Endosialin (#ab204914, Abcam), anti-Cyr61(#1111, ABclonal) and anti-NG2(#A24955, ABclonal).

Matrigel (BD Biosciences) was added to a 48-well plate (150 μL/well). HUVEC cells were required to be serum starved in advance for 6 h and collected after 0.25% trypsin (Invitrogen) digestion. Fifty thousand HUVECs were plated onto Matrigel in 300 μL of PBS/2% FBS per well at the time of plating. The cells were incubated at 37°C with humidified 95% air/5% CO2 for 6 h. The tubes/networks were imaged using Metamorph image analysis software (Molecular Devices, Sunnyvale, CA). Endosialin of HRNVP was knocked-down by si-RNA (GenePharma Co., Shanghai, China) before we collected culture supernatants. Cell proliferation assay was performed by the same way and cell counting kit-8 was purchased from mishubio, Xian, China. Cyr61 combination protein was purchased from Beyotime Biotechnology and the effective concentration was 3μg/mL.

Mice were injected with 2mg/kg lipopolysaccharide to induce systemic blood vessel leakage. 24h later, 1% (weight/volume) Evans blue dye was injected into the peritoneum. Two hours later, tumors of the mice were collected, weighted and treated with formamide (Sigma Aldrich) for 48h at 50°C. Optical density of Evans blue dye absorbance released from tissue was measured at 600nM using a spectrophotometer.

Analysis of single gene differences in TCGA database was prepared for GSEA using DESeq. 2 package (version 1.26.0). In addition, GO and KEGG pathway enrichment analyses were predicted by the DAVID online database (https://david.ncifcrf.gov/). Single-cell data are derived from GSE72056 and GSE174401. Analysis of scRNA-seq datasets was performed primarily using the Seurat package (v4.0.5) in R (v4.1.0) and code was same with our previous article (17).

Results were depicted as mean ± SEM. Comparisons were analyzed by two-tailed Student t test or one-way ANOVA followed by Dunnett’s post hoc test. Differences were considered significant at P<0.05.

Firstly, we integrated two single-cell sequencing datasets. These samples are all from metastatic lesions of melanoma, including lung and brain metastases ( Supplementary Figure S1A ). We identified 10 cell clusters and their characteristic genes ( Figures 1A, B ). During the tumor metastasis, tumor cells have a strong interaction with the microenvironment of the colonization site. Studies have shown that pericytes have a strong role in promoting tumor metastasis, so we analyzed Pericytes in the tumor microenvironment. We found that in metastatic lesions, Pericytes were divided into five subpopulations ( Figure 1C ). Our previous study found that Endosialin was a specific marker of Pericytes, which was also found in melanoma metastasis ( Supplementary Figure S1B ). We further divided the Pericytes into EN-high expression group and EN-low expression group according to the expression degree of Endosialin ( Figures 1D, E ). GO enrichment analysis showed that EN-high Pericytes were associated with extracellular matrix structural constituent and blood vessel remodeling ( Figure 1F ). EN-low Pericytes are associated with cytokine receptor binding and protein kinases activity ( Figure 1G ).

Using the TCGA database, we found that high expression of Endosialin was associated with worse prognosis in melanoma patients ( Figure 2A ). It was further found that Endosialin expression was higher in tumor tissues compared with normal tissues ( Figure 2B ). Next, we compared patients at different periods of disease progression and found that ENDOSIALIN expression was higher in patients at advanced clinical stages compared with patients at earlier stages ( Figure 2C ). In addition, the expression of Endosialin was higher in the metastases compared with the primary tumor ( Figure 2D ). These results suggested that Endosialin may be a driven factor of tumor metastasis to promote tumor progression.

To further explore the functions of Endosialin, we divided melanoma patients into high expression group and low expression group. And then, we performed GO and KEEG enrichment analysis of the differential genes between two groups. The results showed that Endosialin was involved in extracellular matrix receptor interaction and positive regulation of angiogenesis ( Figures 2E, F ). Among the differential genes, the expression of CCN1, which was related to angiogenesis, is down-regulated when Endosialin was highly expressed ( Figure 2G ). These results revealed that Endosialin is involved in angiogenesis.

To explore the function of Endosialin in metastasis, we inoculated melanoma cells (B16F10) into caudal vein of Endosialin knockout (EN ^KO ) mice and wildtype (WT) mice. We found that EN ^KO mice had significantly fewer lung metastases compared with WT mice ( Figure 3A ). Similarly, we used the Endosialin blocking antibody after injection of tumor cells and found that the lung metastases in the anti-EN group were reduced ( Figure 3B ). The successful colonization of tumor cells in metastases depends on the rich vascular structure (2, 18). And Endosialin has been reported to promote vascular maturation (19). Therefore, we examined the vessel density and found that vessels of diameter <50μm increased and vessels of diameter ≥50μm decreased in EN ^KO mice ( Figure 3C ). To further detect the blood perfusion situation of lung metastasis, we injected Evans blue dye to tumor bearing mice, collected the metastasis 2h later and eluted the dye with formamide. We found that the blood perfusion of EN ^KO mice was relatively insufficient compared with WT mice ( Supplementary Figure S2A ). In melanoma, Endosialin is mainly expressed on pericytes ( Figure 3D ). We further examined the changes of all stromal cells in tumor tissues. We found that the number of stromal cells in EN ^KO tumor tissues decreased, which partly explained the changes in tumor microenvironment and vascular maturation disorders ( Figure 3E ). These results indicate that the loss of Endosialin leads to vascular maturation disorder, hypoperfusion of blood flow and, in turn, reduced tumor metastasis.

Based on the changes of tumor vessels in mice and bioinformatics data, we examined whether Endosialin has a regulatory effect on Cyr61. It was reported that Cyr61 played a key role in angiogenesis (4, 20). In EN ^KO mice, the expression of Cyr61 increased ( Figure 4A ). Consistently, after knockdown Endosialin in HRMVP cell line, the phosphorylation of Erk1/2 was downregulated while the expression of Cyr61 was upregulated ( Figure 4B ). Quantitative PCR showed the same results ( Figure 4C ). Following the use of inhibitors of Erk1/2, downregulated phosphorylation of Erk1/2 and upregulated expression of Cyr61 were similarly found ( Figures 4D–E ). Besides, we collected two patients’ tissues. We found that there was a negative correlation between the expression of Endosialin and Cyr61 ( Figure 4F ). These results suggested that Endosialin regulated Cyr61 expression via the Erk1/2 signaling pathway.

We have known that the loss of Endosialin was able to increase small vessels in vivo and Endosialin can regulate the expression of Cyr61. In order to furtherly clarify the effects of Endosialin on endothelial cells, we knocked down the expression of Endosialin on HFL1 and collected the culture supernatant as conditioned medium (CM) ( Figure 5A ). We found that the CM of si-EN groups was able to promote the proliferation of endothelial cells (HUVEC) ( Figure 5B ). Since Endosialin was able to regulate expression of Cyr61, we examined the effects of Cyr61 on endothelial cells. We added the recombinant protein of Cyr61 into blank medium, which was able to promote the proliferation of HUVECs, achieving similar effects as those in the complete medium with 10% serum ( Figure 5C ). We also found that the CM of si-EN groups was able to promote tube forming ability of endothelial cells ( Figures 5D, E ). Besides, the use of Cyr61 recombinant protein in WT mice promoted tumor metastasis and small vessel formation ( Figure 5F ). These results suggested that Endosialin was able to suppress the proliferation of endothelial cells through inhibiting the expression of Cyr61.