Male NOD/SCID mice age of 6 weeks were obtained from SLAC Laboratory Animal Corporation (Shanghai, China). All animal experiments were performed according to animal protocol approved by animal care ethic committee of Fudan University. The xenograft tumor model was established by inoculating PLC/PRF/5 with or without Exo treatment subcutaneously into the groin of mice. Animals were continually monitored for 4 weeks before being euthanized. Tumors were collected and fixed in 4% paraformaldehyde. The tumors were weighted and measured, and the volumes were calculated based on the formula: V = 0.5*long diameter*short diameter².

Hepatic carcinoma cell lines PLC/PRF/5 and MHCC-97H (abbreviated as PLC and 97H in figure label) and normal hepatic cell line L02 were obtained from Cancer Institute of Fudan University (Shanghai, China) and were maintained in DMEM cell culture medium (Gibco, New York, USA) supplemented with 10% FBS (Gemini, West Sacramento, USA) and 1% ampicillin and streptomycin in the incubator at 37°C with 5% CO₂. The EV free FBS was obtained by centrifuging the purchased FBS at 120,000 g 4°C for 16 h, and filtering the supernatant through a 0.22-μm filter.

Primary antibodies used in the Western blot studies are as follows: anti-human Grp75, CD9, OCT4 (CST, Danvers, USA), anti-human ALDH, CD44, CD133 (GeneTex, Irvine, USA), anti-human sytenin1 (Abcam, Cambridge, UK), anti-human CD64 (Proteintech, Wuhan, China), and anti-human GAPDH [Beyotime Institute of Biotechnology (Nanjing, China)]. HRP conjugated secondary antibodies are from Beyotime Institute of Biotechnology (Nanjing, China). Growth factors bFGF and EGF were purchased from Gibco. MicroBCA protein quantification kit was purchased from Thermo Scientific (USA). Cell Counting Kit-8 (CCK-8) was purchased from Dojindo Laboratories (Kumamoto, Japan).

Cells grown to 60% confluency were washed three times with PBS before they are placed in the new medium supplemented with EV free FBS and continued to grow for another 24 h at 37°C with 5% CO₂. The medium is collected and centrifuged at 300 g for 10 min, and the supernatant was further centrifuged at 2,000 g for 20 min. The resulting supernatant is the conditioned medium.

EV and subsequent Exo and MV isolation from CM are based on Kowal et al. (20). Briefly, CM was centrifuged at 10,000 g for 40 min at 4°C. After supernatant was removed, the pellet, which was the MV fraction, was washed with PBS and resuspended in 50–100 µl of PBS. The supernatant from previous centrifugation was centrifuged again at 100,000 g for 90 min at 4°C. After the supernatant was discarded, the pellet, which was the Exo fraction, was washed with PBS and resuspended in 50–100 µl of PBS.

A total of 250 PLC/PRF/5 or 500 MHCC-97H cells were seeded in 24-well plates in 1 ml of DMEM/F12 medium supplemented with 20 ng/ml epidermal growth factor (EGF), 20 ng/ml fibroblast growth factor (FGF), 2% B27, and 1% penicillin. The plate was shaken well before sealed with parafilm and cultured at 37°C, 5% CO2. Replenishment of the medium was made at 0.2 ml per well after 1 week of culture. Sphere count was carried out under the microscope after 2 weeks of culture. Each condition has three triplicate wells.

PLC/PRF/5 and MHCC-97H cells were seeded in a 96-well plate at 1,000 or 2,000/well respectively, and incubated at 37°C and 5% CO₂ for 24 h before various stimulus conditions were applied. Cells were continued to culture for another 48 h before the Cell Counting Kit-8 (CCK8) working solution, made by diluting CCK8 reagent in culture medium at 1:9 ratio, was added at 100 μl/well. Absorbance at 450 nm was measured by a spectrometer after 2 h. Five replicates were included for each culture condition.

PLC/PRF/5 and MHCC-97H cells were seeded in 24-well plates at 125 or 250/well respectively, and incubated at 37°C, 5% CO₂ for 24 h before various stimulus conditions were applied. Cells were continued to culture for another 10 days before the culture medium was removed. Cells were washed with PBS twice before fixed with 4% paraformaldehyde for 10 min and stained with crystal violet. After the plates were washed and air dried, the number of clones with more than 50 cells was counted.

Total protein levels of extracted samples were determined with the Micro BCA Protein Assay Kit (Thermo Scientific). Protein samples were resolved with SDS-PAGE and transferred onto PVDF membrane for Western blot analysis. Primary antibodies were diluted according to the manufacturer’s instruction and incubated with membrane at 4°C, overnight with constant agitation. Secondary antibody incubations were carried out at room temperature for 1.5 h. ECL was used for visualization. Shh ELISA (Abcam) was performed as specified by the manufacturer.

shRNA was introduced by shRNA carrying lentivirus made by GenePharma (Shanghai, China). The shRNA was cloned into GenePhama Supersilencing vector under CMV promoter/enhancer. The vector carries GFP gene for visualization and puromycin resistance gene for selection. Cells were grown in a six-well plate for 24 h to about 50% confluency. Cells were washed with PBS, and viral transfection solution was added together with 3 μl of Polybrene. Cells were incubated at 37°C, 5% CO₂ for 24 h before changing into normal medium. Transfection efficiency was estimated by observing green fluorescence under a fluorescence microscope at 48 h post transfection. Puromycin were added at 48 h post transfection to maintain stable transfected cells. ShRNA sequences were as follows: Shh-shRNA1: GCTCGGTGAAAGCAGAGAACT, Shh-shRNA2: GCCAAGAAGGTCTTCTACGTG.

Total RNA was extracted from cells using Trizol/chloroform method, and precipitated with isopropanol. Reverse transcription was performed according to the manufacturer’s instruction (Takara, Japan). Primers for qPCR (Takara, Japan) are listed in Table 1 .

Statistical analysis was performed using SPSS 20 and GraphPad Prism 8 software. All experiments were performed with three biological replicates. Chi-square test was used in the comparison of the count data. Student’s t-test was used in the comparison of the measurement data between the two groups, while one-way ANOVA was used in the comparison of the measurement data among multiple groups. p-value < 0.05 indicates statistical difference, *<0.05, **<0.01, and ***<0.001.

Exosome can be isolated based on its physical, chemical, and biological property, with methods mainly including ultracentrifugation, polymer precipitation, ultrafiltration, chromatography, and immunoaffinity capture. Among them, ultracentrifugation is more time-consuming. However, with relatively high purity in extraction, it is still a widely used method.

To understand whether EVs would play a role in HCC development, we extracted and separated EVs secreted by HCC cell lines PLC/PRF/5 and MHCC-97H, as well as the normal human liver cell line L02 using ultracentrifugation. Cells were cultured in DMEM supplemented with 10% exosome-free FBS before the isolation process. After a separation protocol containing ultracentrifugation at 10,000 g and 100,000 g, the EV was separated into MV and exosome (Exo) fractions.

The nanoparticle tracking analysis showed that the average particle sizes of the MV fractions from the three cell lines were between 135 and 150 nm in diameter, while the average sizes of the particles from the Exo fractions were between 99 and 114 nm in diameter ( Figure 1A ). Electron microscopy analysis showed that the extracted MV and Exo particles were in typical cup-shape morphology ( Figure 1B ). Analysis of particles and protein levels revealed that both PLC/PRF/5 and MHCC-97H cells secreted more particles as Exo than MVs, and these particles contained more protein in Exo than MVs ( Supplementary Figure S1 ).

The purity of the fractions was analyzed by Western blot using protein markers specific for each fraction. It is shown that MV marker Grp75 was predominantly presented in the cell extract and MV fractions, while the Exo markers CD9, Syntenin-1, and CD63 were predominantly presented in the Exo fractions ( Figure 1C ). These results indicated that we successfully separated Exo and MV fractions through ultracentrifugation process, and this provided us the tools in studying the effect of Exo in subsequent experiments.

To determine whether exosome secreted by cancer cells has any effect on cell growth, we cultured PLC/PRF/5 and MHCC-97H cells in the presence of additional exosome isolated from their respective culture medium. As shown in Figure 2A , addition of Exo into the culture medium promoted PLC/PRF/5 and MHCC-97H growth in a dose-dependent manner. When the Exo concentration was above 5 µg/ml, the cell proliferations were significantly increased. We then studied the colony formation ability. When Exo is added at a concentration of 10 µg/ml in culture medium, the numbers of colonies formed were significantly increased in both PLC/PRF/5 and MHCC-97H ( Figure 2B ). We then investigated whether the addition of exosomes could lead to an increased subpopulation of cancer stem-like cells in culture. The morphology of cells cultured in medium supplemented with Exo were compared with those cultured in normal medium, and we found that the number of spheres in culture increased significantly in both cell lines when isolated Exo was added into culture medium ( Figure 2C ). The expression levels of several cancer stem cell (CSC) markers, CD55, CD133, OCT4, and ALDH1 were also increased at both transcription and protein level ( Figures 2D, E ) when cells were cultured with additional cancer cell-derived exosomes. These findings suggested that the exosome secreted by cancer cells was responsible for promoting cell growth and inducing CSC.

Several studies have shown that Hedgehog signaling pathway plays an important role in HCC growth and development. We decided to investigate whether the effect of exosome on cell growth was through the activation of the Hedgehog pathway.

Two inhibitors of the Hedgehog pathway, Gant-61, which inhibits GLI transcription factor, and Vismodegib, which inhibits SMO receptor, were added into the cell culture, and the effects of these inhibitors on Exo stimulation were studied. Both inhibitors, while themselves had minimal effects on cell proliferation or cell sphere formation, could significantly reverse the stimulating effects of the Exo on cell proliferation and sphere formation on both HCC cell lines ( Figures 3A, B ).

This promoted us to examine the expressions of Hedgehog protein in the two cancer cell lines. Two forms of Hedgehog proteins Shh and Ihh were expressed in these cell lines, with Shh being expressed at a higher level in both cells ( Figure 3C ). Hedgehog proteins were also found in EVs isolated from the cell culture, in which Exo contained much higher levels of both Shh and Ihh than MV. Both cancer cell lines expressed significantly higher Hedgehog protein in the secreted exosomes than the normal hepatocyte cell line L2 (S2), suggesting that the Hedgehog signaling was preferably activated in cancer development. Our results indicated that PLC/PRF/5 and MHCC-97H might use Shh as the main Hedgehog pathway ligand, and Exo would be the preferable way for the cells to secrete the protein.

We performed immunogold labeling with anti-Shh antibody on the exosomes isolated from PLC/PRF/5 and MHCC-97H cells, and as shown in Figure 3D , Shh was localized on the surface of Exo secreted by cancer cells. This indicated that Shh carried by exosome may directly interact with its receptor on cell membrane.

To study whether these Shh carrying exosomes can activate the Hedgehog pathway, we assess the expressions of the Hedgehog pathway target genes in cells stimulated with cancer cell-derived exosomes. We found that in both PLC/PRF/5 and MHCC-97H cells, the mRNA and protein levels of several target genes, including GLI1, PTCH1, Cmyc, and CyclinD1, were all elevated when Exo was added in the culture medium ( Figures 3E, F ).

To determine whether the Shh protein carried in these exosomes was responsible for the stimulating effect, we used shRNA to knock down the expression of Shh, and observed its impact. The Western blot analysis demonstrated that two different shRNA-containing lentivirus constructs successfully reduced the expression of Shh in both PLC/PRF/5 and MHCC-97H cells ( Figure 4A ). As a consequence, the Shh levels in the isolated Exo from respective culture medium were also reduced when shRNA-containing lentivirus was transfected into the cells. We isolated the exosomes from the Shh knockdown cells and compared their effect on the Hedgehog pathway and cell growth with the exosomes isolated from the Shh intact cells (vector transfected). As shown in Figures 4B, C , when the exosomes isolated from the Shh knockdown cells were added to the culture of the PLC/PRF/5 or MHCC-97H cells, they did not stimulate the expression of Hedgehog pathway target genes as the exosomes isolated from the control cells did in both mRNA and protein levels. Moreover, exosomes from the Shh knockdown cells had reduced ability in stimulating the cell proliferation, colony formation, and sphere formation, compared to the exosome isolated from the control cells ( Figures 4D–F ). These studies indicated that Shh played an important role in PLC/PRF/5 and MHCC-97H cancer development, and the cells secreted the protein into exosome to assert regulation.

We demonstrated that PLC/PRF/5 and MHCC-97H could secrete Shh containing exosome to regulate the growth of cells. We decided to investigate whether these exosomes influenced tumor formation in vivo. PLC/PRF/5 was inoculated subcutaneously into NOD/SCID mice at various numbers mixed with PBS or 5 μg/ml of exosome from its own culture; the tumor formation was observed continuously for 4 weeks. The mice were then euthanized and tumors were removed, measured, and weighted. At the inoculation doses of 1 × 10⁶/mouse and 3 × 10⁵/mouse, tumor was formed in all five mice with cells co-injected with PBS or isolated exosomes. However, the average tumor volumes and weights formed by cells mixed with isolated exosomes were significantly higher than those formed by cells mixed with PBS ( Figure 5A ). At the lowest inoculation of 1 × 10⁵/mouse, however, tumor only formed in 40% (2/5) of the mice when cells were mixed with PBS before injection, while it formed in 100% (5/5) of the mice when cells were mixed with cancer cell-derived exosomes ( Figure 5A ). Similarly, the average weight and volume of the tumors were significantly higher formed by cells mixed with exosome than those formed by control cells.

We then studied whether the exosomes isolated from cells with Shh knockdown would abolish the stimulating effect on tumor formation of the exosomes. We inoculated the NOD/SCID mice with PLC/PRF/5 cells mixed with either exosomes isolated from control or Shh knockdown cells. As shown in Figure 5B , at the inoculation doses of 1 × 10⁶ and 3 × 10⁵, PLC/PRF/5 mixed with either exosomes form tumor in all (5/5) mice tested. However, the average volume and weight of the tumors formed by cells mixed with exosomes isolated from untreated cells were significantly higher than those by cells mixed with exosomes isolated from Shh knockdown cells. When mice were inoculated with PLC/PRF/5 at 1 × 10⁵, tumor formed in 60% (3/5) of the mice injected with cells mixed with exosomes derived from normal cells, but none (0/5) with cells mixed with exosomes derived from Shh knockdown cells.

Finally, we explored the clinical relevance of circulating Shh levels in HCC patients. We first compared the plasma Shh levels and plasma Exo-Shh levels from 30 HCC patients to 10 normal healthy donors. We found that the average plasma Shh and plasma Exo-Shh levels were both significantly higher in HCC patients than those in healthy donors ( Figure 6A ).

To understand if there is a correlation between circulating Shh levels and the activation of the Hedgehog pathway in HCC, we performed immunostaining on the sections of the tumor tissue for the expression of GLI-1 and PTCH1. Based on the quantitation of the staining, we categorized them into a low-expression and a high-expression group (S3). We then studied correlation between plasma Exo-Shh levels and the expression levels of GLI-1 and PTCH1. As shown in Figure 6B , plasma Exo-Shh levels were significantly higher in patients with higher expression levels of GLI-1 and PTCH1 than those with lower levels.

We then analyzed the correlations between the plasma Exo-Shh levels with the clinical characteristics, in terms of TNM stages, number of tumors, and histological grades. As shown in Figure 6C , patients with later TMN stages of cancer, multiple tumors, or higher histological tumor grades had significantly higher levels of plasma Exo-Shh than patients with earlier stages of cancer, single tumor, or lower histological grade.

Finally, these 30 HCC patients were followed up for the tumor recurrence after surgery within 18 months. We found that patients with plasma Exo-Shh levels that were higher than 110 pg/ml (n = 15) had a recurrence rate of 46.7% within 18 months after surgery, while patients with plasma Exo-Shh levels lower than 110 pg/ml (n = 15) had a recurrence rate of 13.3%. Recurrence between the two groups reached a statistical difference at the time point of 14th month post-surgery (46.7% vs 6.7%) ( Figure 6D ).