Exosome isolation and labeling

HCC cells (1×10⁶) were cultured in vesicle- depleted medium for 48-72 hours. The medium was gathered and subjected to gradient centrifugation. Briefly, the medium was first centrifuged at 1000g for 10 min, at 3000g for 30 min, at 10,000g for 60 min, at 100,000g for 4 hours. All centrifugal steps were performed at 4 . Exosomes from sera of HCC patients were isolated using ExoQuick exosome precipitation solution (SBI System Biosciences) according to the manufacturer's instructions ²². The extracted exosomes were suspended in 300 μL PBS. The exosomes were quantified by BCA protein assay. A total of 10 µg exosomes were used for each in vitro experiment. For in vivo growth and metastasis assays, 1 µg/g exosomes were applied three times a week. We also used fluorescent dye Dil (Sigma) to label exosomes. Briefly, exosomes were incubated with Dil (1:2000) for 2 hours and then washed with PBS. The endocytosis of recipient cells was visualized using a confocal fluorescence microscope (Zeiss).

Patients

One hundred twenty-four surgery patients diagnosed with E-HCC at Sun Yat-sen University Cancer Center from January 1, 2009 to December 30, 2012 were screened in this study. E-HCC was defined as solitary tumors with diameters of < 5 cm and without vascular invasion ²³. A patient was excluded from the study if he or she had transarterial chemoembolization (TACE), radiotherapy, ablation, or liver transplant before resection. Patients were also excluded if they had no definitive diagnosis or follow-up data. The tumor stage and clinical stage were established using the 2003 Union for International Cancer Control/American Joint Committee on Cancer criterion. Healthy volunteers' blood samples were used as a control. All samples were obtained with informed consent. The study protocol was approved by the Institutional Review Board. The clinical-pathologic characteristic of the E-HCC patients are summarized in Table ​Table11.

Tissue pH measurement

The tissue pH was measured by microelectrode. Briefly, tissues were obtained from HCC patients or xenografted nude mice after surgery. The tissue pH was measured by inserting the glass electrode (LabSen 252, Sanxin, Shanghai, China) into the center of tumor tissues, as directed by the manufacturer's instruction. The center part from each tumor was selected and extracted for analyzing the mRNA expression of GLUT-1, CA9, MMP3, and MMP9 by Q-PCR and their correlation with the pH value in the same region of HCC tissues. The value of tissue pH was measured in triplicate for each sample.

Electron microscopy

The exosomes were examined by transmission electron microscopy as previously described ²⁴. Briefly, the samples were fixed with 2% glutaraldehyde and loaded on to Formvar carbon at room temperature. Subsequently, the samples were negatevely stained with 1% uranyl acetate for 3 minutes at 4 , and dried under an electric incandescent lamp for 10 minutes. Photographs were taken using the JEM-1400 transmission electron microscope (JEOL, Tokyo, Japan) at 120 kV.

RNA isolation and quantitative real-time PCR

Total RNA was extracted by using TRIzol reagent (Invitrogen). cDNA was synthesized using the PrimeScript RT reagent Kit (Promega, Madison, WI, USA) as previously described ²⁵. Real-time PCR was performed using an ABI 7900HT Fast Real-time PCR system (Applied Biosystems, Foster City, California, USA) as previously described ²⁵.

Oligonucleotide transfection, lentiviral packaging

GLUT-1 siRNA, CA9 siRNA, miR-21-5p siRNA, miR-10b-5p siRNA, miR-21 mimics, miR-10b mimics, HIF-1α shRNA and HIF-2α shRNA were synthesized by Kangcheng biotechnology company (Guangzhou, China). The pCDH-CMV-MCS-EF1-coGFP plasmid was used to construct virus particles. This plasmid, together with packaging plasmids pCMV/pVSVG, pRSV/pREV, and pMDLG/pRRE, were transfected into 293FTcells using Lipofectamine 2000 reagent (Invitrogen). After 48 hours, virus particles were harvested from the cell supernatant. SMMC-7721 and Hep3B cells were transfected with centrifuged lentivirus plus 8 mg/ml polybrene (Sigma, St Louis, MO, USA). Oligonucleotide transfection was performed with Lipofectamine 2000 reagent (Invitrogen). U6 snRNA was used as a positive control, and reactions without reverse transcriptase or RNA template were included as negative controls.

PDCM system

The nanoparticles (PDCM) are composed of 1-palmitoyl-2-oleoyl-snglycero-3-phosphocholine (POPC), 1,2-dioleoyl-snglycero-3-phosphoethanolamine (DOPE), cholesteryl hemisuccinate (CHEMS), and cholesteryl-4-((2-(4-morpholinyl) ethyl) amino)-4-oxoburanoate (MOCHOL). The designed nanoparticles have “pH-tunable” characteristic, which facilitates efficient delivery relying on the surrounding pH ²⁶. Preliminary experiments and clinical trials showed good systemic biodistribution of the PDCM system ^26-28. PCDM-GLUT-1, PDCM-CA9, PDCM-miRNA- 21, and PDCM-miRNA-10b were synthesized and assembled by Kangcheng biotechnology company (Guangzhou, China).

miRNA microarray

Sample preparation and miRNA microarrays (Human miRNA Microarray, Release 21.0, Agilent) were performed at Biomedlab Company (Shanghai, China). Briefly, total RNA (2.5 μg) extracted from exosomes was labelled with pCp-DY647 (Dharmacon, Lafayette, CO, USA), hybridized onto microarray, scanned using a LuxScan 10K Microarray Scanner (CapitalBio, Beijing, China) and analyzed using GenePix Pro 6.0 software (Axon Instruments, Foster City, CA, USA). The microarray data have been deposited online under accession number GSE118647.

MTT assay

MTT proliferation assay (Sigma) was performed according to the manufacturer's protocol. A total of 10 µg exosomes was used for each group according to the experimental design. Viability was measured at different time points from 12 h to 72 h after treatment.

Scratch assay

The scratch assay was used to measure cell migration in vitro. In brief, cells were seeded in 60mm dishes and cultured until they formed a fused monolayer. A P200 pipette tip was used to create a scratch following which cells were incubated with different exosomes (10 µg). After 48 h, wound closure was photographed by a microscope. The migration rate was measured by the fraction of cell coverage across the scratch. The migration rate (%) = migration area/total area ×100%.

Invasion assay

Matrigel invasion chambers (BD Biosciences, NJ, USA) were used to measure cell invasion. A total of 10 µg exosomes was used for each group according to the experimental design. Briefly, 1×10⁴ cells were added to the upper chamber, and FBS was added to the lower chamber. After 48 h, the chambers were fixed and stained with 0.05% crystal violet for 2h. The cells on the upper surface were gently scraped. The stained cells were counted and photographed under a microscope (×200).

Immunohistochemistry and immunofluorescence

The slides were dried, dewaxed, rehydrated, and blocked. Next, the sliders were boiled and then incubated with the anti-HIF-1α antibody (1:500, ProteinTech Group, Inc.) and anti-HIF-2α antibody (1:500, ProteinTech Group, Inc.) in a moist chamber. After 3 washes, the slides were incubated with a secondary antibody (Envision, Dako, Glostrup, Denmark) and then stained with DAB (3,3-diaminobenzidine). Nuclei were stained with hematoxylin. The cutoff value (45% positive cells) of HIF-1α/ HIF-2α immunoreactivity was determined by receiver operator curve (ROC) analysis. For immunofluorescence, slides were incubated with rabbit anti-E- cadherin (1:200, Abcam) and mouse anti-vimentin (1:200, Abcam) for 2 hours. Slides were incubated for 1 hour with Alexa Fluor 488 goat anti-mouse IgG (1:800, Cell Signaling Technology) and Alexa Fluor 594 goat anti-rabbit IgG (1:800, Cell Signaling Technology).

Western blot analysis

Western blotting was performed using standard procedures. Briefly, protein samples (25mg) were separated on polyacrylamide-SDS gels and electrophoretically transferred to polyvinylidene difluoride membranes (Millipore). Subsequently, membranes were blocked and incubated with mouse anti-CD81 (1:1000, Abcam), mouse anti-CD63 (1:1000, Abcam), mouse anti-HIF-1α (1:1500, ProteinTech Group, Inc.), mouse HIF-2α (1:1500, ProteinTech Group, Inc.), rabbit anti-PTEN (1500, Abcam), rabbit anti-E-cadherin (1:1000, Abcam), mouse anti-vimentin (1:1000, Abcam), rabbit anti-mTOR (1:1000, Abcam), rabbit anti-AKT (1:1000, Abcam), rabbit anti-pAKT (1:1000, Abcam), and rabbit anti-tubulin (1:1000 Abcam) antibodies. Anti-rabbit or anti-mouse IgG was used as the secondary antibody (1:8000). The signal intensity was evaluated using Quantity One 4.4.0 software.

In vivo growth and metastasis assay

The In vivo growth and metastasis assay were performed as previously described ²⁹ and are illustrated in Supplementary Figure S14. Briefly, for in vivo growth assays, 2×10⁶ SMMC-7721 cells were injected subcutaneously into the flanks of BALB/ C-nu/nu athymic nude mice. The tumor size was measured using Vernier calipers every 7 days. For in vivo metastasis assay, the subcutaneous tumors were resected from growth assays and formed into 1mm³ cubes. These tumor cubes were implanted in the left lobes of the livers of nude mice. Exosomes (three times a week, 1 µg/g for each time) were used for in vivo growth and metastasis assay. On day 50, the liver and lung were dissected, fixed by formalin, and stained with haematoxylin & eosin. Consecutive tissue sections were used to examine pulmonary metastatic nodules. All experimental procedures involving animals were performed according to the institutional ethical guidelines.

Chromatin immunoprecipitation (ChIP) assay

The ChIP Assay Kit (Abcam) was used according to the manufacturer's protocol. Briefly, cells were lysed in cell lysis buffer to generate DNA fragments. Nonspecific IgG (1:200, Sigma), mouse anti-HIF-1α (1:200, Abcam), and rabbit anti-HIF-2α antibodies (1:200, Abcam) were incubated with lysates to precipitate target regions. The promoter fragment (contain HIF response element, HRE) was extracted and visualized by gel. The region of the miR-130b promoter was used as a negative control.

LASSO Logistic regression

The least absolute shrinkage and selection operator (LASSO) is an alternative method for fitting sparse models for regression with high-dimensional predictors. To eliminating variables, the L1 penalty could force coefficient estimates to zero ^{30, 31}. The logistic regression model with LASSO penalty was used to achieve variable shrinkage and selection. Optimal values of λ were determined by repeated cross-validations. The optimal λ is the largest value for which the partial likelihood deviance is within 1-s.e. of the smallest value of partial likelihood deviance. R software version 3.0.1 (R Foundation for Statistical Computing) and the 'glmnet' package were used to perform the LASSO regression analysis.

Exosomes derived from HCC cells cultured under acidic condition promote cell migration and invasion of recipient HCC cells cultured under the normal condition

We isolated extracellular vesicles from the medium of HCC cells (SMMC-7721 and Hep-3B) cultured under pH 6.6 and pH 7.4 by multistep ultracentrifugation. Western blot analysis showed that the nanovesicles expressed exosome markers CD63 and CD81 (Figure ​(Figure2A).2A). Transmission electron microscopy showed diameters of extracellular vesicles ranging from 50 to 200 nm (Figure ​(Figure2C).2C). These nanovesicles were homogeneous in appearance and showed features of exosomes.

Open in a separate window

Figure 2

Acidic HCC cell-derived exosomes promote migration and invasion of the non-acidic recipient HCC cells. A. Western blots of CD63 and CD81 in exosomes derived from SMMC-7721 and Hep3B cells cultured under different pHs. B. Western blot analysis of HIF-1α and HIF-2α in normoxia. C. Transmission electron microscopy of exosomes isolated from the cell culture supernatant. D. Confocal microscopy of the Dil-labeled exosomes (red). E. MTT assays of the proliferation of recipient cells treated with acidic HCC-derived exosomes. F. Invasion of recipient cells treated with acidic HCC-derived exosomes. G. Migration of recipient cells treated with acidic HCC-derived exosomes. Experiments performed in triplicates. Non-Ac exo, the exosomes derived from HCC cells cultured in the non-acidic medium. Ac mock exo, exosomes derived from HCC cells cultured in the acidic medium. Ac ctrl shRNA exo, exosomes derived from HCC cells transfected by empty vector cultured in the acidic medium. Ac HIF-1 KD exo, exosomes derived from HIF-1α KD HCC cells cultured in the acidic medium. Ac HIF-2 KD exo, exosomes derived from HIF-2α KD HCC cells cultured in the acidic medium. Ac HIF-1+2 KD exo, exosomes derived from HIF-1α and HIF-2α KD HCC cells cultured in the acidic medium. *, P < 0.05. **, P < 0.01.

To test whether exosomes derived from HCC cells cultured under acidic condition could influence the biological behavior of recipient cells, we first performed fluorescence microscopy to visualize the endocytosis of exosomes by recipient cells. As shown in Figure. 2D, Dil-labeled exosomes derived from HCC cells cultured under acidic condition were endocytosed by normal HCC cells after incubation for 2 hours. Subsequently, we investigated the effects of acidic condition-derived exosomes on oncogenic properties of non-acidic HCC cells, including cell proliferation, migration, and invasion. MTT, invasion and migration assays showed that acidic HCC- derived exosomes significantly increased cell proliferation, migration, and invasion of recipient cells compared with non-acidic HCC-derived exosomes (Figure ​(Figure2E,2E, 2F, 2G, and Supplementary Figure S6). As previously reported, we verified that acidifying the extracellular milieu of HCC cells could trigger the activation of HIF-1α and HIF-2α in normoxia (Figure ​(Figure2B).2B). Intriguingly, exosomes derived from HIF-1α and HIF-2α knockdown (KD) cells cultured under acidic condition could not increase cell proliferation, migration, and invasion (Figure ​(Figure2E,2E, 2F, 2G, and Supplementary Figure S6).

Identification of candidate functional miRNAs in acidic HCC cell-derived exosomes

It is well known that exosomes could deliver both mRNAs and miRNAs to recipient cells and affect cellular functions ³⁸. To identify exosomal miRNAs that may play an oncogenic role, we performed comparative microarray profiling of miRNAs derived from exosomes from the supernatant of SMMC-7721 cells cultured under normal (pH 7.4) or acidic conditions (pH 6.6) for at least 4 weeks.

Five hundred forty-five and 802 known miRNAs were identified in acidic and non-acidic exosomes, respectively, of which 535 miRNAs were shared by both (Figure ​(Figure3A).3A). We compared the expression levels of miRNAs between acidic and non-acidic exosomes. Using a threshold cutoff of 3-fold and P < 0.05, we discovered that expression of 81 miRNAs was significantly different between the two kinds of exosomes (Supplementary Table S1), of which 70 miRNAs were upregulated in acidic exosomes compared with non-acidic exosomes. Gene ontology (GO) and KEGG analysis showed that genes regulating EMT (epithelial-mesenchymal transition) and PI3K- AKT signaling were among top upregulated clusters. (Figure ​(Figure3B,3B, Supplementary Figure S7 and S8).

Open in a separate window

Figure 3

Identification of functional miRNAs based on miRNA expression profiles of exosomes derived from acidic HCC cells and TCGA dataset. A. The pie chart shows different miRNAs from acidic and non-acidic exosomes. B. Gene ontology (GO) analysis and KEGG analysis of miRNAs identified in (A). C. The volcano plot shows 84 differentially expressed miRNAs from 49 pairs of samples of TCGA dataset. D. 18 miRNAs identified by the LASSO logistic regression model with non-zero coefficients. E. Schematic illustration for the selection of candidate miRNAs.

To further identify the functional miRNAs in acidic HCC-derived exosomes, we analyzed 49 paired HCC and adjacent normal tissues in the TCGA data set to identify differentially expressed miRNAs in these samples. The volcano plot showed that the 84 miRNAs showed more than 2-fold change in the 49 pairs of samples (t-test, all P < 0.0001, Figure ​Figure3C).3C). We used a LASSO (the least absolute shrinkage and selection operator) logistic regression model to analyze the 84 miRNAs and found that 18 had non-zero coefficients (Figure ​(Figure3D3D and Supplementary Figure S9). miR-21 and miR-10b were upregulated in both TCGA dataset and exosome microarray analyses and therefore were considered as the most important functional miRNAs in acidic HCC-derived exosomes (Figure ​(Figure33E).

Acidic condition stimulates exosomal miR-21 and miR-10b expression in HIF-1α and/or HIF-2α-dependent manner

We further confirmed the miRNA microarray results by real-time PCR. Consistent with the microarray results, exosomes secreted from HCC cells cultured under pH 6.6 had higher levels of miR-10b, miR-21, miR-223, and miR-940 than those derived from cells cultured under pH 7.4 condition (Figure ​(Figure4A).4A). We focused on miR-21 and miR-10b for further study because they were extracted from the LASSO logistic model and were higher in acidic exosomes (Figure ​(Figure3E).3E). Under normal condition, cellular and exosomal levels of miR-21 and miR-10b were not influenced by knocking-down either HIF-1α or HIF-2α (Figure ​(Figure4B4B and ​and4C).4C). However, under acidic condition, knocking-down HIF-1α or HIF-2α led to decreased levels of both cellular and exosomal miR-21 and miR-10b (Figure ​(Figure4B4B and ​and4C).4C). We also used an additional HCC cell line Hep3B to confirm these results (Supplementary Figure S10).

Open in a separate window

Figure 4

Acidic microenvironment induces exosomal miR-21 and miR-10b expression through HIF-1α or HIF-2α. A. Comparison of the levels of miR-10b, miR-21, miR-223 and miR-940 from acidic and non-acidic media by RT-PCR. B and C. Comparison of the levels of cellular and exosomal miR-21 and miR-10b in HIF-1α or HIF-2α knockdown cells by RT-PCR. D. Binding of HIF-1α and HIF-2α to the predicted HRE regions in miR-21 and miR-10b promoters by ChIP assays. Input DNA was used as positive control and miR-130b was used as a negative control. E. Expression of cellular miR-21 and miR-10b in SMMC-7721 and Hep3B cells cultured under pH 6.6 and pH 7.4. F. The relationship between exosomal miR-21/miR-10b levels and pH in normal HCC cells and miR-21/miR-10b mimics-transfected HCC cells. Experiments performed in triplicates. *, P<0.05. **, P<0.01. Non-Ac condition, SMMC-7721 cells were cultured in pH 7.4 medium. Ac condition, SMMC-7721 cells were cultured in pH 6.6 medium. HIF-1 KD, HIF-1α KD SMMC-7721 cells. HIF-2 KD, HIF-2α KD SMMC-7721 cells. HIF-1+2 KD, HIF-1α and HIF-2α KD SMMC-7721 cells.

HIF-1α and HIF-2α bind to the HRE sites in promoters. We next performed ChIP assay to validate the binding of HIF-1α and HIF-2α to the HRE regions of miR-21 and miR-10b promoters. As shown in Fig ​Fig4D,4D, anti-HIF-1α or HIF-2α antibodies immunoprecipitated genomic DNAs containing miR-21 and miR-10b HRE-containing promoter regions in cells cultured under pH 6.6 conditions. In contrast, the miR-130 promoter region (without HRE region) did not show recruitment of HIF-1α or HIF-2α under acidic condition. The levels of miR-21 and miR-10b were higher in SMMC-7721 and Hep3B cell lines cultured under acidic condition than non-acidic condition (P < 0.05, Figure ​Figure4E).4E). Also, after transfection of miR-21/10b into both HCC cell lines, we found that the levels of exosomal miR-21 and miR-10b were substantially increased in cells cultured under acidic condition (Figure ​(Figure4F).4F). These data provided evidence that in HCC cells cultured under acidic condition, increased HIF-1α and HIF-2α could directly bind to the HRE-containing promoter regions of miR-21 and miR-10b and up-regulate their cellular expression. The increased cellular expression of miR-21 and miR-10b may result in the elevated levels of these miRNAs in exosomes. However, the precise mechanism for increased loading of these miRNAs in exosomes remains unclear and needs to be investigated. Furthermore, silencing HIF-1α and/or HIF-2α in SMMC-7721 and Hep3B cells under acidic condition resulted in decreased levels of cellular miR-210 (another target of HIF-1α), whereas the levels of exosomal miR-210 were not altered (Supplementary Figure S11). These results indicated complex regulation of expression for individual miRNAs and their loading into exosomes.

miR-21 and miR-10b mediate acidic exosome-induced HCC cell proliferation, migration, and invasion

Knockdown of miR-10b and miR-21 in acidic exosomes significantly decreased cell proliferation, migration, and invasion of recipient cells compared with non-acidic-induced exosomes (Figure ​(Figure55A-​A-5E,5E, Supplementary Figure S12). To study whether cell migration and invasion induced by miR-21 and miR-10b from acidic exosomes depend on HIF-1α or HIF-2α, we overexpressed miR-21 and miR-10b in HIF-1α- or HIF-2α knocked-down cells. Transfection of miR-21 and miR-10b mimics in HIF-1α- or HIF-2α knocked-down cells restored miR-21 and miR-10b levels in exosomes (Supplementary Figure S13). Furthermore, restoration of miR-21 and miR-10b levels in acidic exosomes increased recipient cell proliferation, migration, and invasion (Figure ​(Figure55F-​F-5J).5J). We also examined whether exosomal miR-21 and miR-10b induced EMT-related changes in HCC cells, and found that exosomal miR-21 and miR-10b increased Vimentin and Snail expression while decreased PTEN and E-cadherin levels in treated cells (Figure ​(Figure6A).6A). These results suggested that acidic exosome-induced cell migration and invasion were dependent on miR-21 and miR-10b and these two exosomal miRNAs could also induce EMT in the recipient cells.

Open in a separate window

Figure 5

miR-21 and miR-10b mediated acidic exosome-induced cell proliferation, migration, and invasion. A, F, MTT assay; B, D, G and I, invasion assay; C, E, H and J, scratch assay of cancer cells modified as depicted in the Figure. Experiments performed in triplicates. *, P < 0.05.

Open in a separate window

Figure 6

Acidic exosomal miR-21 and miR-10b promote tumor growth and metastasis in xenograft models. A. miR-21 and miR-10b could bind to 3' UTR of PTEN. Western blot analysis of PTEN, p-AKT, E-cadherin, mTOR, Snail and Vimentin. B. Tumor volumes in PBS, non-acidic exosome, miR-21 KD exosome, miR-10b KD exosome, miR-21 mimics exosome, miR-10b mimics exosome, and acidic exosome groups of cancer cells. C. Lung metastases numbers of groups of cancer cells as in (B). D. Western blot analysis of PTEN, E-cadherin, and Vimentin of cancer cells. E. Immunofluorescence staining of E-cadherin and Vimentin.

Acid-derived exosomal miR-21 and miR-10b promote HCC cell growth and metastasis in a xenograft model

We used a xenograft model and a lung metastasis model to investigate the effects of tumor- derived exosomes on tumor growth and metastasis (Supplementary Figure S14). Overexpression of miR-21 and miR-10 in acid-derived exosomes increased tumor growth and tumor volume. The tumor volume and numbers of lung metastases were significantly increased in mice injected with acid-treated or miR- 21/10b-overexpressing exosomes-treated cells (Figure ​(Figure6B6B and ​and6C).6C). Knock-down of miR-21 and miR-10b expression in acidic exosomes significantly decreased the tumor volume and numbers of lung metastases (Figure ​(Figure6B6B and ​and6C).6C). We also investigated whether exosomal miR-21 and miR-10b regulated EMT in vivo. Consistent with the in vitro experiments, the expression of Vimentin was increased, while the expression of E-cadherin was decreased in acid-treated and miR-21/10b-overexpressing exosomes (Figure ​(Figure6D6D and ​and6E,6E, Supplementary Figure S15).

We thank the TCGA research network for providing the data analyzed in this manuscript. We are grateful to Wenjun He (Department of Medical Statistics and Epidemiology, School of Public Health, Sun Yat-sen University, Guangzhou, China) for supporting part of the data extraction and processing.

This work was supported by grants from the National Key R&D Program of China (2017YFC1309001, 2016YFC1302305), the National Natural Science Foundation of China (81672686, 81730072); Natural Science Foundation of Guangdong Province, China (2015A030313020); Sister Institution Network Fund of the MD Anderson Cancer Center.