Sorafenib inhibits the shedding of major histocompatibility complex class I–related chain A on hepatocellular carcinoma cells by down-regulating a disintegrin and metalloproteinase 9

Authors

  • Keisuke Kohga,

    1. Department of Gastroenterology and Hepatology, Osaka University Graduate School of Medicine, Osaka, Japan
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Tetsuo Takehara,

    1. Department of Gastroenterology and Hepatology, Osaka University Graduate School of Medicine, Osaka, Japan
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Tomohide Tatsumi,

    1. Department of Gastroenterology and Hepatology, Osaka University Graduate School of Medicine, Osaka, Japan
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Hisashi Ishida,

    1. Department of Gastroenterology and Hepatology, Osaka University Graduate School of Medicine, Osaka, Japan
    Search for more papers by this author
  • Takuya Miyagi,

    1. Department of Gastroenterology and Hepatology, Osaka University Graduate School of Medicine, Osaka, Japan
    Search for more papers by this author
  • Atsushi Hosui,

    1. Department of Gastroenterology and Hepatology, Osaka University Graduate School of Medicine, Osaka, Japan
    Search for more papers by this author
  • Norio Hayashi

    Corresponding author
    1. Department of Gastroenterology and Hepatology, Osaka University Graduate School of Medicine, Osaka, Japan
    • Department of Gastroenterology and Hepatology, Osaka University Graduate School of Medicine, 2-2 Yamadaoka, Suita, Osaka 565-0871, Japan
    Search for more papers by this author
    • fax: 81-6-6879-3629.


  • Potential conflict of interest: Nothing to report.

Abstract

The ectodomain of major histocompatibility complex class I–related chain A (MICA) is shed from tumor cells, and may be an important means of evading antitumor immunity. This study investigated the roles of a disintegrin and metalloproteinase 9 (ADAM9) in the shedding of MICA in human hepatocellular carcinoma (HCC). Small interfering RNA–mediated knockdown (KD) of ADAM9 resulted in up-regulation of membrane-bound MICA expression on the HepG2 and PLC/PRF/5 cellular surfaces and down-regulation of soluble MICA levels in their culture supernatant. ADAM9 was cleaved at a site between Gln347 and Val348 of MICA in vitro. We constructed a plasmid of the MICA gene with mutation or deletion of the ADAM9 cleavage site to examine the detailed mechanism of MICA shedding by ADAM9 protease. The results suggested that MICA might be cleaved at the intracellular ADAM9-recognized cleavage site and was further cleaved at the extracellular ADAM9-independent cleavage site in HCC cells, resulting in the production of soluble MICA. Immunohistochemical analysis revealed that ADAM9 was overexpressed in human HCC compared to normal liver tissues. The cytolytic activity of natural killer (NK) cells against ADAM9KD-HCC cells was higher than that against control cells, and the enhancement of this cytotoxicity depended on the MICA/B and NK group 2, member D pathway. Sorafenib treatment resulted in decreased expression of ADAM9, increased expression of membrane-bound MICA expression, and decreased levels of soluble MICA in HCC cells. Adding sorafenib enhanced the NK sensitivity of HCC cells via increased expression of membrane-bound MICA. Conclusion: ADAM9 is involved in MICA ectodomain shedding in HCC cells, and sorafenib can modulate ADAM9 expression. Sorafenib therapy may have a previously unrecognized effect on antitumor immunity in patients with HCC. (HEPATOLOGY 2010.)

Hepatocellular carcinoma (HCC) is one of the leading causes of cancer death worldwide. Chronic liver disease caused by hepatitis virus infection and nonalcoholic steatohepatitis leads to a predisposition for HCC; liver cirrhosis, in particular, is considered to be a premalignant condition.1, 2 With regard to treatment, surgical resection or percutaneous techniques such as ethanol injection and radiofrequency ablation are considered to be choices for curable treatment of localized HCC, whereas transarterial chemoembolization (TACE) is a well-established technique for more advanced HCC.3 The liver contains both a large compartment of innate immune cells (natural killer [NK] cells and NK T cells) and acquired immune cells (T cells),4, 5 but the activation of these immune cells after HCC treatment remains unclear. If such treatments can efficiently activate abundant immune cells in the liver, this could lead to the establishment of attractive new strategies for HCC treatment.

Major histocompatibility complex (MHC) class I–related chain A (MICA) is a ligand for NK group 2, member D (NKG2D) receptors expressed on a variety of immune cells.6 In contrast to classical MHC class I molecules, MICA is rarely expressed on normal cells but frequently on tumor cells.7–10 The engagement of MICA and NKG2D strongly activates NK cells, enhancing their cytolytic activity and cytokine production.11 Thus, the MICA-NKG2D pathway is an important mechanism by which the host immune system recognizes and kills transformed cells.12 In addition to those membrane-bound forms, MICA molecules are cleaved proteolytically from tumor cells and appear as soluble forms in sera of patients with malignancy, including HCC.13–17 The release of soluble MICA/MHC class I–related chain B (MICB) from tumor cells is thought to antagonize NKG2D-mediated immunosurveillance. Recently, members of the metzincin superfamily, such as disintegrin and metalloproteinase (ADAM) proteins have been reported to play essential roles in the proteolytic release of the ectodomain of transmembranous proteins, including MICA, from the cell surface.14, 18 MICA shedding of 293T fibroblast cells and HeLa cervical cancer cells was found to be inhibited by silencing of the ADAM10 and ADAM17 proteases.19 We also demonstrated that ADAM10, but not ADAM17, proteases are associated with MICA shedding in human HCC.20 However, it remains to be determined whether other ADAM proteases can affect MICA shedding.

Sorafenib is a unique multitargeting kinase molecule that inhibits the receptor tyrosine kinases (vascular endothelial growth factor receptor 2 [VEGFR-2], VEGFR-3, Flt-3, platelet-derived growth factor receptor [PDGFR], and fibroblast growth factor receptor 1) as well as Raf serine-threonine kinase in signal transduction. A recent phase III study, the Sorafenib HCC Assessment Randomized Protocol (SHARP), revealed that the median overall survival of sorafenib-treated patients with HCC was significantly higher than that of patients who received the placebo.21 To develop further uses for sorafenib in HCC treatment, its immunological impact in HCC treatment needs to be evaluated.

In this study, we investigated the association of ADAM9 proteases with MICA shedding in human HCC cells. Of importance is the discovery that ADAM9 knockdown (KD) experiments revealed the essential roles of ADAM9 protease in the shedding of MICA molecules. Sorafenib, a multikinase inhibitor that has been recently approved as a new anti-HCC molecular targeted chemotherapy, was effective in down-regulating soluble MICA and up-regulating membrane-bound MICA via inhibition of ADAM9 protease, resulting in enhancing the NK sensitivity of sorafenib-treated HCC cells. This study sheds light on previously unrecognized effects of sorafenib on modulating ADAM9 and MICA shedding, and thus suggests promise for its use in chemoimmunotherapy against human HCC.

Abbreviations

Ab, antibody; ADAM, a disintegrin and metalloproteinase; ELISA, enzyme-linked immunosorbent assay; HCC, hepatocellular carcinoma; HLA, human leukocyte antigen; KD, knockdown; MHC, major histocompatibility complex; MICA, MHC class I–related chain A; mRNA, messenger RNA; NK, natural killer cell; PBS, phosphate-buffered saline; RT-PCR, reverse transcription polymerase chain reaction; siRNA, small interfering RNA.

Materials and Methods

HCC Cell Lines.

Human HCC cell lines HepG2 and PLC/PRF/5 were purchased from the American Type Culture Collection (Manassas, VA) and were cultured with Dulbecco's modified Eagle medium (DMEM) supplemented with 10% fetal bovine serum (GIBCO/Life Technologies, Grand Island, NY) in a humidified incubator at 5% CO2 and 37°C.

Reagents.

Sorafenib was kindly provided by Bayer HealthCare Pharmaceuticals Inc. (Wayne, NJ). The compound was dissolved in 100% dimethyl sulfoxide (DMSO) to a final concentration of 100 mM. The dissolved solution was diluted with DMEM supplemented with 10% heat-inactivated fetal bovine serum (Sigma, St. Louis, MO) to 1–15 μmol/L. HCC cell viability was determined at 72 hours after addition of 1–15 μmol/L sorafenib or DMSO by WST-8 assay using cell count reagent sulforaphane (Nacalai Tesque, Kyoto, Japan) as previously described.20

RNA Silencing.

The small interfering RNA (siRNA) method was used to knockdown ADAM9 as previously described.20 At 24 hours after transfection, the cells were analyzed for specific depletion of the messenger RNA (mRNA) of ADAM9 by real-time reverse transcription polymerase chain reaction (RT-PCR) according to the manufacturer's instructions (Applied Biosystems, Foster City, CA). The following siRNAs were used: ADAM9, 5′-UGUCCAAACACAUUAAUCCCGCCUG-3′; scramble control, 5′-UGUCGCACAAACACUUAACUCCCUG-3′.

Enzyme-Linked Immunosorbent Assay.

HCC cells were cultured with tumor necrosis factor-α protease inhibitor-I (TAPI-I, 50 μmol/L; Calbiochem, San Diego, CA) or sorafenib (1 μmol/mL) for 24 hours and the supernatants were harvested. The supernatants of cultured HCC cells were harvested at 24 hours after transfection with siRNA. The levels of soluble MICA were determined by DuoSet MICA enzyme-linked immunosorbent assay (ELISA) kit (R&D Systems, Minneapolis, MN).

Flow Cytometry.

For the detection of membrane-bound MICA, cells were incubated with anti-MICA antibody (Ab) (Santa Cruz Biotechnology, Santa Cruz, CA) and stained with phycoerythrin-goat anti-mouse immunoglobulin (Ig) (Beckman Coulter, Fullerton, CA) as a secondary reagent and then subjected to flow cytometric analysis using a FACScan flow cytometer (Becton Dickinson, San Jose, CA).

Real-Time RT-PCR.

Total RNA was isolated using the RNeasy Mini Kit (Qiagen K.K., Tokyo, Japan), and was reverse transcribed using SuperScript III First-Strand Synthesis System (Invitrogen, Carlsbad, CA). The mRNA levels were evaluated using ABI-Prism 7900 Sequence Detection System (Applied Biosystems). Ready-to-use assays (Applied Biosystems) were used for the quantification of ADAM9 (Hs00177638_m1), and β-actin (Hs99999903_m1) mRNAs according to the manufacturer's instructions. β-Actin mRNA from each sample was quantified as an endogenous control of internal RNA.

Mass Spectrometry Analysis to Determine the Cleavage Site.

Peptides of 20 amino acid residues partially overlapping each other, covering the α3 domain to the C-terminal end of MICA were synthesized by Sigma. Each peptide substrate (30 μM) was incubated with 50 nM of recombinant ADAM9 in a buffer containing 10 mM HEPES (pH 7.2) and 0.0015% Brij (Sigma). After digestion, the samples were passed over a C18 media (ZipTipC18; Millipore, Billerica, MA), eluted with acetonitrile, and analyzed by matrix-assisted laser desorption/ionization–time of flight/mass spectrometry (MALDI-TOF/MS) to determine the masses of the products and thereby the cleavage site recognized by ADAM9.

Plasmid Construction of pMyc-MICA.

An expression vector of MICA, pcDNA-MICA, was constructed by using specific complementary DNA (cDNA) from the human hepatoma-derived cell line, Huh-7, as described.20 The Myc-tag coding sequence was inserted between the putative leader peptide and the α1 domain of the MICA cDNA using QuikChange Site-Directed Mutagenesis Kit (Stratagene, La Jolla, CA), referred to as pMyc-MICA. For construction of pcDNA-MICA-mut or pMyc-MICA-mut, Val348 and Leu349 were substituted for alanine. pcDNA-MICA-del or pMyc-MICA-del, which expresses MICA (or myc-tagged MICA) truncated at Val348, was generated by introducing a stop codon after Gln347. The stop codon was inserted after Pro298, the C-terminus of the putative α3 domain, to construct soluble MICA expression vectors, pcDNA-MICA-sol or pMyc-MICA-sol. Cells were transfected with the MICA expression vectors using Lipofectamine LTX reagent (Invitrogen). As a control, cells were cotransfected with pEGFP-C1 (Clontech, Mountain View, CA) to monitor the transfection efficiencies.

Immunoprecipitation.

The lysates of cells or tissues were prepared as previously described.20 Immunoprecipitation with anti-c-Myc beads was performed for 1 hour at 4°C. Immunocomplexes were eluted by c-Myc tagged peptide solution (MBL, Woburn, MA). The samples after immunoprecipitation were treated with 250 mU of N-glycosidase F (Roche, Mannheim, Germany) for 3 hours at 37°C.

Western Blotting.

The total cellular protein was electrophoretically separated by sodium dodecyl sulfate-12% polyacrylamide gels and transferred onto polyvinylidene fluoride membrane. The membrane was blocked in Tris-buffered saline-Tween containing 5% skim milk for 1 hour, and then probed with anti-Myc mouse monoclonal antibody (mAb) (Cell Signaling Technology, Danvers, MA), anti-ADAM9 mAb (R&D Systems) at 4°C overnight. Horseradish peroxidase–conjugated anti-rabbit Ab and SuperSignal West Pico System (Pierce, Rockford, IL) were used for the detection of blots.

Liver Tissues and Immunohistochemistry.

Human HCC tissues (n = 11) obtained at surgical resection were used. Informed consent, under a protocol approved by Institutional Review Board, was obtained from all patients before sample acquisition. Liver sections were subjected to immunohistochemical staining using the ABC procedure (Vector Laboratories, Burlingame, CA). The primary Ab used was anti-ADAM9 (R&D Systems). To confirm the specificity of the staining, primary antibodies were incubated with recombinant ADAM9 protein (R&D Systems) for 3 hours and then applied onto liver sections in parallel with staining of the primary Abs as the absorption test.

NK Cell Analysis.

NK cells were isolated from human peripheral blood mononuclear cells by magnetic cell sorting using CD56 MicroBeads according to the manufacturer's instructions (Miltenyi Biotec, Auburn, CA). The cytolytic abilities of NK cells against ADAM9KD/control HCC cells or 0.5 or 1 μmol/L sorafenib-treated HCC cells were assessed by 4-hour 51Cr-releasing assay with or without MICA/B-blocking Ab (6D4; a generous gift from Dr. Veronika Groh and Dr. Thomas Spies, of the Fred Hutchinson Cancer Research Center, Seattle, WA),7 which binds to the α1 and α2 domains of MICA.

Statistics.

All values were expressed as the mean and standard deviation. The statistical significance of differences between the groups was determined by applying the Student t test or two-sample t test with Welch correction after each group had been tested with equal variance and Fisher's exact probability test. We defined statistical significance as P < 0.05.

Results

TAPI-I Blocked the Shedding of MICA in Human HCC.

We added TAPI-I, an α-secretase inhibitor, to human HCC cells and evaluated the membrane-bound MICA and soluble MICA production in human HCC. Both HepG2 cells and PLC/PRF/5 cells expressed membrane-bound MICA and produced soluble MICA in the culture supernatants (Fig. 1A). Membrane-bound MICA expression increased and the production of soluble MICA decreased after TAPI-I treatment in both HepG2 and PLC/PRF/5 cells. These results suggested that the modification of MICA expression on HCC cells might depend on an α-secretase, such as ADAM9, ADAM10, ADAM12, and ADAM17. We had previously investigated the roles of ADAM10 and ADAM17 in the shedding of MICA in human HCC20 and found that ADAM12 was not expressed in human HCC cells (data not shown). In this study, we further investigated the involvement of ADAM9.

Figure 1.

ADAM9 was involved in the shedding of MICA in human HCC. (A) We added TAPI-I (50 μmol/L) to HepG2 or PLC/PRF/5 cells for 24 hours and evaluated the membrane-bound MICA on HCC cells by flow cytometry and soluble MICA production from HCC cells by ELISA. Black histogram, control IgG staining; gray histogram, anti-MICA staining of control HCC cells. Black line, anti-MICA staining of TAPI-I–treated HCC cells; white or black bar, soluble MICA from control HCC cells or TAPI-I–treated HCC cells, respectively. Representative results are shown. Similar results were obtained from three independent experiments. *P < 0.05. (B) HCC cells (HepG2 and PLC/PRF/5) were treated with ADAM9 siRNA or control siRNA, and subjected to analysis of ADAM9 expressions by real-time RT-PCR. (C) The expressions of membrane-bound MICA on HCC cells treated with ADAM9 siRNA (ADAM9KD, black line) or control siRNA (Control, dotted line) were evaluated by flow cytometry. Closed histograms indicate control IgG staining. Soluble MICA production from HCC cells treated with ADAM9 siRNA or control siRNA were evaluated by specific ELISA. *P < 0.05. Representative results are shown. Similar results were obtained from three independent experiments.

ADAM9 Was Involved in MICA Shedding of HCC Cells.

To examine the involvement of ADAM9 in MICA ectodomain shedding, ADAM9 was knocked down in HCC cells using a siRNA-mediated procedure (ADAM9KD). The expression of ADAM9 was clearly suppressed in HepG2 cells and PLC/PRF/5 cells at mRNA levels (Fig. 1B). KD of ADAM9 for both types of HCC cells resulted in increasing membrane-bound MICA and decreasing soluble MICA levels in their culture supernatant (Fig. 1C). These results suggested that ADAM9 is critically involved in the shedding of MICA in HCC cells.

Identification of the ADAM9 Cleavage Site of MICA In Vitro.

Because ADAM9 KD clearly suppressed MICA shedding, we next tried to examine whether ADAM9 is capable of cleaving MICA directly. For this purpose, we carried out an in vitro cleavage assay using recombinant ADAM9 and several synthetic polypeptides which carried the MICA amino acid sequences. After the reaction, the polypeptides were subjected to MALDI-TOF/MS analysis. One of the polypeptides, KTSAAEGPELVSLQVLDQHP, was found to be cleaved by ADAM9. According to the calculated masses, the polypeptide was cleaved between Gln347 and Val348 (Fig. 2A). Based on these data, we constructed a plasmid of Myc-tagged MICA gene with mutation at the ADAM9 cleavage site (“VL” to “AA”, pMyc-MICA-mut; Fig. 2A,B), a plasmid of Myc-tagged MICA gene with a stop codon at Val348 (pMyc-MICA-del, the truncated type of MICA gene; Fig. 2B) and a plasmid of Myc-tagged soluble MICA (pMyc-MICA-sol; Fig. 2B).

Figure 2.

MICA can be a substrate of ADAM9 according to an in vitro cleavage assay. (A) Synthetic polypeptides of MICA were incubated with recombinant ADAM9 and analyzed by MALDI-TOF/MS to identify the cleavage site. One of the synthetic polypeptides, KTSAAEGPELVSLQVLDQHP, was found to be cleaved between Gln347 and Val348. Underlined amino acids Val348 and Leu349 were substituted with alanine as shown in (B) to construct mutant MICA (MICA-mut). (B) Scheme depicting MICA and Myc-tagged MICA expression vectors. CY: cytosolic domain; LP, leader peptide; TM, transmembrane domain. For details, see Materials and Methods. (C) Cleavage of MICA and mutant MICA by recombinant ADAM9. In vitro–translated Myc-MICA and Myc-MICA-mut were incubated with recombinant ADAM9 or left untreated, then analyzed by immunoblot using anti-myc-tag mAb.

Cell-lysates of pMyc-MICA or pcDNA-Myc, a control vector, transfected cells were collected and deglycosylated with tunicamycin. In vitro cleavage assay revealed that the size of full-length MICA was 43 kD, whereas the size of the MICA molecule cleaved by ADAM9 was 39 kD (Fig. 2C, lane 1 and 2), indicating that full-length MICA was an ADAM9 substrate as well as the polypeptide with a partial MICA sequence. To examine whether ADAM9 could directly cleave the identified ADAM9 cleavage site of MICA, in vitro–translated products of pMyc-MICA and pMyc-MICA-mut were treated with ADAM9, followed by immunoblot analysis. The 39-kD product of MICA cleaved by ADAM9 was not detected in the cleavage reaction using pMyc-MICA-mut (Fig. 2C, lane 3 and 4). These results suggested that ADAM9 directly cleaved MICA at the identified ADAM9 cleavage site in vitro.

Ectodomain Shedding of MICA Required a Step of Cytosolic Domain Truncation Mediated by ADAM9.

To examine whether ADAM9 cleavage site was associated with the ectodomain shedding of MICA in HCC cells, we transfected a vector of the MICA gene (pcDNA-MICA), a vector of the MICA gene with mutation at the ADAM9 cleavage site (pcDNA-MICA-mut) or a control vector (pcDNA3) into HepG2 cells and collected the culture supernatants. Soluble MICA levels from pcDNA-MICA transfectants were significantly higher than those from pcDNA3 transfectants. In contrast, transfection of pcDNA-MICA-mut yielded similar levels of soluble MICA as seen with pcDNA3 control transfection (Fig. 3A). Transfection efficacies were similar among all transfectants, as indicated by green fluorescent protein (GFP)-positive rates (Fig. 3A).

Figure 3.

Truncation of cytosolic domain of MICA by ADAM9 is essential for ectodomain shedding. (A) Blockade of ectodomain shedding of MICA by mutation at the ADAM9 recognition site. HepG2 cells were cotransfected with pcDNA3, pcDNA-MICA, or pcDNA-MICA-mut and pEGFP-C1. After 24 hours incubation, the culture media were collected and assayed for soluble MICA by ELISA. Transfection efficiencies were monitored by measuring GFP-positive cell rates by fluorescent-activated cell sorting (FACS). (B) HepG2 cells were cotransfected with pcDNA3, pMyc-MICA, or pMyc-MICA-mut and pEGFP-C1. After 24 hours incubation, the culture media were immunoprecipitated with anti-myc-tag mAb, deglycosylated with glycanase, and then the expression of myc-tagged MICA was detected by immunoblot (upper panel). Expression of myc-tagged MICA and GFP in the cells was confirmed by FACS (lower panel). (C) Western blotting of cell lysate of HepG2 cells and in vitro translation of pMyc-MICA vectors. HepG2 cells were transfected with ADAM9 siRNA (lane 1) or control siRNA transfection (lane 2) followed by transfection of pMyc-MICA. The cell lysates were deglycosylated by tunicamycin as described in Materials and Methods. In vitro translation was carried out using pMyc-MICA-del (lane 3), pMyc-MICA-sol (lane 4), and pMyc-MICA (lane 5). Myc-tagged MICA in the samples was detected by immunoblot using anti-myc-tag mAb. (D) No relevance of ADAM9 found to shedding of MICA lacking the cytosolic domain. HepG2 cells were transfected with pMyc-MICA or pMyc-MICA-del after ADAM9 knockdown. The culture media were immunoprecipitated with anti-myc-tag mAb and deglycosylated, then the expression of myc-tagged MICA was detected by immunoblot. Transfection efficiencies were monitored by measuring GFP-positive cell rates by FACS.

We next transfected expression vectors of Myc-tagged MICA gene (pMyc-MICA), Myc-tagged MICA gene with mutation at ADAM9 cleavage site (pMyc-MICA-mut), or a control vector (pcDNA-Myc) into HepG2 cells and collected the culture supernatants. Immunoprecipitates from those samples with anti-Myc antibody were subjected to western blot analysis after deglycosylation with N-glycanase. Soluble MICA was detected in the supernatants of pMyc-MICA–transfected cells, but not in either pMyc-MICA-mut or pcDNA-Myc–transfected cells (Fig. 3B, upper panel). To verify whether the myc-tagged MICA molecules expressed in the cells were actually transported to the cell surface, we evaluated Myc-tag–positive cells by flow cytometry. Myc-tag–positive rates of pMyc-MICA and pMyc-MICA-mut transfectants were significantly higher than those of pcDNA-Myc transfectants, whereas those of pMyc-MICA transfectants were similar to those of pMyc-MICA-mut transfectants (Fig. 3B). Suemizu et al. have also demonstrated that the “VL” to “AA” mutation did not influence the polarization of MICA expression to the cell surface, which is consistent with our results.22 Taken together, although mutation at the ADAM9 cleavage site did not alter the efficiency of the plasma membrane translocation of MICA, it dramatically inhibited the shedding of MICA, suggesting that the ADAM9 cleavage site has a critical role in the development of soluble MICA. To examine the molecular weight of MICA present in the cells, we transfected pMyc-MICA into control HepG2 or ADAM9KD-HepG2 cells. The whole-cell lysates were immunoprecipitated by anti-Myc Ab and then treated with N-glycanase. In control HepG2 cells, in addition to full-length MICA, two bands with molecular weights of 39 kD and 37 kD were detected (Fig. 3C), whereas neither of them was detected in ADAM9KD-HepG2 cells. These results suggested that ADAM9 protease was required for production of both the 39-kD product and the 37-kD product of MICA in HCC cells. The in vitro translation experiment revealed that the 39-kD product corresponded to ADAM9-cleaved MICA at the aforementioned ADAM9 cleavage site and the 37-kD product corresponded to final soluble MICA proteins formed by the second cleavage of the 39-kD, ADAM9 cleaved product. With respect to these data, two possibilities were raised: (1) ADAM9 activates some protease, which cleaves MICA in the extracellular domain to produce soluble MICA, or (2) 39 kD MICA, which lacks a cytosolic domain, is susceptible to extracellular domain cleavage by some protease. To clarify this, we transfected pMyc-MICA or pMyc-MICA with a stop codon at the ADAM9 cleavage site (pMyc-MICA-del) into control HepG2 or ADAM9KD HepG2 cells. Soluble MICA was detected in the supernatants of pMyc-MICA–transfected control cells, but not of pMyc-MICA–transfected ADAM9KD cells (Fig. 3D). In contrast, pMyc-MICA-del transfection resulted in ectodomain shedding of MICA irrespective of ADAM9 activity. Accordingly, these results suggested that ADAM9 does not directly cleave MICA at the extracellular domain. More importantly, the ADAM9-dependent truncation of cytosolic domain of MICA rendered this molecule susceptible to cleavage to produce soluble MICA.

ADAM9 Is Overexpressed in Human HCC and NK Sensitivity of ADAM9KD HCC Cells.

ADAM9 was detected in all human HCC tissues (N = 11) tested by immunohistochemistry, but not in normal liver tissues (Fig. 4A). The data suggest that overexpression of ADAM9 is a characteristic of human HCC, similar to other malignancies.23 We next evaluated the cytolytic activity of NK cells against HCC cells. The cytolytic activity of NK cells against ADAM9KD-HepG2 or PLC/PRF/5 cells was significantly higher than that against control cells. The cytolytic activities of NK cells against ADAM9KD cells were inhibited by blocking of anti-MICA/B Ab in both HepG2 and PLC/PRF/5 cells, suggesting that the increase of NK sensitivity depended on the increased expression of membrane-bound MICA on ADAM9KD HCC cells (Fig. 4B), although we could not exclude the possible involvement of MICB in this cytotoxicity.

Figure 4.

Expressions of ADAM9 in human HCC tissues and NK sensitivity in ADAM9 KD HCC cells. (A) Immunohistochemical detection of ADAM9 in human HCC tissues (N = 11). Liver sections were stained with the corresponding antibodies (left panels). Primary antibodies were incubated with recombinant ADAM9 protein and then applied to liver sections in parallel as the absorption test (right panels). Representative images are shown. (B) HCC cells (HepG2 and PLC/PRF/5) treated with ADAM9 siRNA or control siRNA were subjected to 51Cr-release assay against NK cells. The cytolytic activity of NK cells against control HCC cells (▪) or ADAM9 KD HCC cells without (♦) or with blocking antibody of MICA/B (6D4) (▴).*P < 0.05 versus the cytolytic activity of NK cells against control HCC cells at each respective E:T ratio. Representative results are shown. Similar results were obtained from three independent experiments.

Sorafenib Inhibits MICA Ectodomain Shedding and Enhanced Susceptibility to NK Cells of HCC Cells.

The above observations led us to investigate whether sorafenib treatment would affect MICA ectodomain shedding in HCC cells. We first examined the cytotoxicity of sorafenib to human HepG2 cells by WST-8 assay. Adding more than 4 μmol/L of sorafenib resulted in a significant decrease in cell growth of HepG2 cells (Fig. 5A). Based on these findings, we used 1 μmol/L of sorafenib to evaluate the biological effect on HepG2 cells without toxicity. ADAM9 expressions in sorafenib-treated HepG2 cells were decreased in a dose-dependent manner at protein levels (Fig. 5B). The mRNA of ADAM9 was also decreased in sorafenib-treated HepG2 cells (Fig. 5B). We previously reported that anti-HCC chemotherapy including epirubicin and doxorubicin reduced ADAM10 expression resulting in inhibiting the shedding of MICA on human HCC cells.20 We also examined ADAM10 expression in sorafenib-treated HepG2 cells. The protein and mRNA expressions of ADAM10 did not change between sorafenib-treated HepG2 cell and nontreated HepG2 cells (Supporting Fig. 1).

Figure 5.

Expression of ADAM9 in sorafenib-treated HCC cells. (A) The cytotoxicity of sorafenib to HepG2 cells was evaluated by WST-8 assay. Cells were treated with different doses of sorafenib (dotted line) or vehicle (DMSO; solid line) for 24 hours, and the viability of the cells was evaluated by WST-8 assay. (B) HepG2 cells were treated with 0.5 or 1 μmol/L sorafenib or vehicle (DMSO) for 24 hours and their protein and mRNA expressions of ADAM9 by western blotting (upper panel) and real-time RT-PCR (lower panel), respectively. Representative results are shown. Similar results were obtained from three independent experiments. *P < 0.05.

Sorafenib treatment also led to an increase in membrane-bound MICA expression and a decrease in soluble MICA production in HepG2 cells in a dose-dependent manner (Fig. 6A). Increased membrane-bound MICA expression and a decrease of soluble MICA were observed in sorafenib-treated control HepG2 cells, but not in ADAM9KD-HepG2 cells (Fig. 6B), suggesting that an increase of membrane-bound MICA expression and a decrease of soluble MICA in sorafenib-treated HepG2 cells depended on ADAM9 expression. NK-mediated effector functions are regulated by a balance between inhibitory and stimulatory signals. NK cells can recognize MHC class I molecules on target cells via surface receptors that signals to suppress NK cell function.24, 25 We also examined the human leukocyte antigen (HLA) class I expressions on sorafenib-treated HepG2 cells by flow cytometry. The expression of HLA class I on sorafenib-treated HepG2 cells was similar to that on nontreated HepG2 cells (Supporting Fig. 2), suggesting that sorafenib did not affect the expression of HLA class I molecule.

Figure 6.

Expression of MICA in sorafenib-treated HCC cells and NK sensitivity in sorafenib-treated HCC cells. (A) HepG2 cells were treated with 0.5 or 1 μmol/L sorafenib or vehicle (DMSO) for 24 hours and their expressions of membrane-bound MICA and the production of soluble MICA in the culture supernatant were evaluated by flow cytometry and specific ELISA, respectively. Closed histograms indicate control IgG staining in flow cytometry. Similar results were obtained from two independent experiments. *P < 0.05. (B) Control HepG2 or ADAM9KD HepG2 cells were treated with 1 μmol/L sorafenib or vehicle (DMSO) for 24 hours, and their expressions of membrane-bound MICA and the production of soluble MICA in the culture supernatant were evaluated by flow cytometry and specific ELISA, respectively. Closed histograms indicate control IgG staining in flow cytometry. Similar results were obtained from two independent experiments. *P < 0.05. (C) The cytolytic activity of NK cells against sorafenib-treated HepG2 cells were evaluated by 51Cr-release assay. Vehicle-treated cells (▴), sorafenib-treated cells (0.5 μmol/mL sorafenib [▪], 1 μmol/mL sorafenib [♦]), 1 μg/mL sorafenib-treated HepG2 cells with blocking antibody of MICA/B (6D4) (●), respectively. *P < 0.05 versus the cytolytic activity of NK cells against vehicle-treated HepG2 cells at each E:T ratio. Representative results are shown. Similar results were obtained from three independent experiments.

We next evaluated whether the sorafenib treatment could also modify the NK sensitivity of human HCC cells. The cytolytic activities of NK cells against sorafenib-treated HepG2 cells were significantly higher than those against nontreated HepG2 cells (Fig. 6C). The cytolytic activity against sorafenib-treated HepG2 cells was decreased to the control levels by adding anti-MICA blocking antibody. These results demonstrated that adding sorafenib enhanced the NK sensitivity of HepG2 cells via increased expression of membrane-bound MICA. The sorafenib-treated PLC/PRF/5 HCC cells also showed similar results to those obtained from sorafenib-treated HepG2 cells (data not shown).

Discussion

MICA shedding is thought to be the principal mechanism by which tumor cells escape from NKG2D- mediated immunosurveillance.13 In this study, we demonstrated that ADAM9 was overexpressed in human HCC tissues and that ADAM9 knockdown resulted in increased expression of membrane-bound MICA, decreased production of soluble MICA, and up-regulation of NK sensitivity of human HCC cells. These results point to ADAM9 as a possible therapeutic target for inhibiting MICA shedding, thereby increasing immunity against HCC.

We identified the ADAM9 cleavage site of MICA in vitro, which is located at the intracellular domain of MICA. ADAM9 protease is usually located in the extracellular area, but we revealed that ADAM9 protease is required for the production of not only the 37 kD soluble MICA but also the 39 kD MICA in HCC cells. Based on our present data, it is speculated that ADAM9 protease may enable intracellular cleavage of MICA protein by activating some intracellular protease which can recognize a similar ADAM9-cleavage site of MICA or by direct cleavage of MICA by activating ADAM9 while the intracellular domain of MICA shifts to the extracellular area by a flip-flop mechanism such as that observed with lipids.26 Further study is needed to clarify the detailed mechanism of the intracellular cleavage. On the other hand, ADAM9 does not directly cleave MICA at the extracellular domain, and the ADAM9-dependent truncation of the cytosolic domain of MICA rendered this molecule susceptible to cleavage to produce soluble MICA. These results suggested that 39 kD MICA, which lacks a cytosolic domain, is susceptible to extracellular domain cleavage by some unidentified protease. Interestingly, this unidentified protease is independently activated after ADAM9 activation. This is the first report to show the involvement of ADAM9 in the shedding of MICA in cancer cells, which might offer new insights of the detailed escape mechanism of human HCC cells from the immune-surveillance system.

One of the important findings of the present study is that sorafenib, a new molecular targeted anticancer drug, could remodel HCC cells by down-regulating ADAM9 expressions, thereby inhibiting MICA ectodomain shedding and enhancing sensitivity to NK cells. Liu et al. demonstrated that the antitumor activity of sorafenib in human HCC might be attributed to inhibition of tumor angiogenesis via blocking of VEGF receptor or PDGF receptor and direct effect on HCC cell proliferation/survival through a Raf kinase signaling–dependent and/or Raf kinase signaling–independent mechanism.27 However, early clinical study revealed that sorafenib treatment did not inhibit the progression of HCC tumor, although sorafenib prolonged the median overall survival of patients with advanced HCC.21, 28 This might be partly because sorafenib may not be distributed to HCC tissues enough to induce apoptosis of HCC cells. The ADAM family proteins, which are highly expressed in some tumors, play a role in secreting growth factors, such as heparin-binding epidermal growth factor, and migration of cells. This study is the first to demonstrate that clinically available molecular targeted anticancer drugs have the ability to modulate the expression of ADAM family proteins and NK sensitivity of tumor cells even if HCC cells were treated with a nontoxic dose of sorafenib. Sorafenib seemed to suppress ADAM9 expression at a transcriptional level, but the precise mechanism of this suppression is not yet known. Because sorafenib enhances NK sensitivity of HCC cells, if liver NK cells are efficiently activated during sorafenib treatment, an additional antitumor effect against HCC cells could be expected. We previously demonstrated that immune modulators such as α-galactosylceramide can efficiently activate liver innate immune cells including NK cells.29, 30 The combination therapy of anti-HCC molecular targeted therapy and immunotherapy targeting activation of NK cells might improve the antitumor effect against unresectable HCC and the prognosis of patients with HCC.

In spite of recent progress and early successes reported for HCC therapies, there remains significant room for improvement, especially with respect to advanced liver cancer. We have shown here that ADAM9 plays essential roles in MICA shedding in human HCC cells and that anti-HCC molecular targeted therapy enhances NK sensitivity of HCC cells via inhibition of the activity of ADAM9 protease followed by modification of MICA expression. These findings indicate that modulation of MICA shedding mediated by ADAM9 might represent a particularly promising approach to suppressing tumor growth and promoting regression in patients with HCC.

Ancillary