Interferon Alfa Receptor Expression and Growth Inhibition by Interferon Alfa in Human Liver Cancer Cell Lines

Type I interferon (IFN) receptor consists of two chains (Hu-IFN- a R1 and Hu-IFN- a R2), and Hu-IFN- a R2 takes a soluble (Hu-IFN- a R2a), short (Hu-IFN- a R2b), or long (Hu-IFN- a R2c) form. We examined the expression of type I IFN receptor, the growth-suppression effect of IFN- a , and their relationship in 13 liver cancer cell lines. With reverse-transcription polymerase chain reaction (RT-PCR) analysis, the expressions of Hu-IFN- a R1, Hu-IFN- a R2a, and Hu-IFN- a R2c were confirmed in all cell lines, and that of Hu-IFN- a R2b in 12 cell lines. All cell lines expressed mRNAs of a transcriptional activator, interferon regulatory factor (IRF)-1, and its antagonistic repressor (IRF-2). Flow cytometry revealed weak expression of Hu-IFN- a R2 on the cell surface in 12 cell lines. The soluble-form protein of Hu-IFN- a R2 was detected at varying levels in culture supernatants of all cell lines with enzyme-linked immunosorbent assay (ELISA). Cell proliferation was suppressed in proportion to the dose of human natural IFN- a at 96 hours of culture, but it was not clearly related to the expression of Hu-IFN- a R2 protein on the cell surface. Investigations on the morphology, DNA, and cell cycle presented four growth suppression patterns as a result of IFN- a : 1) induction of apoptosis and blockage of cell cycle at the S phase (9 cell lines); 2) blockage at the S phase (2 cell lines); 3) induction of HCC nodules; KYN-2 and HAK-6, from surgically resected moderately to poorly differentiated HCC nodules; and KYN-3, HAK-4, and HAK-5, from peritoneal effusion of HCC patients with moderately to poorly differentiated, poorly differentiated, and sarco- matous HCCs, respectively. HAK-1A and HAK-1B were established from a single HCC nodule showing a three-layered structure with a different histological grade in each layer. HAK-1A cells the outer layer of the original tumor, the inner layer.

Interferon alfa (IFN-␣) has been shown to possess antiviral activity, antiproliferative activity, and various immunoregula-tory activities including: 1) stimulation of cytotoxic activities of lymphocytes and macrophages, and of natural killer cell activity; and 2) induction of class I major histocompatibility complex antigens. 1 The effects of IFN-␣ are mediated through interaction with the specific cell-surface receptor, type I IFN receptor. This receptor consists of two chains, Hu-IFN-␣R1 and Hu-IFN-␣R2, which can be present in different forms. [2][3][4][5][6] The Hu-IFN-␣R1 chain is present as either the full chain (Hu-IFN-␣R1) or a splice-variant (Hu-IFN-␣R1s) lacking exons 4 and 5. Hu-IFN-␣R2 chain exists in soluble, short, and long forms (Hu-IFN-␣R2a, Hu-IFN-␣R2b, and Hu-IFN-␣R2c, respectively). [2][3][4] Most likely, the Hu-IFN-␣R1 and Hu-IFN-␣R2c chains represent the predominantly active form. 2 Binding of the receptor and IFN-␣ induce transcription of IFN-inducible genes through the activation of the Jak/signal transducer and activator of transcription (STAT) signaling pathway. [7][8][9] Interferon regulatory factor (IRF)-1, a transcriptional activator, and its antagonistic repressor, IRF-2, have been identified as regulators of type I IFN (mainly IFN-␣ and IFN-␤) and IFN-inducible genes. [10][11][12] IRF-1 has recently been shown to inhibit cell proliferation, induce apoptosis, and manifest antioncogenic activities, 10,13-17 while IRF-2 has the oncogenic potential. 10 The IRF-1 gene itself is IFN-inducible and may thus be one of the critical target genes mediating IFN action. 11 Antivirus activity of IFN-␣ has attracted a great deal of attention, and IFN-␣ has been applied in treatment for hepatitis B virus (HBV)-and hepatitis C virus (HCV)-related chronic hepatitis in several countries (reviewed in Gutterman 18 ). In the liver of HCV-infected patients, expressions of Hu-IFN-␣R1 and Hu-IFN-␣R2 chains were investigated in terms of mRNA level, and the relationship between their expression levels and response to IFN-␣ therapy was reported. 19,20 Although IFN-␣ has been proven to have a curative potential in treatment of HBV-and HCV-associated chronic liver diseases, its effect on hepatocellular carcinoma (HCC), which is a common and often fatal complication of HBV-and HCV-related chronic liver diseases, 21 is not well known. Clinical trials of IFN-␣ in treatment of HCC did not achieve consistent results: one study showed beneficial effects, 22 and the other studies did not show significant antitumor effects. 23,24 In contrast, IFN-␣ has been shown to be useful for the treatment of several malignant diseases, including hairy-cell leukemia and chronic myelogenous leukemia (reviewed in Gutterman 18 ).
Experimental studies showed that IFN-␣ can inhibit the growth of various normal and malignant cells in vitro by inducing cell-cycle changes (e.g., induction of G 0 /G 1 arrest and prolongation of the S phase) [25][26][27][28][29][30][31][32][33][34] and/or apopto-sis. 17,33,[35][36][37][38] To date, antiproliferative effects of IFN-␣ against a few HCC cell lines have been reported in vitro 39,40 ; however, the mechanism of growth inhibition by IFN-␣ has not been studied in HCC cells in detail. Furthermore, no studies have been conducted to clarify whether cancer cells express type I receptors, or whether IFN-␣ suppresses the proliferation of cancer cells in proportion to the expression of the receptors on the cell surface. It is important to deepen the understanding of the action of IFN-␣ on HCC cells, because some patients with HBV-and HCV-related liver diseases may already have a small, clinically undetectable HCC during IFN-␣ therapy. In the present study, we examined: 1) mRNA expressions of type I IFN receptor and IRFs; 2) protein expression of Hu-IFN-␣R2; 3) in vitro growth-inhibitory effects and the mechanism of actions of IFN-␣ in human liver cancer cells, i.e., 11 HCC cell lines and 2 combined hepatocellular and cholangiocarcinoma (CHC) cell lines; and 4) the relationship between (2) and (3).
Flow Cytometry. For cell-surface staining of Hu-IFN-␣R2, cell suspensions (4 ϫ 10 5 cells per tube) were washed once with a washing-buffer (10 mmol/L phosphate-buffered saline [PBS] [pH 7.4]; 0.2% bovine serum albumin; 0.1% NaN 3 ) and reacted with 10 µL of anti-Hu-IFN-␣R2 antibody (final concentration, 10 µg/mL) on ice for 1 hour. Cells were then washed twice with the washing buffer, incubated with 4 µL of fluorescein isothiocyanate-conjugated goat anti-mouse immunoglobulin on ice for 30 minutes in a dark place, washed twice with the washing buffer, fixed in 4% paraformaldehyde on ice for 10 minutes, washed once with the washing buffer, and analyzed by using a FACScan (Becton Dickinson Immunocytometry Systems USA).
Enzyme-Linked Immunosorbent Assay. Cultured cells (4-14 ϫ 10 5 cells per well) were seeded on 100-mm dishes (Falcon 3003), cultured for 24 hours, and then the medium was renewed. After 72 hours, the amount of the soluble form of Hu-IFN-␣R2 in the supernatant was measured by using enzyme-linked immunosorbent assay kits manufactured by Cellular Technology Institute, Otsuka Pharmaceutical Co., Ltd. Cultured cells were washed with PBS twice and collected by a scraper, and the amount of the cellular proteins was determined by using the BCA protein assay reagent (Pierce, Rockford, IL). The amount of the soluble form of the Hu-IFN-␣R2 was assessed after making the correction for the amount of the cellular protein. Each experiment was repeated at least three times, and each experiment was duplicated.

Effect of IFN-␣ on the Proliferation of HCC and CHC Cell Lines.
Effect of natural human IFN-␣ on cultured cell proliferation was investigated by using colorimetric assays with 3-(4,5-dimethylthiazol-2yl-yl)-2,5diphenyl tetrazolium bromide cell growth assay kits (Chemicon, Temecula, CA). Cultured cells (1.5 to 8 ϫ 10 3 cells per well) were seeded on 96-well plates (Falcon, Becton Dickinson Labware, Tokyo, Japan), cultured for 24 hours, and then the medium was replaced by a fresh 100-µL medium alone or medium containing IFN-␣ (1,2,4,8,16,32,64,128, 256, 512, or 1,024 U/mL). After 24, 48, 72, or 96 hours, 100 µL of 3-(4,5-dimethylthiazol-2yl-yl)-2,5diphenyl tetrazolium bromide, 160 µg/mL, was added to each well, cultured for 4 hours, the supernatant was removed, and 100 µL of 40 mmol/L HCl/dimethylsulfoxide was added to each well. Viable cell numbers were estimated by measuring the absorbance with an Easy Reader EAR 400 (SLT Lab Instruments, Salzburg, Austria) by setting the test wavelength at 570 nm and the reference wavelength at 630 nm. Six to eight samples were used in each experiment, and each experiment was repeated at least twice to confirm the reproducibility of the test results.
Morphological Observation. In all experiments, cells were observed daily under a phase-contrast microscope (Nikon, Tokyo, Japan). For light-microscopic observation, cultured cells were seeded on Lab-Tek Tissue Culture Chamber Slides (Nunc, Inc., Roskilde, Denmark), cultured with IFN-␣ (250, 500, or 1,000 U/mL) or without IFN-␣ for 72 hours, fixed in Carnoy' s solution for 10 minutes, and then stained with hematoxylin-eosin.
Quantitative Analysis of Fragmented DNA Induced by IFN-␣. Cells (1 to 4 ϫ 10 5 cells/dish) were seeded on 60-mm dishes, cultured for 24 hours, and then the medium was replaced by a fresh medium alone or medium containing 1,000 U/mL IFN-␣. After 3 days, the amount of fragmented DNA was quantified by using a technique described previously in detail. 50 The percentage of fragmented DNA was calculated as follows: percent of fragmented DNA ϭ supernatant absorbance/(supernatant absorbance ϩ pellet absorbance) ϫ 100. The average value was calculated from at least two independent experiments (n ϭ 7 to 12 in total). Agarose Gel Electrophoresis of DNA. DNAs were isolated from cells cultured with or without 1,000 U/mL IFN-␣ for 72 hours by using Sepa Gene nucleic acid isolation kits (Sanko Junyaku Co. Ltd., Tokyo, Japan). Floating cells and attached cells were collected and washed twice in ice-cold PBS, and the pellet in 100 µL DNA extraction buffer (10 mmol/L Tris-HCl [pH 8.0], 100 mmol/L ethylenediaminetetraacetic acid, 20 µg/mL RNase A) was incubated for 1 hour at 37°C. DNA was then isolated according to the manufacturer' s protocol. A portion (20 µg) of the isolated DNA was electrophoresed in a 1.6% agarose gel containing 0.5% ethidium bromide, visualized by using an ultraviolet illuminator, and photographed.

Sodium Dodecyl Sulfate-Polyacrylamide Slab Gel Electrophoresis and
Western Blotting. Seven of the 10 cell lines on which IFN-␣ induced apoptosis was selected, and the relationship between the induction of apoptosis by IFN-␣ and the appearance of antiapoptotic proteins (Bcl-2 and Bcl-xL) or proapoptotic proteins (Bak and Bax) of the Bcl-2 family, were investigated. After 72 hours of culture with or without 1,000 U/mL IFN-␣, cells were incubated in lysis buffer (50 mmol/L Tris-HCl, 150 mmol/L NaCl, 1 mmol/L ethylenediaminetetraacetic acid, 1% Nonidet P-40, 0.1% sodium deoxycholate, 0.1% sodium dodecyl sulfate, 1 µg/mL aprotinin, 0.5 µg/mL leupeptin, 1 mmol/L phenymethylsulfonyl fluoride) on ice for 30 minutes. The mixture of the cells and the lysis buffer was centrifuged for 30 minutes (15,000 rpm, 4°C). The amount of protein in the supernatant was measured as described above, and the supernatant was mixed with Laemmli' s sample buffer and boiled for 5 minutes. Samples (50 µg per lane) were electrophoresed in the 12.5% sodium dodecyl sulfate-polyacrylamide slab gel, and transferred to a polyvinylidine difluoride transfer membrane (Immobilon-P, Millipore Corporation, Bedford, MA) by using the Trans-blot SD semidry transfer cell (Bio-Rad, Richmond, CA). Membranes were blocked in 5% skim milk in 20 mmol/L Tris-HCl (pH 7.6), 137 mmol/L NaCl, and 0.1% Tween 20 (TBS-T) for 1 hour at 37°C, and then probed with anti-Bcl-2 (dilution, 1:200), anti-Bcl-xL (1:1,000), anti-Bak (1:1,400), or anti-Bax (1:500) antibodies in TBS-T containing 5% skim milk at 4°C overnight. After washing with TBS-T, the membranes were incubated with peroxidase-conjugated goat IgG fraction to rabbit IgG (1:48,000) or peroxidase-conjugated rabbit IgG fraction to mouse IgG (1:34,000) in TBS-T containing 5% skim milk at room temperature for 1 hour, washed several times, and then developed using an enhanced chemiluminescence detection system (Amersham, Buckinghamshire, UK) and exposed to Hyperfilm-MP (Amersham). To verify the relative protein levels in each lane, the membranes were reprobed with anti-␤-actin antibody (1:5,000), and ␤-actin levels were used as an internal control. The intensities of the bands were quantified densitometrically using the NIH Image 1.61 software program. Bcl-2, Bcl-xL, Bak, and Bax levels were normalized against ␤-actin levels. The normalized levels of Bcl-2, Bcl-xL, Bak, and Bax were comparatively analyzed between the cells treated with 1,000 U/mL of IFN-␣ and those without the treatment.
Cell-Cycle Analysis. Cell lines were cultured with IFN-␣ (250 or 1,000 U/mL) or without IFN-␣ for 24, 48, 72, and 96 hours, labeled with 10 µmol/L BrdU for 30 minutes, fixed in 70% cold ethanol at 4°C overnight, stained with anti-BrdU and propidium iodide (Sigma Chemical Co.), and then analyzed using a FACScan. Staining was performed using the modified technique described elsewhere. 51 Briefly, fixed cells were washed with PBS, subjected to double-strand DNA denaturation treatment with 2 mol/L HCl at room temperature for 30 minutes, neutralized with 0.1 mol/L Na 2 B 4 O 7 , washed twice with PBS, 0.5% Tween 20, and 0.5% BSA (PBS-T), incubated for 30 minutes at room temperature with 20 µL anti-BrdU antibody, washed with PBS-T, incubated for 30 minutes at room temperature with 4 µL fluorescein isothiocyanate-conjugated goat anti-mouse immunoglobulin, and washed with PBS-T. DNA was counterstained with 5 µg/mL of propidium iodide for at least 20 minutes before flow cytometric analysis. Percentage of cells in the G 1 , S, and G 2 /M phase was calculated from a dot or contour plot obtained after flow cytometric analyses of double-stained cells.
Statistics. Significance in differences was estimated by using the Student' s t test (two-tailed).

Expression of Hu-IFN-␣R1, Hu-IFN-␣R2, IRF-1, and IRF-2 mRNAs
in the Cell Lines. In all 13 cell lines, a 765-bp band that corresponds to the PCR product of Hu-IFN-␣R1 was detected (Fig. 1A). In the experiments on Hu-IFN-␣R2 expressions using primer pair 3 and 4, a 713-bp band corresponding to the PCR product of Hu-IFN-␣R2b was detected in all cell lines except KMCH-2; 481-and 350-bp bands corresponding to the PCR products of Hu-IFN-␣R2c and Hu-IFN-␣R2a, respectively, were found in all cell lines. The band of Hu-IFN-␣R2c showed the strongest intensity (Fig. 1B). In KMCH-2 (a CHC cell line), three extra bands appeared in addition to those of Hu-IFN-␣R2c and Hu-IFN-␣R2a. In the analysis for the extracellular domain of Hu-IFN-␣R2 using primer pair 5 and 6, the expected 642-bp band was detected in all cell lines; however, in the KMCH-2 cell line, a band was detected at approximately 480 bp in addition to the normal (intact) band (Fig. 1C). Direct sequence analysis of this short band revealed that it represented 474-bp PCR products with complete deletion of exon 7 (data not shown), and this result suggested that two of the three extra bands found in the analysis of Hu-IFN-␣R2 using primer pair 3 and 4 could be the products of Hu-IFN-␣R2c and Hu-IFN-␣R2a, in which 169 bp was skipped from each band (i.e., 312 bp and 181 bp, respectively). Besides these bands, there was one at approximately 420 bp in the KMCH-2 cell line, as well as a quite weak band at approximately 600 bp in all cells except KMCH-2. However, their specificity has not yet been clarified, and this should be examined in future studies.

Expressions of Hu-IFN-␣R2 on the Cell Surface in Each Cell Line.
Hu-IFN-␣R2 expression on the cell surface was detectable at a low level in all cell lines except HAK-4, whose expression level was equivalent to that in negative controls. The positive cell rate was higher than 10% in 2 cell lines, i.e., 20.8% in HAK-1B and 13.2% in KMCH-1; in the range between 5 and 10% in 3 cell lines; and Յ5% in 8 cell lines.

Detection of the Soluble Form of Hu-IFN-␣R2 in the Culture
Supernatants. In all 13 cell lines, the soluble form of Hu-IFN-␣R2 was detected in the spent medium at 72 hours of culture, but the highest amount was approximately 600 pg/mL. When the amount of the soluble form of Hu-IFN-␣R2 was corrected with the amount of cell proteins at the time when the spent medium was collected, the mean level was the highest (1,782.6 Ϯ 77.6 pg/mg protein) in KIM-1, and the lowest (59.8 Ϯ 7.0 pg/mg protein) in HAK-4. KIM-1 expressed Hu-IFN-␣R2 on the cell surface at an equivalently low level to KYN-1 and HAK-6, but its expression of soluble Hu-IFN-␣R2 was at the highest level. If KIM-1 was excluded, there was a relatively good correlation between the expression of Hu-IFN-␣R2 on the cell surface and that of soluble Hu-IFN-a R2 in the spent medium.

Effects of IFN-␣ on the Proliferation of Liver Cancer Cell Lines.
Cell proliferation was clearly suppressed in a dose-dependent manner after 48 hours of the culture in all cell lines except KYN-2, HAK-2, and HAK-3, in which dose-dependent suppression was observed after 72 hours. When chronological changes in the ratio of viable cell numbers in the cultures with 1,024 U/mL of IFN-␣ to the number in the cultures with medium only were monitored, the ratio in the 5 cell lines (KIM-1, KYN-2, KYN-3, HAK-4, and HAK-5) decreased until 72 hours of culture, but it was at a plateau or increased at 96 hours. In the other 8 cell lines, cell proliferation was continuously suppressed with time, though sensitivity was different between the cell lines. Proliferation of all cell lines was suppressed in a dose-dependent manner at 96 hours, and the suppression was significant in the range of 1 to 1,024 U/mL of IFN-␣ in all cells except HAK-5, and in the range of 8 to 1,024 U/mL in HAK-5 (P Ͻ .05-.0001). In particular, the ratio in the 5 cell lines (KYN-1, HAK-1A, HAK-1B, HAK-6, and KMCH-1) decreased to 50% or lower when 1,024 U/mL of IFN-␣ was added to the cultures, and their 50% inhibitory concentration was 291. cells was markedly larger in all cell lines except KYN-3, HAK-4, and KMCH-2 than in the controls, though the number varied among the cell lines. When these cells were cultured for 72 hours with or without IFN-␣ (250, 500, or 1,000 U/mL), the frequency of the cells showing characteristic features of apoptosis (e.g., cytoplasmic shrinkage, chromatic condensation, and nuclear fragmentation) tended to increase with the increase of IFN-␣ level ( Fig. 2A and 2B). Quantitative analysis of fragmented DNA revealed that the rate of fragmented DNA (appearance of apoptosis) was significantly higher in the cultures with IFN-␣ than those without IFN-␣ in all cell lines except KYN-3, HAK-4, and KMCH-2. Agarose gel electrophoresis of DNAs showed the appearance of a clear ladder of fragmented chromosomal DNA that consisted of the multiples of units comprising 180 to 190 bp in the IFN-␣-treated KIM-1, KYN-1, KYN-2, HAK-1A, and KMCH-1. In HAK-1B, the ladder was intensified in the IFN-␣-treated cells. In HAK-3 and HAK-6, the ladder was present, though the intensity was weak. Two (HAK-2 and HAK-5) of the 10 cell lines that had a significant increase in DNA fragmentation by IFN-␣ did not show a clear ladder.

Expression of Apoptosis-Related Proteins in IFN-␣-Treated and
Untreated Cells. Expression of Bcl-2 was confirmed in the 5 cell lines (except KYN-2 and HAK-6) (Fig. 3A), though the level in KYN-1 was low. IFN-␣ treatment enhanced the expression by 1.6 times in KIM-1 and 2.6 times in HAK-1B, while the enhancement of the expressions in the other 3 cell lines was lower than 1.5 times. Expression of Bcl-xL protein was confirmed in all cell lines, and it was enhanced by 1.5 times or more in 4 cell lines, i.e., 1.9 times in KIM-1, 1.5 times in KYN-1, 2.5 times in HAK-1B, and 1.8 times in KMCH-1, and the expressions in the other cell lines were either mildly enhanced or not changed (Fig. 3A). Expression of Bak protein was enhanced by 1.5 times in HAK-6 and 1.7 times in KMCH-1, but the expression in the other cell lines increased only slightly. Expression of Bax increased, but the enhancement was lower than 1.5 times in all cells (Fig. 3B).
Effects of IFN-␣ on Cell Cycle. In 11 of the 13 cell lines (excluding HAK-6 and KMCH-2), the ratio of the cells in the S phase increased, and the ratio of the cells in the G 2 /M phase slightly decreased or did not change, compared with those in the control cells (no treatment) at any time point during 24 to 96 hours after the addition of IFN-␣ 250 or 1,000 U/mL to the cultures. In 8 of the 11 cell lines (excluding KIM-1, KYN-2, and HAK-2), the ratio of the cells in the S phase tended to increase in a dose-dependent manner (Fig. 4A). In HAK-6, in which the most apparent growth-suppression effect of IFN-␣ was observed, the cells tended to be in the G 2 /M phase in a time-and dose-dependent manner (Fig. 4B). In KMCH-2, in which apoptosis was not induced but cell proliferation was suppressed by IFN-␣, the cells tended to be in the G 1 phase in a time-and dose-dependent manner (Fig. 4C). Table 1 summarizes expressions of Hu-IFN-␣R2 proteins on the cell surface and in the spent medium, and the effects of

DISCUSSION
The present study investigated the mRNA expression of type I IFN receptor in 13 human liver cancer cell lines, each of which possesses different morphological characteristics. Human type I IFN receptor consists of two chains, i.e., Hu-IFN-␣R1 and Hu-IFN-␣R2, 2-6 and a recent study suggests that the Hu-IFN-␣R2 is the binding subunit, while the Hu-IFN-␣R1 is a necessary unit to form high-affinity receptors. 3 In regard to Hu-IFN-␣R2 expression, cloning of the gene for Hu-IFN-␣R2 revealed that it produces four different transcripts encoding three different polypeptides that are generated by exon skipping, alternative splicing, and different use of polyadenylation sites. 4 One polypeptide is the soluble form of the receptor (Hu-IFN-␣R2a), and the other two are transmembrane proteins with identical extracellular and transmembrane domains but with divergent cytoplasmic tails of 67 (Hu-IFN-␣R2b) and 251 amino acids (Hu-IFN-␣R2c). 3,4 Transfection experiments revealed that expression of the Hu-IFN-␣R2c, but not Hu-IFN-␣R2b, was required for normal interferon binding, activation of the Jak/STAT signal transduction pathway, IFN-inducible gene expression, and antiviral response. 3,4 Our RT-PCR analysis revealed the expressions of Hu-IFN-␣R1, Hu-IFN-␣R2c, and Hu-IFN-␣R2a in all cell lines, and that of Hu-IFN-␣R2b in all cell lines except KMCH-2. These findings demonstrated that liver cancer cells express, at least at mRNA levels, both Hu-IFN-␣R1 and Hu-IFN-␣R2c, which are necessary for IFN-␣ to express its effects. In KMCH-2, expressions of variant mRNAs of Hu-IFN-␣R2a and Hu-IFN-␣R2c, which would be produced by the skipping of exon 7, were found in addition to the expressions of the normal mRNAs. This indicates the presence of a KMCH-2-specific alternative-splicing mechanism; however, the precise mechanism and meaning of these variant expressions should be investigated further.
Hu-IFN-␣R2 expression at a protein level was then examined with flow cytometry. As a result, generally weak expression of Hu-IFN-␣R2 was detected on the cell surface in all cell lines except HAK-4. IFN-␣ suppressed proliferation of all cell lines in a dose-dependent manner at 96 hours of culture, but there was no clear relationship between the amount of Hu-IFN-␣R2 expression (positive cell rate) on the cell surface and the suppressive effect of IFN-␣.
The true reasons for this are unknown at present, but five possible explanations could be suggested. First, the antibody used in the present study is thought to recognize the number of binding sites of Hu-IFN-␣R2; however, for the expression of IFN-␣ effects, not only the number of binding sites, but also the binding affinity of type I IFN receptor to IFN-␣ is important. Binding affinity could be modulated by the expressions of Hu-IFN-␣R1 and additional membrane proteins; therefore, Hu-IFN-␣R2 expression would not be related to the growth-suppression effects of IFN-␣. Second, soluble Hu-IFN-␣R2 could act as an antagonist and neutralize the effects of IFN-␣. Soluble Hu-IFN-␣R2 was detected at various levels in culture media of all cell lines, and there was a relatively clear relationship between the amount of Hu-IFN-␣R2 protein expressed on the cell surface and that of soluble Hu-IFN-␣R2 protein, when KIM-1 was excluded from the analysis. This indicated that the majority of the cell lines express Hu-IFN-␣R2 on the cell surface and in the soluble form at a certain ratio, but some cell lines overexpress soluble A B FIG. 3. Western blot analysis of antiapoptotic proteins, Bcl-2 and Bcl-xL (A), and proapoptotic proteins, Bak and Bax (B), in 7 liver cancer cell lines that were sensitive to IFN-␣-mediated apoptosis. The cells were incubated with medium alone (Ϫ) or 1,000 U/mL of IFN-␣ (ϩ) for 72 hours, and then used in the analysis. Equal amounts (50 µg) of protein were subjected to electrophoresis in the 12.5% sodium dodecyl sulfate-polyacrylamide slab gel, and transferred to polyvinylidine difluoride transfer membranes. The membranes were then probed with anti-Bcl-2, anti-Bcl-xL, anti-Bak, anti-Bax, or anti-␤-actin. The intensities of the bands were quantified densitometrically using the NIH Image 1.61 software program. Bcl-2, Bcl-xL, Bak, and Bax levels were normalized against ␤-actin levels used as an internal control. The normalized levels of Bcl-2, Bcl-xL, Bak, and Bax were comparatively analyzed between the cells treated with IFN-␣ and those without the treatment.
Hu-IFN-␣R2. However, it is questionable whether soluble Hu-IFN-␣R2 protein would act as an antagonist because the highest level of the soluble Hu-IFN-␣R2 in the present study was approximately 600 pg/mL, which is within the normal range of urinary concentration in humans (0.1 to 1.0 ng/mL) 52 or lower than the blood level in healthy human subjects (1.76 Ϯ 0.74 ng/mL). 53 Third, the antibody used in the present study recognized not only Hu-IFN-␣R2c, which mediates the activation of the Jak/STAT pathway, but also Hu-IFN-␣R2b, which does not take a direct function in the activation of the Jak/STAT pathway. Fourth, sensitivity to IFN-␣ could be decreased as a result of the abnormalities of the signaling molecules of the STAT or IRF family of transcription factors. In regard to IRF-1, there have been reports on gene deletion and rearrangement at the DNA level, 13 and the suppression of normal IRF-1 mRNA expression as a result of the increased exon skipping 49 in preleukemic and leukemic cells. RT-PCR analysis revealed that our 13 cell lines express both IRF-1 and IRF-2 mRNAs, and in regard to IRF-1, the expression of intact IRF-1 products was higher than those of the exon-skipped forms of IRF-1. The human IRF-1 gene is mapped on chromosome 5q31.1. In our previous chromosomal analysis for the 13 cell lines, only KYN-2 lacked one chromosome 5, 43 and this was thought to result in the lack of one IRF-1 allele. This could lower the sensitivity of KYN-2 to the growth-suppression effects of IFN-␣, because IRF-1 manifests antiproliferative properties and restrains cell growth according to the balance between IRF-1 and its antagonistic repressor, IRF-2. 10,15 Fifth, the effects of IFN-␣ could be decreased as a result of the abnormalities of IFN-␣-inducible genes that relate to cell growth suppression. In the present study, growth-suppression mechanisms induced by IFN-␣ varied widely, and even among the cell lines that showed the same pattern, the suppression level was different. Therefore, we hypothesize that each cell line presents different abnormalities or varying degrees of abnormalities in genes that relate to the regulation of apoptosis and cell cycle; these then result in different growth-suppression mechanisms of IFN-␣ or different sensitivity to IFN-␣, and as a result, the growth-suppression effect of IFN-␣ was not related to Hu-IFN-␣R2 expression on the cell surface.
To date, IFN-␣ has been reported to induce apoptosis in various neoplastic cells other than liver cancer cells and in normal cells, 33,35-37 but IFN-␣ has also been reported to suppress apoptosis in some cells. 54,55 In the present study, morphological and DNA analyses revealed that IFN-␣ could induce or enhance apoptosis in 10 of the 13 cell lines (except KYN-3, HAK-4, and KMCH-2). This suggests that IFN-␣ induces apoptosis in liver cancer cell lines at a relatively high frequency. Interestingly, cell lines with a larger amount of fragmented DNA do not always form a clearer DNA ladder, and this suggested that DNA ladder formation associated with apoptosis is the phenomenon limited to the individual cell line.
The mechanism of IFN-␣-mediated apoptosis has not yet been clarified. In 7 of the 13 cell lines, in which IFN-␣ induced apoptosis, we examined whether the four members of the Bcl-2 family, which is the important regulator of apoptosis (i.e., Bcl-2, Bcl-xL, Bak, and Bax), 56 were involved in IFN-␣-mediated apoptosis. As a result, the expressions of these four proteins were not thought to be the direct inducer of IFN-␣-mediated apoptosis. Additionally, normal p53 gene expression is not necessary for apoptosis induced by IFN-␣, because mutations at codon 242 in the p53 gene were found in HAK-1A and HAK-1B 45 that showed IFN-␣-mediated apoptosis. This point also agreed with the previous findings in the other cell lines. 33 Recent studies reported the involve- ment of the up-regulation of the Fas (CD95)-Fas ligand system as a mechanism of IFN-␣-mediated apoptosis. 35,38 We previously investigated Fas expressions and anti-Fasmediated apoptosis in 6 of the 13 cell lines. 50 Because there was no clear relationship between IFN-␣-mediated apoptosis and anti-Fas-mediated apoptosis at least in these 6 cell lines, the Fas-Fas ligand system would not take a central role in the IFN-␣-mediated apoptosis in the 13 liver cancer cell lines. A recent study reported the presence of the IRF-1-dependent apoptosis induction system, and the possible involvement of direct transactivation of IL-1␤-converting enzyme (caspase-1) gene by IRF-1. 16 In addition, in the squamous cell carcinoma cell line, IRF-1 expression was reported to correlate with the retinoic acid-and IFN-induced apoptosis. 17 Therefore, IRF expressions and the activation levels of the caspase family should be further investigated in IFN-␣-mediated apoptosis.
As the effects of IFN-␣ on the cell cycle of various normal and tumor cell lines, induction of G 1 arrest has been well reported. [25][26][27]34 Additional effects reported are the S phase prolongation or S phase block, 27-30 G 2 /M arrest, 31 and others. 32 In the present study, blockage of cell cycle at the S phase was found at the highest frequency. The other effects observed were the induction of blockages in the G 1 and G 2 /M phases in 1 cell line each. Known mechanisms in regard to IFN-␣-induced G 1 arrest are the inhibition of p21/WAF1 and cyclin E-and cyclin D1-dependent kinase 2 activities, suppression of retinoblastoma (Rb) phosphorylation, lowered E2F DNA binding activity, and suppression of c-myc expression. 25,[57][58][59][60][61][62][63] In the liver cancer cells, abnormalities would be present in some of these induction systems. Recently, Qin et al. 28 reported proper IFN signaling and loss or inactivation of the normal G 1 checkpoint conferred by the Rb protein as a mechanism of IFN-␣-and IFN-␤-mediated accumulation of the cells in the S phase. However, Hobeika et al. 63 and Sangfelt et al. 60 reported the induction of G 1 arrest in cells without a normal Rb gene. Therefore, not only a wild-type Rb protein, but also another abnormality in the molecules that regulate cell cycle could be involved.
IFN-␣ was reported to decrease the frequency of HCC in HCV-related cirrhotic patients, 64,65 and among HCV-related chronic hepatitis patients who received IFN-␣ treatment, frequency of HCC was reported to be lower in responders in whom alanine aminotransferase levels were normalized than in nonresponders. 66 It is unknown how HCV contributes to HCC development; however, when hepatocytes continuously receive damage and have replication, the frequency of genetic alteration would increase, and this would lead to the development of HCC. 65,66 On the other hand, the mechanisms by which IFN-␣ suppresses liver cancer development have not been clarified, but possible mechanisms are: 1) the reduced incidence of genetic abnormalities accumulated during cell division, which results from the slowed cell cycle in hepatocytes as a result of the sedation of inflammation; 2) activation of the immune system (e.g., increase of natural killer cell activity); and 3) direct suppressive effects on cancer cells. 65 The present study showed that IFN-␣ directly suppresses the proliferation of liver cancer cells by inducing apoptosis and the inhibition of cell-cycle progression. However, it is still unknown whether IFN-␣ prevents in vivo oncogenesis by expressing these effects in the very-early-stage, clinically undetectable cancer cells and by suppressing their growth. It is because the peak serum level of IFN-␣ is only 23 to 53 U/mL with the dose of IFN-␣ used in the intramuscular injection (500 1,000 million units) for type C chronic hepatitis, 67,68 and this level is lower than the 50% inhibitory concentration of HAK-6, i.e., 86.3 U/mL, which showed the †The amount of the soluble Hu-IFN-␣R2 in the supernatants was measured by using enzyme-linked immunosorbent assay after 72 hours of culture. The amount was assessed after making the correction for the amount of the cellular protein.
‡Figures show the ratio of viable cell number in cultures with 1,024 U/mL of IFN-␣ to the number in control cultures without IFN-␣ at 96 hours of culture. ¶Degree of apoptosis was estimated as follows: the difference between fragmented DNA percentage in cells cultured with or without IFN-␣ was calculated and classified into 5 levels: Ϫ, no significant difference; ϩ, Ͻ5%; ϩϩ, 5%-10%; ϩϩϩ, 10%-20%; and ϩϩϩϩ, Ͼ20%.
§Blockage of cell cycle at S phase. **Blockage of cell cycle at G 2 /M phase. † †Blockage of cell cycle at G 1 phase. ‡ ‡Mean Ϯ SD.
highest sensitivity to IFN-␣. In regard to this point, Gutterman 18 stated that it is highly likely that IFN applied early in the stages of tumor evolution could have a very important clinical effect, whereas its activity in advanced stages in which multiple genetic aberrations are present, would be minimal. Therefore, the sensitivity of early-stage HCC to IFN-␣ could not be estimated from the sensitivity of the cell lines that possess a larger number of gene abnormalities and higher proliferation capability. Among clinical trials that administered IFN-␣ to advanced HCC patients, Lai et al. 22 used a quite higher dose of IFN-␣ than in the other studies, 23,24 and they obtained higher response rates. 22 Therefore, such a low dose of intramuscular IFN-␣ used in treatment for chronic hepatitis could not induce apoptosis or cell-cycle arrest in advanced HCC. If the route of administration is changed and a higher dose of IFN-␣ is administered to the cancer lesion, e.g., by continuous administration into tumor-feeding arteries, a growth-suppression effect would be expected. In addition, concomitant administration of IFN-␣ could magnify the antitumor effect of a drug that acts specifically on the cells in the S phase, because IFN-␣ induces accumulation of S-phase cells at a high frequency.