TIS21 negatively regulates hepatocarcinogenesis by disruption of cyclin B1–Forkhead box M1 regulation loop

A functional and biochemical interaction of TIS21/BTG2/PC3 with Forkhead box M1 (FoxM1), essential transcription factor for hepatocyte regeneration and a master regulator of mitotic gene expression, was explored. Growth of hepatocellular carcinoma (HCC), developed by a single injection of diethylnitrosamine (DEN), was the same in both the TIS21+/+ and TIS21−/− mice until 6 months, whereas it was significantly higher in the TIS21−/− mice at 9 months. Expression of TIS21 was significantly lower in both human and murine HCCs than in the surrounding tissues. Forced expression of TIS21 impaired growth, proliferation, and tumorigenic potential of Huh7 cells. At the mechanistic level, TIS21 inhibited FoxM1 phosphorylation, a required modification for its activation, by reducing cyclin B1–cdk1 activity, examined by in vitro kinase assay and FoxM1 mutant analyses. These observations were further confirmed in vivo by the reciprocal control of TIS21 expression and FoxM1 phosphorylation in the diethylnitrosamine‐induced HCCs and TIS21−/− mouse embryonic fibroblast (MEF), in addition to increased expression of cyclin B1 and cdk1 activity. Conclusion: TIS21 negatively regulated hepatocarcinogenesis in part by disruption of the FoxM1–cyclin B1 regulatory loop, thereby inhibiting proliferation of transformed cells developed in mouse and human livers. (HEPATOLOGY 2008.)

F irst isolated from SW3T3 cells treated with 12-Otetradecanoylphorbol-13-acetate, 1 TIS21 (12-Otetradecanoylphorbol-13-acetate-inducible sequences 21) belongs to early growth response genes 2 and the antiproliferative gene family, 3 along with BTG1, 4 BTG3, 5 and Tob 6 genes. TIS21 /BTG2/PC3 represents orthologs of mouse, human, and rat, 7 respectively. TIS21 down-regulates expression of cyclin E and cdk4 independent of pRB 8 ; however, PC3 inhibits G 1 -S progression by inhibiting cyclin D1 expression dependent on pRB. 9 Genetic ablation of PC3 prevents G 2 arrest in ES cells p53-dependently, 10 and prolonged activation of BTG2 by 12-Otetradecanoylphorbol-13-acetate induces G 2 /M arrest of U937 cells p53-independently. 11 Therefore, we suggested TIS21 /BTG2/PC3 as a pan-cdk inhibitor. 12 There are only a few reports on the potential role of TIS21 /BTG2/PC3 as a tumor suppressor: The first report was on the thymic carcinoma developed in SV40TAg-transgenic mice and human cancer cells, such as NCIH69, U937, and A549 cells. 13 All of the counterpart normal tissues, thymus, lymphoid, macrophage, and lung are constitutive TIS21 expressers. 14,15 The importance of BTG2 /TIS21/PC3 in cancer has been shown in prostate carcinogenesis, 16 clear cell renal cell carcinoma and its cancer cell lines, 17 and medulloblastoma. 18 These findings support TIS21 /BTG2/PC3 as a tumor suppressor.
Forkhead box M1 (FoxM1) is a transcription factor known to be essential for normal hepatocyte regeneration and a master regulator of mitotic gene expression, expressed in all replicating cells. [19][20][21] Phosphorylation and recruitment of p300/CBP coactivators to FoxM1 is required for FoxM1 activation. 22 FoxM1 phosphorylation is initiated by cyclin-cdk complexes in early G 1 and continues to G 2 and M phases of the cell cycle. 22,23 Based on the putative consensus phosphorylation sites, cdk1, cdk2, and mitogen-activated protein kinase have been suggested to be involved in FoxM1 phosphorylation, 24 and cyclin E/cdk2, cyclin A/cdk2, and cyclin A/cdk1 complexes induce FoxM1 phosphorylation and increase transcriptional activity. 25 Based on the report 20 that cyclin B1 and cdc25B are target genes of p-FoxM1 and hepatocyte mitosis is associated with expression of cdc25B phosphatase and cyclin B1 accumulation, increased cyclin B1 may generate more cyclin B1-cdk1 complex and further activates FoxM1 in a positive-feedback loop, thus resulting in hepatocyte regeneration.
We already reported that TIS21 inhibited cdk1 activity and worked as a pan-cdk inhibitor in cancer cells. 11,12 In the current study, therefore, we attempted to investigate a role of TIS21 in hepatocarcinogenesis, focusing on the regulation of FoxM1 activity. Proliferation of hepatocellular carcinoma (HCC) was increased in the TIS21 Ϫ/Ϫ mice, whereas forced expression of TIS21 significantly reduced cell growth and in vitro tumorigenicity of Huh7 cells, as well as G 2 /M arrest of the cell cycle. Furthermore, expression of cyclin B1 and FoxM1 phosphorylation were markedly increased in mouse HCCs and TIS21 Ϫ/Ϫ mouse embryonic fibroblast (MEF). We propose here that TIS21 negatively regulates hepatocarcinogenesis via disruption of the cyclin B1-FoxM1 regulatory loop, resulting in inhibition of cyclin B1 transcription.

Induction of HCC and Preparation of MEF.
Diethylnitrosamine (DEN) (Sigma Chemical Co.; 20 mg/kg body weight) was injected intraperitoneally once to 12-day-old wild-type and TIS21 Ϫ/Ϫ mice. 26 The mice were sacrificed at 3, 6, and 9 months after the injection. Saline was injected as the control treatment. TIS21 ϩ/ϩ MEF and TIS21 Ϫ/Ϫ MEF were prepared from 13.5 dayold embryos of the wild-type and TIS21 Ϫ/Ϫ mice and then cultured in Dulbecco's modified Eagle's medium with 10% fetal bovine serum in a 37°C incubator with 5% CO 2 in air.
Tumor Samples From Patients. During the period of 2003 to 2005, HCCs and surrounding control tissues were obtained from 15 patients at Ajou University Hospital after surgical resection with informed consent. The hard and firm tumor tissues were trimmed free of normal tissue and snap frozen in liquid nitrogen immediately after resection according to the specimen regulation of Ajou University Hospital. No patient in the current study received chemotherapy or radiation therapy before the surgery. Preparation of Cell Lines. Huh7 cells were transfected with TIS21 in pcDNA3-HA using Lipofectamine (Invitrogen, San Diego, CA), and selected with 975 g/mL Geneticin 418 (Gibco BRL, Bethesda, MD) for 3 weeks. Adenoviruses of TIS21 (Ad-TIS21) and FoxM1B (Ad-FoxM1) were prepared by transfection of complementary DNAs (cDNAs) of TIS21 and FoxM1B inserted into replication-defective E1-adenoviral vectors and E3adenoviral vectors, and then amplified in 293 cells. Adenovirus of bacterial ␤-galactosidase (Ad-LacZ) was prepared for the control experiment.
Clonogenic and Soft Agar Colony-Forming Assays. Huh7-TIS21 and Huh7-V cells were prepared by transfection of Huh7 cells with TIS21 and pcDNA3-HA, respectively. Cells (4 ϫ 10 2 ) were seeded in a 35-mm plate and maintained for 9 days. Clones containing more than 50 cells were counted as positive after staining with crystal violet. For soft agar colony-forming assay, the cells (5 ϫ 10 3 ) were plated in 0.6% agarose solution and then layered on 1.0% agarose bed. Colonies larger than 125 m were counted as positive in 2 weeks.
Cyclin-Associated Kinase Assay. Employing anti-cyclin B1 and anti-cyclin A antibodies, cdk1, and cdk2 immunocomplexes were prepared from Huh7 cells infected with either Ad-TIS21 or Ad-LacZ viruses. The cyclinassociated kinase activity was measured by incubating with histone H1 (5.0 g) as a substrate and [␥-32 P]adenosine triphosphate (ATP) (2.5 Ci/reaction) for 30 minutes at 30°C, and then terminated by adding 6 ϫ sodium dodecyl sulfate sample buffer. Phosphorylation of the substrate was examined by autoradiography after sodium dodecyl sulfate polyacrylamide gel electrophoresis. To further confirm inhibition of cdk1 activity by TIS21 in vitro, rTIS21 protein (5.0-10.0 g) was added to the reaction mixture, and the kinase activity was assayed.
To investigate whether TIS21 can inhibit FoxM1 phosphorylation in vitro, immunoprecipitation (IP)-kinase assays were performed: Cell lysates of Huh7, infected with Ad-FoxM1B virus (100 moi) for 2 days, were immunoprecipitated overnight with anti-FoxM1 antibody, and the precipitate was used as a substrate, while using cyclin B1 immunocomplexes as a kinase source. Kinase assay was carried out in the presence of glutathione S-transferase TIS21 fusion protein (rTIS21) or glutathione S-transferase (GST) (0-10 g) for 30 minutes at 30°C.

FoxM1 Mutants by Site-Directed Mutagenesis.
Wild-type FoxM1 (wtFoxM1) cDNA was isolated from Huh7 cells, cloned into pCMV-Flag vector (Sigma), and used for FoxM1 mutants construction (Supplementary  Table 1). FoxM1⌬C, C-terminal deletion mutant, was prepared by PCR with ⌬C-sense and ⌬C-antisense primers with wtFoxM1 7cDNA as template. To prepare N-terminal deletion mutant of FoxM1 (FoxM1⌬N), ⌬N-sense and ⌬N-antisense primers were employed with the wtFoxM1 template. For FoxM1 single mutant (FoxM1⌬N-S 507 A, FoxM1⌬N-S 657 A, FoxM1⌬N-T 585 A, and FoxM1⌬N-T 596 A) preparation, the ⌬N-sense S 507 A and ⌬N-antisense primers, ⌬N-sense and megaprimer 1 primers, ⌬N-sense and megaprimer 2 primers, and ⌬N-sense and megaprimer 3 primers were used, respectively, with the wtFoxM1 template. FoxM1 double mutants (FoxM1⌬N-S 507, 657 A and FoxM1⌬N-T 585, 596 A) were amplified with ⌬N-S 507 A and ⌬N-antisense primers with the FoxM1⌬N-S 657 A template, and ⌬N-sense and megaprimer 3 primers with the FoxM1⌬N-T 585 A template, respectively. FoxM1 triple mutants (FoxM1⌬N-T 585,596 A,S 657 A; FoxM1⌬N-S 507,657 A,T 596 A; FoxM1⌬N-S 507,657 A,T 585 A and FoxM1⌬N-S 507 A,T 585,596 A) were amplified with ⌬N-sense and megaprimer 1 primers the with FoxM1⌬N-T 585,596 A template, ⌬N-sense S 507 A and megaprimer 1 primers with the FoxM1⌬N-T 596 A template, ⌬N-sense S 507 A and megaprimer 4 primers with the wtFoxM1 cDNA template, and ⌬N-sense S 507 A and ⌬N-antisense primers with the FoxM1⌬N-T 585, 596 A template, respectively.
Cyclin B1 Promoter Analysis. Reporter construct containing cyclin B1 promoter (Ϫ950 base pairs) was generated in basic pGL3 vector (Promega, Madison, WI), and transfected to Huh7 cells using lipofectamine. Thymidine kinase promoter-driven Renilla luciferase plasmid (Promega) was employed as control DNA. Transfection activity was measured by TD 20/20 luminometer (Turner BioSystems, Sunnyvale, CA) according to the instructions for the Dual-Luciferase Reporter Assay System (Promega). All transfection experiments and luciferase assays were carried out in triplicate and repeated more than twice.

Rapid Growth of HCC in Livers of TIS21 ؊/؊ Mice.
No spontaneous tumor developed in either TIS21 ϩ/ϩ or TIS21 Ϫ/Ϫ mice until 9 months by saline treatment; however, single injection of DEN developed HCCs in both wild-type and null mice (Fig. 1A). Severity of the HCCs in TIS21 ϩ/ϩ mice was less than that of the TIS21 Ϫ/Ϫ  mice. The number and size of tumors increased markedly in the TIS21 Ϫ/Ϫ mice compared with the TIS21 ϩ/ϩ ( Table 1).

Loss of TIS21 Expression in Liver Tumors of Mouse and Human.
When TIS21 expression in liver was measured by RT-PCR, the level was variable in the salinetreated mice, whereas it was much less in every case of HCC obtained at 9 months after a single injection of DEN (Fig. 1B). It was approximately 30% of the control level, examined by real time PCR (Fig. 1C). Expression of BTG2 /TIS21 was found in 1 of 15 human HCC cases, as opposed to 7 of 15 surrounding tissues (Fig. 1D). These data indicate that TIS21 /BTG2 expression is significantly reduced in tumor compared with its surrounding tissue and suggest possible involvement of TIS21 in hepatocarcinogenesis.
Inhibition of Tumorigenesis of Huh7 Cells by TIS21. When the role of TIS21 in HCC cells was investigated, TIS21 messenger RNA (mRNA) and protein expressions were found in Huh7-TIS21 cells, but not in Huh7-V, by RT-PCR and immunoblot analyses ( Fig.   2A), and growth of Huh7-TIS21 was significantly lower than that of Huh7-V (Fig. 2B). In contrast, however, growth of TIS21 Ϫ/Ϫ MEF was much higher than that of TIS21 ϩ/ϩ MEF, suggesting regulation of tumor cell growth by the TIS21 gene. To evaluate the effect of TIS21 on tumorigenicity, clonogenic and soft agar colony-forming abilities were examined; Huh7-TIS21 revealed one third the clonogenicity (Fig. 2C) and soft agar colony-forming abilities (Fig. 2D) of the Huh7-V. These data strongly suggest that TIS21 might negatively work in hepatocarcinogenesis.

Regulation of Cyclin B1 Expression by TIS21.
Based on the reduced cell growth and in vitro tumorigenicity of Huh7-TIS21 cells, changes of cyclin proteins were evaluated in the TIS21 expressers by immunoblot analyses; mRNA and protein expression of cyclin B1, but not cyclin A and cyclin D1, were markedly reduced in Huh7-TIS21 cells, whereas pRb and FoxM1 expressions were slightly decreased (Fig. 3A). In contrast, expressions of cyclin A and cyclin B1 were higher in TIS21 Ϫ/Ϫ MEF than in TIS21 ϩ/ϩ (Fig. 3B), further confirming a role of TIS21 in regulation of cyclin B1 expression. To test a possible effect of stable transfection of TIS21 on the reduced cyclin B1 expression, we performed Ad-TIS21 virus infection and found reduced protein and mRNA levels of cyclin B1 and G 2 /M arrest of the cells (Fig.  3C,D). These data support our hypothesis that expression of TIS21 inhibits cell growth and tumorigenicity of HCC cells by inhibiting cyclin B1 expression.
Inhibition of cdk1 Activity In Vivo and In Vitro by TIS21. Based on our earlier report that TIS21 binds to cdk1, 11 interaction of TIS21(HA) with cyclin B1 was examined. Thus, Huh7 cell lysates infected with either Ad-LacZ or Ad-TIS21 viruses for 2 days were subjected to IP and immunoblot analyses using anti-cyclin B1 and anti-HA antibodies. As expected, cyclin B1 was bound to TIS21(HA), and vice versa (Fig. 4A). To evaluate regulation of cyclin-associated kinase activity by TIS21, a kinase assay was performed with [␥-32 P]ATP and histone H1 after IP with either anti-cyclin B1 or anti-cyclin A antibodies. Phosphorylation of histone H1 by cdk1 in the Ad-TIS21-infected cells was found to be inhibited by 70% (Fig. 4B), whereas cyclin A-associated activity was inhibited by only 10%. To prove whether TIS21 directly inhibited the activity, in vitro kinase assay and autoradiography were performed in the presence of rTIS21 protein (0-10 g); approximately 40% of cdk1 activity was inhibited by adding rTIS21 protein, but not GST (Fig. 4C).
The results indicate that TIS21 inhibits in vivo and in vitro cdk1 activity via direct interaction with cyclin B1-cdk1.

Fig. 2. Reduced tumorigenicity of Huh7 cells after induction of TIS21 expression. (A) RT-PCR and immunoblot analyses of
Huh7-V and Huh7-TIS21 cells, revealing the absence and presence of TIS21 expression, respectively. (B) Significant reduction of Huh7-TIS21 growth over that of Huh7-V cells. Conversely, growth of TIS21 Ϫ/Ϫ MEF was significantly higher than that of TIS21 ϩ/ϩ MEF. The same experiment was repeated 3 times and is presented as mean Ϯ standard deviation. Note significant inhibition of tumorigenicity of Huh7-TIS21 compared with Huh7-V cells, evaluated by in vitro clonogenicity (C) and soft agar colony-forming assay (D).

Inhibition of FoxM1 Phosphorylation and Its
Transcriptional Activity by TIS21. Based on reports that cdk1 and cdk2 are involved in FoxM1 phosphorylation and increase its transactivation activity, 22,23 the possibility of regulating FoxM1 phosphorylation by TIS21 was investigated. FoxM1 was isolated by IP of Huh7-TIS21 and Huh7-V cell lysates, and the degree of p-FoxM1 was measured by immunoblot analyses. FoxM1 phosphorylation was significantly reduced in Huh7-TIS21 compared with Huh7-V (Fig. 5A). To further investigate whether TIS21 regulates FoxM1 phosphorylation, in vitro kinase assay was performed with IP of cyclin B1associated kinase and FoxM1. FoxM1 phosphorylation was concentration-dependently inhibited in the presence of rTIS21, indicating direct inhibition of the phosphorylation by TIS21 (Fig. 5B). To examine the phosphorylation residues in FoxM1, deletion mutant analyses were extensively performed using IP and immunoblot analyses with FoxM1, Flag, p-Ser, and p-Thr antibodies (Fig. 5C): FoxM1 phosphorylation on Ser and Thr residues was inhibited by Ad-TIS21 infection (the 1 st row); however, it was lost in both the Ad-LacZ-infected and Ad-TIS21infected cells after transfection of C-terminal deletion mutant of FoxM1 (FoxM1⌬C, the second row), suggesting that the phosphorylated residues are on the C-termi-nus. When N-terminal deletion mutant (FoxM1⌬N) was expressed, p-Ser and p-Thr were detected in the Ad-LacZ-infected but not in Ad-TIS21-infected cells (the third row), suggesting inhibition of phosphorylation by TIS21. When the Ser and Thr residues were singly mutated to Ala (S 507 A, S 657 A, T 585 A, and T 596 A), phosphorylation was same as that of FoxM1⌬N; however, double (S 507,657 A, T 585,596 A) and triple mutants (T 585,596 A,S 657 A; S 507,657 A,T 596 A; S 507,657 A,T 585 A; S 507 A,T 585,596 A) revealed loss of phosphorylation in both the Ad-LacZ and Ad-TIS21 cells, suggesting the phosphorylation on the two Ser and the two Thr residues (the fourth to thirteenth rows). Mutant analyses indicate that TIS21 inhibits phosphorylation of the two Ser and two Thr residues of FoxM1. When cDNA of FoxM1B was transfected to Huh7 cells, promoter activity of cyclin B1 increased 2.7 times from that of the vector-transfected cells. However, when TIS21 cDNA was cotransfected together with FoxM1B, the promoter activity was significantly reduced, as compared with the FoxM1B alone-treated group (Fig.  5D), which suggests that TIS21 down-regulates the transcriptional activity of FoxM1B.
To investigate a possibility of direct interaction of TIS21 with FoxM1, IP with anti-FoxM1 and anti-HA antibodies, and immunoblot with anti-HA and anti- FoxM1 antibodies, respectively, were performed. The results confirmed that TIS21 and FoxM1 do not directly interact with each other (Supplementary Fig. 1)

In Vivo Phosphorylation of FoxM1 in DEN-Treated Mice and TIS21 ؊/؊ MEF.
To examine whether expression of cyclin B1 and cdk1 activities were changed in TIS21 Ϫ/Ϫ hepatocytes, immunoblot analyses were performed with liver homogenates of TSI21 ϩ/ϩ and TIS21 Ϫ/Ϫ mice, and significant increases in cyclin B1 expression (Fig. 6A) and cdk1 activity (Fig. 6B) were found in the TIS21 Ϫ/Ϫ mice compared with the wildtype.
To evaluate the activity of FoxM1 in normal liver and HCC, we measured p-FoxM1 in mouse liver treated with either DEN or saline. Expression of p-FoxM1 was low in TIS21 ϩ/ϩ mice, but increased twice after DEN treatment (Fig. 6C). The p-FoxM1 was already increased in TIS21 Ϫ/Ϫ mice regardless of DEN injection. Not only in liver but also in MEF, expressions of cyclin B1, p-FoxM1, and cdk1 activity were significantly higher in TIS21 Ϫ/Ϫ than TIS21 ϩ/ϩ , evidenced by immunoblot, IP-immunoblot analyses, and autoradiography (Fig. 6D). Expression of FoxM1 was the same; however, p-Ser in FoxM1 was significantly higher in TIS21 Ϫ/Ϫ MEF than in wild-type. These data indicate negative regulation of FoxM1 phosphorylation in vivo by TIS21.

Discussion
HCC is the 3 rd most common cancer in Korean men. 27 However, liver transplantation remains the only viable treatment for HCC. 28 Even though cancer is not a disease regulated by a single gene expression, malignancy of HCC is significantly reduced by regulating TIS21 expression; TIS21 inhibits cdk1 activity via interaction with the cyclin B1/cdk1 complex, which in turn inhibits FoxM1 phosphorylation and transcription of cyclin B1 (Fig. 7). TIS21 forms a negative regulation loop with FoxM1 and cyclin B1. Indeed, TIS21 expression was significantly reduced in and cyclin B1 to TIS21-HA. Huh7 cells were infected with Ad-LacZ and Ad-TIS21 viruses, and the cell lysates were immunoprecipitated with anti-cyclin B1 or anti-HA antibodies and then immunoblotted with anti-HA or anti-cyclin B1 antibodies, respectively. Note mutual interaction only in the Ad-TSI21-infected cells. (B) In vivo inhibition of cyclin B1-associated kinase activity by TIS21. Huh7 cells infected with LacZ or TIS21 adenoviruses were precipitated with anti-cyclin B1 and anti-cyclin A antibodies, and the immunoprecipitates were subjected to in vitro kinase assay with histone H1 and [␥-32 P]ATP as substrates. For IP, Ad-TIS21 (1.0 mg) and Ad-LacZ (500 g) cell lysates were used. Panels from top to bottom show immunoblot of the immunoprecipitates, Coomassie blue stain of histone H1, and autoradiography with relative activities. Note significant decrease of cyclin B1-associated, but not cyclin A-associated, kinase activity in the Ad-TIS21-infected cells. (C) In vitro inhibition of cyclin B1-associated kinase activity by rTIS21. IP-kinase assay was performed with or without rTIS21 protein, described in (B). Inhibition of cyclin B1-associated kinase activity was over 40% by GST-TIS21, but not by GST alone.
human and mouse HCC compared with the surrounding tissue. Thus, induction of TIS21 appears to be an attractive alternative to handle HCC by genetic manipulation.
TIS21 did not bind to FoxM1 (Supplementary Fig. 1), but rather indirectly regulated FoxM1 activity by inhibiting cyclin B1-cdk1 activity. Nevertheless, size and number of HCCs were not significantly different between FoxM1⌬NS 507 A,T 585,596 A), were prepared, in addition to FoxM1⌬C, deletion of C-terminal target residues. The FoxM1 mutants and wild-type constructs were transfected to Huh7 cells for 12 hours, and then adenovirus (LacZ and TIS21) infection was performed. The cell lysates were subjected to IP-immunoblot analyses 3 days after transfection. FoxM1 constructs could be phosphorylated; however, Ad-TIS21 infection significantly inhibited FoxM1 phosphorylation on any of the S 507 , T 585 , T 596 , and S 657 residues. The data confirmed that TIS21 inhibited FoxM1 phosphorylation at all the proposed sites. (D) TIS21 significantly inhibited transcriptional activity of FoxM1B. When 3 different DNAs were transfected to Huh7 cells, FoxM1B increased cyclin B1 promoter activity by 2.7 times that of the promoter alone. However, cotransfection of TIS21 inhibited FoxM1B-increased promoter activity by 1.7 times (P Ͻ 0.01). Data show mean Ϯ standard deviation from 6 independent experiments. TIS21 ϩ/ϩ and TIS21 Ϫ/Ϫ mice until 6 months after a single injection of DEN. This is in accordance with an earlier report 29 and can be explained by a previous report 30 ; kinetics between doses of DEN and HCC development show a linear response, indicating only 1 hit, whereas time-response requires 4 hits to develop HCC, regardless of the dose used. Therefore, it takes more than 6 months to develop HCC after a single injection of DEN in both To investigate the effect of TIS21 on cyclin B1 expression, we performed immunoblot analyses with liver homogenates of TIS21 ϩ/ϩ and TIS21 Ϫ/Ϫ mice and found significantly increased expression of cyclin B1 in the TIS21 Ϫ/Ϫ mice. (B) Activity of cdk1 was higher in the TIS21 Ϫ/Ϫ hepatocytes than TIS21 ϩ/ϩ , examined by IP-kinase assay and autoradiography. (C) Immunoblot analyses revealed an in vivo difference of FoxM1 phosphorylation in the HCCs of TIS21 ϩ/ϩ and TIS21 Ϫ/Ϫ mice. Note the more p-FoxM1 in mice with DEN injection than in the saline group in both the TIS21 ϩ/ϩ and TIS21 Ϫ/Ϫ mice. 1 and 2 represent individual variations. However, p-FoxM1 expression in TIS21 Ϫ/Ϫ livers was already higher even in the saline-treated mice. The right panels showed relative expressions obtained from 8 mice by densitometer. (D) Increases of cyclin B1 expression and cdk1 activity were confirmed by RT-PCR, immunoblot, and IP-kinase assay in the TIS21 Ϫ/Ϫ MEF compared with wild-type. IP-immunoblot analysis revealed the presence of p-FoxM1 in the TIS21 Ϫ/Ϫ MEF, but not in TIS21 ϩ/ϩ . TIS21 ϩ/ϩ and TIS21 Ϫ/Ϫ mice. However, a lack of the TIS21 gene may provide a growth advantage to the transformed cells, thus forming more HCCs in the TIS21 Ϫ/Ϫ mice at 9 months. A possible mechanism of the reduced expression of TIS21 in HCC may be selection of the clone transformed by DEN injection. Moreover, transfection of TIS21 to Huh7 cells significantly reduced in vitro tumorigenicity and tumor cell growth. These findings strongly suggest in vivo and in vitro effects of TIS21 on the negative regulation of HCC. In addition to TIS21, lack of the antiproliferative gene Tob also induces spontaneous tumors in liver, lung, and lymph nodes, 31 supporting a tumor suppressor role of TIS21.
We propose that TIS21 works as an important regulator of HCC growth in mouse and human by inhibiting FoxM1 phosphorylation by cyclin B1/cdk1. The hypothesis is well supported by other reports that FoxM1 is an essential transcription factor for liver regeneration 20,32 and that cyclin-mediated/cdk-mediated phosphorylation of FoxM1 is required for transactivation potential of FoxM1 22,24 and tight regulation of cyclin B1 expression by FoxM1. 33 A close interaction between TIS21 binding to cyclin B1, FoxM1 phosphorylation, and transcription of cyclin B1 is suggested here based on the following evidence: When Fig. 3A is compared with Fig. 3C, cyclin B1 expression was significantly lower in the Huh7-TIS21 (TIS21-stable expresser) than in the Ad-TIS21 cells (TIS21-transient expresser). This might be attributable to a transient effect of adenovirus in Ad-TIS21 cells, as opposed to continuous selection of Huh7-TIS21 cells with Geneticin 418. Moreover, it was increased in the TIS21 Ϫ/Ϫ MEF compared with TIS21 ϩ/ϩ (Fig. 3B).
The observation that TIS21 partially inhibited FoxM1-induced transcriptional activity can be supported by an earlier study on the FoxM1 phosphorylation at Ser and Thr residues to maintain transcriptional activity of FoxM1. [22][23][24][25] Because synthesis and degradation of cyclin B1 are essential for progression of mitosis, 34,35 indirect regulation of FoxM1 activity by TIS21 appears to be a very effective and clever way to inhibit carcinogenesis.
Based on the reports that FoxM1 protein is highly expressed, especially in all replicating cells and cancer tissues, 19,20,36 and that sustained expression of TIS21 induces G 2 /M arrest by its binding to cdk1, 11 inhibition of FoxM1 phosphorylation by TIS21 could be another function of TIS21 as a cell cycle inhibitor. Epidermal growth factor-induced cell death starts with TIS21 phosphorylation and Pin-1 binding in U937 cells under enforced expression of TIS21, which results in mitochondrial cell death. 12 Therefore, epidermal growth factor-induced growth inhibition of various cancer cells, such as A431, HN6 37 , breast cancer, 38 and U937 cells, 39 could be explained. Here, we propose that loss of TIS21 expression in mouse liver by a single injection of DEN may be 1 of the critical hits. 29 In conclusion, TIS21, a pan-cell cycle regulator, inhibits FoxM1 activation by binding to the cyclin B1-cdk1 complex, thereby reducing cell growth, proliferation, and clonogenicity of transformed cells containing one critical hit by DEN injection. Fig. 7. Reciprocal regulations of TIS21, cyclin B1/cdk1 activity, FoxM1 activation, and transcription of cyclin B1 in Huh7 cancer cells, mouse livers, and MEF cells. TIS21 inhibits cyclin B1-cdk1 activity by interaction with cdk1, which in turn reduces FoxM1 phosphorylation. p-FoxM1 induces transcription of cyclin B1; therefore, TIS21 negatively regulates the FoxM1-cyclin B1 loop by inhibiting the cyclin B1-cdk1 complex.