P27^Kip1, regulated by glycogen synthase kinase-3β, results in HMBA-induced differentiation of human gastric cancer cells

The gastric cancer cell line SGC7901 was obtained from the cell bank in the Chinese Academy of Sciences and cultured as previously described [24]. SGC7901 cells were infected with an adenovirus encoding the activated form of Akt (Ad-Akt) or the adenoviral control vector encoding β-galactosidase (β-gal) at a multiplicity of infection (MOI) of 10 pfu/cell. After infection with vectors for one hour, followed by replacement of medium and incubation for a further 24 h, cells were treated in the presence or absence of HMBA, and protein and RNA were extracted for western and Northern blotting, respectively.

HMBA, TSA, SB-415286 and LiCl were purchased from Sigma Chemical Company, USA. Adenovirus vectors encoding β-gal and the myristoylated active form of Akt (Ad-Akt) were purchased from Cell BioLabs, USA. The vector encoding the catalytically active mutant of GSK3β (HA-GSK-3βCA) was purchased from Addgene. Non-targeting control siRNA and the SMARTpool for targeting GSK-3β were purchased from Dharmacon, USA. All probes were labeled with a Biotin Random Prime DNA Labeling Kit (Pierce).

Rabbit anti-Akt, anti-phospho-Akt (Ser473) and antiphospho-GSK-3β (Ser9) were purchased from Cell Signaling, USA. Mouse anti-GSK-3β, mouse anti-p27^Kip1, mouse anti-Top IIb and mouse anti-p21 ^Waf1 were purchased from BD Biosciences, USA. Mouse anti-phospho-GSK-3 (Tyr278/Tyr216) was obtained from Upstate, USA. Rabbit anti-cleaved PARP antibody was purchased from Abcam, USA. Anti-proton pump was bought from MBL International (USA). Polyclonal anti-CDK2, anti-CDK4, anti-α-tubulin and anti-caspase-3 were obtained from Santa Cruz Biotechnology (USA).

Nuclear and cytosolic fractions were extracted using a NE-PER Nuclear and Cytoplasmic Extraction Reagents kit (Pierce, USA). Cytosolic protein (80 mg) or nuclear protein (20 mg) was resolved on a 10% polyacrylamide gel and transferred to PVDF membranes as previously described [25]. Filters were incubated for one hour at room temperature in blotting solution. Membranes were incubated overnight at 4°C with primary antibodies followed by blotting with a horseradish peroxidase-conjugated secondary antibody for one hour, and visualized using an ECL detection system.

Total RNA was prepared using TRIzol reagent (Invitrogen). Samples were run on 1.2% agarose/formaldehyde gels and transferred to supported nitrocellulose. Membranes were hybridized with a biotin labeled gastric proton pump cDNA probe. After hybridization with the GAPDH probe, a loading control, membranes were washed and signals were detected using an ECL detection system. RNase protection experiments were performed using the RPA-III kit from Ambion and the RiboQuant MultiProbe RNase Protection Assay System was utilized to detect multiple specific mRNAs. ³²P-labeled antisense RNA probes were prepared using the Human Apoptosis hCC-2 and hCYC-1 Template Sets and hybridizations were performed according to the manufacturer's protocol.

Gastric cancer cells were harvested using trypsin. Cells were collected, washed twice with ice-cold PBS and fixed in ice-cold 70% ethanol. After being washed twice with ice-cold PBS, resuspended in PBS containing 100 U/ml RNase A and incubated at 37°C for 30 min, cells were stained with PI (20 mg/ml) and analyzed using FACScan (Becton Dickinson, San Jose, CA, USA), as previously described [25].

The activities of CDK2, CDK4 and GSK-3β were measured as previously described [26, 27]. Briefly, CDK2, CDK4 or GSK-3β was immunoprecipitated from cytosolic (100 mg of protein) or nuclear (25 mg of protein) extracts. Kinase activity was measured by incubating immunoprecipitated CDK2, CDK4 or GSK-3β in 40 ml of kinase buffer with 4 mg recombinant Snail protein (to measure GSK-3β-associated kinase activity), 5 mg of histone H1 (to measure CDK2-associated kinase activity) or retinoblastoma protein (to measure CDK4-associated kinase activity) at 30°C for 30 min. The samples were processed as described in previous reports [28].

SGC7901 cells accumulated at the G0/G1 cell cycle checkpoint and differentiated into an enterocyte-like phenotype after treatment with HMBA [29]. GSK-3β contributes to the inhibition of cell cycle progression in differentiating cells [20, 30]. Therefore, whether GSK-3β plays a role in HMBA-induced SGC7901 cell cycle inhibition was investigated. As demonstrated in Figure 1a, treatment with HMBA induced cells to accumulate at the G0/G1 cell cycle checkpoint. Treatment with lithium chloride (LiCl), which inhibits GSK-3β in a Mg²⁺ competitive manner [31], increased the proportion of cells in the S phase. Treatment with a combination of LiCl and HMBA reversed HMBA-mediated G1 cell arrest. Similar results were obtained after treatment with SB-415286, a potent inhibitor of GSK-3β [32] (Additional file 1). These results suggest that GSK-3β could play a role in HMBA-induced G1 arrest. To determine whether HMBA resulted in cell death during the 24 h treatment period, protein was extracted to assess whether there was increased PARP cleavage and/or active caspase-3. As demonstrated in Figure 1b, there was no increase in PARP cleavage and active caspase-3 until 48 h after HMBA treatment. An important early event in the terminal differentiation of cells is their withdrawal from the cell cycle [6]. Since GSK-3β is documented to play a role in cell cycle arrest [33], it was postulated that inhibition of GSK-3β could inhibit differentiation. Therefore the effects of GSK-3β inhibitors on the induction of HMBA-mediated gastric proton pump expression, a marker of gastric differentiation, were examined. SGC7901 cells were pre-treated with LiCl (Figure 1c and 1d) or SB-415286 (Figure 1e and 1f) at various concentrations for one hour, and then treated with HMBA for 24 h. LiCl inhibited HMBA-induced gastric proton pump expression in a dose-dependent manner. Consistent with these results, SB-415286 blocked gastric proton pump protein and mRNA expression, which was induced by HMBA. Taken together, these results indicate that GSK-3β plays an important role in HMBA-mediated gastric cell differentiation.

GSK-3β is inactivated when it is phosphorylated downstream of Akt [34]. Therefore, it would be predicted that activation of Akt by PI3-kinase would be associated with inhibition of GSK-3β and, subsequently, inhibition of gastric cell differentiation. To test this hypothesis, SGC7901 cells were infected with Ad-Akt or a control vector. Infection with Ad-Akt increased expression of phosphorylated Akt, Akt and phosphorylated GSK-3β protein (Figure 2a), consistent with previous results demonstrating that GSK-3β acts as a substrate of Akt. As demonstrated in Figure 2b, infection of SGC7901 cells with the Ad-Akt adenoviral vector alone had no effect on gastric proton pump and mRNA expression. However, infection with the Ad-Akt vector resulted in inhibition of gastric proton pump mRNA expression induced by HMBA compared with HMBA and infection of the control (β-gal) adenovirus, suggesting that signaling through the PI3-kinase/Akt pathway regulates gastric cell differentiation induced by HMBA treatment.

To test whether GSK-3β was influenced by HMBA treatment, GSK-3β activity was determined by measuring the phosphorylation of recombinant Snail, a well-characterized substrate of GSK-3β [35, 36]. GSK-3β is located in the cytosolic and nuclear compartments of cells, but predominantly in the cytoplasm during the G1 phase. Therefore, nuclear and cytoplasmic proteins were fractionated from control and HMBA-treated cells and examined for GSK-3β activity. HMBA treatment resulted in an increase in the activity of nuclear GSK-3β (Figure 3a), and GSK-3β inhibition attenuated HMBA-mediated G1 arrest, indicating a role for GSK-3β in HMBA-induced cell cycle arrest. Ser9 phosphorylation of GSK-3β decreases GSK-3β activity, whereas Tyr216 phosphorylation increases GSK-3β activity [37]. To analyze the mechanisms underlying increased GSK-3β activity caused by HMBA treatment, Ser9-phosphorylated and Tyr216-phosphorylated GSK-3β protein expression was determined using western blotting. HMBA treatment increased nuclear expression levels of total GSK-3β and Tyr216-phosphorylated GSK-3β without affecting their expression in the cytosol (Figure 3b). Interestingly, HMBA treatment increased Ser9-phosphorylated GSK-3β protein expression in both the cytosolic and nuclear fractions. Similar results were obtained following treatment with other HPCs, SAHA and EMBA (Additional file 2). In addition, HPC increased the activity of GSK-3β in the nucleus as demonstrated by in vitro kinase assays (Additional file 3). These results suggest that HPC increases nuclear GSK-3β activity irrespective of phosphorylation at Ser9.

Progression through G1 is dependent on CDK2 and CDK4 [38, 39]. To determine whether GSK-3β regulation of HMBA-mediated G1 cell cycle arrest occurs via CDK2 or CDK4 inhibition, SGC7901 cells were pre-treated with LiCl (Figure 4a) or SB-415286 (Figure 4b) and then subjected to combination treatment with HMBA for 24 h before immunoprecipitation assays were carried out on the cell lysates using anti-CDK2 or anti-CDK4 antibodies, respectively. In addition, CDK2 or CDK4 activity was determined using an in vitro kinase assay. Treatment with HMBA alone inhibited CDK2 activity (top, left panel) but increased CDK4 activity (top, right panel); inhibition of GSK-3β using LiCl or SB-415286 significantly attenuated HMBA inhibition of CDK2 activity (bottom). Treatment with HMBA increased the expression and activity of GSK-3β in the nucleus. Taken together, these results suggest that GSK-3β contributes to HMBA-mediated G1 cell cycle arrest through inhibition of CDK2.

To examine the mechanisms underlying HMBA mediated G1 arrest and CDK2 inhibition further, cell cycle-regulatory mRNA expression was analyzed using RPA assays. Treatment with HMBA increased P27^Kip1 mRNA expression but decreased p53, p57, P15 (Figure 5A right), cyclin A and cyclin D1 mRNA expression (Figure 5A left). However, treatment with LiCl increased p21^Waf1 mRNA expression but did not affect the expression levels of other genes. Similar results were obtained when cells were treated with SB-415286 (Additional file 4). These results suggest that GSK-3β regulation of HMBA-mediated cell cycle arrest does not involve the transcriptional regulation of cell cycle-related genes.

To analyze the mechanisms underlying GSK-3β-associated cell cycle arrest further, the expression of p21^Waf1 and p27^Kip1 proteins in SGC7901 cells treated with HMBA in the presence or absence of LiCl or SB-415286 was examined. Addition of LiCl (Figure 5B right) or SB-415286 (Figure 5B left) attenuated the induction of p27^Kip1 but not p21^Waf1 protein expression, suggesting that p27^Kip1 participates in the cell cycle transitions regulated by GSK-3β. When p27^Kip1 accumulates in the nucleus it binds to CDK2, inhibiting its activity, and eventually induces cell cycle arrest. Furthermore, HMBA (10 mM) increased p27^Kip1 protein expression in the cytosol and nucleus from 0 to 24 h after treatment (Figure 6a). HMBA (0-5 mM) increased p27^Kip1 expression after 24 h in cytosolic fractions and after 48 h HMBA (5-10 mM) increased p27^Kip1 expression in nuclear fractions (Figure 6b). Addition of LiCl (Figure 6c) or SB-415286 (Figure 6d) blocked HMBA-increased p27^Kip1 nuclear expression without affecting p27^Kip1 expression in the cytosol, suggesting specific regulation of nuclear p27^Kip1 expression by GSK-3β. To demonstrate the role of GSK-3β in the regulation of nuclear p27^Kip1 expression further, cells were transfected with siRNA directed against GSK-3β (Figure 6e). RNAi-mediated suppression of GSK-3β was confirmed by immunoblotting and attenuated nuclear p27^Kip1 induction by HMBA without affecting cytosolic p27^Kip1 induction. To confirm the role of GSK-3β in the regulation of nuclear p27^Kip1 expression, SGC7901 cells were transfected with a vector encoding the activated form of GSK-3β (GSK-3β-CA) or an empty control vector. Cytosol and nuclear proteins were extracted and western blotting was performed to determine p27^Kip1 expression. Transfection of SGC7901 cells with the GSK-3βCA plasmid resulted in increased p27^Kip1 in the nuclear fraction (Figure 6F) without affecting p27^Kip1 levels in the cytosol. Over-expression of the active form of GSK-3β was confirmed using western blotting and in vitro kinase assays using Snail protein as the substrate (Figure 6G). Taken together, these results indicate that GSK-3β participates in the regulation of the cell cycle through the specific regulation of nuclear p27^Kip1 protein expression.

HMBA increased p27^Kip1 protein expression, inhibited CDK2 activity and increased CDK4 activity (Figure 4). Extracts from control or HMBA-treated cells were immunoprecipitated to investigate p27^Kip1 binding to CDK2 and CDK4. As demonstrated in Figure 7a, HMBA treatment increased the level of p27^Kip1 in the complexes immunoprecipitated with anti-CDK2 but not in complexes immunoprecipitated with anti-CDK4. Re-probing the filters with anti-CDK2 and anti-CDK4 antibodies confirmed that the immunoprecipitates from control and HMBA-treated cells contained the same levels of CDK2 and CDK4. Therefore, HMBA appears to cause a selective increase in p27^Kip1 binding to CDK2. To analyze whether inhibition of GSK-3β affects the association of p27^Kip1 with CDK2, SGC7901 cells were pre-treated with LiCl or SB-415286 and subjected to combination treatment with HMBA for 24 h; whole-cell extracts were immunoprecipitated. Treatment with GSK-3β inhibitors, LiCl (Figure 7b) or SB-415286 (Figure 7c) blocked p27^Kip1 binding to CDK2. These results suggest that GSK-3β is essential for HMBA-induced increased p27^Kip1 binding to CDK2. To confirm the role of GSK-3β in the regulation of p27^Kip1 association with CDK2, SGC7901 cells were transfected with a vector encoding the activated form of GSK-3β or Ad-β-gal. Whole-cell protein was extracted and immunoprecipitated. As presented in Figure 7d, transfection of SGC7901 cells with the GSK-3β-CA vector resulted in an increased level of p27^Kip1 in the complexes immunoprecipitated with anti-CDK2 compared with transfection of the control plasmid. This suggests that GSK-3β is not only necessary for HMBA-mediated p27^Kip1 binding to CDK2, but also sufficient to increase the association of p27^Kip1 with CDK2 in SGC7901 cells.