5-Aza-2′-deoxycytidine Activates the p53/p21Waf1/Cip1 Pathway to Inhibit Cell Proliferation*

Abstract

In addition to its demethylating function, 5-aza-2′-deoxycytidine (5-aza-CdR) also plays an important role in inducing cell cycle arrest, differentiation, and cell death. However, the mechanism by which 5-aza-CdR induces antineoplastic activity is not clear. In this study, we found that 5-aza-CdR at limited concentrations (0.01–5 μm) induces inhibition of cell proliferation as well as increased p53/p21Waf1/Cip1 expression in A549 cells (wild-type p53) but not in H1299 (p53-null) and H719 cells (p53 mutant). The p53-dependent p21Waf1/Cip1 expression induced by 5-aza-CdR was not seen in A549 cells transfected with the wild-type human papilloma virus type-16 E6 gene that induces p53 degradation. Furthermore, deletion analysis and site-directed mutagenesis of the p21 promoter reveals that 5-aza-CdR induces p21Waf1/Cip1 expression through two p53 binding sites in the p21 promoter. Finally, 5-aza-CdR-induced p21Waf1/Cip1 expression was dependent on DNA damage but not on DNA demethylation as demonstrated by comet assay and bisulfite sequencing, respectively. Our data provide useful clues for judging the therapeutic efficacy of 5-aza-CdR in the treatment of human cancer cells.

Article

As demethylating agents, 5-aza-cytidine and 5-aza-2′-deoxycytidine (5-aza-CdR)1 have been extensively used for epigenetic research (14). Both demethylating agents are incorporated into DNA where they bind DNA methyltransferase (DNMT) in an irreversible, covalent manner, thus sequestering the enzyme and preventing maintenance of the methylation state (57). Consequently, silenced genes induced by hypermethylation are re-expressed after treatment with these demethylating agents.

Originally, 5-aza-cytidine and 5-aza-CdR were developed as anticancer agents (5, 8) and have been shown to have significant cytotoxic and antineoplastic activities in many experimental tumors (912). 5-Aza-CdR is reported to be noncarcinogenic and incorporates into DNA but not RNA or protein (13, 14). 5-Aza-CdR has been found empirically to have more potent therapeutic effects than 5-aza-cytidine in cell culture and animal models of human cancers.

However, 5-aza-CdR-induced cytotoxicity may not be linked to its demethylating function (3, 1517). In addition, the therapeutic effects of 5-aza-CdR in the treatment of different human cancer cells are conflicting. 5-Aza-CdR appears to be beneficial in the treatment of human leukemias (9, 18, 19), myelodysplastic syndromes (20, 21), and hemoglobinopathies (22, 23). On the other hand, there has been less positive experience in the effectiveness of 5-aza-CdR for the treatment of human solid tumors (10, 24). Therefore, it is possible that one or more critical factors may be involved in regulating the cellular response to 5-aza-CdR treatment that vary in different cell types.

p53 is a very important tumor suppressor gene and is reported to be abnormal in more than 50% of human cancers (25). Chemotherapeutic agents frequently act through the mechanism of DNA damage, and p53 plays an important role in the induction of cell cycle arrest and apoptosis in response to DNA damage (26). 5-Aza-CdR has also shown anticancer activity that may be related to its ability to induce DNA damage (15, 27). Based on the scenario mentioned above, it is hypothesized that 5-aza-CdR may induce DNA damage, thereby activating p53, which in turn increases p21Waf1/Cip1 expression, leading to the inhibition of cell proliferation.

To confirm the role of p53 in the 5-aza-CdR-induced inhibition of cell proliferation, human lung cancer cells with different p53 status were selected as the targets for this study. As an important downstream target of p53 activation, p21Waf1/Cip1 plays a critical role in inhibiting cell proliferation; therefore, p21Waf1/Cip1 expression upon treatment of 5-aza-CdR in cells with different p53 status was given special attention in this study. Present data indicated that 5-aza-CdR is a strong DNA-damaging agent, and 5-aza-CdR induces inhibition of cell proliferation by activating the p53/p21 pathway.

EXPERIMENTAL PROCEDURES

Cells and Treatments—Human lung cancer cell lines A549, H1299, and H719 were grown in RPMI 1640 supplemented with 10% fetal bovine serum (heat-inactivated at 56 °C for 45 min) and penicillin/streptomycin, in a humidified, 5% CO2 atmosphere and 37 °C incubator. These cells were treated with 5-aza-CdR (0.01–20 μm; Sigma) for up to 72 h. Fresh medium containing 5-aza-CdR was added every 24 h. The treated cells then were washed with phosphate-buffered saline, placed in drug-free medium, and harvested at 24 h after incubation at 37 °C.

Western Blotting—Protein expression was detected by Western blotting as previously described with minor modifications (12, 16). Briefly, the cells were harvested with a scraper and then washed with cold phosphate-buffered saline once. The cells were then lysed in lysis buffer (50 mm Tris-HCl, 250 mm NaCl, 5 mm EDTA, 50 mm NaF, 0.15% Igepal CA-630, and 1.5 mm phenylmethylsulfonyl fluoride). Equal amounts of proteins (100–150 μg) were size-fractionated on 9–15% SDS-PAGE. The antibodies used are anti-p21Waf1/Cip1 (F-5; Santa Cruz; 1 μg/ml), anti-p53 (DO-1; Oncogene Research Products; 0.5 μg/ml), and α-tubulin (Oncogene Research Products; 0.3 μg/ml).

Generation of Stable Clone and Transfection—Wild-type human papilloma virus type-16 E6 gene (HPV E6) was a gift from Dr. H. Ding (Department of Radiology, Ohio State University) (28). The HPV E6 gene was inserted into pCMV-neo, and the pCMV-neo-E6 was transfected into A549 cells by using a transfection kit (Qiagen) according to the manufacturer's instructions. The stable clone of A549-pCMV-neo-E6 (A549-E6) was maintained in medium containing G418 at 500 μg/ml.

Methylation-specific PCR and Methylation Detection—DNA was extracted and then treated with bisulfite as previously described with minor modifications (29). Briefly, genomic DNA (1 μg) in a volume of 50 μl was denatured by NaOH (final concentration, 0.275 m) for 10 min at 42 °C. The denatured DNA was then treated with 10 μl of 10 mm hydroquinone and 520 μlof3 m sodium bisulfite at 50 °C overnight. The primers for p21 were as follows: forward primer, 5′-GGG AGG AGG GAA GTG TTT TT-3′, and reverse primer, 5′-ACA ACT ACT CAC ACC TCA ACT-3′. The PCR conditions were initiated with a denaturing step of 95 °C for 10 min, followed by 36 cycles of 96 °C for 30 s, 53 °C for 20 s, and 72 °C for 20 s and were concluded with 72 °C for 7 min. The PCR products were purified with a purification kit (Qiaquick) and then incubated with HhaI (50 °C) for 2 h and TaqI at 65 °C for 2 h, respectively. Digested DNA was then size-fractionated via polyacrylamide gel electrophoresis to detect the methylation status.

Bisulfite Sequencing—DNA was treated with bisulfite and purified for PCR as described above. The PCR products were gel-extracted (Qiagen) and ligated into a plasmid vector, pCR2.1-TOPO, using the TA cloning system (Invitrogen). Plasmid-transformed bacteria TOP10 F′ were cultured overnight, and the plasmid DNA was isolated (Qiagen). At least 10 separate clones were chosen for sequence analysis.

Transient Transfection and Measurement of Relative Luciferase Activity—Vectors used for transfection include pWWP-Luc and pWP101-Luc (30, 31). The human wild-type p21 promoter luciferase fusion plasmid, pWWP-Luc, was made from a 2.4-kb genomic fragment of p21 promoter containing the transcriptional start site and two p53 binding sites and then subcloned into the luciferase reporter vector, pGL-3Ba-sic. pWP101-Luc does not contain any p53 binding site. A549 cells transfected with the pWWP-Luc or the pWP101-Luc were treated with 5-aza-CdR (0.5–5 μm, 24 h) and then harvested to analyze the relative luciferase activity.

Site-directed Mutagenesis—Mutant p21 promoter constructs were generated using a site-directed mutagenesis kit (QuikChange; Stratagene, La Jolla, CA). pWWP-Luc (containing two p53 binding sites) was used as the mutagenesis template. The p53 recognition elements consist of four tandem PuPuPuC (A/T) pentamers (32). The first p53 binding site, GAACA (33) (-2234 to -2230 relative to the translational start site) was replaced with GAAAC, and the second binding site, AGACT (33) (-1344 to -1340 relative to the translational start site) was replaced with AGAAT following the manufacturer's directions.

Comet Assay for Detecting DNA Strand Breaks—The comet assay, also called the single-cell gel electrophoresis, was performed as described previously (34). In brief, fully frosted microscopic slides were covered with 110 μl of 0.5% normal melting agarose at 60 °C. The slides were immediately covered with a coverslip and then kept at 4 °C for 15 min to allow the agarose to solidify. About 105 cells of 5-aza-CdR treated or untreated cells in 40 μl of phosphate-buffered saline were mixed with equal amount (40 μl) of 1% lower melting agarose to form a cell suspension. After gently removing the coverslip, the cell suspension was pipetted onto the first agarose layer, spread using a coverslip, and maintained at 4 °C for 15 min to allow it to solidify. After removal of the coverslips, the slides were immersed in fresh prepared cold lysing solution (2.5 m NaCl, 100 mm Na2EDTA, 10 mm Tris, pH 10.0, 1% sodium sarcosinate) with 1% Triton X-100 for 40 min at 4 °C. The slides were then placed in a horizontal gel electrophoresis tank filled with fresh electrophoresis solution (1 mm Na2EDTA, 300 mm NaOH, pH 13.0) for 10 min. The slides were then placed in Tris buffer (0.4 m Tris, pH 7.5) for 15 min twice (neutralizing the excess alkali) after electrophoresis at 4 °C. The slides were then stained with 75 μl of propidium iodide (5 μg/ml) for 30 min.

The slides were examined at 600× magnification, and the pictures were taken under a fluorescence microscope (TCS; Leica, Mannheim, Germany). To score the percentage of DNA in the tail, the image analysis system was used (Q550CW; Leica). The percentage of comet tail area (DNA tail area/total DNA area) and the comet tail length (from the center of the DNA head to the end of the DNA tail) were analyzed in 50 cells for one slide.

RESULTS

Dose-dependent Inhibition of Cell Proliferation by 5-Aza-CdR—In this study, human lung cancer cell lines A549 (wild-type p53), H1299 (p53-null), and H719 (mutant p53) were treated with 5-aza-CdR at different concentrations for 72 h (Fig. 1A) and with different intervals of 5-aza-CdR treatment at 1 μm, respectively (Fig. 1B). The cell viability was determined by MTT assay. A dose- and duration-dependent inhibition of cell proliferation was observed only in A549 cells but not in H1299 and H719 cells (Fig. 1, A and B). As shown in Fig. 1A, for example, cell viability was decreased to 77% of the untreated control when A549 cells were treated with 5-aza-CdR even at very low dose (0.078 μm) and 51% or 39% of the untreated control when cells were treated with 5-aza-CdR at 1.25 or 5 μm, respectively. However, 5-aza-CdR was unable to induce an obvious decrease in cell viability in H1299 cells or H719 cells until the concentrations were well above 1 μm (Fig. 1, A and B). Cell viability was decreased in H1299 and H719 cells only when they were treated at very high concentrations (10–20 μm).

Inhibition of cell proliferation is a reflection of cell cycle arrest that is mainly controlled by proteins from the INK4 family and the CIP/KIP family of cyclin-dependent kinase inhibitors (35). A549 cells have been reported to have a homozygous deletion of CDKN2a (36). Therefore, we hypothesized that p21Waf1/Cip1 expression may be a critical factor for the 5-aza-CdR-induced inhibition of cell proliferation in A549 cells. Fig. 1C indicates that increased p21Waf1/Cip1 expression is a dose-dependent effect of 5-aza-CdR treatment in A549 cells but not in H1299 cells. The increased expression of p21Waf1/Cip1 was observed in the A549 cells even when treated at a very low concentration (0.01 μm) of 5-aza-CdR. However, 5-aza-CdR even at a very high concentration (10 μm) was unable to induce increased expression of p21Waf1/Cip1 in H1299 cells (Fig. 1C). Because p53 is a mutant in H719 cells (30), increased p21Waf1/Cip1 expression upon 5-aza-CdR treatment was similarly not observed in H719 cells in this study (data not shown).

5-Aza-CdR-induced Inhibition of Cell Proliferation Is Dependent on p53/p21Waf1/Cip1 Pathway—To investigate whether 5-aza-CdR-induced p21Waf1/Cip1 expression is through activation of p53, a wild-type HPV E6 gene was inserted into pCMV-neo. After transfection and selection with G418, the pCMV-neo-E6 stable clone (A549-E6) was established. For testing the functionality of the stable clone, the A549-E6 cells were irradiated with γ-rays, and then p53 and p21Waf1/Cip1 levels were determined by Western immunoblot analysis. As shown in Fig. 2A, the p53 level in A549-E6 cells was much lower than that in A549 cells after exposure to 2 or 8 grays of γ-rays. Similarly, p21Waf1/Cip1 levels in the A549 and the A549-E6 cells after irradiation showed the same trend as the p53 expression (Fig. 2B). These data suggested that the expression of HPV E6 in the A549-E6 clone is sufficient to degrade the p53 level in the transfected A549 cells. Next, A549 and A549-E6 cells were treated with 5-aza-CdR at different concentrations for 72 h, and then MTT assay was performed to test cell viability in a fashion analogous to the experiments demonstrated in Fig. 1A. As shown in Fig. 2C, cell viability in the A549-E6 was much higher than that in the A549 cells after treatment with 5-aza-CdR at different concentrations. These results support the hypothesis that cell viability after treatment with 5-aza-CdR is related to wild-type p53 in the A549 cells. To investigate the reasons for the difference between A549 and A549-E6 cells, p53 and p21Waf1/Cip1 expression was evaluated by Western immunoblot (Fig. 2, D–E). Clearly, both p53 levels as well as p21Waf1/Cip1 levels were significantly increased in A549 cells after treatment with 5-aza-CdR (Fig. 2, D–E). This was not seen in the A549-E6 cells, even when they were treated with 5-aza-CdR at higher concentrations (up to 10 μm; Fig. 2E). Although p21Waf1/Cip1 expression is slightly increased in the A549-E6 cells upon treatment with 5-aza-CdR, the effect is very attenuated when compared with the parental A549 cells (Fig. 2E).

For further confirmation of the critical role of p53 in the 5-aza-CdR-induced p21Waf1/Cip1 expression, a full-length p21 promoter construct (pWWP-Luc) that included two p53 binding elements and a truncated p21 promoter (pWP101-Luc) that lacked any p53 binding site (Fig. 3A) were transiently transfected into A549 cells, and relative luciferase activity (RLU) was evaluated after treatment with 5-aza-CdR in both transfectants. As shown in Fig. 3B, the RLU in the pWWP-Luc-transfected A549 cells was much higher than that in the pWP101-Luc-transfected A549 cells after treatment with 5-aza-CdR. The 5-aza-CdR-induced increase in the RLU was dose-dependent. The RLU in pWWP-transfected cells, for example, increased 3-fold when treated with 5-aza-CdR at 5 μm compared with the same transfectants treated at 0.5 μm (Fig. 3B). The RLU, however, was not significantly increased in the pWP101-transfected A549 cells after 5-aza-CdR treatments (Fig. 3B). Because there is no endogenous p53 in H1299 cells, the RLU was not increased in the pWWP-transfected H1299 cells after treatment with 5-aza-CdR (Fig. 3C).

To determine which p53 binding site of the p21 promoter was important for the 5-aza-CdR treatment, pWWP-Luc containing the mutated first p53 binding site (Mut-1) and second p53 binding site (Mut-2) were constructed (Fig. 3A) and then transfected into A549 cells that were subsequently treated with 5-aza-CdR. As shown in Fig. 3D, the RLUs decreased from 3.4 to 1.42 and 1.51, respectively, after 5-aza-CdR treatment when the Mut-1 or Mut-2 transfectants were compared with wild-type pWWP-transfected cells (Fig. 3D). No difference in the RLU between the Mut-1 and Mut-2 transfected cells was observed after 5-aza-CdR treatment (Fig. 3D). The data suggest that both p53 binding sites in the p21 promoter are important in inducing p21Waf1/Cip1 expression in response to 5-aza-CdR treatment.

5-Aza-CdR Activates the p53/p21 Pathway through DNA Damage—Because 5-aza-CdR is a DNA methyltransferase inhibitor, it was necessary to rule out the possibility that the p21 promoter is fully or partially methylated in A549 cells. To detect the methylation status of the p21 promoter, we performed combined bisulfite restriction analysis and bisulfite sequencing of the p21 promoter from A549 and H1299 cells. The combined bisulfite restriction analysis confirmed that the CpG island of the p21 promoter (-233 to +2; Fig. 3A) was unmethylated (Fig. 4). Bisulfite sequencing analysis confirmed these findings (Table I). Ten separate clones were selected for the sequencing. The promoter regions of the p21 gene in A549 and H1299 cells were shown to be almost totally unmethylated in the 24 CGs analyzed by bisulfite sequencing (1 of 240 CGs in A549 cells and 3 of 240 CGs in H1299 cells are methylated, respectively; Table I).

To investigate whether 5-aza-CdR plays a direct role in damaging DNA, comet assay was performed as well. A549 cells were treated with 5-aza-CdR at 0.1, 1, and 5 μm for 72 h and then harvested for this assay. As shown in Fig. 5, dose-dependent DNA damage was observed after 5-aza-CdR treatment. Compared with the untreated control (Fig. 5A), 5-aza-CdR even at a low concentration (0.1 μm) induced DNA damage, as indicated by the presence of a DNA tail (Fig. 5B). Greater DNA tail area and longer DNA tail length (a distance from DNA head to the end of DNA tail) show more extensive DNA damage. These 5-aza-CdR-induced DNA-damaging features are observed more frequently in the cells treated with a higher concentration of 5-aza-CdR than that with a lower concentration of 5-aza-CdR (Fig. 5, compare C and D with A). The quantitative data of 5-aza-CdR-induced DNA damage as determined by comet assay are shown in Table II. For example, the percentage of DNA tail area relative to the total area and the DNA tail length are increased 2.6- and 1.4-fold, respectively, when cells were treated with 5-aza-CdR at 0.1 μm compared with that in the untreated cells (Table II). Similarly, in the cells treated with 5-aza-CdR at 5 μm, the percentage and the length are increased 4.2- and 3.8-fold, respectively (Table II). These results suggest that 5-aza-CdR inhibits cell proliferation by damaging DNA, which causes activation of the p53/p21Waf1/Cip1 pathway.

DISCUSSION

Increased expression of p21Waf1/Cip1 after inhibition of DNA methyltransferase has been reported by several investigators (3739). To date, at least two separate mechanisms explain this effect. The first mechanism involves a demethylating function. 5-Aza-CdR, for example, was reported to bind to DNMT and inactivate the enzyme (40), inducing a re-expression of p21Waf1/Cip1 in cells that are hypermethylated in the promoter of the p21 gene (30, 41, 42). A second mechanism for enhanced p21Waf1/Cip1 expression is independent of DNA methylation. For instance, a significant increase in p21Waf1/Cip1 expression was observed in human cancer T24 cells following treatment with DNMT antisense oligonucleotides (43) or DNMT antagonist (37) and in normal human fibroblasts with 5-aza-CdR (44), but the promoter region of p21 in these cells is totally unmethylated. These data indicate that inhibition of DNMT itself, unrelated to methylation status, may activate p21Waf1/Cip1 expression. Consistent with these reports, in the present study, the 5-aza-CdR-induced p21Waf1/Cip1 expression in A549 cells is not associated with DNA methylation because the promoter region of p21 is almost completely unmethylated (Fig. 4 and Table I).

As one of the principle downstream effectors of p53, p21Waf1/Cip1 is a cyclin-dependent kinase inhibitory protein and plays a role in preventing cyclin E/Cdk2 and cyclin A/Cdk2 kinase from promoting cell cycle progression (45). The ability of p53 to induce cell cycle arrest in the G1 phase in response to DNA damage is largely dependent on p21Waf1/Cip1 expression (26). Therefore, the p53/p21Waf1/Cip1 pathway confers the damaged cells enough time to repair DNA and ensure that the cell cycle progresses correctly. There is conflicting data as to whether expression of p21Waf1/Cip1 in cells after inhibition of DNMT is dependent on p53 changes. 5-Aza-CdR induced an elevated p53 level in human colon cancer cells, but p21Waf1/Cip1 expression is not completely dependent on p53 status. In their experimental system, Karpf et al. (27) showed that p21Waf1/Cip1 expression induced by 5-aza-CdR was not only observed in human colon tumor cells HCT (p53+/+) but also seen in HCT (p53-/-) cells, although the degree of elevated p21Waf1/Cip1 is different in both cell lines. Milutinovic also reported that DNMT antagonists or antisense oligonucleotides can induce rapid expression of p21Waf1/Cip1 in A549 cells, but p53 levels are not changed (37). In this case, the DNMT antagonist or antisense oligonucleotides are not incorporated into DNA; therefore direct DNA damage with subsequent enhanced p53 expression does not occur. However, in the present study, p21Waf1/Cip1 expression in A549 cells after 5-aza-CdR treatment in limited doses is totally dependent on p53 expression (Fig. 2, D–E). This p53-dependent p21Waf1/Cip1 expression is demonstrated by the following evidence: 1) Elevated p21Waf1/Cip1 expression after 5-aza-CdR treatment occurred only in A549 cells (with wild-type p53) but not in H1299 cells (p53-null) (Fig. 1C); 2) 5-aza-CdR-induced p21Waf1/Cip1 expression in A549 cells is much decreased when p53 levels were diminished by transfection with an HPV E6 gene that promotes p53 degradation (Fig. 2, D–E); and 3) deletion and mutation analysis showed increased relative luciferase activities after 5-aza-CdR treatment in cells with transfected full-length p21 promoter (with intact p53 binding sites) but not in those with a truncated p21 promoter (without p53 binding sites) or full-length p21 promoter (with mutant p53 sites) (Fig. 3, B and D). Therefore, it is reasonable to hypothesize that the antineoplastic effect of 5-aza-CdR may activate p53 and subsequently induce an increase in p21Waf1/Cip1 expression through which cell proliferation is inhibited.

Increases in p53/p21Waf1/Cip1 levels in mammalian cells are often observed when cells are exposed to DNA-damaging agents including irradiation (46, 47), UV light (48, 49), and chemicals (50). Upon exposure to these DNA-damaging elements, p53 is activated, taking part in post-translational modifications including phosphorylation and acetylation (26, 51, 52). Previously, 5-aza-CdR was reported to be a potential DNA-damaging agent in embryonic cells, and its cytotoxicity was related to DNMT itself and not the secondary demethylation of genomic DNA (15). 5-Aza-CdR was incorporated into DNA, and then the spontaneous degradation of the incorporated analog may result in DNA damage (11, 53). It has also been shown that 5-aza-CdR, when incorporated into DNA, covalently links with DNMT, an event that could cause DNA damage because of the structural instability at its incorporation sites (54) or by obstructing DNA synthesis (15, 55). Consistent with these reports, our data for the first time demonstrates directly that 5-aza-CdR is a DNA-damaging agent when assayed by comet assay (Fig. 5 and Table II).

Interestingly, growth inhibition and cytotoxicity induced by 5-aza-CdR at higher doses (>20 μm) are not dependent on p53 status because the cells tested are mostly dead after 5-aza-CdR treatment (data not shown). This p53-dependent inhibition of cell proliferation by a limited dose of 5-aza-CdR treatment has a great potential advantage in the treatment of different human cancers. The major reported side effect of 5-aza-CdR when pushed to the maximum tolerated dose is hematopoietic toxicity (10). A more limited dose of 5-aza-CdR treatment may decrease these toxic effects on normal tissue and cells. In addition, the limited dose of 5-aza-CdR required to induce effects in wild-type p53 cells may explain why different human cancer cells appear to have different sensitivities to 5-aza-CdR treatment. p53 status may be a critical factor for judging the antineoplastic effects of 5-aza-CdR in cancer cells. Although there is insufficient data to show a relationship between p53 status and efficacy of 5-aza-CdR in patients, several studies suggest that p53 status may be a key predictive factor for the efficacy of 5-aza-CdR treatments (10, 24, 56). For example, 5-aza-CdR appears to be very effective in the treatment of adult human chronic myelogenous leukemias, which generally have wild-type p53 (56, 57) but is less demonstrably effective in the treatment of solid tumors, such as human lung cancers, which typically have mutant p53 (24, 58). Our data suggest that the future clinical development of 5-aza-CdR may depend on genetic factors such as the p53 status of the treated tumor.

Acknowledgments

We thank Drs. T. Sakai and H. Ding for providing us with the vectors used in this study. We also thank Dr. Y. Shang and Quanhui Zhen, Shan Chen, and Zhe Li for technical help. We appreciate Dr. X. B. Yao (University of Science and Technology of China) for support and encouragement during the course of this study.

Footnotes

References