Bmi1 Functions as an Oncogene Independent of Ink4A/Arf Repression in Hepatic Carcinogenesis

Bmi1 is a polycomb group proto-oncogene that has been implicated in multiple tumor types. However, its role in hepatocellular carcinoma (HCC) development has not been well studied. In this article, we report that Bmi1 is overexpressed in human HCC samples. When Bmi1 expression is knocked down in human HCC cell lines, it significantly inhibits cell proliferation and perturbs cell cycle regulation. To investigate the role of Bmi1 in promoting liver cancer development in vivo, we stably expressed Bmi1 and/or an activated form of Ras (RasV12) in mouse liver. We found that while Bmi1 or RasV12 alone is not sufficient to promote liver cancer development, coexpression of Bmi1 and RasV12 promotes HCC formation in mice. Tumors induced by Bmi1/RasV12 resemble human HCC by deregulation of genes involved in cell proliferation, apoptosis, and angiogenesis. Intriguingly, we found no evidence that Bmi1 regulates Ink4A/Arf expression in both in vitro and in vivo systems of liver tumor development. In summary, our study shows that Bmi1 can cooperate with other oncogenic signals to promote hepatic carcinogenesis in vivo. Yet Bmi1 functions independent of Ink4A/Arf repression in liver cancer development. (Mol Cancer Res 2009;7(12):1937–45)


Introduction
Bmi1, a member of the mammalian polycomb group of multimeric transcriptional repressors, is involved in the regulation of development, stem cell self-renewal, cell cycle, and senescence (1). Bmi1 was first identified as a c-myc cooperating oncogene in murine B-cell lymphomas (2). Subsequent studies have revealed that Bmi1 is required by both normal and leukemic hematopoietic stem cells to maintain their proliferative capacity (3,4). In addition, Bmi1 has been shown to be important for self-renewal of neural stem cells (5), and its expression is essential for the tumorigenicity of MycN-induced neuroblastoma (6). Studies have found that Bmi1 induces telomerase activity and subsequently immortalizes mammary epithelial cells (7). Perhaps the most prominent link between Bmi1 and tumor development is its inhibition of the Ink4A/Arf locus, which results in the regulation of cell senescence and proliferation (8,9).
Deregulation of Bmi1 expression has been reported in multiple tumor types, including non-small cell lung carcinoma, colon carcinoma, medulloblastoma, metastatic melanoma, and nasopharyngeal carcinoma (10)(11)(12)(13)(14). Upregulation of Bmi1 in human hepatocellular carcinoma (HCC) has also been reported (15,16). In a recent study, Chiba et al. (17) showed that silencing Bmi1 expression decreased the side population (SP) cells in HCC cell lines. These SP subpopulation cells are considered to harbor cancer stem cell like properties (18). However, the exact role of Bmi1 during HCC pathogenesis remains unclear. There are currently no in vivo models, which show that Bmi1 functions as an oncogene and directly contributes to HCC pathogenesis.
In this article, we describe that Bmi1 is overexpressed in human HCC samples. Bmi1 expression is also required for HCC cell proliferation in vitro. Notably, we established a novel mouse model for Bmi1 and show that Bmi1 can cooperate with activated Ras signaling to promote hepatic carcinogenesis in vivo. However, expression analysis suggests that Bmi1 functions independent of its ability to repress Ink4A/Arf tumor suppressor genes. Our data therefore provide solid evidence for a functional role of Bmi1 in liver cancer pathogenesis.

Bmi1 Is Overexpressed in Human HCC Samples
In our previous studies, we used genomic approaches, including cDNA microarray and array-based comparative genomic hybridization, to characterize molecular variations in human HCC (19)(20)(21). We identified 703 genes, which are highly expressed in human HCC (21). One of these upregulated genes is Bmi1. From this microarray study, Bmi1 expression is upregulated in human HCC compared with nontumor liver tissues (P = 2 × 10 −6 , after Bonferoni correction; Fig. 1A).
To verify this observation, we performed real-time reverse transcription-PCR (RT-PCR) analysis for Bmi1 expression in an independent liver tumor sample set that have not been previously assayed in microarray studies. Again, we observed upregulation of Bmi1 in HCC samples (P < 0.001; Fig. 1B). In two recent studies, the overexpression of Bmi1 in human HCC samples was shown at protein levels (15,16).
Because p16Ink4A and p14Arf have been considered to be major targets of Bmi1 during tumor development, we investigated whether there is any correlation between Bmi1 and Ink4A/Arf expression in human HCC. On the cDNA microarrays, there was one probe corresponding to the COOH-terminal sequences of Ink4A/Arf, which hybridized to both p16Ink4A and p14Arf. Our analysis of this microarray data found no correlation between Bmi1 and total Ink4A/Arf expression (R = −0.094; Fig. 1C). We next assayed the expression of Bmi1, p16Ink4A, and p14Arf individually using real-time RT-PCR in 19 human HCC and 4 nontumor liver tissues. Again, we found no correlation between the expression values of Bmi1 and p16Ink4A or p14Arf ( Fig. 1D and E).
Altogether, these data show that Bmi1 is upregulated in human HCC, suggesting that Bmi1 may play a role in HCC pathogenesis. However, Bmi1 expression does not seem to be correlated with the expression of Ink4A/Arf tumor suppressor genes in human HCC samples.

Stable shRNA-Mediated Knockdown of Bmi1 Inhibits Cancer Cell Growth In vitro
Our expression analysis suggests a potential role for Bmi1 during liver tumorigenesis. We therefore decided to study the functional significance of Bmi1 in hepatocarcinogenesis. We found that Bmi1 protein is highly expressed in human HCC cell lines (data not shown). To investigate whether Bmi1 is required during liver cancer development, we stably knocked down its expression using lentiviral shRNA in human HCC cell lines. To better study the relationship of Bmi1 expression and genetic alternations in human HCCs, we chose three HCC cell lines (SK-Hep1, Huh7, and Hep3B) with different genetic variations in α-fetoprotein, p53, p16Ink4A, and p14Arf (Supplementary  Table S1).
We found that despite different genetic backgrounds and Ink4A/Arf status, silencing of Bmi1 inhibits growth of all three HCC cell lines (Supplementary Fig. S1; Fig. 3A). Furthermore, the expression of cell cycle genes, such as Cdc2, Cdc20, and Bub1 (22), are significantly downregulated in Bmi1/pLKO.1infected HCC cells (Supplementary Fig. S2; Fig. 3B). In addi-tion, bromodeoxyuridine (BrdUrd) labeling revealed a decreased proliferative rate (Supplementary Fig. S3; Fig. 3C), whereas activated caspase 3 assay showed a slight increase in apoptosis when Bmi1 expression is silenced (data not shown). Finally, cell cycle analysis suggests that loss of Bmi1 perturbs cell cycle regulation and leads to G 2 -M accumulation (Supplementary Table S2 and Fig. S4; Table 1; Fig. 3D). There is also an increase of sub-G 1 phase cells in Bmi1/pLKO.1-infected cells, providing further support of increased cell apoptosis when Bmi1 expression is inhibited ( Supplementary Fig. S4 and Table S2; Fig. 3D).
In summary, the studies support that Bmi1 expression is required for in vitro growth of human HCC cell lines.
Loss of Bmi1 Does Not Lead to Significant Increased Expression of p16Ink4A or p14Arf in HCC Cell Lines One of the major mechanisms of Bmi1-induced tumor development is its function as a potent inhibitor of CDKN2A that encodes two major proteins: p16Ink4A and p14Arf (p19Arf in mice; refs. 8, 23). We first determined whether Bmi1 knockdown affects the mRNA expression of CDKN2A genes using real-time RT-PCR. SK-Hep1 cells have a deletion of Ink4A/ Arf locus, whereas Huh7 cells have strong promoter methylation of p16Ink4A. Therefore, p16Ink4A expression is virtually undetectable in these two cell lines, regardless of whether Bmi1 is downregulated ( Fig. 2A). In addition, the loss of Bmi1 expression in transfected Hep3B cells does not seem to affect the expression of p16Ink4A. Likewise, we found that silencing Bmi1 expression does not lead to upregulation of p14Arf in Huh7 and Hep3B cells, whereas p14Arf expression is absent in SK-Hep1 cells ( Fig. 2A). We next assayed p16Ink4A protein expression in these HCC cell lines (Fig. 2B). Consistent with real-time RT-PCR results, p16Ink4A is undetectable in SK-Hep1 and Huh7 cells, whereas there is little change of p16Ink4A protein levels in Bmi1/pLKO.1-infected cell Hep3B cells (Fig. 2B).
In summary, our data showed that Bmi1 is required for HCC cell proliferation; however, the effect of Bmi1 in promoting HCC cell growth is independent of Ink4A/Arf status.

Overexpression of Bmi1 Cooperates with Activated Ras to Induce HCC in Mice
We next determined whether Bmi1 can function as an oncogene by establishing a mouse model. We reasoned that it is unlikely that Bmi1 alone is sufficient to induce liver cancer formation in vivo. Therefore, we searched for other signaling pathways that may be able to cooperate with Bmi1 to promote hepatic carcinogenesis. We chose activated Ras as the second signal, based on studies that have shown that Bmi1 is capable of cooperating with activated Ras to transform cells in vitro (24,25). In addition, Ras/mitogen-activated protein kinase (MAPK) signaling is activated in all human HCC samples (26). Therefore, it represents a critical genetic alteration present in human HCC. Furthermore, studies from our and other laboratories have found that activated Ras alone is not sufficient to induce HCC formation in mice (27,28).
We applied hydrodynamic transfection to stably express Bmi1 (with COOH-terminal V5 tag) and/or an activated form of N-ras (RasV12) into mouse hepatocytes. These animals were then monitored and sacrificed at specific time points or when moribund. We found that whereas overexpression of RasV12 (n = 15) or Bmi1 (n = 5) alone was not sufficient to promote liver tumor development, the coexpression of Bmi1 and RasV12 induced liver tumors in 78.6% (11 of 14) of the mice between 15 and 30 weeks postinjection (Fig. 4A). Tumors tend to be multifocal, sometimes with over 100 tumor nodules scattered around the entire liver (data not shown; Fig. 4B).
Histologic examination of liver tumor samples induced by Bmi1/RasV12 showed that tumors consisted of neoplastic cells with frequent trabecular disorganization, which are characteristic of HCC (Fig. 4D). In most cases, the tumor cells appear to be well differentiated. Real-time RT-PCR analysis revealed high expression of HCC-specific marker α-fetoprotein ( Fig. 4C), further confirming the tumors to be of hepatocellular origin.  We next examined the tumor nodules for expression of injected Bmi1 (with a COOH-terminal V5 tag) and RasV12. Using anti-V5 antibody, we observed that all tumor cells showed positive nuclear staining of Bmi1 (Fig. 4D). Sporadic expression of Bmi1 was also detected in the hepatocytes of surrounding nontumor liver. RasV12 is indicated by elevated protein levels (Fig. 4E). Because activated Ras is a potent inducer of MAPK signaling, we investigated the activity of RasV12 by assaying for the presence of phospho-ERK. Both Western blot and immunohistochemical analyses detected strong expression of phospho-ERK in the tumors ( Fig. 4D and E).
Altogether, these results support that Bmi1 and RasV12 can cooperate to induce HCC in vivo.

Molecular Characterization of Bmi1/RasV12 Induced HCC
We then investigated the molecular features of Bmi1/ RasV12-induced tumors to determine whether these traits resemble phenotypes observed in human HCC. We first assayed for cell proliferation in Bmi1/RasV12 tumor samples. Our detection of proliferative marker, Ki67, suggested the tumor cells to be highly proliferative (Fig. 5A). We also observed increased expression of cyclin E1 in liver tumor samples, whereas there is little variability in the expression of cell cycle regulator, cyclin D1 (Fig. 5B). In addition, we found that these tumors exhibited elevated levels of cell cycle inhibitor p21 (Fig. 5B), which is likely to be a feedback response to the activated Ras signaling. Furthermore, antiapoptotic marker survivin and cell-cell adhesion marker E-cadherin were also found to be upregulated in liver tumor samples (Fig. 5B). The upregulation of E-cadherin is consistent with welldifferentiated tumor histology and is frequently observed in certain mouse models of HCC (29,30). The occurrence of angiogenesis during liver carcinogenesis can be distinguished by the expression of endothelial markers, like PODXL1 (31). Although PODXL1 is not typically expressed by normal liver sinusoidal endothelial cells, this marker is frequently present in the endothelial cells of liver tumors (31). Therefore, we analyzed our samples for PODXL1 and observed that only endothelial cells within tumor nodules stained positive for this marker (Fig. 5A). Furthermore, these HCC samples also highly expressed angiogenic factor Ang2 (Fig. 5B).
Overall, our data suggests that Bmi1/RasV12-expressing tumors resemble a subset of human HCC characterized by the deregulation of factors involved in proliferation, apoptosis, and angiogenesis.  Upregulation of p16Ink4A/p19Arf Expression in Bmi1/ RasV12 Induced Liver Tumors Bmi1 has been shown to cooperate with RasV12 to transform murine embryonic fibroblast cells via inhibition of Ink4A/Arf locus (24). We therefore investigated whether this regulation is a mechanism by which Bmi1 and activated Ras promote tumorigenesis in vivo. We examined the expressions of p16Ink4A and p19Arf in our samples by quantitative RT-PCR. We have used multiple primers against p16Ink4A and p19ARF, and we found that in all cases, both p16Ink4A and p19ARF can only be detected after >30 cycles of PCR, indicating that p16Ink4A and p19ARF are expressed at very low levels in normal liver tissues. In contrast, p16Ink4A expression is upregulated ∼5-fold and p19Arf expression is upregulated ∼50-fold in Bmi1/RasV12 tumor samples (Fig. 6).

Regulation of Ink4A/Arf Expression by RasV12 and Bmi1 in Mouse Hepatocytes
The upregulation of p16Ink4A and p19Arf in Bmi1/RasV12 tumor samples is quite surprising because Bmi1 is a known inhibitor of Ink4A/Arf locus. One of the possibilities is that Ras is a potent inducer of p16Ink4A and p19Arf. The upregulation of Ink4A/Arf may be why activated Ras alone is not sufficient to induce HCC formation in vivo. It is possible that the partial inhibition of Ras-induced Ink4A/Arf expression by Bmi1 is what eventually leads to hepatic carcinogenesis. If this is the case, it is likely that Bmi1/RasV12 tumor cells may have somewhat elevated expression of p16Ink4A and p19Arf compared with normal liver. However, this hypothesis is only possible if RasV12 can strongly induce p16Ink4A and p19Arf expression in hepatocytes. We therefore investigated  First, we generated adenovirus encoding activated Ras. Adenoviral infection has been shown to be able to transfect 100% of mouse hepatocytes at multiplicity of infection of 10 (32). We infected mouse primary hepatocytes with either control adenovirus encoding EGFP (AD-EGFP), or activated Ras (AD-RasV12-HA). Western blot analysis showed the expression of RasV12 in infected cells (Fig. 7A). Using quantitative RT-PCR, we found that p16Ink4A expression is repressed, whereas p19ARF expression remains unchanged in primary mouse hepatocytes after AD-RasV12-HA infection (Fig. 7A). Next, we transfected primary mouse hepatocytes with plasmids encoding activated N-Ras (RasV12-HA) and/or Bmi1 (Bmi1-V5) using Targefect-Hepatocyte transfection reagents, which have a transfection efficiency of 50% for primary hepatocytes. The expression of RasV12 and Bmi1 are indicated by Western blot analysis (Fig. 7B). We found that although the expression of p16Ink4A is downregulated by RasV12 transfection, there is very little change in the expression of p19Arf. Bmi1 upregulates p16Ink4A expression and inhibits p19Arf expression in this condition (Fig. 7B). Cotransfection with Bmi1 and RasV12 showed similar impact on p16Ink4A while having little effect on p19Arf (Fig. 7B). Next, we assayed p16Ink4A protein expression in these mouse hepatocytes. We found that p16Ink4A protein is undetectable in primary mouse hepatocytes in all these conditions (Fig. 7A).  Ras has been shown to induce cell senescence in certain cell types (33,34), but the induction of cell senescence in hepatocytes by Ras has not been reported. It is possible that Bmi1 cooperates with RasV12 to promote HCC pathogenesis by overcoming Ras-mediated induction of senescence. We therefore investigated whether RasV12 or Bmi1 regulates cell senescence in primary hepatocytes. We transfected primary mouse hepatocytes with EGFP, RasV12, Bmi1, or RasV12 and Bmi1, and assayed for cell senescence. We found no evidence that either RasV12 or Bmi1 can modulate senescent status in primary hepatocytes ( Supplementary Fig. S5 Table S3).
In summary, our data do not support activation of Ink4A/Arf by Ras or inhibition of Ink4A/Arf by Bmi1 overexpression in hepatocytes. The experiments therefore indicate that Bmi1 cooperates with RasV12 to promote HCC pathogenesis in an Ink4A/Arf-independent manner.

Discussion
There is increasing evidence supporting Bmi1 as an important oncogene in tumor development. Upregulation of Bmi1 expression has been reported in multiple tumor types. Studies also showed that Bmi1 expression is required for in vitro cell proliferation in Ewing Sarcoma, lung cancer, and medulloblastoma cells (35)(36)(37), whereas overexpression of Bmi1 enhances cell survival in epidermis (38) and prostate cancer cells (39). Using Bmi1 knockout mice, studies showed that Bmi1 expression is required for the tumorigenesis of leukemia and lung cancer in vivo (4,40). However, there is still little evidence whether Bmi1 overexpression can directly contribute to carcinogenesis, especially in solid tumors. Proper mouse models need to be established to address this critical question. In our current study, we showed that Bmi1 is overexpressed in human HCC samples and required for HCC cell growth in vitro. More importantly, we established a novel mouse model that shows that Bmi1 can cooperate with activated Ras to promote HCC pathogenesis in mice. Our study therefore provides pivotal data supporting Bmi1 as an oncogene and its role in hepatic carcinogenesis.
In this study, we used activated Ras to mimic the activation of Ras/MAPK pathway and in combination with Bmi1 to induce HCC in our mouse model. Although there is ubiquitous activation of Ras/MAPK signaling in human HCC, Ras mutations are, in fact, very rare (41,42). Deregulation of other factors, including tumor suppressor genes Spry2 and RASSF1, overexpression of H-Ras, as well as overactivation of receptor tyrosine kinases, such as epidermal growth factor receptor and c-Met, have been implicated in human HCC, all of which result in upregulation of this pathway (26). Therefore, combination of these genetic alternations with Bmi1 overexpression in mouse models will provide additional in vivo models to mimic human HCC pathogenesis. Interestingly, we assayed the expressions of H-Ras and RASSF1A in human HCC samples, and found there to be no correlation between their expressions and Bmi1 expression (Supplementary Fig. S6). Clearly, future experiments will be needed to determine whether the expression of other factors involved in the activation of Ras/MAPK pathways, such as c-Met or epidermal growth factor receptor, are associated with upregulation of Bmi1 expression during HCC pathogenesis.
An important implication of our study is that the tumorigenicity of Bmi1 during HCC pathogenesis is independent of its ability to repress Ink4A/Arf expression. First, we showed there is no correlation between the expressions of Bmi1 and p16Ink4A or p14Arf in human HCC samples. Second, we found that the downregulation of Bmi1 inhibits HCC cell growth independent of Ink4A/Arf status. Finally, we showed that in mouse tumor cells induced by Bmi1/RasV12, there is no downregulation of p16Ink4A or p19Arf expression. Consistent with our observation, several recently published reports have revealed an Ink4A/Arf-independent role for Bmi1 during tumor pathogenesis. Bruggeman et al. (43), for instance, showed that Bmi1 controls mouse glioma development in an Ink4a/ Arf-independent manner. Bmi1 knockdown significantly inhibits cell growth in both wild-type and p16Ink4A null Ewing Sarcoma (35), and medulloblastoma cell lines (37). Thus, although inhibition of Ink4A/Arf tumor suppressor gene expression has been widely considered to be the key mechanism of the oncogenic activity of Bmi1, more recent data suggest a critical role of an Ink4A/Arf-independent mechanism for Bmi1 during carcinogenesis.
Clearly, the next step in the characterization of molecular mechanisms of Bmi1 is to identify novel targets and/or pathways regulated by Bmi1 during HCC pathogenesis, and investigate how they cooperate with activated Ras/MAPK signaling to induce liver cancer formation. Some potential targets of Bmi1 have been identified in human cancer cell lines. For example, hTert is thought to be a major target in Bmi1-induced immortalization of mammary epithelial cells (7). NID1, a gene related to cell adhesion, has been implicated as a Bmi1 target in Ewing Sarcoma cells (35). Signaling molecules including BMP5, transforming growth factor β2, and Notch2 have been found to be regulated by Bmi1 in medulloblastoma cell lines (37). It would be of interest to determine whether the expression of these factors is modulated by Bmi1 during HCC pathogenesis. A recent study indicated that the loss of Bmi1 results in the increase of reactive oxygen species and subsequent stimulation of the DNA damage response pathway (44). Activation of the DNA damage response pathway has been found to be an important barrier to tumorigenesis. In our recent studies, we observed upregulation in the mRNA of p53 and ATM genes in Bmi1/RasV12-induced liver tumor samples. 8 Clearly, it would be important to further characterize the expression of these genes in tumor samples at protein levels. Analysis of the regulation of this pathway by Bmi1 in liver may identify an additional function for Bmi1 during the development of HCC.

Human Tissue Samples and RNA Preparation
Samples of tumor and nontumor liver tissues were collected from liver resections at The University of Hong Kong. Tissues were frozen in liquid nitrogen within 0.5 hour after they were resected. Total RNA was extracted using Trizol (Invitrogen). This study was approved by the Ethics Committee of the University of Hong Kong and the Internal Review Boards from University of California at San Francisco.
Cell Culture, Lentiviral Infection, Cell Proliferation, BrdUrd Labeling, Caspase-3 Activity, and Cell Cycle Assays All human HCC cell lines were purchased from American Type Culture Collection except Huh7, which were kindly provided by Dr. Ben Yen of University of California at San Francisco, San Francisco, CA. The cells are cultured in DMEM plus 10% fetal bovine serum. Lentivirus was generated and used to infect HCC cells. Three days postinfection, cells were expanded and selected with 1 μg/mL puromycin for 3 d and harvested for protein or RNA analysis. To assay cell proliferation rate, equal number of cells were seeded in six-well plates and counted 3 to 4 d postseeding. Cell cycle analysis was done by flow cytometry after propidium iodide staining, and the results were analyzed using FlowJo. BrdUrd labeling was as described (46) and Caspase-3 activity was measured using Caspase-Glo3/7 Assay kit (Promega).

Hepatocyte Isolation, Transfection, Adenovirus Infection, and Cell Senescence Assay
Primary hepatocyte isolation was done using standard collagenase perfusion method as described (47). The hepatocytes were transfected with plasmids using Targefect-Hepatocyte reagents (Targeting Systems) per manufacturer's instruction. Ad-H-RasV12 was kindly provided by Dr. Judy Meinkoth of the University of Pennsylvania, Philadelphia, PA (48). Adenovirus was amplified and titered by Vector Biolabs and used to infect primary hepatocytes at 50 multiplicity of infection. Hepatocytes were harvested 30 h posttransfection or infection. Hepatocyte senescence was determined using senescence β-galactosidase staining kit (Cell Signaling Technology).

Mouse Hydrodynamic Transfection and Monitoring
Wild-type FVB/N mice were used in this study. The hydrodynamic transfection procedure are as described previously (45). The injected mice were monitored weekly and sacrificed between 14 to 30 wk postinjection. All mice were housed, fed, and treated in accordance with protocols approved by the committee for animal research at the University of California at San Francisco.

Histology and Immunohistochemistry
Animals were euthanized and their livers were removed and rinsed in PBS. Samples collected from the livers were either frozen in dry ice for RNA and protein extraction or fixed overnight in freshly prepared cold 4% paraformaldehyde. Fixed tissue samples were embedded in paraffin. Five-micron sections were placed on slides and stained with H&E. Immunohistochemistry was done as described (28). Antibodies and dilutions were as follows: anti-V5, 1:1000 (Invitrogen); antiphospho-ERK, 1:100 (Cell Signaling Technology); anti-PODXL1, 1:200 (Applied Genomics); and anti-Ki67, 1:150 (Lab vision).

Real-time RT-PCR
Total RNA was extracted from frozen liver tissues using Trizol (Invitrogen) and digested with DNase I to remove any genomic DNA contamination. Sybergreen-based real-time RT-PCR was carried out as described (21), and rRNA was used as internal control. Transcript quantification was done in triplicate for every sample and reported relative to rRNA. The primer pairs used are listed in Supplementary Table S4.

Statistical Analysis
The Pearson's correlation coefficient (R) was used to determine the correlations between gene expression values, and P value was determined using SPSS statistical program. Student's t test was used to evaluate statistical significance among experimental groups. Values of P < 0.05 were considered to be significant.

Disclosure of Potential Conflicts of Interest
No potential conflicts of interest are disclosed.