PRMT5 competitively binds to CDK4 to promote G1-S transition upon glucose induction in hepatocellular carcinoma

Although cancer cells are known to be “addicted” to glucose, the effect of glucose in proliferation of these cells remains elusive. Here, we report that upon glucose induction, protein arginine methyltransferase 5 (PRMT5) exerts a profound effect on the G1-S cell cycle progression via directly interacting with cyclin dependent kinase 4 (CDK4) in hepatocellular carcinoma (HCC). Upregulation of both PRMT5 and CDK4 predicts more malignant characteristics in human HCC tissues. Mechanistically, glucose promotes the interaction between PRMT5 and CDK4, which leads to activation of CDK4-RB-E2F-mediated transcription via releasing CDKN2A from CDK4. Moreover, the PRMT5 competitive inhibition of the interaction between CDK4 and CDKN2A is important for glucose-induced growth of HCC cells. Furthermore, the CDK4 mutant R24A weakly binds to PRMT5, inhibiting HCC cell cycle progression and tumor growth. Thus, our findings uncover a critical function for PRMT5 and CDK4 and provide an improved therapeutic strategy against HCC.


INTRODUCTION
Liver cancer, especially hepatocellular carcinoma (HCC), causes high morbidity and mortality relative to other cancers [1]. Major precipitating factors for hepatocarcinogenesis include hepatitis B virus (HBV) or hepatitis C virus (HCV) infection, chemical cytotoxicity and aberrant hepatic metabolism [2][3][4]. Although hepatectomy has been identified as the main treatment strategy for HCC, poor prognosis still frequently exists [5]. Therefore, the need to explore underlying therapeutic options and molecular targets is particularly urgent.
Since aberrant progression of the cyclin dependent kinase (CDK)-driven cell cycle is one of the hallmarks of cancers [6], many studies have focused on these cell cycle regulators as potential targets in HCC. Among the CDKs, cyclin dependent kinase 4 (CDK4) plays a specific role in tumorigenesis and development [7]. Combined with cyclin D (CCND), CDK4 controls cancer cells from the G1 phase to S phase for DNA synthesis and expedites cell proliferation [8]. Of these CCNDs (cyclin D1, D2 and D3), CCND1 is the most pivotal for the interaction with CDK4 in hepatic cell cycle progression [9]. On the other hand, some CDK inhibitory proteins (CKIs), such as CDKN2A, CDKN2B, CDKN2C and CDKN2D, inhibit the function of the CDK4 and CCND complex in cell cycle regulation [8,10]. Overall, the CDK4 and CDKN2A (p16 INK4a as the major isoform) complex is more critical for growth suppression in hepatoma cells [11] or other types of tumor cells [12]. With CCND1 stimulation and CDKN2A inhibition, CDK4 phosphorylates the tumor suppressor retinoblastoma protein (RB) at serine780 and serine795 [13], allowing pRB to activate the transcriptional factor E2F [14]. The transcriptional functions of E2Fs have been fully determined. E2F1, E2F2 and E2F3a are perceived as transcriptional activators, while E2F3b, E2F4, E2F5, E2F6 and E2F7 are considered to be repressors [15]. RB interacts with most activators of E2Fs (E2F1, E2F2, E2F3
Protein arginine methyltransferase 5 (PRMT5) is an enzyme which can transfer methyl groups to the arginine residues of histones and some nonhistone proteins, and its methyltransferase activity is necessary for tumor cell proliferation [18]. PRMT5 has been considered as a potential target for cancer due to its function in tumor cell cycle regulation. For example, PRMT5 depletion leads to apoptosis and cell cycle arrest via methylation of tumor suppressor p53 [19]. Furthermore, PRMT5 can upregulate levels of cell cycle regulators in lung cancer, such as CDK4/6 and CCND1/D2/E1 [20]. Although one study has shown cyclin D1/CDK4 to phosphorylate MEP50 and then promote PRMT5 methyltransferase activity [21], the concrete interaction between PRMT5 and CDKs in HCC cell cycle regulation still needs to be addressed.
Here, we confirmed that glucose is indispensable for PRMT5 to facilitate HCC cell growth. Under the high glucose condition, PRMT5-depleted cells were more sensitive to a CDK4 inhibitor. Importantly, we identified a direct glucose-induced interaction between PRMT5 and CDK4. Through that interaction, PRMT5 inhibited the interaction between CDK4 and CDKN2A and then activated the CDK4-RB-E2F pathway in HCC cells under glucose induction. Furthermore, we revealed that the CDK4 mutant R24A weakly bound with PRMT5 and inhibited HCC cell cycle progression. As a result, the protein levels of PRMT5 and CDK4 were found to positively correlate in HCC and stimulate HCC cell proliferation.

Protein levels of PRMT5 and CDK4 are positively correlated, which predict more malignant characteristics in human HCC tissues
To identify the role of PRMT5 and CDK4 in HCC, we analyzed 75 pairs of human HCC and adjacent tissues by immunohistochemistry (IHC). As shown in Figure  1A, PRMT5 proteins were detected in almost all HCC cells, and quantification of the staining on a scale of 0 to 12 showed that 62 out of 75 (83%) human HCC tissues displayed high PRMT5 expression levels compared with the adjacent normal tissues (Table 1 and Figure 1B). By statistical analysis of clinicopathological parameters of these 75 HCC patients, PRMT5 protein levels were observably correlated with HCC tumor stage (P = 0.029). However, patient sex, age, degree of tumor differentiation and other parameters had no observable relationship with PRMT5 expression (Table 1). Analogously, IHC analysis also revealed that CDK4 proteins were markedly detected ( Figure 1C) in HCC tissues and highly expressed in 46 (61%) cases (Table 1 and Figure 1D). The tumor size and tumor stage, but not other parameters, correlated with tumor CDK4 expression (P<0.05, Table 1 =. Moreover, we divided the 75 HCC samples into four groups: high CDK4 and high PRMT5 expression, low CDK4 and low PRMT5 expression, high CDK4 and low PRMT5 expression, low CDK4 and high PRMT5 expression (the representative samples were shown in Supplementary Figure S1). The statistical analysis of clinicopathological parameters showed a correlation between CDK4 and PRMT5 co-expression level and tumor size or tumor stage (P < 0.05, Table 2). Furthermore, we detected a correlation (Pearson r = 0.6651, P < 0.001, Figure 1E) between the staining scores of CDK4 and PRMT5 expressed in HCC tissues. Thus, these results indicated that the protein levels of PRMT5 and CDK4 are positively correlated in human HCC tissues, which predict more malignant characteristics.

Glucose-induced PRMT5 promotes HCC cell proliferation
Since PRMT5 is essential for tumor cell cycle and proliferation, we investigated the role of PRMT5 in HCC cell cycle regulation by flow cytometry. As shown in Figure 2A, a significant increase in the proportion of HepG2 cells with PRMT5 knocked down by siRNA were in the G1 phase, while and those in the S phase were decreased (similar results in PRMT5 knockdown by shRNA in HuH-7 cells, Supplementary Figure S2). Tumor cells need massive glucose uptake, and previous studies have shown that PRMT5 stimulates hepatic glucose metabolism [22,23]. Therefore, we examined whether PRMT5 promotes HCC cell proliferation by relying on glucose induction. Compared with the control HuH-7 cells, PRMT5 knockdown by shRNA obviously suppressed the proliferation of HuH-7 cells in the high glucose condition ( Figure 2B and 2C). However, in the low glucose condition, both the control-shRNA and PRMT5-shRNA HuH-7 cell groups presented slower growth than in those in the high glucose condition ( Figure  2B and 2C). These results indicated that PRMT5 promoted HCC cell proliferation under glucose induction.

PRMT5 depletion sensitizes HCC cells to CDK4 inhibitor upon glucose induction
In the high glucose condition, HepG2 cells with both PRMT5 and CDK4 knocked down by siRNA showed slower growth than those with only PRMT5 or CDK4 interference ( Figure 2D). Moreover, using a small molecule CDK4 inhibitor fascaplysin, we also observed a more repressive effect in shPRMT5 HepG2 cells than control cells ( Figure 2E). We also compared the effect of fascaplysin on cell viability of control or shPRMT5 HepG2 cells and observed that the PRMT5 knockdown cells had a lower cell survival rate with fascaplysin ( Figure 2F). Furthermore, to examine the effects of PRMT5 and CDK4 in cell cycle regulation, control and shPRMT5 HepG2 cells cultured in high glucose were exposed to different concentrations of fascaplysin for 2 h. With the increase of fascaplysin concentration, a visible increase of shPRMT5 treated HepG2 cells in the G1 phase was observed, while control cells showed no such increase ( Figure 2G). However, in low glucose, no noticeable change in either control or shPRMT5 treated cells in the G1 phase was observed in the 2-h treatment ( Figure 2H). Thus, our results suggested that PRMT5 and CDK4 promoted HCC cell proliferation and cell cycle progression upon glucose induction, and PRMT5 depletion sensitized HCC cells to the CDK4 inhibitor.

PRMT5 activates CDK4-pRB-E2F-mediated transcription in high glucose condition
Next, we examined the role of PRMT5 on CDK4 function. In the high glucose condition, PRMT5 knockdown HepG2 cells displayed a weaker phosphorylation of RB serine780 ( Figure 3A); however, it was not significantly changed in the low glucose condition ( Figure 3A). Since RB serine780 phosphorylation activates the CDK4-RB-E2F pathway, we examined whether PRMT5 would affect its function. Using Western blot assay and quantitative realtime PCR (qRT-PCR), we analyzed the protein and mRNA levels of RB-E2F downstream genes between control and PRMT5-siRNA treated HepG2 cells in various glucose conditions. The results indicated that the protein levels of CCNE1 and CDC6 in high glucose-cultured siPRMT5 cells were decreased ( Figure 3B). And the mRNA levels of most of these downstream genes (CCNE1, CDC6, MCM3, CDK1, MYBL2, ORC1) were more significantly down-regulated in siPRMT5 treated cells cultured in high glucose medium ( Figure 3C). Thus, our results demonstrated that PRMT5 promoted cell cycle progression through CDK4-RB-E2F transcriptional activation in the high glucose condition.

PRMT5 interacts with CDK4
In a previous study, CDK4-CCND1 interaction was shown to activate PRMT5 methyltransferase through the regulatory factor MEP50 [21]. However, our study suggests an interaction exists between PRMT5 and CDK4. This interaction was confirmed using a GSTfused CDK4 recombinant protein to pull down purified Flag-tagged PRMT5. As a negative control, another cell cycle regulator CDK2 which could not interact with PRMT5 was used ( Figure 4A). Moreover, exogenous HA-tagged PRMT5 and endogenous PRMT5 expressed in HEK293T cells could be co-immunoprecipitated by purified Flag-tagged CDK4 ( Figure 4B and 4C).
Conversely, exogenous Flag-tagged CDK4 and endogenous CDK4 could be immunoprecipitated by purified HA-tagged PRMT5 ( Figure 4D and 4E). As PRMT5 has been reported an interaction with anti-Flag antibody [24,25], we performed the endogenous co-IP assay (Supplementary Figure S3A) and exogenous Flag-PRMT5 precipitating HA-CDK4 assay (Supplementary Figure S3B), and confirmed PRMT5 could interact with CDK4 in the absence of anti-Flag antibody. Next, HEK293T cells and HCC cell lines HepG2 and HuH-7 were transfected with eGFP-PRMT5 (green) and mCherry-CDK4 (red), and yellow signals indicating their co-localization were detected both in cytoplasm and nucleus in the merged images ( Figure 4F). Furthermore, the interaction between PRMT5 and CDK4 was found to be enhanced in HepG2 cells cultured in high glucose ( Figure 4G). As PRMT5-MEP50 complex was essential for PRMT5 function, next we tested whether CDK4 could interact with MEP50. Co-IP assays confirmed the interaction between CDK4 and MEP50 ( Figure  4H). However, glucose did not significantly influence the interaction between CDK4 and MEP50 ( Figure 4I). Together, our results suggested that PRMT5-MEP50 complex interacted with CDK4, and high glucose condition mainly enhanced PRMT5-CDK4 interaction not MEP50-CDK4.

PRMT5 inhibits interaction between CDK4 and CDKN2A upon glucose induction
The mechanism of PRMT5 and CDK4 interaction was further explored. Although PRMT5 is a symmetric arginine dimethylase, monomethylation (MMA) and    Flag-PRMT5 overexpressed in HKT293T cell lysates were performed. B, C, D, E. HEK293T cells were transfected with Flag-CDK4 or HA-PRMT5 (or both) for 48 h. The interaction between exogenous (or endogenous) PRMT5 and CDK4 was examined by coimmunoprecipitation assays followed by Western blot. F. HepG2, HuH-7 and HEK293T cells were co-transfected with eGFP-PRMT5 and mCherry-CDK4 for 48 h. The eGFP (green) and mCherry (red) signals were acquired by a fluorescence microscope at 40× magnification (Scale bar, 10 μm). G. HepG2 cells were co-transfected with HA-PRMT5 and Flag-CDK4 for 24 h in normal conditions and then cultured in no/low/high glucose media for 24h. The interaction was examined by co-immunoprecipitation assays followed by Western blot.
H. HEK293T cells were transfected with Flag-CDK4 (or empty vector) and HA-MEP50 for 48 h. The interaction between exogenous MEP50 and CDK4 was examined by co-immunoprecipitation assays followed by Western blot. I. HepG2 cells were co-transfected with HA-MEP50 and Flag-CDK4 for 24 h in normal conditions and then cultured in no/low/high glucose media for 24h. The interaction was examined by co-IP assays followed by Western blot. symmetric dimethylation (SDMA) of CDK4 through the interaction were not detected ( Figure 5A). Using the methyltransferases inhibitor adenosine-2', 3'-dialdehyde (Adox, which inhibits SAH hydrolase and accumulates SAH, resulting in inhibiting methyltransferases) (Supplementary Figure S4A) and PRMT5 enzymatic mutation Δmut (Supplementary Figure S4B), we also did not observe any change of MMA or SDMA modification of CDK4. Because CDK4 activation depends on CCNDstimulative binding and CDKN2A-suppressive binding [7,8], we needed to determine whether PRMT5 would influence CDK4-CCND or CDK4-CDKN2A binding. Thus, we co-transferred Flag-tagged CDK4, Myc-6×His-tagged p16 INK4a (the major isoform of CDKN2A) and different quantities (0, 0.4 and 4 μg) of HA-tagged PRMT5 into high/low glucose cultured HEK293T cells, and the cell lysis solution was immunoprecipitated with an anti-Flag M2 antibody. Interestingly, p16 INK4a weakly bound to CDK4 in the presence of PRMT5 in the high glucose condition ( Figure 5B). However under the low glucose condition, PRMT5 could not affect the interaction between CDK4 and p16 INK4a ( Figure 5B). Moreover, using the same experimental approach, we observed that CDK4-CCND1 binding was not influenced by PRMT5 (Supplementary Figure S4C). These data suggested that upon glucose induction, PRMT5 inhibited the interaction between CDK4 and CDKN2A.

PRMT5 and CDKN2A compete for CDK4 binding
To identify sequence motifs of CDK4 required for PRMT5 binding, we divided the CDK4 protein into four sections according to its secondary structure (Uniprot database): amino acids 1-70, 71-139, 140-215 and 216-303 ( Figure 5C). GST alone and GST fusion proteins to CDK4 fragments were produced in Escherichia coli BL21 and then purified (Supplementary Figure S4D). GST-pull down assays showed the ability of HA-PRMT5 to interact with CDK4 full-length (1-303 aa) and CDK4 1-70 aa proteins ( Figure 5D), which includes the ATPbinding domain, cyclin-binding domain and some inhibitory sites [8]. Since PRMT5 was needed for CDK4 1-70 aa binding and inhibited the binding between CDK4 and CDKN2A, we next examined whether PRMT5 and CDKN2A competitively interact with CDK4 1-70 aa. We purified 6× His-tagged PRMT5 and p16 INK4a in vitro and then incubated GST-fused CDK4 1-70 aa with Histagged p16 INK4a and increasing concentrations of Histagged PRMT5. As the levels of PRMT5 were increased, the protein levels of p16 INK4a pulled down by CDK4 1-70 aa decreased ( Figure 5E). Similarly, the protein levels of PRMT5 pulled down by CDK4 1-70 aa also decreased as the levels of p16 INK4a were increased ( Figure 5F). These results suggested that PRMT5 and CDKN2A competed for binding with CDK4 in the 1-70 aa region.

PRMT5-binding mutant CDK4 R24A inhibits HCC cell cycle progression
Previous studies have shown that the arginine 24 residue within the CDK4 1-70 amino acid region is essential for CDKN2A-binding [26], and mutations R24C and R24H of CDK4 were shown to inhibit protein binding with CDKN2A in melanoma [27][28][29]. Here, we constructed the Flag-tagged CDK4 R24A mutant ( Figure 6A) and validated that R24A had a lower level of binding with p16 INK4a than the wild-type (WT) protein (Supplementary Figure S5A). Interestingly, the interaction between PRMT5 and CDK4 R24A was also diminished ( Figure 6B). However, the CCND1-misbinding mutant CDK4 Δ50-56, lacking the 50-56 aa sequence PISTVRE ( Figure 6A), still maintained the majority of the interaction with PRMT5 ( Figure 6B). To determine whether R24A could function in the presence of PRMT5, cell growth and cell cycle progression in HuH-7 and HepG2 cells were analyzed. As shown in Supplementary Figure  S5B and S5C, compared to CDK4/PRMT5, R24A/ PRMT5 significantly inhibited HCC cell proliferation. The flow cytometry analysis also showed that R24A/ PRMT5 increased the percentage of cells in the G1 phase relative to CDK4/PRMT5 ( Figure 6C). Furthermore, we investigated whether CDK4 R24A also affected the CDK4-RB-E2F pathway. The phosphorylation level of RB serine780 was decreased in HepG2 cells expressing R24A/PRMT5 ( Figure 6D). Similarly, R24A/PRMT5 caused down-regulation of the mRNA (Supplementary Figure S5D) and protein ( Figure 6E) levels of RB-E2F downstream genes in comparison to CDK4/PRMT5. Taken together, these results suggested that CDK4 arginine24 was indispensable for PRMT5 function, and the CDK4 R24A mutant inhibited HCC cell growth and cell cycle progression.

CDK4 R24A inhibits HCC tumor growth
To determine whether the CDK4 mutant R24A also would inhibit HCC growth in vivo, we conducted xenograft tumor studies. Nude mice were injected with 5 × 10 6 HepG2 cells stably expressing CDK4 WT and PRMT5 (or stably expressed CDK4 R24A and PRMT5). The consistent expression levels of PRMT5 and CDK4 (WT and R24A) in this two cell lines were validated ( Figure 6E). As shown in Figures 6F and S5E, tumors in mice injected with CDK4 WT/PRMT5 HepG2 cells were larger than those in mice injected with CDK4 R24A/PRMT5 HepG2 cells. Furthermore, injection of HepG2 cells transfected with CDK4 WT/PRMT5 also resulted in a larger tumor width ( Figure 6G) and tumor weight ( Figure 6H) than those with CDK4 R24A/ PRMT5. Our results suggested that the weak binding of CDK4 R24A with PRMT5 inhibited HCC tumor growth. www.impactjournals.com/oncotarget  B. HEK293T cells were co-transfected with Flag-CDK4 WT (R24A or Δ50-56) and HA-PRMT5 for 48 h. The interaction was examined by co-immunoprecipitation assays followed by Western blot. C. HepG2 cells stably expressing CDK4 WT/PRMT5 and CDK4 R24A/PRMT5 were analyzed by FACS, and then percentages of cells in the G1 phase were determined using Modfit LT software and tested with t-test. *P < 0.05. D. HepG2 cells were transfected with HA-empty vector, HA-PRMT5, HA-PRMT5/Flag-CDK4 or HA-PRMT5/Flag-CDK4 R24A for 48 h. The phosphorylation level of RB serine780 was examined by Western blot. E. Protein levels of CCNE1 and CDC6 in HepG2 cells stably expressing CDK4 WT/PRMT5 and CDK4 R24A/PRMT5 were examined by Western blot. F. Nude mice were injected with 5 × 10 6 HepG2 cells stably expressing CDK4 WT/PRMT5 (upper row) and CDK4 R24A/PRMT5 (lower row). Tumors isolated from mice were imaged. G. Average widths of tumors were recorded on various days after injection, and P < 0.05 was considered significant. H. Weights of tumors were recorded when the mice were sacrificed. Data are presented as the mean ± SEM and tested with t-test, *P < 0.05. www.impactjournals.com/oncotarget DISCUSSION PRMT5 and CDK4 are highly expressed in HCC.
In our study, we demonstrated the glucose-induced facilitation of HCC cell growth by PRMT5, which can promote the G1/S cell cycle transition. Moreover, cell proliferation assays showed that shPRMT5 HepG2 cells were more sensitive to fascaplysin, and the FACS analysis suggested higher fascaplysin sensitivity for PRMT5-depleted cells in the G1/S transition under the high glucose condition. In other words, under glucose induction, PRMT5-depleted HCC cells showed CDK4 instability and high CDK4 inhibitor sensitivity. On the other hand, PRMT5 was found to interact with and affect the function of CDK4. With high glucose, PRMT5 competitively inhibited the interaction between CDK4 and CDKN2A, leading to CDK4-RB-E2F transcriptional activation. Furthermore, relative to the coexpression of CDK4 WT with PRMT5, that of the CDK4 mutant R24A with PRMT5 resulted in the arrest of HCC cell cycle and proliferation. In conclusion, the interaction between PRMT5 and CDK4 rendered a tumor-promoting effect.
The relationship between PRMT5 and CDKs in previous studies have been identified in that PRMT5 plays a role in regulating symmetric dimethylation of histones through the interaction of methylosome protein 50 (MEP50) with CDK4 [21], CDK8 or CDK19 [30]. For example, the previous works have shown that CCND1/CDK4 phosphorylates the PRMT5 co-factor MEP50 and activates PRMT5 methyltransferase in CUL4 transcriptional suppression. Those observations suggest a critical oncogenic role of PRMT5 and CDK4 in tumorigenesis. Here, we provide considerable evidence of a positive role for PRMT5 and CDK4 in tumor cell cycle regulation. Our study clearly described the non-methyltransferase activity of PRMT5 by competitively preventing CDK4 from binding CDKN2A, thereby increasing the CDK4-RB-E2F transcriptional activity. Although we affirmed MEP50 interacted with CDK4, the interaction could not be significantly induced by glucose. That means glucose may promote the formation of PRMT5/MEP50-CDK4 complex, however, it mainly enhances the link of PRMT5-CDK4 not MEP50-CDK4 in a molecular level. Thus, our results support a supplemental function for CDK4-PRMT5 transcriptional repression in contributing to tumor growth.
Commonly, PRMT5 contributes to tumorigenesis and cell cycle imbalance as an arginine methyltransferase. PRMT5 methylates arginine residues of histones or nonhistones leading to transcriptional silencing, including the cell cycle regulators. For example, acting as a tumor suppressor, E2F1 methylated by PRMT5 can increase tumor cell growth and inhibit apoptosis [31]. However, E2F1 plays an accelerative role in cell cycle regulation of tumor growth, which is dependent on pRB and CDKs. Here, we provide evidence for PRMT5 functioning in a non-enzymatic manner. The competitive binding of CDK4-CDKN2A or PRMT5-CDK4 is alternatively decided by glucose induction. Thus, it turns the pRB-E2F1 pathway into a transcriptionally activated state or depressed state. The results suggest a crucial mechanism for PRMT5 in cell cycle regulation contributing to oncogenesis.
There are varieties of factors that affect cell proliferation, in which an important one is regulation of cell cycle progression. Cell cycle activators promote cell cycle progression, the ratio of apoptosis cells decrease and then cells grow faster. This is a cascade amplification process. In our study, PRMT5 as an upstream regulator promoted cell cycle through CDK4 in performance at G1-S transition. We observe a limited change of the ratio of G1 and S phase when PRMT5 was short-term knockdown, but over time the effect of cell proliferation could present a large variation. On the other hand, PRMT5 can also promote cell growth in other ways, such as accelerating cell metabolites biosynthesis [32], and cell cycle regulation is not the only pathway. Therefore, this is why cell cycle effect looks not dramatic but proliferation effect looks significant.
As a major nutrient, glucose is critically required for cellular biosynthesis and energy supply. Compared with normal cells, tumor cells metabolize more glucose to lactate via glycolysis, a phenomenon called the Warburg effect [33]. Due to the liver acting as the main metabolic organ, HCC is an aggressive cancer requiring high glucose uptake. In our study, PRMT5 was shown to facilitate HCC cell growth induced by high-level glucose, similar to results of numerous prior studies on PRMT5 function. Despite the oncogenic role of PRMT5, the underlying mechanisms of PRMT5 in HCC are not fully understood. Surprisingly, in our study, PRMT5 directly bound to CDK4 in rapidly growing cells such as HEK293T and HepG2 cells. Interestingly, this whole system is glucose-driven. Glucose has been considered an actuator for the cell cycle, and a high level of it controls the cell cycle progression by increasing CCNA, CCNB1 and CCND1 [34]. In our study, we found that the combination of PRMT5 and CDK4 presented a larger impact on HCC cell cycle progression under the high glucose condition. On the other hand, CDK4 can maintain intercellular glucose homeostasis by controlling glucose and lipid metabolism [35] or repressing gluconeogenesis in a CCND1-dependent manner [36]. Taken together, results of previous studies and our current work have revealed PRMT5 and CDK4 as a crucial link for cell cycle and glucose metabolism regulation.
Since results of the GST-pull down and coimmunoprecipitation assay suggested an interaction www.impactjournals.com/oncotarget between PRMT5 and CDK4, we predicted that PRMT5 impacts HCC cell proliferation via the effect of binding with CDK4. CDK4 is well known as an oncogene and a significant clinical marker in HCC progression [37]. Moreover, some CDK4-specific inhibitors have been applied in research and clinical trials. Fascaplysin is a marine bis-indole alkaloid and shows specific CDK4 inhibitory activity with an IC50 of 0.35 μM, while its activity was observed to be low against other CDKs [38]. Another CDK4 inhibitor PD0332991, which can significantly suppress proliferation of hepatoma cells, has been evaluated in a clinical phase II study of liposarcomas [39,40]. In this study, effects of fascaplysin on HCC cell growth and G1/S transition arrest were more significant in the absence of PRMT5. The results imply a more efficient therapeutic strategy for CDK4 inhibitors in the treatment of HCC.
In addition, CDK4 mutations have been reported in several types of cancer. The phosphorylation of CDK4 at threonine172 was found to be essential for CDK4-CCND activating binding, while the CDK4 T172A mutant displayed very weak activity [41,42]. CDK4 R24C and R24H were found in human melanoma as tumor activating mutations, but they showed no effect on sporadic pituitary adenomas, insulinomas or Leydig cell tumors [43]. In this study, our data showed an important function of the CDK4 mutant R24A in HCC cells. In comparison to WT CDK4, the R24A mutant did not influence the change in G1 phase in HCC cells (data not shown). However, in the presence of PRMT5, R24A significantly arrested G1/S transition and HCC cell growth. Although R24A bound less strongly to CDKN2A, the weak interaction between PRMT5 and R24A showed more evidently that CDK4 R24A suppresses HCC cell cycle progression in a PRMT5-dependent manner. Thus, an important mechanism was discovered for the roles of PRMT5 and CDK4 in HCC.
In summary, we have demonstrated that PRMT5 is a conditional regulator of CDK4 in HCC. In the high glucose condition, PRMT5 was shown to emphatically interact with CDK4, releasing it from CDKN2A and leading the CDK4-RB-E2F axis to an active state for tumor growth (Figure 7). Our study may provide a strategy for improving CDK4-targeting treatments against HCC.

Immunohistochemistry
All experimental procedures using human tissues were approved by the Human Ethics Committee of Institute of Health Sciences. The samples were obtained by surgical resection from patients who gave consent at Renji Hospital. The tissue sections were treated with 3% H 2 O 2 to block endogenous peroxidase activity and incubated with primary antibodies (anti-CDK4 at 1:2000 and anti-PRMT5 at 1:2000). The sections were then incubated with an HRP-conjugated secondary antibody, followed by treatment with diaminobenzidine chromogen to visualize the signals. Staining of CDK4 and PRMT5 was scored by two independent researchers based on the distribution and intensity of signals. The distribution (percentage of stained tissue area) was scored as 0 (0-5%), 1 (5-25%), 2 (25-50%), 3 (50-75%), 4 (75-100%). The intensity of staining was scored as 0 (no staining), 1 (weak), 2 (moderate) and 3 (strong). The final scores of CDK4 or PRMT5 expression in HCC and adjacent tissues were on a scale of 0-12. Samples were defined as having high expression (score > 6) or low expression (score ≤ 6). Pearson's test was applied to determine the correlation between CDK4 and PRMT5 using SPSS 20 software.

Quantitative real-time PCR
Total RNA was isolated using a Trizol kit (Omega, GA, USA), and cDNA was synthesized using a cDNA synthesis kit (Takara, Otsu, Japan). Quantitative real-time PCR was performed using the SYBR Green PCR Master Mix (Takara, Otsu, Japan) on the StepOnePlus Real-Time PCR System (Applied Biosystems, USA). The primers used in this study are listed as following:

Cell proliferation assays
HepG2 or HuH-7 cells were plated in 24-well plates with a density of 1 × 10 4 cells per well. Twenty-four hours after transfection or fascaplysin treatment, cells were harvested every 24 h, and cells numbers were counted using a hemocytometer.

Colony formation assay
12-well plates were covered with 0.4% agarose gel (Gibco), and then cells in 0.375% agarose gel were overlaid on the top of the 0.4% gel. 200 cells were suspended per well and cultured. After 3 weeks, the cells were stained with 0.1% crystal violet. The picture of each well was photographed and the colony number was calculated.

Flow cytometry
Trypsin digested cells were washed with 1 × PBS, fixed in cold 70% ethanol for 24 h and then incubated with RNase A (Invitrogen) for 30 min at 37°C and 50 μg/ ml propidium iodide (Sigma-Aldrich) for 15 min at 4°C shielded from light. Cells were analyzed with a Gallios (Beckman) fluorescence activated cell sorter.