GCN-5/PGC-1α signaling is activated and associated with metabolism in cyclin E1-driven ovarian cancer

Introduction

Approximately 220,000 women are affected by malignant epithelial ovarian tumor each year and ovarian cancer (OvCa) is the fifth commonest cause of death in women with cancer in the United States [1]. High grade serous ovarian cancer (HGSOC) consists of 70% of OvCa cases that decease [2]. HGSOC at late stage responds poorly to platinum-based chemotherapy, which is the standard-of-care. Rapid relapse and unfavorable overall survival are critical clinical plights [3].

HGSOC is highly heterogeneous with different genetic subtypes conferring distinct prognosis. Besides well-established truncal TP53 mutations, HGSOC is also characterized with high frequency of CCNE1 amplification and resulting upregulation (~30%). This genotype is termed Cyclin E1-driven HGSOC [4]. Cyclin E1 is encoded by CCNE1 and forms complex with Cyclin Dependent Kinase 2 (CDK2). The complex plays critical role in regulation of G1/S phase and incurs in part the uncontrolled cell cycle progression in cancer context. Unlike other genotypes, Cyclin E1-driven HGSOC are insensitive to platinum-based chemotherapy and resistance surfaces even at initiation. Thus, such genotype is associated with sooner relapse and worsened prognosis [5]. Notwithstanding exploratory administration of CDK2 in Cyclin E1-driven HGSOC showing promising efficacy, future resistance remains unsolved [6].

Our previous reports indicate that Cyclin E1-driven ovarian cancer is sensitive to inhibition of both CDK2 and glucose metabolism. We also report that G6PC may stay at the hub of glucose metabolism and cell cycle mediation. Those findings shed light on novel function of Cyclin E1-CDK2 complex, which may in our speculation mediate crosstalk between cell cycle regulation and metabolism, a mechanism that may contribute to resistance to cell cycle based inhibitors like chemotherapy or CDK2 inhibition [7, 8]. In a recent study focusing on gluconeogenesis, CITED2 was reported to inhibit the acetylation of PGC-1a by blocking its interaction with the acetyltransferase general control of amino acid synthesis 5–like 2 (GCN5) and the acetylation of PGC-1a was GCN5-dependent. Together with our findings, we speculate that the GCN-5/PGC-1α may also play a role in ovarian cancer [9].

In the current study, we focus on the mechanism of resistance to CDK2 inhibition in Cyclin E1-driven HGSOC. With in vitro and in vivo modeling of the genotype, we aim at breaking bottleneck of treatment dilemma of OvCa and our findings may hold promise to precision targeted therapy for OvCa.

Results

In the current study, we first aim to validate our speculation whether CCNE1 amplified HGSOC has altered metabolic pathway via GCN5/PGC-1α/G6PC axis. We first analyzed expressions of genes enriched in CCNE1 amplified cases (~22% of cases) and found that genes significantly enriched were at abundance (Figure 1A) and that the subunit of G6PC, the G6PC3 was significantly upregulated in CCNE1 amplified cases (Figure 1B). This was not surprising as the sequencing dataset we selected only encompassed G6PC3 subunit of the gene, yet considering the functional redundancy we proceeded with in vitro probing. Functional annotation of enriched genes showed that except for cell cycle pathway significantly altered which was established, the second most significantly altered pathway was metabolism (Figure 1C). We then studied correlation between expressions of genes within the signaling axis using the mRNA expression data from TCGA cohort. Using partial correlation simulation, we observed significant positive correlation between CCNE1 and G6PC3 expression, indicating potential positive regulation between the two. We also noticed positive correlation between expressions of KAT2A (GCN5) and G6PC3, also indicating potential positive regulation between the genes. No correlation was noticed for CDK2 or PGC-1α with CCNE1 (Figure 1D). Here we showed in clinical tissue samples, echoing our previous work that metabolism plays a role in CCNE1 amplified OvCa and G6PC may play a role therein linking GCN5 and PGC-1α.

We next performed in vitro studies to elucidate mechanistic regulation. Both OVCAR-3 and A2780 OvCa cells were confirmed to be CCNE1 amplified. By silencing CDK2 and GCN5 in both cell lines (Figure 1E–1F), we found that KD of GCN5 resulted in decreased G6PC and increased PGC-1α level in both cell lines, respectively (Figure 1G–1H). CDK2-KD significantly lowered expression of G6PC in OVCAR-3 cells but not in A2780 cells (Figure 1G). CDK2-KD significantly increased expression of PGC-1α in both cell lines (Figure 1H). We then investigated whether negative regulation of CDK2 and GCN5 on PGC-1α was due to acetylation. We found that both genetic and pharmaceutical (MB-3) inhibition resulted in significant decrease in acetylation of PGC-1α in both cell lines (Figure 2A–2B). Silencing of CDK2 also resulted in significant decrease in acetylation of both PGC-1α and Rb (Figure 2C–2D), supporting the decreased mRNA expression of PGC-1α could be due to inhibition via acetylation by upstream CCNE1/CDK2/GCN5 signaling. We next examined metabolic profiles and found silencing of GCN5 genetically or pharmaceutically had similar effect to CDK2-KD by significantly decreasing glucose uptake (Figure 2E), increasing lactate production (Figure 2F), and decreasing SDH activity (Figure 2G).

We next tried to identify hierarchy of the elements. We first found that silencing of CCNE1 had minute impact on CDK2 level (Figure 3A), in reminiscence of a recent report that overexpression and amplification of CCNE1 is dependent on CDK2 activity in Cyclin E1 driven cancer cell lines. We then silenced CDK2 in both cell lines and found a corresponding decrease in CCNE1 level in OVCAR-3 cells but not as apparent in A2780 cells, indicating individualized mechanism in crosstalk between CCNE1 and CDK2 (Figure 3B). Also, CDK2 silencing resulted in decreased GCN5 level in both cell lines (Figure 3B). Silencing of GCN5 showed minimal impact on CDK2 level yet resulted in increased PGC-1α level in both cell lines (Figure 3C). Silencing of PGC-1α also showed minimal impact on GCN5 level (Figure 3D). We thus speculate that the Cyclin E1-CDK2 complex mediates GCN5 which subsequently inhibits PGC-1α by acetylation.

We next studied whether inhibiting GCN5 could have additive effect to CDK2 inhibitor. As the Cyclin E1-CDK2 complex plays pro-tumorigenic role in a variety of mechanisms, inhibition of GCN5 is not expected to substitute CDK2 inhibition despite its downstream position in the signaling. In both cell lines, we found that inhibitory effect of Dinaciclib was similar to that of GCN5 silencing, whereas combination therapy further inhibited cell proliferation significantly (Figure 4A–4B). The cell cycle analyses revealed that both Dinaciclib and GCN5-KD resulted in increased G1 population in both cell lines, whereas Dinaciclib almost completely blocked cell entering G2 phase (Figure 4C–4D). Combination therapy augmented G1 phase population leaving only minimal population in G2/M phase (Figure 4C–4D). Combination therapy also significantly increased cell apoptosis as compared to either blockade in both cell lines (Figure 4E–4F). We next studied inhibitory effect on migratory ability of cells and found that combination therapy lead to significantly additive inhibition in cell migration and invasion as compared to GCN5-KD or Dinaciclib alone (Figure 4G–4H). Combination therapy also conferred significantly additive inhibition to anchorage-independent growth in both cell lines (Figure 4I). Finally, we performed xenograft experiments and found, intriguingly, GCN5-KD did not alter tumor growth compared with control in both cell lines (Figure 4J–4K). However, combination therapy of Dinaciclib and GCN5-KD conferred significant inhibition of tumor growth compared with either therapy alone (Figure 4J–4K). No toxicity was observed in mice throughout the experiment period.

Discussion

In the current study, we have provided evidence that CCNE1 drives metabolic alteration in ovarian carcinoma with activated GCN5/PGC-1α signaling. The hypothesis of this study was proposed based on 2 previous studies by our group showing unique metabolic shift in ovarian cancer [7, 8]. We first discovered association between G6PC and CDKN1B in HGSOC and that Warburg effect was not the prominent metabolic shift in HGSOC. In our later study, we showed that Cyclin E1-dirven HGSOC was dependent on tricarboxylic acid cycle upon CDK2 inhibition, further supporting a connection between cell cycle control and glucose metabolism in this genotype. We thus proposed that Cyclin E1-CDK2 complex may stand at the hub of mediating both cycle and metabolism, and while the mechanism of former has been well studied, we here focused on its signaling in glucose metabolism.

Crosstalk between cell cycle and glucose metabolism plays critical role yet has not been well elucidated. Cyclins are a series cell cycle-dependent proteins controlling checkpoint passage via their activity and quantity. The dynamic change in Cyclin amount ensures no excessive proliferation and adequate time for DNA repair under normal circumstances [10]. Cyclin E1 belongs to cyclin family and binds to CKD2 as a compound which phosphorylates downstream proteins, playing an important role in G1 phase and G1-S transition of mammalian cell cycle progression. Due to the amplification of genes, Cyclin E1 RNA is overexpressed in 29.5% and CKD2 in 6.5% of ovarian tumors tested, making it another hallmark of ovarian cancer besides BRCAness [11]. So far metabolic mediation of Cylcin E1-CDK2 has not been reported besides our group. However, similar mechanistic analyses have shown that Cyclin D1-CDK4 complex can regulate glucose metabolism in hepatic cells via cell cycle-independent pathways [12]. Also, loss of cell cycle controlling gene CDKN2A, another commonly deleted gene in cancer, can break glucose equilibrium and cause increased insulin secretion by pancreatic β cells [13]. We therefore postulate that the signaling between Cylcin E1-CDK2 and downstream G6PC is via GCN5/PGC-1α axis, which has been established to mediate glucose metabolism [9].

By elucidating in part the signaling, we also evaluated the therapeutic potential of GCN5 inhibition alone and in combination with CDK2 inhibitor. Genetic inhibition of GCN5 has been shown to suppress human hepatocellular carcinoma and leukemia [14, 15]. However, the substrates reported to be acetylated are different by GCN5 in those reports. In the current study, we show that GCN5 acetylates PGC-1α and leads to alteration of glucose metabolism. Though we revealed in part the interaction between GCN5 and PGC-1α, Co-IP is now being performed to validate the findings. We also noted that although pharmaceutical GCN5 inhibition shows potent effect alone or in combination with CDK2 inhibition, it is currently unsuitable for in vivo use as the toxicity was strong (data not shown) and that’s the reason we used genetic silencing in animal models for xenograft tumor. While silencing GCN5 shows effect, reducing toxicity in pharmaceutical development remains challenging. We also noticed that genetic silencing of GCN5 alone showed only showed inhibitory effect in vitro but not in vivo. We speculate that the complexity of microenvironment which is absent in vitro may have mitigated effect of GCN5 inhibition, possibly via signaling bypass, while combination therapy that blocks more axes may have rendered shunting signaling unresponsive, given the relative short period of xenograft growth.

Given that CCNE1 is amplified in a variety of cancers, targeting tumors with this genotype is a hotspot of investigation. Rb can be phosphorylated to promote the progression of G1 when the biosynthetic activities are at a high rate, and the over-phosphorylated Rb will accelerate the transition from G1 to S phase [16]. Moreover, it is reported that the time point to entry S phase is dependent on cyclin E levels [17]. Cyclin E-CDK2 can also phosphorylate Smad3 and inhibit the transcriptional activity so as to shorten the cell cycle progression [18]. More importantly, cyclin E is observed to be overexpressed in many types of tumor cells, including breast, lung, bladder and colon cancer [19]. Based on those knowledge, we tested acetylation Rb as well and found that GCN5 or CDK2 knockdown reduced acetylation of Rb. Targeting CDK2 has thus far been the only targeting approach for CCNE1 amplified cancer. A recent screen on synthetic lethality also shows that amplification of CCNE1 is dependent on CDK2 activity [20]. Also, as we have now observed metabolic shift in Cyclin E1-driven OvCa, we are now performing a metabolic profile analysis to comprehensively pinpoint metabolic vulnerability following CDK2 and GCN5 monotherapy or combined. Those data may help overcome drug resistance to CDK2 inhibitors. To sum up, our study showed that targeting GCN5 could inhibit metabolic mediation of the CCNE1-CDK2 complex and combined inhibition with CDK2 blockade held promise for drug development.

Materials and Methods