Cancer cells are metabolically stressed during tumour progression due to limited tumour vascularity and resultant nutrient, growth factor and oxygen deficiency that can induce cell death and inhibit tumour growth. We demonstrate that Rab25, a small GTPase involved in endosomal recycling, that is genomically amplified in multiple tumour lineages, is a key regulator of cellular bioenergetics and autophagy. RAB25 enhanced survival during nutrient stress by preventing apoptosis and autophagy via binding and activating AKT leading to increased glucose uptake and improved cellular bioenergetics. Unexpectedly, Rab25 induced the accumulation of glycogen in epithelial cancer cells, a process not previously identified. Strikingly, an increase in basal ATP levels combined with AKT‐dependent increases in glucose uptake and glycogen storage allowed maintenance of ATP levels during bioenergetic stress. The clinical relevance of these findings was validated by the ability of a Rab25‐dependent expression profile enriched for bioenergetics targets to identify patients with a poor prognosis. Thus, Rab25 is an unexpected regulator of cellular bioenergetics implicated as a useful biomarker and potential therapeutic target.
The paper explained
Despite progress in molecular medicine, cancer remains one of the leading causes of death. This is due to a failure to diagnose disease at an early curable stage as well as to the lack of effective therapeutic approaches. Alterations in the RAB small GTPases and the vesicle recycling processes they control contribute to multiple human diseases including cancer. RAB25, located in the 1q22 amplicon prevalent in many cancers, contributes to the aggressiveness of breast and ovarian and likely other cancer lineages. However, the molecular mechanisms underlying the effects of Rab25 on cancer pathophysiology remain unclear.
To explore the mechanisms by which Rab25 alters cancer cell behaviour, we identified a RAB25 transcriptome. Importantly, the Rab25 signature is transferrable allowing interrogation of tumour samples and identification of patients with a poor prognosis. The Rab25 transcriptome is highly enriched in genes involved in cellular metabolism and survival. Indeed, Rab25 potently increases cancer cell survival under nutrient stress via increased AKT activation and subsequent glucose uptake, glycogen storage and maintenance of cellular bioenergetics. The data implicate an unexpected Rab25‐induced increase in glycogen stores providing an energy source used during cell stress as a novel mechanism by which Rab25 could contribute to increased tumour aggressiveness and worsened patient outcomes.
Given the importance of Rab GTPases in regulating critical cellular functions, it is not surprising that altered expression or mutation of Rab proteins and their interacting partners are associated with human disease. Rab25 and its binding partners are dysregulated in a number of tumour lineages. Functional studies demonstrate an impact of Rab25 on cell growth, proliferation, apoptosis, migration and invasion and in vivo tumourigenicity in mouse models, as well as an association with clinical outcomes. An understanding of the role of Rab25 in tumourigenesis will hasten the evolution of Rab25, its interacting partners and downstream targets as novel biomarkers and therapeutic targets.
Cancer cells characteristically require higher levels of nutrients and energy metabolism than normal cells, due to increased rates of cell proliferation that demand synthesis of proteins, lipids and nucleotides as well as energy in the form of adenosine 5′‐triphosphate (ATP; Vander Heiden et al, 2009). However, as cancers grow and metastasize, they frequently outpace their vascular supply resulting in both nutrient and oxygen deprivation. ATP, the main cellular energy source, is produced primarily from two sources, glycolysis and mitochondrial respiration (Krebs cycle and oxidative phosphorylation), with oxidative phosphorylation providing the majority of ATP. Cancer cells, with respect to their normal counterparts, demonstrate increased glycolytic capacity even in the presence of adequate oxygen supplies (Moreno‐Sánchez et al, 2007) providing energy as well as increased production of cellular building blocks such as amino acids and fatty acids (Vander Heiden et al, 2009).
RAB GTPases comprise the largest subfamily of small GTPases and play a master role in regulating intercellular vesicle transport and trafficking of proteins (Agarwal et al, 2009). Numerous studies have implicated aberrations in Rab‐regulated vesicle recycling in the initiation and progression of multiple cancer lineages (Mosesson et al, 2008). We have previously shown that RAB25 is a candidate oncogene located within the 1q22 amplicon in ovarian and breast cancers (Cheng et al, 2004). Rab25, which is primarily expressed in epithelial cells (Goldenring et al, 1993), is a member of the Rab11 subfamily with its closest homologues Rab11a and Rab11b (Agarwal et al, 2009). Functional studies demonstrate that the Rab11 subfamily is involved in transcytosis, endocytic sorting and transport across polarized epithelial cells as well as in apical vesicle recycling (Tzaban et al, 2009).
RAB25 is amplified and over‐expressed in many human cancers. Increased RAB25 mRNA levels have been reported in ovarian, prostate cancer, transitional cell carcinoma of the bladder, invasive breast cancer, liver cancer and Wilms tumour (Agarwal et al, 2009), suggesting a pathological role of Rab25 in development or progression of multiple tumour lineages. In addition, we and others have demonstrated that Rab25 increases anchorage‐independent growth, survival in response to growth factor deprivation, suppresses apoptosis and increases tumour development, while down‐regulation of Rab25 by RNAi transfection significantly inhibits ovarian cancer growth in vivo (Cheng et al, 2004; Fan et al, 2006). Recent studies demonstrated that expression of Rab25 in rat intestine epithelial cells leads to transformation and cancer development (Lapierre et al, 2011). Further, Rab25 genomic amplification and increased mRNA expression are both associated with poor prognosis in a subset of ovarian and breast cancers (Cheng et al, 2004). Together, these data strongly implicate Rab25 in tumour initiation and aggressiveness. However, Rab25 has also been suggested to act as a tumour suppressor in claudin‐low breast cancers and in bowel epithelium (Cheng et al, 2010; Nam et al, 2010). Thus, Rab25 may function as an oncogene or a tumour suppressor depending on the cellular context and the extracellular environment of the tumour.
Analysis of transcriptional profiles has emerged as a powerful tool in elucidating the molecular mechanisms by which a genetic aberration contributes to cancer development (Nevins & Potti, 2007). To explore the mechanisms by which Rab25 expression alters cancer cell function, we identified a RAB25 transcriptome in vitro. Importantly, the Rab25 signature is portable from cell lines allowing interrogation of tumour samples. The transcriptome was highly enriched in genes implicated in cellular metabolism indicating a novel role of Rab25 in regulating cellular metabolism that we explore further herein.
The Rab25 transcriptome is enriched in metabolism‐associated genes leading to decreased cellular sensitivity to nutrient stress
Transcriptional profiling revealed a Rab25‐dependent gene expression signature containing 1128 probes representing 908 known genes [Fig S1A–C and Table S1 of Supporting information, raw data was submitted to Gene ExpressionOmnibus (GSE28299)]. Based on GeneOntology Biological Process Classification, metabolism‐associated genes constituted the largest category, with over 47% of the Rab25‐dependent gene expression signature being implicated in cellular metabolism, including lipid, amino acid and carbohydrate metabolism (Fig S1D and Table S2 of Supporting information). Based on this potential novel role for Rab25 in regulating cellular metabolism and the increased glycolytic dependence of cancer cells, it was of particular interest to evaluate whether Rab25 alters cell survival during nutrient stress by culturing cells under glucose (deprivation or by addition of 2‐deoxyglucose; 2DG) or amino acid (EBSS) deprivation conditions. As growth factors are required for efficient uptake and utilization of nutrients (Edinger & Thompson, 2002), growth factor deprivation provides a potent mimic of decreased cellular energy balance and increased cellular stress.
Expression of RAB25 (to levels present in tumours with amplified RAB25) in ovarian cancer A2780 and HEY cells or IOSE29ht immortalized normal ovarian epithelium cells, all of which express low endogenous levels of RAB25, significantly decreased cell death (p < 0.01) under growth factor and nutrient deprivation conditions, as measured by DNA fragmentation and flow cytometry of fluorescent‐tagged deoxyuridine triphosphate nucleotide incorporation (Fig 1A and B). In contrast, decreasing Rab25 expression with siRNA sensitized cells to nutrient withdrawal‐induced cell death (Fig 1B). To further evaluate the physiological relevance of these findings, we examined the effect of decreasing Rab25 expression in ovarian OVCAR3 and breast MCF7 cancer cell lines, which express high endogenous Rab25 levels (Cheng et al, 2005). Stable expression of shRNA specific to Rab25 significantly reduced the expression of Rab25 protein in these cells (Fig S2 of Supporting information). Similar to the results obtained from HEY cells, reducing Rab25 expression by shRNA led to increased cell death after nutrient withdrawal in both OVCAR3 and MCF7 cells (Fig 1B).
Under conditions of reduced nutrient availability most cells undergo autophagy, degrading cytoplasmic organelles to provide substrates for energy metabolism (Meijer & Codogno, 2004). To elucidate the contribution of autophagy to Rab25‐mediated resistance to metabolic stress in cancer cells, the formation of autophagosomes, the final step in the autophagy cascade, was assessed by measuring the levels of the endogenous autophagosomal marker LC3, microtubule‐associated protein1 light chain 3, by Western blotting (WB) after 4 and 6 h of nutrient withdrawal (Mizushima, 2004). Surprisingly, LC3‐II protein fragment levels were lower in Rab25‐expressing A2780 and HEY cells than in control pcDNA‐transfected cells after serum and glucose withdrawal (Fig 1C), suggesting a change in autophagic activity. Consistent with the WB data, expression of Rab25 markedly decreased autophagosome formation (Fig 1D) in Rab25‐expressing A2780 cells (29 ± 7 autophagosome/cells) compared to control cells (70 ± 11 autophagosomes/cells) after both glucose and serum withdrawal for 4 h (p < 0.00057) as assayed by electron microscopy (EM), a quantitative and definitive method for detection of autophagy (Mizushima, 2004). However, prolonged nutrient withdrawal (24 h) was sufficient to induce autophagy in Rab25‐transfected cells to levels similar to those in parental cells, indicating that the autophagic machinery is not compromised by the expression of Rab25 but rather that autophagy induction is delayed.
The Rab25‐mediated reduction in autophagy was noted in multiple cell lines including human osteosarcoma U2OS cells (Fig S3A of Supporting information), suggesting the effects of Rab25 on autophagy are generalizable. Autophagy requires the coordinate activity of multiple ‘autophagy‐related’ proteins (Atg) including Atg6/Beclin1 that has been implicated in human breast, ovarian and prostate tumours cancer (Karantza‐Wadsworth et al, 2007). Atg6/Beclin1 heterozygous mutant mice are tumour‐prone implicating autophagy in tumourigenicity (Jin & White, 2007). However, expression of Atg6/Beclin1 was not significantly changed in response to Rab25 expression (Fig S3B of Supporting information).
5′‐AMP‐activated protein kinase (AMPK) is a major physiological sensor of intracellular energy levels. An increase in intracellular AMP/ATP ratio activates AMPK to maintain cellular energy balance (Kahn et al, 2005). As expected, following nutrient withdrawal, AMPK phosphorylation (pAMPK) increased as measured by both reverse phase protein arrays (RPPA; Fig 1E) and WB analysis (Fig 1F). However, the increase in pAMPK levels was markedly lower in Rab25‐transfected cells following glucose and growth factor withdrawal (Fig 1E). AMPK, when activated, also phosphorylates acetyl CoA carboxylase (ACC) resulting in inhibition of energy‐consuming fatty acid synthesis. AMPK phosphorylation levels paralleled ACC phosphorylation (pACC), as measured by RPPA (Fig 1G) and WB (Fig 1F). In keeping with the effect of Rab25 on AMPK phosphorylation, ACC phosphorylation was markedly decreased in cells expressing Rab25 (Fig 1F). Furthermore, Rab25 knockdown using RNAi sensitized cells to nutrient stress as indicated by increases in both basal‐ and nutrient deprivation‐induced AMPK and ACC phosphorylation (Fig 1H).
Rab25 expression alters ATP levels
How does Rab25 expression reduce bioenergetic stress and delay the onset of autophagy under nutrient deprived conditions? Unexpectedly, Rab25 expression in cells grown in complete medium (RPMI1640 + 5% FBS) results in a consistent and statistically significant elevation of basal ATP levels compared to pcDNA‐transfected cells (Fig 2A). The effects of Rab25 on ATP levels are even more striking during nutrient stress. Expression of Rab25 in A2780 or HEY ovarian cancer cells results in maintenance of higher ATP levels in the absence of glucose and FBS (growth factor deprivation) over at least 6 h and to a lesser degree over 24 h of culture (Fig 2B). Hence, the elevated ATP level in Rab25‐expressing cells likely lowers the AMP/ATP ratio leading to decreased AMPK activation. Compatible with a role for Rab25 in maintaining ATP levels under nutrient stress, down‐regulation of Rab25 in cells expressing endogenous Rab25 by shRNA (Fig 2C) or by RNAi (Fig S4A of Supporting information) significantly decreased ATP levels (p < 0.01).
The ability of Rab25 to maintain ATP levels following nutrient withdrawal without activation of autophagy could be due to the utilization of internal nutrient reserves such as glycogen, triglycerides or proteins, which can be rapidly metabolized for energy production (Greenberg et al, 2006). Cellular glycogen but not triglyceride or protein levels (data not shown) were significantly higher in Rab25‐expressing cells grown in complete medium (Fig 2D). This is striking as glycogen has not been proposed to represent a major component of energy stores in epithelial cancer cells. Down‐regulation of Rab25 expression using siRNA (left panel) or shRNA (right panel) reduced basal glycogen levels suggesting a role for Rab25 in controlling cellular glycogen production or storage. Following nutrient deprivation, glycogen levels fell rapidly, compatible with conversion to glucose for ATP synthesis (Fig 2E), suggesting that glycogen may contribute to the maintenance of cellular energy levels during metabolic stress in Rab25 expressing cells (Fig 2B).
Glycogen is converted to glucose‐1‐phosphate by glycogen phosphorylase (GP) and then to glucose‐6‐phosphate before entering the glycolytic pathway to produce ATP (Roach, 2002). We thus determined the effect of preventing conversion of glycogen to glucose‐1‐phosphate, by GP inhibition, on ATP levels. A2780 ovarian cancer cells expressing Rab25 or pcDNA‐transfected cells were pretreated with 30 µM Bay U6571 GP inhibitor (GPi) or 200 ng/ml oligomycin (an ATP synthase inhibitor) for 4 h in complete media before nutrient withdrawal. Incubation of cells with the GPi in complete media did not significantly alter basal ATP levels (Fig 2F), presumably due to the high concentration of glucose in the media that can be utilized as an energy source. Strikingly, the presence of GPi essentially abrogated the ability of Rab25 to maintain cellular ATP levels after glucose deprivation (Fig 2F), indicating that the conversion of endogenous glycogen to glucose‐1‐phosphate plays a critical role in the ability of Rab25 to maintain elevated ATP levels during glucose deprivation. Blocking mitochondrial ATP production by oligomycin significantly decreased cellular ATP levels under all conditions suggesting that a significant component of the ATP production in Rab25‐expressing cells is through oxidative phosphorylation (Fig 2F). The ability of the GPi to block glycogenolysis and decrease ATP levels during glucose deprivation was confirmed in Rab25‐expressing HEY (Fig 2G) and MCF7 cancer cells (Fig S4B of Supporting information). Thus the elevated glycogen stores induced by Rab25 appear to maintain ATP levels during the early phase of metabolic stress. In agreement with the changes in cellular ATP levels, blocking glycogen catabolism by pretreating cells with the GPi elevated ACC phosphorylation above control levels in the absence of the GPi (Fig S4C of Supporting information). Functionally, addition of the glycolysis inhibitor 2DG induced cell death in both OVCAR3 and MCF7 cells (Fig 2H). Importantly, reducing Rab25 expression by shRNA‐mediated knockdown increased cell death induced by low dosage 2DG (2–10 µM). Thus, blocking glycogenolysis and subsequent glycolysis reduce the ability of Rab25‐expressing cells to maintain survival during early nutrient stress, implicating utilization of glycogen as an important energy source (Fig 2H).
Rab25 regulates cancer cell bioenergetics through an AKT‐dependent pathway
Glycogen synthase (GS), which converts glucose into glycogen for storage, is inhibited by phosphorylation at several sites by GS kinase‐3 (GSK3; Woodgett, 1994). GSK3, in turn, is inhibited by phosphorylation by AKT following PI3K pathway activation. We have previously reported that Rab25 expression is associated with a concomitant increase in AKT activation through a yet unexplored mechanism (Cheng et al, 2004). Consistent with our previous observations, AKT S473 phosphorylation, reflecting AKT activation, was markedly increased in Rab25‐expressing cells leading to increased GSK3 phosphorylation and consequent decreases in GS phosphorylation in the presence or absence of serum (Fig 3A). To determine whether the activation of AKT and subsequent inhibition of GSK3 and consequent activation of GS play a role in determining intracellular glycogen levels, we examined the effect of inhibition of PI3K or GSK3 by PI103, a potent and selective inhibitor of class I PI3K (Raynaud et al, 2007) or GSK3i (SB‐216763), respectively, on cellular glycogen levels. Cells were placed in complete media containing PI103 or GSK3i for 2 h to allow PI103 or GSK3i to enter cells, followed by a 2 h incubation in glucose‐ and foetal bovine serum (FBS)‐free medium to induce a reduction in endogenous glycogen levels. WB demonstrated the efficacy of PI103 and GSK3i in inhibiting phosphorylation of AKT and GSK3, as well as decreasing phospho‐GS levels (Fig 3B). The GPi Bay U6571 was included as a control for the specificity of effects of PI103 and GSK3i.
Treating A2780 pcDNA‐transfected ovarian cancer cells with GSK3i, thus releasing inhibition of GS activity as a consequence of GSK3‐mediated phosphorylation, increased cellular glycogen content to levels observed in Rab25‐expressing cells (Fig 3C). In contrast, GSK3i had minimal effects in Rab25‐expressing cells compatible with GSK3 inhibition as a consequence of Rab25‐induced AKT activation (Fig 3C). Similar results were observed in HEY and IOSE80ht cells where GSK3i increased glycogen levels (p < 0.05) in pcDNA‐transfected cells, whereas, a significant change in glycogen levels was not observed in Rab25‐expressing cells (Fig 3C). Thus, increased AKT activity with subsequent inhibition of GSK3 in Rab25‐expressing cells likely contributes to increased GS activity and glycogen accumulation.
PI103 did not alter the modest glycogen accumulation that occurred in pcDNA‐transfected HEY or IOSE80ht cells returned to nutrient‐ and FBS‐replete media (Fig 3D). In both HEY and IOSE80ht cells, Rab25 expression markedly increased the accumulation of glycogen in cells released from starvation (p < 0.01). When cells were subjected to a second nutrient stress, glycogen levels fell rapidly (Fig 3D) suggesting that the glycogen stored during the short incubation in glucose‐ and serum‐replete media provides an immediate source of energy. The role of the PI3K–AKT pathway in maintaining cellular ATP levels in Rab25 expressing cells was assessed using PI103. The ability of Rab25 to induce glycogen accumulation (Fig 3D) and maintain ATP levels (Fig 3E) under stress was compromised in the presence of PI103, compatible with Rab25 increasing glycogen stores and maintaining ATP levels during nutrient stress through Rab25‐mediated PI3K pathway activation (Fig 3D and E).
These observations were further evaluated in OVCAR3 and MCF7 cells using shRNA‐mediated stable knockdown of Rab25 expression, which decreased basal glycogen levels (Fig 3F and Fig S5A of Supporting information). The specific AKT inhibitor (MK2206) significantly decreased glycogen levels further and indeed overrode the effects of Rab25 shRNA supporting a role for AKT activation in Rab25‐induced glycogen accumulation (Fig S5A of Supporting information). When cells were cultured in serum‐free conditions, wherein, AKT activation by exogenous stimulation is absent, cells expressing Rab25 maintained elevated basal glycogen levels, whereas, Rab25 shRNA induced a decrease in glycogen levels (Fig 3F). MK2206 markedly decreased glycogen levels and abrogated the effect of Rab25 shRNA (Fig 3F). Similar results were obtained in T47D cells treated with Rab25 siRNA or with MK2206 (Fig S5B of Supporting information).
Culturing OVCAR3 and MCF7 cells in serum‐free medium led to a decrease in cell survival (Fig 3G). Knocking down Rab25 levels augmented the cell death (Fig 3G). Inhibition of PI3K or AKT activity with PI103 or MK2206, respectively, resulted in similar levels of cell death regardless of the presence of Rab25 shRNA supporting a Rab25/PI3K/AKT‐dependent mechanism for Rab25‐mediated cell survival (Fig 3G). Addition of lonidamine (LND) or 3‐bromopyruvate (3BrPy), two potent glycolysis inhibitors (Pelicano et al, 2006), also significantly decreased cell viability supporting the dependence on glucose as an energy source for cell survival.
Cellular glycogen synthesis depends on intracellular glucose. Glucose uptake, assessed using tritium‐labelled 2DG in cells maintained in Krebs–Ringers phosphate solution (i.e. growth factor and glucose depleted environment), was significantly higher in Rab25‐expressing cells than in parent cells (Fig 4A). Similarly down‐regulation of Rab25 decreased glucose uptake (Fig 4B). AKT activation has been demonstrated to increase glucose uptake (Tong et al, 2009). Indeed, treating Rab25 over‐expressing HEY cells with MK2206 effectively inhibited cellular glucose uptake regardless of the presence or absence of Rab25 (Fig 4B). Similarly, AKT‐specific siRNA (Fig S5C of Supporting information) decreased glucose uptake, supporting a critical role for AKT in Rab25‐induced glucose uptake. MK2206 further reduced glucose uptake (Fig 4B) in the presence or absence of Rab25 or AKT siRNA further supporting a role for AKT Rab25 function. Likewise, knockdown of endogenous Rab25 or inhibiting AKT activity in OVCAR3 and T47D cells using specific shRNA or MK2206, respectively, significantly reduced glucose uptake (p < 0.001) confirming a role for AKT in regulation of glucose uptake by Rab25 (Fig 4C).
In rat cardiomyocytes, Rab11, the closest Rab family member to Rab25 (by sequence homology), colocalizes with glucose transporter (GLUT) 4‐containing vesicles and redistributes to the plasma membrane with GLUT4 (Watson et al, 2004). We, therefore, reasoned that Rab25 could contribute to GLUT trafficking. As GLUT1 is the primary GLUT expressed in ovarian cancer cells (Rudlowski et al, 2004), we examined the localization of GLUT1 and Rab25 by immunofluorescence staining under nutrient stress and nutrient‐replete conditions in A2780 cells (Fig 4D). When cells were cultured in the absence of glucose and growth factors, GLUT1 co‐localized with Rab25 in a perinuclear distribution (Fig 4D). Addition of FBS as a source of growth factors and glucose for 5 min resulted in redistribution of a portion of GLUT1 from a perinuclear location to the cytoplasmic membrane. Importantly, Rab25 colocalized with GLUT1 during the translocation process (as indicated by arrows) suggesting that GLUT1 may represent a Rab25 cargo.
High AKT activity has been reported to render cells more susceptible to death induced by glucose withdrawal or treating with a glycolysis inhibitor (Elstrom et al, 2004; Fan et al, 2010; Kurtoglu et al, 2007). Thus, Rab25 expression, with subsequent increases in AKT activation and glucose uptake, would be expected to increase sensitivity to 2DG‐induced cell death. However, as noted above, Rab25 expression decreased sensitivity of HEY cells to low doses of 2DG (2–10 µM), whereas, high doses of 2DG (20–40 µM) led to similar levels of cell death in parental and Rab25‐transfected HEY cells (Figs 2H and 4E). Decreasing expression of Rab25 or activity of AKT by siRNA specific to Rab25 (Figs 2H and 4F) or MK2206 (Fig 4E), respectively, increased cellular sensitivity to 2DG‐induced cell death. Interestingly, 2DG has been reported to activate the AKT signalling pathway and protect cancer cells from death (Zhong et al, 2008, 2009). To further evaluate the relationship between Rab25 and AKT in the response to 2DG, we examined AKT activation in OVCAR3 and MCF7 cells. 2DG induced an increase in AKT phosphorylation in both cell lines in a dose‐ and time‐dependent manner similar to the results reported by Zhong and colleagues. Down‐regulation of Rab25 levels reduced AKT activation, compatible with Rab25 regulating AKT activation (Fig S5D and E of Supporting information). As expected, culturing OVCAR3 or MCF7 cells in 2DG decreased cell viability, a process that was augmented by siRNA to Rab25 (Fig 4G). Importantly, the addition of AKT or PI3K inhibitors further decreased cellular viability independently of the presence or absence of Rab25 siRNA compatible with the PI3K/AKT pathway mediating the effects of Rab25 on cellular viability (Fig 4G).
Interaction of Rab25 with AKT is required for effects of Rab25 on cellular metabolism
A protein complementation‐based assay (see Materials and Methods Section of Supporting information for details), wherein, reconstitution of fluorescence is dependent of protein–protein binding bringing two fragments of a GFP‐like molecule into proximity, demonstrated the association of Rab25 with AKT (Fig 5A). Transient transfection of: (1) AKT‐IFPN (intensely fluorescent protein amino‐terminal half) and RAB25‐IFPC (intensely fluorescent protein carboxy‐terminal half) or (2) AKT‐IFPC and RAB25‐IFPN fusion proteins in HEY ovarian cancer cells resulted in fluorescence indicating association of AKT and Rab25 protein in the PCA complex (Fig 5A upper panel, double transfected cell indicated by arrow). No fluorescent cells were observed in other control conditions including (AKT‐IFPC and IFPN), (AKT‐IFPN and IFPC) as well as (Rab25 IFPC and ACTN4‐IFPN) indicating a specific binding of Rab25 with AKT (Fig S6 of Supporting information). We further confirmed the association of Rab25 with AKT using PCA in AKT‐IFPN stably expressing HeLa cells by transient transfection of a Rab25‐IFPC construct (Fig 5A, lower panel). 3‐phosphoinositide‐dependent protein kinase‐1 (PDK1), a well‐established AKT binding partner, was included as a positive control. In support of an interaction between AKT and Rab25, AKT and Rab25 could be co‐immunoprecipitated (IP) in ovarian cancer cells expressing Rab25 (Fig 5B).
To elucidate the molecular mechanism by which Rab25 interacts with AKT, Rab25 deletion mutants were designed (Fig 5C). Expression of these deletion mutants was demonstrated by WB and immunofluorescence staining using anti‐GFP antibody (Fig 5C). Wild‐type Rab25‐YFP localizes to the perinuclear region similar to endogenous Rab25 (Fig 5C). Deletion of 43 amino acids from the C‐terminal (Rab25‐C‐del) of Rab25, containing the putative geranylgeranyl transferase II acceptor motif CCISL, results in a diffuse distribution of Rab25 throughout the cytosol, while addition of a CCISL motif to Rab25‐C‐del (Rab25 C‐del‐CC) restores the normal Rab25 perinuclear distribution implicating the CCISL motif in Rab25 localization. Deletion of 41 amino acids from the N‐terminus of Rab25 (Rab25 N‐del) did not affect localization (Fig 5C). However, no observable fluorescent complex formed when any of the mutants were coexpressed with AKT as assayed by the PCA methodology (data not shown). Thus, a full length Rab25 able to localize to the correct compartment appears necessary to interact with AKT. Strikingly, transient expression of each of the mutants that are unable to bind AKT, in contrast to full length Rab25 YFP that binds AKT, into ovarian cancer cells did not alter cellular glycogen levels or glucose uptake (Fig 5D), supporting the contention that interaction of Rab25 with AKT is likely required for the ability of Rab25 to alter glucose uptake and glycogen storage.
Elevated Rab25 levels in patient samples are associated with altered cellular bioenergetics
To determine whether the effects of Rab25 on cellular stress observed in cell lines also occurred in human tumours, we measured Rab25, total AKT, total AMPK, total ACC, total GS, pAKT (T308), pAMPK, pACC and pGS protein levels in tumour lysates from 667 ovarian cancer specimens using RPPA. As expected, due to ACC being an AMPK target, a positive correlation (Spearman, p = 1.007161e−32) was observed between pAMPK and pACC protein levels in ovarian cancer tissues. This positive correlation also served to validate the ability to accurately assess pAMPK and pACC in patient samples. Recapitulating the in vitro data, there was a highly significant inverse correlation between Rab25 levels and pAMPK (p = 0.00015), pACC (p = 7.93e−12) and pGS (p = 0.00001) as well as a positive correlation (p = 1.34e−9) between Rab25 and pAKT in patient samples. We did not examine the potential correlation between pGSK3 and Rab25 as available phosphoGSK3 antibodies detect both alpha and beta forms of GSK3 as well as p90RSK mitigating its utility in RPPA. Furthermore, we observed a positive correlation (p = 0.032) between cellular glycogen content and expression of Rab25 in 31 ovarian patient tumour samples (Fig 6A). Thus, the effects of Rab25 on cellular metabolism in vitro are recapitulated in ovarian cancer in the patient.
To assess the potential clinical relevance of the Rab25‐dependent transcriptional profile and the effect of Rab25 on cellular metabolism, we classified serous ovarian cancers into Rab25‐like and pcDNA‐like based on their expression patterns in two large‐independent publically available datasets (Dressman et al, 2007 and Tothill et al, 2008, see External Datasets Section of Supporting information for details) using approaches described previously (Lee et al, 2006). Using linear discriminant analysis, leave‐one‐out cross validation, and cell lines with and without Rab25 expression as the training set, a Rab25‐dependent gene signature – consisting of 62 probes for 53 genes (Table S3 of Supporting information) from the original 1167 probesets (Fig S1 of Supporting information) – was used to classify ovarian cancers as Rab25‐like in the Tothill et al (Fig 6B) and Dressman et al (Fig 6C) datasets. Rab25‐like cancers were associated with significantly higher rates of relapse (Fig 6B) and poorer survival (Fig 6C). Furthermore, the Rab25 signature was an independent prognostic factor in multivariate analysis (Fig 6B and C), in keeping with a role for Rab25 in modulating aggressiveness of ovarian cancer. We previously demonstrated that Rab25 levels alone predict survival only in the subset of patients with very high and very low levels of Rab25 (Cheng et al, 2004). In contrast, the Rab25 signature was able to predict outcomes for all ovarian cancer patients indicating that the Rab25‐dependent signature contains important information related to ovarian cancer pathophysiology not captured by Rab25 levels alone. Intriguingly, this signature is not only highly portable from cell lines to clinical ovarian tumour samples, but also proved to be a strong predictor of survival in breast cancer samples (Fig 6D) using two‐independent datasets (Pawitan et al, 2005 and Chin et al, 2006, see External Datasets Section of Supporting information for details), further supporting the role of Rab25 in aggressiveness of cancers.
Collectively, our data implicate Rab25 in glucose and glycogen metabolism. The effects of Rab25 appear to be mediated, at least in part, through an interaction between Rab25 and AKT leading to increased glucose uptake, which combined with AKT‐induced inhibition of GSK3 and increase in GS activities results in higher basal glycogen and elevated ATP levels. The increased glycogen stores, in turn, provide an alternative energy source during early nutrient stress reducing sensitivity to nutrient stress and increasing cell survival (Fig 6E).
RAB25 is amplified as part of the 1q22 amplicon in multiple cancer lineages. In support of an oncogenic function for Rab25, we have previously demonstrated worsened outcomes for ovarian and breast cancer patients whose tumours have markedly elevated Rab25 levels (Cheng et al, 2004). Further, increased Rab25 levels markedly increased tumour growth in xenograft models. Similarly, knockdown of Rab25 expression by RNAi decreases growth of ovarian cancer xenografts (Fan et al, 2006). Rab25 promotes α5β1 integrin‐containing vesicle trafficking to the membrane of pseudopodal tips, thereby, facilitating cell invasion (Caswell et al, 2007) potentially contributing to the aggressiveness of ovarian cancer. A recent study further demonstrated that overexpression of Rab25 in rat intestinal cells can lead to microtubule‐dependent transformation in vitro and tumour formation in vivo (Lapierre et al, 2011). However, other studies have suggested that Rab25 may act as a tumour suppressor under some contexts. Loss of Rab25 in some triple‐negative cell lines (Cheng et al, 2010) promotes tumourigenicity, and a recent study has reported an increase in colonic tumour incidence in Rab25‐deficient mice bred into an APCmin/+ genetic background (Nam et al, 2010). Indeed, we have previously demonstrated that Rab25 is not sufficient to transform immortalized but non‐tumourigenic cells (Cheng et al, 2004) and similarly, Rab25‐knockout mice do not exhibit significant anatomical and pathological abnormalities (Nam et al, 2010). Consequently, Rab25 may not play a role in tumour initiation either as a primary oncogene or tumour suppressor gene, but rather may function as a regulator of tumour progression in already transformed cells. Thus, the effects of Rab25 appear to be context‐dependent. Rab25 may exert different effects dependent on the intrinsic expression patterns in different cell lineages, the presence of different underlying genomic aberrations, the levels of specific Rab25 molecular partners, the cargoes being transported in Rab25‐containing vesicles or the response of cells to activation of the signal transduction pathways controlled by Rab25. Context‐dependent effects on tumour initiation and progression are not without precedence with both autophagy (Eisenberg‐Lerner & Kimchi, 2009) and TGF beta (Chaudhury & Howe, 2009) demonstrating tumour inhibiting and promoting activity in different circumstances. In the present study, we manipulated Rab25 levels with a combination of enforced expression and knockdown to elucidate the mechanisms underlying Rab25 function. We identified a Rab25‐associated gene signature using A2780 ovarian cancer cells in vitro. The Rab25 transcriptome was capable of identifying ovarian cancer patients with a poor prognosis with high confidence. Importantly, we observed a novel role of Rab25 in cellular survival and, in particular, in regulating cellular metabolism.
Increased glucose consumption is a characteristic of malignant cells. However, early studies noted in vivo that the amount of glucose uptake in tumours was too high to be utilized solely through glycolysis (Gullino et al, 1967). Our current data show that increased Rab25 expression decreases sensitivity to nutrient stress. Rab25‐expressing cells demonstrated an unexpected and unprecedented increase in cellular glycogen stores. The elevated glycogen content in cells with high Rab25 levels is recapitulated in vivo in patient tumour samples. Glycogen stores have not previously been proposed as a viable energy source that could contribute to the ability of epithelial cancer cells to withstand nutrient stress. Indeed other than clear cell tumours where glycogen stores are prominent, the ability to store and utilize glycogen as an energy source has not been implicated in epithelial tumourigenesis and represents a novel observation in the context of cancer biology.
Our data suggest that Rab25 may increase glycogen stores and ATP levels, in part, by increasing glucose uptake. GLUT1 is over‐expressed in multiple solid tumours and high GLUT1 levels have been associated with poor patient survival (Rudlowski et al, 2004). Strikingly, Rab25 colocalizes with GLUT1 suggesting a potential role for Rab25 in regulating translocation of GLUT1 to the membrane thus enhancing glucose uptake. Rab25‐mediated activation of AKT may further enhance glucose uptake. The increased glucose uptake under nutrient‐rich conditions in Rab25‐expressing cells is associated with elevated glycogen stores, the mobilization of which is sufficient to explain, at least in part, the maintenance of ATP levels during acute bioenergetic stress.
GSK3 is a key and potent regulator of GS activity. Expression of Rab25 increased AKT activity (Cheng et al, 2004), which led to increased GSK3 phosphorylation, and subsequent inhibition of GSK3 activity as demonstrated by decreased GS phosphorylation. Together, this cascade provides an additional mechanism by which Rab25 could increase glycogen synthesis resulting in glycogen accumulation. Indeed, inhibition of PI3K decreased the ability to Rab25 to induce glycogen accumulation and inhibition of GSK3 recapitulated the ability of Rab25 to increase glycogen storage.
A full length functional Rab25 including the C‐terminal geranylgeranylated motif is required to regulate cancer cell bioenergetics, at least as compared to either N‐ or C‐terminal deletion mutants. Newly synthesized Rab proteins must be geranylgeranylated at C‐terminal cysteine residues prior to anchoring to cellular membranes (Pereira‐leal & Seabra, 2000). In agreement with this requirement, deletion of the C‐terminus of Rab25 (Rab25‐C‐del) including the putative geranylgeranylated motif (CCISL) led to mislocalization of Rab25. Addition of the CCISL motif to the C‐terminal truncation mutant was sufficient to restore normal Rab25 localization. Deletion of the N‐terminus of Rab25 did not disturb subcellular localization compatible with the CCISL motif playing a key role in Rab25 localization. However, none of the Rab25 mutants retained the ability to interact with AKT, or to increase ATP, glycogen content and glucose uptake compatible with these processes being dependent on AKT activation. It is likely that the deletion mutants of Rab25 caused a change in availability of important functional motifs. For instance, the Rab25‐N‐del mutant lacks several functional domains, including RabSF1, RabSF2 and RabF1, the structural β‐sheet and α‐helix as well as 2‐conserved G box (GDP/GTP‐binding motif elements) that are implicated in GTP and downstream effector binding (Colicelli, 2004; Itzen & Goody, 2011; Pereira‐leal & Seabra, 2000). Likewise, the deletion of the C‐terminus of Rab25 deletes domains that interact with the Rab escort protein and Rab GDP‐dissociation inhibitor (Itzen & Goody, 2011; Pereira‐leal & Seabra, 2000). Additional fine mapping may further elucidate the functional domains of Rab25 required for the interaction of Rab25 with AKT and functional effects of Rab25. Overall, the coordinate effects of Rab25 mutation on AKT binding and ATP levels, glycogen content and glucose uptake suggests that the ability of Rab25 to bind and regulate AKT plays a critical role in the ability of Rab25 to regulate cellular bioenergetics, a contention supported by the PI3K, AKT and GSK3 inhibitor studies. In particular, it is worth noting that the addition of PI103 or MK2206 abolished the ability of Rab25 to increase cellular viability supporting the dependence of Rab25 activity on a functional PI3K/AKT pathway.
It has been reported that cells with high level of AKT activity are more sensitive to glucose withdrawal or glycolysis inhibition induced cell death induced (Elstrom et al, 2004; Fan et al, 2010; Kurtoglu et al, 2007). Inhibiting glycolysis by 2DG has been demonstrated to promote nutrient stress‐induced cell death and to enhance the therapeutic efficacy of chemotherapeutic drugs (Simons et al, 2009). Interestingly, we have observed a decrease sensitivity of Rab25 expressing cells to low doses of 2DG (2–10 µM) despite elevated AKT activity in these cells. Hence, it is possible that under the submaximal dose of 2DG (10 µM), Rab25 may alter metabolism through pathways parallel to Akt to influence the outcome of cellular responses. Further, the discrepancy may also due to the nature of different cells or systems used in these studies. In addition, other studies have suggested that inhibition of glycolysis is not sufficient to explain the actions of 2DG on cell survival (Ralser et al, 2008). Further, 2DG has been shown to activate AKT and prevent cell death (Zhong et al, 2008; Zhong et al, 2009). Indeed, lung cancer cells with low levels of AKT activity were more susceptible to 2DG‐induced cell death (Wangpaichitr et al, 2008). In the present study, we observed an increase in AKT activation in Rab25‐expressing cells after 2DG administration, and inhibition of AKT by MK2206 increased cell death in response to 2DG, implicating AKT in preventing 2DG‐induced cell death. AKT has been suggested to protect cells from 2DG‐induced cell death by negatively regulating AMPK activity (Hahn‐Windgassen et al, 2005; Memmott & Dennis, 2009; Priebe et al, 2011). Compatible with this observation, pAMPK and pACC are decreased in the presence of Rab25. Activation of mitochondrial‐bound hexokinase (Gottlob et al, 2001), or by inhibiting N‐linked glycosylation under normoxia, independent from glycolytic inhibition (Kurtoglu et al, 2007) have also been proposed as alternative mechanisms by which AKT protects cells from 2DG‐induced death. Together, these studies suggest that a complex intracellular response governs the outcome of 2DG cytotoxicity.
It is noteworthy that expression of Rab25 delays induction of autophagy, which can function as an alternative form of non‐apoptotic cell death under conditions of nutrient deprivation. The ability of Rab25 to increase basal ATP levels and to maintain ATP levels under early glucose‐ and serum‐deprived conditions likely plays a major role in preventing the induction of autophagy. Defects in autophagy have been associated with increased tumourigenesis (Degenhardt et al, 2006; Liang et al, 1999). Metabolic stress can drive tumour cells with both defects in autophagy and apoptosis into necrotic cell death (Degenhardt et al, 2006). Diverting cells to necrotic cell death induces inflammatory cell infiltration and cytokine signalling (Proskuryakov et al, 2003), which by analogy with the wound healing process, may facilitate cell proliferation and angiogenesis favouring tumour growth (de Visser et al, 2006). In addition, defects in autophagy may allow accumulation of chromosomal instability and genomic mutation, as demonstrated by the higher rates of gene amplification and DNA damage observed in renal (Mathew et al, 2007) and mammary epithelial cells in response to metabolic stress when both apoptosis and autophagy were suppressed (Karantza‐Wadsworth et al, 2007). However, following prolonged bioenergetic stress, ATP levels decrease in Rab25‐expressing cells, AMPK becomes activated and cells eventually enter autophagy. Thus, Rab25 improves cellular bioenergetics sufficiently to delay the decrease in intracellular energy levels that induce entry into autophagy.
In conclusion, high levels of Rab25 appear to confer a functional advantage to cancer cells allowing cellular survival under acute nutrient stress conditions. In addition, a significant proportion of ovarian and breast cancers display a Rab25‐associated gene expression signature, consisting primarily of genes involved in cellular metabolism, that predicts a poor outcome. Together, these data indicate that Rab25 induces resistance of cancer cells to nutrient stress, contributing to increased aggressiveness and poorer survival of patients with ovarian and breast cancer with high levels of Rab25, and identifies Rab25 as well as its functional consequences as potential therapeutic targets in tumours that express high levels of Rab25.
MATERIALS AND METHODS
Patient samples and cell culture
All patient samples and information were collected under IRB approved (LAB01‐144) and HIPPA compliant protocols in MD Anderson Cancer Center. All patient samples contained greater than 80% tumour on histology. HEY, A2780 and SKOV3 ovarian cancer cells were maintained in RPMI 1640 containing 5% FBS. Immortalized ovarian surface epithelial cells IOSE29ht and IOSE80ht were maintained in MCDB105/M199 (50:50) containing 5% FBS as described previously (Cheng et al, 2004). ON‐TARGET plus siRNAs specific to Rab25 and AKT were purchased from Dharmacon Inc. (Lafayette, CO). Rab25 shRNA was purchased from Origene (Rockville, MD).
Detection of autophagy
Autophagic vacuoles were detected by fluorescence microscopy labelled with LC3‐GFP constructs. Fluorescent images were captured by fluorescence microscopy (Nikon Eclipse TE 2000‐E, Germany) and analysed using the program IPLab (Scanalytic Inc). EM analysis of autophagy was carried out in High Resolution Electron Microscopy Facility at the MD Anderson Cancer Center core.
Detection of glucose uptake, cellular ATP levels, cellular glucose, glycogen and protein assays
Glucose uptake was measured by incubating cells with 2‐H3‐deoxyglucose. Cellular ATP level was determined using an ATP Bioluminescence Assay Kit CLS II (Roche Applied Science, Indianapolis, IN) and normalized to cellular protein levels. Total cellular protein content was assessed using a BCA assay. Glucose and glycogen levels were assessed using glucose assay kit (Cat # K606‐100) and glycogen assay kit (K646‐100), respectively, obtained from BioVision Inc (Mountain View, CA) according to the manufacturer's protocol.
Cell death assays
Apoptotic cells were measured using paraformaldehyde fixed cells with an APO‐DIRECT kit (Phoenix Flow Systems) with flow cytometry or the Cell Death Detection ELISAPLUS (Roche Applied Science) according to the manufacturer protocol. Cell viability was detected using CellTiter‐Blue® Cell Viability Assay (Promega, Madison, WI).
Protein expression profiling
Experiment results obtained were statistically evaluated by ANOVA or Student's t‐test. Differences were considered significant for two‐sided p < 0.05 or as designated. We estimated the survival probabilities using the Kaplan–Meier method, and compared survival curves using the log‐rank test. Statistical analysis was performed by GraphPad Prism V.4 (San Diego, CA) or SPSS 16.0 (Chicago, IL).
For more information
Gene expression analysis tool; BRB Array Tool:
Gene Annotation Enrichment analysis; DAVID Bioinformatics Resources:
Bioenergetics; Cellular Bioenergetics Forum:
Protein complementation Assay: PMCID: PMC1636760
Reverse phase protein array: PMID: 17041095, PMID: 17599930, PMID: 19102773, PMID: 19176552
The overall study was conceived and designed by KWC, RA and GBM, with important contributions from SM, JSL, MC and JWG; KWC, RA, SM, MC performed the experiments; KWC, RA and JSL analysed the data; JSL and JWG contributed reagents or analysis tools; KWC, RA and GBM wrote the paper.
Detailed facts of importance to specialist readers are published as ”Supporting Information”. Such documents are peer‐reviewed, but not copy‐edited or typeset. They are made available as submitted by the authors.
This work was supported by Breast SPORE Career Development Award and Department of Defense Breast Idea Award W81XWH‐06‐1‐0488 to KWC, Cancer Research UK Clinician Scientist Fellowship C2757/A5902 to RA, National Institutes of Health SPORE (P50‐CA83639) and PPG CA64602 to GBM and JWG, P30 grant CA16672‐28 and Cancer Center Core Grant CA16672 and Ovarian Cancer Research Fund to GMB.
Supporting information is available at EMBO Molecular Medicine online.
The authors declare that they have no conflict of interest.
FundingNational Institutes of Health SPORE P50‐CA83639
- Copyright © 2012 EMBO Molecular Medicine