Oncogenes can create metabolic vulnerabilities in cancer cells. We tested how AKT (herein referring to AKT1) and MYC affect the ability of cells to shift between respiration and glycolysis. Using immortalized mammary epithelial cells, we discovered that constitutively active AKT, but not MYC, induced cell death in galactose culture, where cells rely on oxidative phosphorylation for energy generation. However, the negative effects of AKT were temporary, and AKT-expressing cells recommenced growth after ∼15 days in galactose. To identify the mechanisms regulating AKT-mediated cell death, we used metabolomics and found that AKT-expressing cells that were dying in galactose culture had upregulated glutathione metabolism. Proteomic profiling revealed that AKT-expressing cells dying in galactose also upregulated nonsense-mediated mRNA decay, a marker of sensitivity to oxidative stress. We therefore measured levels of reactive oxygen species (ROS) and discovered that galactose-induced ROS exclusively in cells expressing AKT. Furthermore, ROS were required for galactose-induced death of AKT-expressing cells. We then confirmed that galactose-induced ROS-mediated cell death in breast cancer cells with upregulated AKT signaling. These results demonstrate that AKT but not MYC restricts the flexibility of cancer cells to use oxidative phosphorylation.
This article has an associated First Person interview with the first author of the paper.
Many cancers preferentially use glycolysis for survival and proliferation, even in the presence of oxygen, a phenomenon known as aerobic glycolysis or the Warburg effect. Increased glycolytic activity is thought to help satisfy the rapacious demands of highly proliferative cancer cells for biosynthetic precursors including lipids, proteins and nucleic acids. However, this altered metabolism can leave tumors vulnerable to metabolic disruptions such as starvation of substrates including glucose, asparagine, glutamine, methionine, serine and others (Chalishazar et al., 2019; Gao et al., 2019; Graham et al., 2012; Joly et al., 2020; Maddocks et al., 2013; Wise et al., 2008; Yuneva et al., 2007). Therefore, understanding the interplay between oncogenes and metabolism is essential to understand how to design therapeutic strategies targeting tumor metabolism (Galluzzi et al., 2013; Martinez-Outschoorn et al., 2017).
The altered metabolism of tumor cells has been directly linked to the same oncogenes that drive tumorigenesis. In breast cancer, both the phosphoinositide 3-kinase (PI3K)/AKT signaling pathway (herein AKT is referring to AKT1) and the transcription factor MYC are frequently hyperactivated (The Cancer Genome Atlas Network, 2012). Deregulated PI3K/AKT signaling can result from multiple mechanisms including PIK3CA mutation, PTEN loss or mutation, or high AKT3 expression. Hyperactivated PI3K/AKT signaling, which is prominent in the luminal subtype of breast cancer, results in increased glycolytic flux by altering the localization and activity of glycolytic enzymes including glucose transporters, hexokinase and phosphofructokinase (Simons et al., 2012). The oncogenic transcription factor MYC is typically hyperactivated by high-level DNA amplification, especially in the basal subtype of breast cancer. Like AKT, MYC also exerts broad effects on the metabolism of tumor cells including having roles in glycolysis, glutamine metabolism, nucleotide biosynthesis and other metabolic processes (Stine et al., 2015). However, although both AKT and MYC promote aerobic glycolysis, these oncogenes can exert differential effects on metabolism in prostate cancer and pre-B cells (Fan et al., 2010; Priolo et al., 2014).
The galactose culture system is a classic technique for shifting the metabolism of mammalian cells from glycolysis to aerobic respiration. In glucose-containing media, mammalian cells can utilize both glycolysis and oxidative phosphorylation (OXPHOS) to generate ATP. However, when galactose is substituted for glucose, cells must rely on OXPHOS for energy generation. Most mammalian cells exhibit a flexible metabolic state that can dynamically shift between glycolysis and OXPHOS. Because galactose culture forces oxidative metabolism, it has proved useful in the identification of inborn errors of metabolism (Iannetti et al., 2018; Robinson et al., 1992), drugs that redirect metabolism to glycolysis (Gohil et al., 2010), genes essential for OXPHOS (Arroyo et al., 2016), and the role of aerobic glycolysis in T cell effector functions (Chang et al., 2013). However, to our knowledge, the galactose culture system has not been used to investigate the metabolic vulnerabilities induced by oncogenes.
Here, we report that non-tumorigenic, immortalized mammary epithelial cells expressing constitutively active AKT (i.e. myristoylated AKT1; mAKT) undergo rapid cell death when switched from glucose to galactose culture. In contrast, cells expressing either RFP (i.e. negative control) or MYC readily proliferated in galactose. Cell death in AKT-expressing cells was, however, short-lived, and cells expressing AKT recommenced cell growth after ∼15 days in galactose culture. To understand the different phenotype of AKT-expressing cells, we performed metabolomic and proteomic analyses using mass spectrometry, and found evidence that AKT-expressing cells in galactose were dying due to oxidative stress. We therefore used the reactive oxygen species (ROS) scavenger catalase to validate that ROS were required for galactose-induced death of AKT-expressing cells. Importantly, we tested breast cancer cell lines with constitutively active PI3K/AKT signaling and found that these cells also exhibited ROS-mediated cell death in galactose culture. Taken together, our ‘multi-omic’ analysis has identified a metabolic state-dependent lethality, namely the reduced ability of cells with constitutive activation of PI3K/AKT signaling to switch between glycolysis and respiration.
AKT but not MYC temporarily induces cell death in oxidative culture
To investigate the effect of oncogenes on the ability of human cells to survive and proliferate in oxidative culture, we expressed either constitutively active AKT (i.e. myristoylated AKT1, or mAKT), the transcription factor MYC, or the negative control red fluorescent protein (RFP) in immortalized but non-tumorigenic MCF-10A human mammary epithelial cells. We then switched these cells from glucose culture to galactose culture, which forces mammalian cells to rely on oxidative phosphorylation rather than glycolysis for energy generation (Robinson et al., 1992). Following the switch to galactose culture, cells expressing either RFP or MYC grew about half as fast as compared to those in glucose culture but did not exhibit significant cell death (Fig. 1A; Fig. S1A,B). In contrast, cells expressing mAKT exhibited significant cell death and declining cell numbers for three passages in galactose culture. In the first 5 days following the switch from glucose to galactose, we confirmed that cells expressing mAKT but not RFP or MYC, declined in both viable cell number and overall viability (Fig. S1C–E). Staining with annexin V and propidium iodide (PI) additionally revealed that mAKT- but not MYC-expressing cells underwent necrotic cell death after 3 h of galactose culture (Fig. S1F).
Interestingly, after three passages (∼15 days) in galactose culture, mAKT-expressing cells (hereafter mAKT cells) reversed this phenotype and recommenced proliferation. However, even after resuming proliferation, mAKT cells exhibited significantly slower growth in galactose relative to glucose than did RFP- or MYC-expressing cells (Fig. 1B). Cell viability remained high in mAKT cells proliferating in galactose (Fig. S1B). Next, using liquid chromatography-mass spectrometry (LC-MS) metabolomics, we assessed the effect of galactose culture on aerobic glycolysis by measuring extracellular lactate secretion. We found that galactose culture significantly reduced the abundance of secreted lactate in both the short (24 h) and long term (∼5 passages or ∼25 days) for all three cell types (Fig. 1C). Notably, mAKT cells in glucose exhibited the highest rate of lactate secretion, consistent with reports that mAKT can increase aerobic glycolysis (Elstrom et al., 2004; Priolo et al., 2014). Next, we tested whether galactose culture affected the expression of the exogenous oncogenes and found that galactose culture induced a slight decrease in the amount of total AKT and AKT phosphorylated on serine 473 (pSer473-AKT) (Fig. 1D). Conversely, expression of MYC was unaffected by galactose culture. Quantitation of the ratio of pSer473-AKT to total AKT by both western blotting and phospho-flow demonstrated a ∼20% reduction in mAKT-expressing MCF-10A cells in galactose culture (Fig. 1E; Fig. S1G,H). Taken together, these data demonstrate that the oncoproteins mAKT and MYC exert different effects on the ability of non-tumorigenic MCF-10A cells to proliferate in oxidative culture conditions (i.e. galactose) at short times (<3 passages) but that cells expressing either oncoprotein can proliferate in oxidative culture at longer times.
We next tested whether MCF-10A mAKT cells that had acquired the ability to proliferate in galactose could retain this ability if temporarily removed from galactose. We thus switched RFP and mAKT cells from long-term galactose culture to glucose culture for 5 passages (∼25 days) and then compared the growth of these cells in galactose against cells which had never been exposed to galactose (i.e. naïve). RFP-expressing cells grew similarly regardless of whether they had been previously cultured in galactose (Fig. 1F). In contrast, mAKT cells that had been previously cultured in galactose exhibited an enhanced ability to grow in galactose. This demonstrates that MCF-10A mAKT cells selected for growth in galactose retain the ability to proliferate in galactose even after removal from galactose.
mAKT cells exhibit extensive metabolic adaptation in galactose culture
To elucidate the metabolic mechanisms regulating the differential phenotypes of RFP-, mAKT- and MYC-expressing cells in galactose culture, we profiled cells using stable isotope tracing LC-MS metabolomics (Fig. S2A). Because cells in galactose culture upregulate glutamine anaplerosis (Gohil et al., 2010; Reitzer et al., 1979), we first labeled cells with [U-13C]-L-glutamine for 24 h in either glucose, short-term galactose culture (24 h) or long-term galactose culture (∼5 passages) and then undertook LC-MS metabolomics (Fig. S2B and Table S1). Relative to RFP-expressing cells, mAKT- but not MYC-expressing cells cultured in glucose exhibited an increased percentage of M0 isotopomers for the TCA cycle intermediates citrate or isocitrate, aconitate, α-ketoglutarate, succinate, fumarate and malate (Fig. 2A,B; Fig. S3), indicating that mAKT cells were utilizing less glutamine to fuel the TCA cycle. Upon switching to galactose culture, all cells exhibited increased percentages of fully labeled isotopomers (e.g. M6 citrate/isocitrate), indicating increased flux of glutamine-derived carbon through the TCA cycle. Additionally, in galactose culture at both short and long times, all MCF-10A cells exhibited an increased percentage of M5 isotopomer in citrate/isocitrate, indicating increased reductive carboxylation flux. Taken together, [U-13C]-L-glutamine stable isotope tracing indicated that all cells increased glutamine oxidation and reductive carboxylation upon switching from glucose to galactose, but that mAKT cells experienced larger changes than either RFP- or MYC-expressing cells because of their more aerobic glycolytic basal state in glucose.
Next, we tested how MCF-10A cells expressing oncogenes adapt to galactose culture by labeling cells with [U-13C]-galactose in short-term (24 h) and long-term galactose culture (∼5 passages) followed by LC-MS metabolomics (Fig. S2C and Table S2). Examining the fractional incorporation of 13C, we found that mAKT cells exhibited a small but significant reduction in 13C fractional incorporation compared to RFP- and MYC-expressing cells in the glycolytic intermediates 3-phosphoglycerate (3PG) and phosphoenolpyruvate (PEP), indicating slower flux from galactose into glycolysis (Fig. 2C,D). At longer times, however, mAKT cells reversed this difference and exhibited 13C fractional incorporation to levels similar to RFP- and MYC-expressing cells, indicative of increased flux from galactose into glycolysis. In addition, we found that mAKT cells significantly increased the 13C fractional incorporation from galactose into the amino acids glutamate and alanine (Fig. 2E,F), TCA cycle intermediates, including citrate/isocitrate and fumarate (Fig. 2G,H), and the redox-buffering molecules reduced glutathione (GSH) and oxidized glutathione (GSSG) (Fig. 2I,J). In contrast, RFP- and MYC-expressing cells exhibited smaller increases in 13C fractional incorporation in glutamate, alanine, citrate/isocitrate and fumarate, and decreased 13C fractional incorporation into GSH and GSSG. Overall, the changes in galactose-derived 13C fractional incorporation for mAKT cells were much greater than for RFP- or MYC-expressing cells. Taken together, this demonstrates that AKT-expressing cells exhibited significantly larger adaptations to galactose culture than do RFP- or MYC-expressing cells, consistent with the observation that AKT-expressing cells initially die in galactose culture before recommencing proliferation.
Short-term galactose culture increases glutathione metabolism in mAKT cells
To further understand the metabolic profile that occurs in MCF-10A cells expressing oncogenes when switched from glucose to galactose, we next analyzed metabolite pool sizes (Table S3). Unsupervised hierarchical clustering of metabolite pool sizes was performed, segregating samples based on the culture condition (glucose, short-term galactose and long-term galactose culture) rather than by protein (RFP, mAKT and MYC) (Fig. 3A). No obvious grouping of functionally related metabolites was apparent from the hierarchical clustering of metabolites. Therefore, to identify metabolic pathways affected by galactose culture in each cell type, we employed Metabolite Set Enrichment Analysis (MSEA) (Kankainen et al., 2011; Persicke et al., 2012; Priolo et al., 2014). First, we analyzed the relative enrichment of all KEGG pathways comparing short-term galactose to glucose culture. To identify differentially enriched pathways, we plotted the MSEA results on a volcano-style plot (Fig. 3B). This analysis revealed that glutathione metabolism was significantly enriched in short-term galactose mAKT- but not RFP- or MYC-expressing cells (Fig. 3C; Fig. S4A,C). Next, we conducted a similar analysis comparing metabolite pool sizes in short-term and long-term galactose-cultured cells. Again, we found that glutathione metabolism was significantly enriched in short-term galactose mAKT- but not RFP- or MYC-expressing cells (Fig. 3D,E; Fig. S4B,D). Taken together, this MSEA suggests that glutathione metabolism is specifically upregulated in mAKT cells when initially switched to galactose culture.
Proteomic analysis reveals enriched nonsense-mediated mRNA decay in mAKT cells in short-term galactose culture
Next, to further characterize the mechanisms underlying the differential phenotype of mAKT cells in galactose culture, we performed label-free quantitative LC-MS proteomics. We analyzed two independent biological experiments of MCF-10A cells expressing either RFP, mAKT or MYC for glucose, short-term galactose and long-term galactose culture. Across both experiments, we identified and quantified 2460 proteins, 1356 of which were quantified in both biological replicates (Fig. 4A; Table S4). To identify global differences between samples, we conducted principal component analysis (PCA) using the proteins identified in both biological replicates. PCA revealed a clear separation across samples and consistent trends across the two experiments. Notably, in both replicates, short-term galactose culture induced a positive shift on PC1 (48.4% of variation for experiment 1) for all cell types (Fig. 4B; Fig. S5A). Long-term galactose culture, in contrast, generally exhibited a negative shift on PC1 relative to short-term galactose culture. Notably, the PC1 shift for mAKT cells was significantly larger than for either RFP- or MYC-expressing cells. To understand the proteins driving separation on PC1, we analyzed the PC1 loadings vector via 1D annotation enrichment using the Reactome Pathway Database (Cox and Mann, 2012; Fabregat et al., 2018). Among the enriched pathways, we found that two nonsense-mediated mRNA decay (NMD) pathways were enriched in the positive direction of the PC1 loadings vector, in addition to several other mRNA-related pathways (Fig. 4C; Fig. S5B). To better understand how these pathways were regulated by galactose culture, we examined NMD enhanced by the exon–junction complex (EJC), a surveillance pathway that eliminates mRNAs containing premature translation stop codons (Goetz and Wilkinson, 2017). Visualization of proteins from this pathway on a heatmap revealed that their basal expression was low in mAKT cells cultured in glucose and that short-term galactose culture induced higher expression (Fig. 4D; Fig. S5C). In long-term galactose culture, the levels of NMD proteins were again reduced to low expression in mAKT cells. Taken together, this proteomic analysis suggests that mAKT cells in short-term galactose culture dramatically upregulated NMD.
mAKT but not RFP- or MYC-expressing cells exhibited oxidative stress in short-term galactose culture
In mammalian cells, activation of NMD can sensitize cells to oxidative stress (Martin and Gardner, 2015). Because our metabolomic analysis (Figs 3 and 4) demonstrated that mAKT cells in short-term galactose culture upregulated glutathione metabolism, we hypothesized that mAKT cells were dying from increased oxidative stress following the switch from glucose to galactose culture. We thus measured the levels of ROS using the fluorescent ROS probe DCF-DA. MCF-10A cells expressing either RFP, mAKT or MYC were switched from glucose to galactose culture for 3 h and ROS levels were measured by flow cytometry. In RFP- and MYC-expressing cells, the switch to galactose culture did not significantly alter ROS levels (Fig. 5A,C). In mAKT cells, however, 3 h of galactose culture significantly increased the levels of ROS (Fig. 5B). Next, we investigated whether the increase in ROS was sustained following the switch from glucose to galactose. In MCF-10A-mAKT cells, we found that ROS levels were consistently elevated in mAKT cells for at least 48 h after exposure to galactose (Fig. 5D; Fig. S6A). To further understand the sources of ROS generation, we next used mitoSOX, a fluorogenic dye that measures superoxide production in the mitochondria. Similar to results found with DCF-DA staining, we observed that 3 h of galactose culture increased mitochondrial ROS in mAKT but not MYC cells (Fig. 5E,F). We did not measure levels of mitochondrial ROS in RFP-expressing cells because of the overlap between the fluorescent signals of mitoSOX and RFP (both ∼580 nm). Taken together, this data supports that constitutively active AKT but not MYC induces elevated ROS in short-term galactose-cultured MCF-10A cells.
Galactose culture-induced cell death can be rescued by ROS scavengers
Having identified that mAKT cells experience oxidative stress in short-term galactose culture, we next tested whether ROS were functionally involved in the cell death caused by galactose culture. We cultured MCF-10A cells expressing either RFP, mAKT or MYC in medium without glucose, medium with galactose, or medium with galactose plus the ROS scavenger catalase. For MCF-10A cells expressing either RFP or mAKT, glucose starvation resulted in significant cell death (Fig. 6A; Fig. S6B). Interestingly, cells expressing MYC were protected from glucose starvation-induced cell death, mirroring observations in normal human fibroblasts and glioma cells (Wise et al., 2008; Yuneva et al., 2007). In galactose culture, however, only MCF-10A cells expressing mAKT exhibited cell death. In these mAKT cells, supplementation with the ROS scavenger catalase rescued cells from galactose culture-induced cell death. Thus, ROS induced by galactose culture are required for galactose culture-induced cell death.
Next, we sought to determine whether our results in MCF-10A cells expressing mAKT were also reflected in breast cancer cell lines. We chose two cell lines reported to exhibit constitutively active AKT signaling, MDA-MB-436 (MB436) and Hs578t (Choi et al., 2018; Yi et al., 2013). Western blotting confirmed that both cell lines exhibited AKT activation even after serum starvation (Fig. 6B). Next, we tested the effect of glucose starvation and galactose culture on these cell lines, and found that both cell lines exhibited significant cell death upon glucose starvation and short-term galactose culture (Fig. 6C; Fig. S6B). In addition, for both cell lines, the ROS scavenger catalase rescued cells from galactose culture-induced cell death. Taken together, these results show that breast cancer cell lines with constitutively active AKT signaling experience ROS-mediated cell death when cultured in galactose medium, similar to our MCF-10A mAKT cells.
The altered metabolism of tumors has long been proposed as a therapeutic target. Defining the metabolic vulnerabilities induced by specific oncogenes is crucial for the design and stratification of therapeutics targeting tumor metabolism (Galluzzi et al., 2013; Martinez-Outschoorn et al., 2017). Here, using the galactose culture system, which forces mammalian cells to rely on OXPHOS instead of glycolysis for energy generation (Arroyo et al., 2016; Chang et al., 2013; Gohil et al., 2010; Iannetti et al., 2018; Robinson et al., 1992), we have uncovered a metabolic state-dependent lethality, namely the restricted ability of cells with constitutively active PI3K/AKT signaling to switch between glycolysis and OXPHOS. Through multi-omic analysis, we identified that the galactose-induced cell death exhibited by mAKT cells was accompanied by increased glutathione metabolism (Fig. 3) and increased expression of NMD proteins (Fig. 4), which is a hallmark of sensitivity to oxidative stress. Based on these results, we found that MCF-10A cells expressing mAKT cultured in galactose exhibited increased ROS levels (Fig. 5) and ROS-mediated cell death (Fig. 6). Importantly, we replicated this result in breast cancer cell lines with activated PI3K/AKT signaling. Taken together, these results reveal a novel metabolic state vulnerability induced by PI3K/AKT signaling.
Our findings in MCF-10A cells, a non-tumorigenic, but immortalized, breast cancer cell line, confirm reports that constitutive activation of PI3K/AKT signaling forces cells to rely on aerobic glycolysis (Elstrom et al., 2004; Fan et al., 2010). Stable isotope tracing with [U-13C]-L-glutamine demonstrated that these cells exhibited less glutamine anaplerosis than either RFP- or MYC-expressing cells when cultured in glucose (Fig. 2A,B). Interestingly, MCF-10A cells expressing the negative control RFP were sensitive to glucose deprivation, and mAKT expression did not further sensitize cells to glucose starvation (Fig. 6A). However, RFP-expressing cells were able to switch from glycolysis to aerobic respiration, as evidenced by their survival and growth in galactose medium (Figs 1 and 6). mAKT cells, in contrast, were initially unable to metabolize galactose, leading to ROS-mediated cell death even though PI3K/AKT activation has been shown to increase resistance to oxidative stress through upregulation of glutathione biosynthesis (Lien et al., 2016). In addition, our observation that MYC did not sensitize MCF-10A cells to galactose culture supports previous reports that AKT and MYC differentially alter metabolism, including the creation of metabolic vulnerabilities in glycolysis and mitochondrial bioenergetics, respectively (Fan et al., 2010; Priolo et al., 2014). Similarly, our data confirm that MYC activation protects against glucose deprivation (Wise et al., 2008; Yuneva et al., 2007), although MYC expression in MCF-10A cells did not increase oxidative metabolism relative to RFP-expressing cells, as seen in other cell types (Murphy et al., 2013). Taken together, the differential sensitivity of AKT- and MYC-expressing cells suggests that these oncoproteins alter metabolism through distinct mechanisms even though both can upregulate aerobic glycolysis (Broecker-Preuss et al., 2017; Simons et al., 2012; Stine et al., 2015).
Although mAKT cells initially died in galactose culture, these cells recommenced proliferation after ∼15 days. At present, it is not known if this represents metabolic remodeling to process galactose (i.e. acquired resistance) or the emergence of a pre-existing subpopulation of galactose-metabolizing cells (i.e. clonal selection). Notably, mAKT cells that had been selected in galactose retained the ability to proliferate in galactose even after an extended period of glucose culture (∼25 days) (Fig. 1F). This suggests that a subpopulation of galactose-metabolizing cells could stably co-exist within a population of cells cultured in glucose. If a galactose-resistant subpopulation did exist, one could hypothesize that this subpopulation would exhibit less ROS generation upon the switch from glucose to galactose culture. However, using flow cytometry, we failed to observe the emergence of any such sub-populations during the first 48 h of galactose culture (Fig. 5D; Fig. S6A). It is notable, however, that the AKT-expressing cells proliferating in galactose exhibited slower growth relative to in glucose than either RFP- or MYC-expressing cells (Fig. 1B). Thus, even the ‘galactose-resistant’ mAKT cells exhibited a disadvantage in oxidative culture, which may explain why proliferating mAKT cells exhibited a substantially different galactose metabolism than RFP- or MYC-expressing cells (Fig. 2E,F). Notably, because long-term culture in galactose only slightly reduced AKT expression and phosphorylation (Fig. 1D,E), the changes that enable proliferation of mAKT cells in galactose likely occur downstream of AKT itself. These changes could include increased mitochondrial content and aerobic metabolism, as has been observed when human myotubes, which have low mitochondrial oxidative potential, are cultured in galactose (Aguer et al., 2011; Kase et al., 2013). Regardless, the fact that mAKT cells reliably and rapidly demonstrated resistance to galactose suggests that eradication of tumors with constitutive PI3K/AKT signaling will require the therapeutic targeting of another, compensatory, pathway to prevent re-emergence. Notably, at this time, we have been unable to generate galactose-resistant clones of either MB436 or Hs578t, suggesting that these breast cancer cells harbor oncogenic lesions in addition to PI3K/AKT signaling that more permanently restrict the flexibility to utilize OXPHOS.
Another possible mechanism by which mAKT cells adapt to galactose culture might be through suppression of NMD (Fig. 4). In general, oxidative stress suppresses NMD in order to enhance the ability of cells to survive ROS toxicity (Goetz and Wilkinson, 2017; Martin and Gardner, 2015). However, when NMD is activated, cells become more sensitive to oxidative stress (Martin and Gardner, 2015). Thus, the galactose-induced cell death of mAKT cells may be due to increased ROS levels coinciding with increased sensitivity to oxidative stress due to NMD upregulation. In fact, there might be a direct link between AKT signaling and NMD upregulation, as insulin signaling in HeLa cells upregulates NMD through increased binding of UPF1, the master regulator of NMD, to mRNA transcripts (Park et al., 2016). In our MCF-10A mAKT and breast cancer cell lines, the mechanisms by which AKT signaling in galactose culture induces the upregulation of NMD proteins are currently under investigation.
Notably, the myristoylated constitutively active form of AKT1 (mAKT) used here is not found in tumors. Future studies using more realistic models of PI3K/AKT activation, including PI3K point mutants and homozygous PTEN deletion, will be required to fully understand the role of AKT-induced metabolic state inflexibility. However, it is interesting to note that MCF-10A mAKT cells and the breast cancer cell line MB436 exhibited very similar levels of AKT activation and ROS-mediated cell death in galactose culture (Fig. 6). Additionally, given the known differences between AKT isoforms, future studies will be required to test whether AKT2 and AKT3 can restrict metabolic flexibility in a similar manner to AKT1 (Gonzalez and McGraw, 2009). Finally, future integration of transcriptomic and phospho-proteomic profiling with our proteomic and metabolomic profiling may reveal additional mechanisms of metabolic adaptation to galactose culture.
In summary, the deregulation of PI3K/AKT and MYC in breast cancer motivates further research into how these oncogenes generate oncogene-dependent metabolic vulnerabilities. Like AKT and MYC, many other frequently altered oncoproteins including BRAF (Kaplon et al., 2013), ERBB2 (Zhao et al., 2009), KRAS (Gaglio et al., 2011; Yun et al., 2009) and VHL (Chan et al., 2011) alter the balance between glycolysis and OXPHOS. However, it remains to be tested whether these oncogenes will affect the flexibility of cells to shift between glycolysis and OXPHOS as does AKT. Regardless of the oncogene, it will be crucial to understand how oncogenic events define the response to fluctuations in nutrient availability, oxygen tension and pH within the tumor microenvironment (Lyssiotis and Kimmelman, 2017). In addition, given the flexibility and adaptability of cancer metabolism, inhibition of a single molecular target (e.g. glycolysis in tumors with hyperactivated AKT signaling) may not prove sufficient for tumor eradication. As such, combinational therapies that generate synthetic lethality in tumors also need to be investigated in the context of oncogene dependence (Clark et al., 2012; Dörr et al., 2013; Joly et al., 2020; O'Neil et al., 2017). Taken together, our findings highlight the importance of oncogene-dependent metabolic vulnerabilities in cancer cells and suggest that therapies targeting tumor metabolism will need to be appropriately paired with tumor genetic profiles.
MATERIALS AND METHODS
MCF-10A human mammary epithelial cells were obtained from American Type Culture Collection (ATCC, obtained in November 2016). Cells were cultured in DMEM/Ham's F-12 supplemented with 5% horse serum, 100 ng/ml cholera toxin, 20 ng/ml epidermal growth factor, 10 µg/ml insulin, 500 ng/ml hydrocortisone, 10,000 units/ml penicillin G, 10 mg/ml streptomycin sulfate and 25 µg/ml amphotericin B, and either 25 mM glucose or galactose. MDA-MB-436 (MB436) and Hs578t cells were gifts from Dr Michael Press (Department of Pathology, University of Southern California, Los Angeles, CA, obtained in January 2019). MB436 and Hs578t cells were cultured in DMEM supplemented with 10% FBS, 10,000 units/ml penicillin G, 10 mg/ml streptomycin sulfate and 25 µg/ml amphotericin B. Dialyzed FBS was used for all experiments where MB436 and Hs578t cells were cultured with galactose. All cells were grown in a 5% CO2, 37°C and humidified incubator and were used within 30 passages of thawing. Cell counting and viability were assessed using Trypan Blue staining with a TC20 automated cell counter (Bio-Rad).
RFP, myristoylated AKT1 (mAKT) and MYC were cloned into the pDS-FB-neo retroviral vector (a gift from Thomas Graeber, Department of Molecular and Medical Pharmacology, University of California, Los Angeles, CA) and verified by Sanger sequencing. Retrovirus was prepared in 293T (a gift from Pin Wang, Mork Family Department of Chemical Engineering and Materials Science, University of Southern California, Los Angeles, CA) cells by co-transfection with viral packaging plasmids. Following infection, MCF-10A cells were selected with 1 mg/ml G418. Following selection, cells were maintained in media with 500 µg/ml G418.
Reactive oxygen species measurement
MCF-10A were treated in respective media for 3 h. For total ROS determination, cells were incubated with 5 µM of DCF-DA (Biotium #10058) for 30 min at 37°C MCF-10A prior to trypsinization. The cells were washed with PBS twice, lifted with TrypLE, and resuspended in PBS for an approximate final concentration of 106 cells/ml. For mitochondrial ROS determination, cells were assessed by use of the MitoSOX™ Red Mitochondrial Superoxide Indicator (ThermoFisher M36008) according to the manufacturer's instruction. Samples were analyzed on a Miltenyi Biotec MACSQuant flow cytometer with FITC channel (488 nm excitation/520 nm emission) or phycoerythrin (PE) channel (585 nm exication/578 nm emission) to measure fluorescence, and data were processed and analyzed with the flowCore (1.48.1) R package.
Annexin V and PI staining
mAKT- and MYC-expressing MCF-10A cells were switched from glucose culture to galactose culture for 3 h. Cells were then subjected to trypsinization and pelleting by centrifugation (500 g for 3 min). Cell pellets were resuspended in 500 μl of binding buffer, then stained with 5 μl of Annexin V–FITC and 5 μl of PI stock (BioVision 10013-436) for 5 min in darkness according to the manufacturer's instruction. The probed samples were analyzed on a Miltenyi Biotec MACSQuant flow cytometer with FITC channel (488 nm excitation/520 nm emission) or APC channel (652 nm excitation/660 nm emission) to measure fluorescence, and data were processed and analyzed with FlowJo™ (v.10.6.1).
Phospho-flow measurement of AKT and phospho-AKT
MCF-10A cells were treated in respective media for 3 h before trypsinization. The suspended cells were centrifuged (3000 g for 2 min) and washed twice with 1 ml of PBS. Cell pellets were gently mixed with 100 µl of Citofix/Cytoperm (BD Biosciences #554714) and incubated at 4°C for 15 min for membrane permeabilization. Then cell suspensions were diluted with 100 µl of wash buffer (BD Biosciences #554714) and pelleted. An additional 150 µl of wash buffer was added and cells pelleted by centrifugation for washing. Cells were then resuspended in 50 µl of wash buffer containing anti-phosphoSer473-AKT1 APC-fluorochrome (1:50; eBioscience™, #17971542) or 1:50 anti-total-AKT (Cell Signaling Technology 9272) and incubated for 30 min at 4°C in darkness. The anti-total-AKT stained cells were further stained with 1:100 dilution of secondary antibody (Thermo Fisher Scientific, #A-10931) in wash buffer for 10 min at 4°C in darkness. Cells were washed twice with 150 µl of PBS and analyzed on a Miltenyi Biotec MACSQuant flow cytometer with APC channel (652 nm excitation/660 nm emission) to measure fluorescence. Data were processed and analyzed with FlowJo™ (v.10.6.1).
Cells were lysed in modified RIPA buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 50 mM β-glycerophosphate, 0.5 mM NP-40, 0.25% sodium deoxycholate, 10 mM sodium pyrophosphate, 30 mM sodium fluoride, 2 mM EDTA, 1 mM activated sodium vanadate, 20 µg/ml aprotinin, 10 µg/ml leupeptin, 1 mM DTT and 1 mM PMSF). Whole-cell lysates were resolved by SDS-PAGE on 4–15% gradient gels and blotted onto nitrocellulose membranes (Bio-Rad). Membranes were blocked for 1 h and then incubated with primary antibodies overnight and secondary antibodies for 2 h. Blots were imaged using the Odyssey Infrared Imaging System (Li-Cor). Primary antibodies used for western blot analysis were: anti-AKT (1:500; Cell Signaling Technology 9272), anti-phospho-Ser473-AKT (1:500; Santa Cruz Biotechnology sc-7985-R), anti-c-MYC (1:500; Cell Signaling Technology 9402), anti-β-actin (1:1000; Proteintech 66009-1-lg), IRDye 800CW-conjugated goat anti-mouse IgG (VWR 926-32210), and IRDye 680RD-conjugated goat anti-rabbit IgG (926-68071). Band intensities were quantified using Li-Cor Image StudioTM Lite (V.5.0).
MCF-10A cells were plated on six-well plates at the density of 7333 cells/cm2. After 24 h, medium was removed, cells were washed twice with 2 ml of PBS, and 1 ml of medium was added to cells. Medium contained either [U-13C]-L-glutamine or [U-13C]-galactose (Cambridge Isotope Laboratories). After 24 h, the culture plates were cooled on ice, medium was aspirated, and the cells were washed with 1 ml of cold ammonium acetate. Upon aspirating the ammonium acetate, metabolites were extracted with 1 ml of −80°C methanol. The methanol cell suspension was scraped and transferred to Eppendorf tubes, and the cell suspension was centrifuged (21,100 g for 5 min) at 4°C. The supernatants was transferred to new Eppendorf tubes, and the pellet was re-extracted with another 350 μl of −80°C methanol. The second methanol extraction was spun down (21,100 g for 5 min), and the supernatant was pooled with the first extraction. Metabolites were speed-vac dried, resuspended in LC-MS grade water, and sent for LC-MS analysis.
Samples were randomized and analyzed on a Q Exactive Plus hybrid quadrupole-Orbitrap mass spectrometer coupled to an UltiMate 3000 UHPLC system (Thermo Scientific). The mass spectrometer was run in polarity switching mode (+3.00 kV/−2.25 kV) with an m/z window ranging from 65 to 975. Mobile phase A was 5 mM NH4AcO, pH 9.9, and mobile phase B was acetonitrile. Metabolites were separated on a Luna 3 µm NH2 100 Å (150×2.0 mm) column (Phenomenex). The flow rate was 300 µl/min, and the gradient was from 15% A to 95% A in 18 min, followed by an isocratic step for 9 min and re-equilibration for 7 min. All samples were run in biological triplicates. Metabolites were detected and quantified as area under the curve based on retention time and accurate mass (≤8 ppm) using the TraceFinder 3.3 (Thermo Fisher Scientific) software. Raw data were corrected for naturally occurring 13C abundance (Moseley, 2010). Intracellular data was normalized to the cell number at the time of extraction.
MCF-10A cells dishes were placed on ice and washed with PBS. Cells were then scraped and pelleted by centrifugation (2400 g for 3 min). The cell pellets were lysed by probe sonication in 8 M urea, 50 mM Tris-HCl pH 7.5, 1 mM activated sodium vanadate, 2.5 mM sodium pyrophosphate, 1 mM β-glycerophosphate and 100 mM sodium phosphate. The above procedures were performed at 0–4°C. Insoluble cell debris were filtered by 0.22 µm syringe filter. Protein concentration was measured with a BCA assay (Pierce, PI23227). Lysates were reduced with 5 mM DTT, alkylated with 25 mM iodoacetamide, quenched with 10 mM DTT, and acidified to pH 2 with 5% trifluoracetic acid. Proteins were then digested to peptides using a 1:100 trypsin-to-lysate ratio by weight. Tryptic peptides were desalted by reverse phase C18 StageTips and eluted with 30% acetonitrile. The eluents were vacuumed dried, and 250 ng/injection was submitted to LC-MS. We performed two independent biological replicates, and each experiment was subjected to two technical LC-MS injections.
The samples were randomized and injected into an Easy 1200 nanoLC ultra high-performance liquid chromatography coupled with a Q Exactive quadrupole-Orbitrap mass spectrometer (Thermo Fisher Scientific). Peptides were separated by a reverse-phase analytical column (PepMap RSLC C18, 2 µm, 100 Å, 75 µm×25 cm). Flow rate was set to 300 nl/min at a gradient from 3% buffer B (0.1% formic acid, 80% acetonitrile) to 38% B in 110 min, followed by a 10-min washing step to 85% B. The maximum pressure was set to 1180 bar (118 MPa) and column temperature was maintained at 50°C. All samples were run in technical duplicate. Peptides separated by the column were ionized at 2.4 kV in the positive ion mode. MS1 survey scans were acquired at the resolution of 70,000 from 350 to 1800 m/z, with a maximum injection time of 100 ms and AGC target of 1e6. MS/MS fragmentation of the 14 most abundant ions were analyzed at a resolution of 17,500, AGC target 5e4, maximum injection time 65 ms, and normalized collision energy of 26. Dynamic exclusion was set to 30 s and ions with charge +1, +7 and >+7 were excluded.
MS/MS fragmentation spectra were searched with Proteome Discoverer SEQUEST (version 2.2, Thermo Scientific) against in silico tryptic digested Uniprot all-reviewed Homo sapiens database (release June 2017, 42,140 entries) plus all recombinant protein sequences used in this study. The maximum missed cleavages was set to two. Dynamic modifications were set to oxidation on methionine (M, +15.995 Da) and acetylation on protein N-terminus (+42.011 Da). Carbamidomethylation on cysteine residues (C, +57.021 Da) was set as a fixed modification. The maximum parental mass error was set to 10 ppm, and the MS/MS mass tolerance was set to 0.02 Da. The false discovery threshold was set strictly to 0.01 using the Percolator Node validated by q-value. The relative abundance of parental peptides was calculated by integration of the area under the curve of the MS1 peaks using the Minora LFQ node.
Data analysis and statistics
The Proteome Discoverer peptide groups abundance values were normalized to median values of the corresponding samples. After normalization, the missing values were imputed using the K-nearest neighbor algorithm (Webb-Robertson et al., 2015). The optimized number of neighbors was determined to be n=10. The protein copy numbers were assessed using intensity-based absolute quantification (iBAQ) (Schwanhäusser et al., 2011). Proteomics data analysis was performed in Microsoft Excel, R (version 3.4.2) and Perseus (version 184.108.40.206).
Metabolite set enrichment analysis
MCF-10A intracellular pool sizes were ranked based on log2 fold change, and enrichment analysis was run with the unweighted the statistic using the Broad Institute's GSEA java applet (https://www.gsea-msigdb.org/gsea/index.jsp) against all KEGG metabolic pathways. Statistical significance was assessed by 5000 permutations of the ranked list.
We would like to thank Dr Pin Wang for assistance with flow cytometry experiments.
Conceptualization: D.Z., N.A.G.; Methodology: D.Z., A.D., N.A.G.; Validation: D.Z., N.A.G.; Formal analysis: D.Z.; Investigation: D.Z., J.H.S., M.P.J., S.T.P., M.A.M., A.D.; Data curation: D.Z., J.H.S., M.P.J., N.A.G.; Writing - original draft: D.Z., J.H.S., M.P.J., N.A.G.; Writing - review & editing: D.Z., J.H.S., M.P.J., N.A.G.; Supervision: N.A.G.; Project administration: N.A.G.; Funding acquisition: N.A.G.
This work was supported by The Margaret E. Early Medical Research Trust, The Rose Hills Foundation, The USC Provost's Office and The Viterbi School of Engineering. D.Z. was supported by a Mork Family Doctoral Fellowship. J.H.S. and M.P.J. were supported by USC Provost's Undergraduate Research Fellowships.
The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository (Perez-Riverol et al., 2019) with the dataset identifier PXD015122.
Peer review history
The peer review history is available online at https://jcs.biologists.org/lookup/doi/10.1242/jcs.239277.reviewer-comments.pdf
The authors declare no competing or financial interests.