AMPK Activation and Metabolic Reprogramming by Tamoxifen through Estrogen Receptor–Independent Mechanisms Suggests New Uses for This Therapeutic Modality in Cancer Treatment

Tamoxifen is the most widely used nonsteroidal selective estrogen receptor modulator (SERM) for adjuvant therapy of estrogen receptor–positive (ER⁺) breast cancer. Patients on standard treatment receive tamoxifen at 20 mg/day for 5 years during, or shortly after surgical removal of the primary tumor and treatment with ionizing radiation or cytotoxic chemotherapy. Tamoxifen therapy has reduced disease recurrence by half and overall mortality by one third in ER⁺ breast cancer patients. However, the development of tamoxifen resistance remains a significant clinical problem (1, 2).

The antitumor effects of tamoxifen are attributed primarily to its antiproliferative activity via competitive inhibition of the ER in breast tissue (2). Several studies have previously described non-ER–dependent effects of tamoxifen, such as the inhibition of protein kinase C (PKC; ref. 3) and the sensitization of ER-negative (ER⁻) tumors to chemotherapeutic agents (4). Furthermore, during the past two decades, more than 25 clinical trials have been published reporting the effects of “high dose” tamoxifen (i.e., doses above those needed to inhibit ER activity) ranging from 80 to 720 mg/day (5) to treat non-breast cancers, including glioma (6), melanoma (7), and others (8, 9), as a single agent or in combination with chemotherapeutics. Although, of these, phase I and II trials showed variable clinical benefit, they clearly demonstrated the safety of high-dose tamoxifen in diverse patient populations. Pharmacodynamic analyses indicated that tamoxifen could reach plasma concentrations as high as 8 μmol/L (8), and additional preclinical studies indicated that tamoxifen accumulates in tumor tissue at 60 to 70 times the plasma concentration (10). Therefore, further studies on “high-dose” tamoxifen are clinically relevant and may result in expanded indications for this affordable and well-tolerated chemotherapeutic.

The chemical properties of tamoxifen as a lipophilic weak base contribute to its high partition into lipid bilayers, and tamoxifen has been previously described to have inhibitory effects on mitochondrial respiratory rate (11) and membrane potential (12). Cells in which oxidative phosphorylation (OXPHOS) is inhibited by pharmacologically targeting members of the mitochondrial respiratory chain, or due to lack of oxygen (hypoxia), exhibit a compensatory increase in glycolysis. For example, metformin, a diabetes drug showing promise as a cancer therapeutic, was recently demonstrated to inhibit mitochondrial complex I and to upregulate glycolysis (13). A defining characteristic of many cancers is the Warburg effect, which denotes an elevated aerobic glycolytic rate, irrespective of any hypoxic burden, as well as an increased dependence on ATP and macromolecule intermediates to support rapid cell growth and division (14). Tumor promotion by certain oncogenes can alter the metabolic characteristics of cells as a requirement for transformation (15). As a result, strategies that target these vulnerable metabolic pathways have shown significant therapeutic promise. For example, both BRAF and KRAS mutations promote glycolysis (16, 17). Resistance to targeted therapies limits the therapeutic options in cancers such as melanoma and pancreatic cancer that are driven by these pathways. The drug-resistant populations are thus dependent on elevated OXPHOS, and drugs that target mitochondrial metabolism, including metformin, have potential utility in targeting such resistant populations (16, 18–20).

One of the consequences of OXPHOS inhibition, especially in metabolically active cancer cells, is the activation of the AMP-activated protein kinase (AMPK), the major regulator of cellular energy homeostasis. AMPK is activated when the AMP:ATP ratio increases, signaling a loss in cellular energy charge. AMP binds to the γ subunit of AMPK, inducing a conformational change that prevents the dephosphorylation of the catalytic α subunit of AMPK by phosphatases. This enhances phosphorylation of AMPK by upstream kinases, including LKB1 and CaMKK2. Once activated, AMPK restores cellular energy levels by promoting catabolic and inhibiting anabolic processes (21). The role of AMPK in cancer is quite complex, as it has been reported to function contextually as both a tumor suppressor and promoter (22).

On the basis of these studies, we investigated whether tamoxifen had an impact on cellular metabolism and metabolic signaling. Here, we report the ability of tamoxifen to inhibit mitochondrial oxygen consumption, increase glycolysis, and alter fatty acid metabolism. We also observe rapid (<10 minutes) and acute activation of AMPK in vitro and in vivo in an ER-independent manner. We demonstrate that activation of the AMPK pathway by tamoxifen promotes cell death in triple-negative breast cancer cells in an AMPK isoform–specific manner. Moreover, we demonstrate that cells treated with low micromolar doses of tamoxifen exhibit increased dependence on glucose metabolism and that pharmacologic inhibition of these processes is synergistically lethal with tamoxifen. Overall, these data suggest that tamoxifen functions as a potent regulator of tumor metabolism and AMPK activity and that these metabolic effects can be exploited with combination therapies to increase tumor cytotoxicity.

Tamoxifen, fulvestrant, methyl succinate, and spautin-1 were purchased from Sigma Aldrich. The following rabbit polyclonal antibodies were obtained from Cell Signaling Technology: AMPKα, AMPKα1, phos-AMPKα, phos-ACC, acetyl-CoA carboxylase (ACC), phos-p70s6K, phos-s6, phos-4EBP1, cleaved PARP, cleaved caspase-3, and LC3B. ERα antibody was purchased from GeneTex, AMPKα2 antibody was purchased from Millipore, and siRNA against PRKAA2 was purchased from Dharmacon.

Cell lines were cultured in DMEM supplemented with 10% FBS, penicillin/streptomycin, and HEPES at 37°C humidified 5% CO₂ atmosphere. MCF7, MDA-MB-231, and 4175 cell lines were a gift from Andy Minn (University of Pennsylvania, Philadelphia, PA), LCC1 cells were a gift from Robert Clarke (Georgetown University, Washington DC), and WM9 and 1205Lu cells were a gift from Meenhard Herlyn (The Wistar Institute, Philadelphia, PA). Cell lines were authenticated through genetic testing by ATCC.

Oxygen consumption was measured using a YSI 5300a Biological Oxygen Monitor. Electrodes were equilibrated in media for 10 minutes. Cells were trypsinized and suspended at 2 × 10⁷ cells/mL and injected into the chamber at a 1:10 dilution. After the oxygen consumption rate was recorded, drug was injected into the chambers at 10-μmol/L increments, allowing the consumption rate to be determined after each injection. Total protein per chamber was quantified by Lowry assay.

Liver was removed from a sacrificed mouse and homogenized in 8 volumes of sucrose buffer [0.25 mol/L sucrose, 0.02 mol/L KCl, 0.005 mol/L MgCl₂, and 0.01 mol/L KH₂PO₄ (pH = 7.4)]. Homogenate was spun at 600 × g for 5 minutes at 4°C. The supernatant was then transferred to a new tube and centrifugation was repeated. Cleared supernatant was spun at 9,000 × g for 10 minutes at 0°C, and the pellet was resuspended in sucrose buffer. The following conditions were used for respiration measurements: complex 1 respiration: sucrose buffer + 0.1 mmol/L ADP, 5 mmol/L glutamate, and 5 mmol/L malate; complex 2 respiration: sucrose buffer + 0.1 mmol/L ADP and 5 mmol/L sodium succinate.

For more details, see Supplementary Methods.

For more details, see Supplementary Methods.

Knockout cell lines were generated using the CRISPR-Cas9 system as described previously (23). sgRNAs were designed to target the first exon of the genes of interest using the CRISPR design tool at http://crispr.mit.edu/. Two sgRNA sequences were chosen for each gene of interest (Supplementary Table S1). Briefly, sgRNAs were cloned into CRISPR-Cas9 plasmid. pSpCas9(BB)-2A-Puro (PX459) was a gift from Feng Zhang (Addgene plasmid # 48139). Cells were transfected using lipofectamine 2000 and carrier DNA. Media were changed 24 hours after transfection, and cells were selected with puromycin for 48 hours. Remaining cells were grown up, and clones were selected. Clones were screened for gene excision or mutation by DNA gel and expression of the protein of interest by Western blot analysis. Knockout was successful using both PRKAA1 sgRNAs 1 and 2. Knockout of ESR1 expression required combined transfection of sgRNA 1 and 2.

AMP and ATP were quantified using a previously published method (24). Briefly, 2–4 × 10⁶ cells were plated in 10-cm² dishes and treated with tamoxifen. Cells were then pelleted and nucleotides extracted with perchloric acid. Extracts were analyzed by ion pair reverse-phase high-performance liquid chromatography (HPLC). Analysis was carried out on Jasco HPLC. Separation was performed on SUPELCOSIL C18 Column (15 cm × 4.6 mm, 3-μm particle size) in combination with a Supelguard Ascentis C18 guard column (2 cm × 4.0 mm) from Sigma.

Liquid chromatography–selected reaction monitoring (LC-SRM)/MS analysis of acyl-CoA thioesters was performed as described previously (25). CoAs were separated using a Reversed Phase HPLC Phenomenex Luna C18 Column (2.0 mm × 150 mm; particle size, 3.5 μm) with 5 mmol/L ammonium acetate in water as solvent A, 5 mmol/L ammonium acetate in acetonitrile (ACN)/water (95:5; v/v) as solvent B, and ACN/water/formic acid (80:20:0.1; v/v/v) as solvent C. Samples were analyzed using an API 4000 triple-quadrupole mass spectrometer (Applied Biosystems) in the positive electrospray mode. Samples (10 μL) were injected using a Leap CTC Autosampler (CTC Analytics), where they were maintained at 4°C.

Extractions were performed as described in detail previously (25, 26). For more details, see Supplementary Methods.

Media were aspirated and cells were quenched by the direct addition of 1-mL −80°C 4:1 methanol:water (v/v) to the cell culture dish. Plates were placed at −80°C for 20 minutes, scraped, and transferred into Eppendorf tubes. Samples were pulse sonicated on ice for 30 seconds at a rate of 1 pulse/sec prior to centrifugation at 16,000 × g at 4°C for 10 minutes. The supernatants were then transferred to clean glass tubes and evaporated to dryness under nitrogen. Dried residues were resuspended in 100 μL of mobile phase A for LC/MS analysis. For labeling studies, cells were grown in media omitting glucose supplemented with 1 mg/mL [¹³C₆]-glucose. Details of the LC/MS-MS analysis have been described in detail previously (27).

Frozen tissue was weighed and homogenized by electric homogenizer on ice in 4:1 methanol:water (v/v). Samples were then centrifuged at 16,000 × g at 4°C for 10 minutes. Ten microliters of the resulting supernatant was analyzed by reverse-phase LC/MS-MS, utilizing teniposide as the internal standard for absolute quantification.

For more details, see Supplementary Methods.

BODIPY lipid probes were purchased from Molecular Probes. Staining solution was made up at 2 μg/mL in PBS. After drug treatments, cells were fixed using 4% paraformaldehyde and stored at 4°C. Neutral lipid staining was performed by pelleting the cells and resuspending in staining solution and incubated for 10 minutes at room temperature. Cells were washed twice with PBS, suspended in 500 μL PBS, and strained. Cells were analyzed for mean BODIPY stain intensity by flow cytometry (FL1-H).

Cells were plated in 60-mm dish and treated for experimental conditions. Cells were incubated with methionine/cysteine–free media for 30 minutes. Hot labeling media made up at 0.075 mCi/mL [³⁵S]-methionine/cysteine. Cells were incubated with labeling media for 30 minutes. Cells were washed with cold PBS and harvested for protein. Protein assay preformed on cold samples and equal amounts of [³⁵S]-labeled protein were resolved on SDS-PAGE. Gel was washed with fixing solution (water with 20% methanol, 10% acetic acid) for 30 minutes, washed with deionized water, and then washed with enlightening solution (PerkinElmer) for 30 minutes. Gel was dried overnight using a Bio-Rad gel drying apparatus. Gel was exposed to autoradiography film at −80°C and film was developed.

Four-week-old female nu/nu mice were purchased from Charles River Laboratories and housed in the University of Pennsylvania animal facility maintained on a 12-hour light/dark cycle and given free access to water and food ad libitum. 4175 cells (2 × 10⁶) suspended in 50 μL of a 50/50 mixture of PBS and Matrigel Matrix Basement Membrane (Corning) were injected into the mammary fat pad.

Tamoxifen (20 mg) was dissolved in 200-μL ethanol and 800-μL peanut oil. Ethanol was evaporated off overnight. Mice were dosed via intraperitoneal injection once per day for 3 days at 100 mg/kg.

For more details, see Supplementary Methods.

All statistical analyses were performed using a two-tailed Student t test assuming homoscedasticity. P < 0.05 was considered statistically significant.

Although tamoxifen has been reported to affect mitochondrial electron transport and membrane potential, its effects on oxygen consumption have not been adequately described. A Clark-type electrode was used to measure oxygen consumption of cells. Tamoxifen, at low micromolar doses, inhibited oxygen consumption in a dose-dependent manner in both ER⁺ MCF7 cells and triple-negative MDA-MB-231 cells (Fig. 1A). Tamoxifen also inhibited oxygen consumption in a panel of 6 non-breast cancer cell lines (Supplementary Table S2). The active metabolite of tamoxifen, 4-hydroxy-tamoxifen (4OH-Tam) also inhibited oxygen consumption, whereas the ER antagonist fulvestrant, which inhibits ER activity by blocking ER dimerization and promoting degradation of the ER (28), did not have a significant effect (Supplementary Fig. S1A). This suggests that inhibition of ER signaling is not responsible for the rapid inhibition of oxygen consumption in response to tamoxifen treatment. To further rule out the involvement of ER signaling in tamoxifen-mediated inhibition of oxygen consumption, we used the CRISPR-Cas9 system to knock out ERα from LCC1 cells, a variant of the MCF7 line, which expresses functional ER but is not dependent on its signaling for proliferation (29). The ability of tamoxifen to inhibit oxygen consumption in the LCC1-crspER cells was no different from the parental LCC1 cells. Moreover, fulvestrant treatment, which leads to the downregulation of total ER protein levels, had no impact on the ability of tamoxifen to inhibit oxygen consumption in MCF7 cells (Supplementary Fig. S1B), further supporting the notion that tamoxifen inhibits oxygen consumption independently of ER signaling.

To determine whether the effect of tamoxifen on oxygen consumption was due to a direct interaction with mitochondria, we permeabilized MCF7 and MDA-231 cells using digitonin. Tamoxifen inhibited oxygen consumption of mitochondria by approximately 60% in both permeabilized cell lines. Respiration could be rescued by supplementation with electron transport chain complex II substrate succinate but not with complex I substrates, glutamate and malate (Fig. 1D). These effects of tamoxifen on mitochondria respiration are similar to what is seen with rotenone, a known complex I inhibitor (Fig. 1D). In mitochondria isolated from mouse liver, tamoxifen also inhibited oxygen consumption in the presence of glutamate and malate but not in the presence of succinate (Supplementary Fig. S1C). This indicates that the tamoxifen-mediated inhibition of oxygen consumption occurs through a direct effect on mitochondria, upstream of complex II. The addition of methyl succinate, a cell-permeable analogue of succinate, to media, reduced the ability of tamoxifen to inhibit oxygen consumption in intact breast cancer cells (Fig. 1E) and significantly increased cell survival of MCF7 cells treated with tamoxifen (Fig. 1F). These results suggest that inhibition of mitochondrial function upstream of complex II mediates the effects of tamoxifen on oxygen consumption and downstream effects.

To determine whether inhibition of oxygen consumption by tamoxifen alters cellular energy charge, we quantified intracellular AMP and ATP pools following tamoxifen treatment. Tamoxifen increased the AMP:ATP ratio in MDA-231 cells and MCF7 cells (Fig. 2A), the major activating mechanism of AMPK, as soon as 5 minutes after tamoxifen treatment, and this ratio further increased following one hour of treatment. This increase in the AMP:ATP ratio correlated with concomitant increase in tamoxifen-induced AMPK phosphorylation of the downstream signaling target ACC and dephosphorylation of the mTOR targets P70S6K, S6, and 4EBP1 in both MCF7 (Fig. 2B) and MDA-MB-231 cells (Fig. 2C). To investigate whether ER signaling is necessary for tamoxifen activation of the AMPK pathway, we used pharmacologic and genetic methods to block ER signaling. Pretreatment with fulvestrant did not inhibit tamoxifen activation of AMPK (Fig. 2D). Furthermore, activation of AMPK in LCC1-crspER was comparable with that in LCC1 parental cells (Fig. 2E), indicating that the effects of tamoxifen on AMPK signaling are also ER independent.

AMPK activation is known to increase glycolytic activity in an attempt to restore cellular ATP pools (21). To investigate whether this occurred after tamoxifen treatment, we measured glucose uptake from the media in MDA-231 and MCF7 cells. Glucose uptake increased in a dose-dependent manner with tamoxifen treatment in both cell lines (Fig. 3A). To examine whether glycolysis was also upregulated by tamoxifen, we analyzed the incorporation of [¹³C] from [¹³C₆]-glucose into acetyl-CoA and lactate by LC/MS-MS. Following tamoxifen treatment, we observed an increase in [¹³C] labeling in these products, indicating an increase in glycolytic activity (Fig. 3B and C). The activation of AMPK (P-AMPK) was enhanced under glucose deprivation conditions (Fig. 3D) and treatment with the glycolytic inhibitor 2-deoxyglucose (2DG; Supplementary Fig. S1E) in both MCF7 and MDA-231 cells, underscoring the importance of the glycolytic switch for ATP generation during tamoxifen treatment. Interestingly, under glucose deprivation conditions, supplementation with methyl succinate rescues MDA-231 cells from tamoxifen treatment, whereas it does not significantly improve MDA-231 survival in complete media (Supplementary Fig. S1D). This effect further highlights the importance of glycolysis during tamoxifen treatment.

Products of glucose metabolism are used for the synthesis of fatty acids, which are of particular importance in cancer due to the need to generate excess membranes for rapid proliferation. Moreover, the oxidation of fatty acids in the mitochondria can replenish cellular ATP levels during times of metabolic stress (30). To examine whether tamoxifen also impacted fatty acid metabolism, we preformed metabolite analysis of fatty acyl-CoA species by LC/MS-MS. This analysis showed an accumulation of acetyl-CoA and malonyl-CoA following tamoxifen treatment (Supplementary Fig. S2A). We also observed a reduction in succinyl-CoA and other medium chain acyl-CoA species (C3:0, C6:0), consistent with reports of the effects of other complex I inhibitors (31) and an accumulation of longer chain (C16-C20) acyl-CoA species (Supplementary Fig. S2A). The addition of [¹³C]-glucose or [¹³C]-glutamine resulted in no labeling of the long-chain CoA metabolites, indicating that these fatty acyl-CoAs are not synthesized from glucose or glutamine de novo in response to tamoxifen treatment. Activation of the AMPK pathway is also known to promote β-oxidation of fatty acids as a strategy to increase ATP pools. The degradation of lipid droplets by autophagy, termed lipophagy, increases the supply of free fatty acids, which are then transported into the mitochondria for oxidation (30). We also observed a rapid activation of autophagy upon tamoxifen treatment (Fig. 3E). BODIPY stain, which measures total neutral lipid content in the cell, decreased modestly after 4 hours of tamoxifen treatment, by 14%, in cells treated with tamoxifen (Supplementary Fig. S2B and S2C). These data suggest that the accumulation of the long chain acyl-CoA species results from breakdown of lipid droplets.

To further dissect the biologic role of AMPK in the cellular response to tamoxifen, we used the CRISPR-Cas9 system to knock out AMPKα1. We observed that loss of AMPKα1 resulted in a compensatory increase in AMPKα2 expression (Fig. 4A) and had increased survival in response to tamoxifen treatment (Fig. 4C). This effect was recapitulated in MCF7 cells, where AMPKα1 inhibition by shRNA resulted in reduced levels of cleaved PARP and increased survival upon tamoxifen treatment compared with shNT-transfected cells (Supplementary Fig. S3A and S3B). To evaluate the impact of complete inhibition of AMPK catalytic activity on cellular response to tamoxifen, we used siRNA to knock down AMPKα2 in both wild-type MDA-231 cells and in crspAMPKα1 cells. Loss of AMPKα2 in MDA-231 cells resulted in increased cell death upon tamoxifen treatment (Fig. 4C), whereas knockdown of AMPKα2 in the AMPKα1 CRISPR knockout cells resulted in further increase in tamoxifen-induced cytotoxicity.

Interestingly, knockdown of either isoform of AMPKα resulted in loss of phospho-ACC upon tamoxifen treatment, whereas loss of AMPKα1 alone ameliorated the inhibition of the mTOR pathway following tamoxifen treatment. Induction of autophagy remained unchanged upon AMPKα1 inhibition; however, knockdown of both isoforms substantially reduced autophagy levels as seen by a reduction in LC3 processing (Fig. 4A). To further investigate the role of AMPK isoforms in the autophagic response of breast cancer cells to tamoxifen treatment, we pretreated cells with E64d and pepstatin A to prevent LC3-II recycling and monitored autophagic flux by immunoblot. Although we saw reduced levels of LC3 in both siAMPKα2 and crspAMPKα1 conditions, MDA-231 cells require both AMPKα1 and AMPKα2 to be knocked down to significantly reduce autophagy induction with tamoxifen treatment (Fig. 4B). These data suggest that AMPK activation promotes autophagy in response to tamoxifen treatment. To determine whether the AMPK isoform–specific effects on the mTOR pathway impacted upon cellular translation, we pulse labeled MDA-231 cells and crspAMPKα1 cells with [³⁵S]-cysteine and methionine. Translation was inhibited in both cell lines, albeit to a substantially lesser extent in the crspAMPKα1 cells upon tamoxifen treatment (Fig. 4D and Supplementary Fig. S3E).

Notably, although the dual AMPKα1/2 knockdown showed the highest levels of tamoxifen toxicity as determined by modified MTT assay, the levels of apoptosis markers cleaved PARP and cleaved caspase-3 were quite low. This could be due to more rapid processing of these markers under these conditions as indicated by complete loss of total PARP in these cells. Caspase-3/7 activity assay suggests a decrease in apoptosis upon AMPKα1 knockout, whereas Annexin V and propidium iodide staining support an increase in apoptosis with AMPKα2 knockdown (Supplementary Fig. S3C). However, the data from these apoptosis assays were not statistically significant and showed only a small trend toward different levels of apoptosis dependent on AMPK status. Crystal violet staining of cells 7 days following tamoxifen treatment also shows that AMPKα1 knockout rescues cells from tamoxifen treatment, whereas AMPKα2 knockdown promotes cell death (Supplementary Fig. S3D). Therefore, although increased apoptosis may contribute to reduced viability of AMPKα2 knockout cells following tamoxifen treatment, additional mechanisms leading to reduced overall survival are likely involved.

To determine whether tamoxifen activates the AMPK pathway in vivo, we injected 4175 cells, a variant of the MDA-MB-231 triple-negative breast cancer cell line (32) into the mammary fat pad of athymic nude female mice. After 3 weeks of tumor growth, 100 mg/kg tamoxifen was administered once daily for 3 days. Mean tamoxifen and metabolite levels of 4OH-Tam and N-desmethyl-tamoxifen quantified by LC/MS-MS in the tumor tissue were 19.07, 3.97, and 33.64 ng/mg respectively (Fig. 5A). These data indicate that the concentrations of tamoxifen needed to activate the AMPK pathway in vitro are achievable in vivo. Mice treated with tamoxifen also had significantly higher levels of phospho-AMPK and displayed downregulation of the mTOR target phospho-4-EBP1 (Fig. 5B and C). These results are in agreement with AMPK pathway activation seen in vitro.

Dual blockade of required metabolic pathways to target tumor cells is gaining significant traction in preclinical and clinical settings. In terms of metabolic targeting, it has been postulated that concurrently targeting adaptive responses to metabolic reprogramming can lead to significant enhancement of the therapeutic window for multiple chemotherapeutic modalities (33, 34). As we demonstrated that tamoxifen blocks OXPHOS and concurrently increases glucose uptake and utilization, it is likely that effective strategies for metabolic targeted therapies may require the inhibition of both OXPHOS and glycolysis simultaneously. 2DG, a nonmetabolizable analogue of glucose and 3-bromopyruvate (3BP), an alkylating agent and inhibitor of hexokinase-2 and GAPDH, induced a dose-dependent decrease in glucose uptake, as expected (Fig. 6A and C). More importantly, combined treatment with relatively nontoxic doses of each of these inhibitors (Supplementary Fig. S1F) with tamoxifen significantly enhanced tamoxifen-induced cytotoxicity in MCF7 cells and in MDA-231 cells (Fig. 6B and D). Notably, the sensitization to tamoxifen was larger in the MDA-231 cells, which display a higher basal rate of glucose uptake (Fig. 3A) and lower basal OXPHOS level (Fig. 1B).

We have demonstrated that beyond its antiestrogenic effects, tamoxifen exerts a pronounced effect on tumor metabolism via a non-ER, mitochondrial-specific pathway. Specifically, through the inhibition of oxygen consumption upstream of complex II, the increase in AMP/ATP ratio leads to the activation of AMPK, enhancement of glycolytic activity, and loss of neutral lipid stores. We also discovered a previously unreported dichotomy in the role of AMPK isoforms in tumor survival following metabolic stress. Whereas AMPKα1 primarily appears to promote cell death through inhibition of the mTOR pathway and translation, AMPKα2 promotes survival by a yet unidentified mechanism. We have also identified a strategy in which these metabolic perturbations induced by tamoxifen can be exploited for therapeutic benefit through combination therapy with glycolytic inhibitors

Although the clinical benefit of tamoxifen in ER⁺ breast cancer patients has been mainly attributed to the inhibition of ER activity, tamoxifen has previously been shown to have multiple effects independent of ER. We have further characterized the effects of tamoxifen on mitochondria by describing the dose-dependent effects on oxygen consumption to occur independent of ER. We have also demonstrated that AMPK activation occurs in both ER⁺ and ER⁻ breast cancer cells and that it is independent of ER signaling. In addition, tamoxifen activated the AMPK pathway in an orthotopic model of triple-negative breast cancer. Although we treated these animals for a short time at a high dose, we would anticipate similar effects at a lower dose over a longer period of time, as tamoxifen has been shown to accumulate in tumor tissues to levels up to 60 to 70 times the plasma concentration (10).

The non-ER–dependent effects of tamoxifen described here have the potential to expand the therapeutic window for tamoxifen into ER⁻ BC and non-breast cancers. Furthermore, most ER⁺ BC related deaths have been attributed to metastatic disease, which is often tamoxifen resistant (2). Although the mechanism by which tamoxifen resistance occurs remains unclear, it is thought to involve alterations in ER itself and alterations in the levels of coregulators or downstream effectors in the signaling pathway (2). Our results suggest that it would be valuable to further explore whether combination therapies that take advantage of the metabolic effects of tamoxifen could restore tamoxifen sensitivity or prevent resistance.

The precise target for tamoxifen in the mitochondrial respiration chain is not currently clear. The evidence presented in this article supports the hypothesis that tamoxifen inhibits complex I in the electron transport chain (see Fig. 7). The ability of complex II substrate succinate to restore oxygen consumption in tamoxifen-treated cells points to the inhibition of mitochondrial respiration upstream of complex II. Moreover, the inability of complex I substrates glutamate and malate to restore any oxygen consumption in permeabilized breast cancer cells indicates that complex I may be the site of tamoxifen action (Fig. 7). However, we cannot completely rule out possible effects of tamoxifen on upstream processes, such as the TCA cycle, that contribute electron donors to the respiratory chain. Another implication of our findings with translational potential is the fact that inhibition of oxygen consumption is a validated strategy to increase tumor pO2 and decrease hypoxia, a major cause of cancer resistance to genotoxic therapy, including radiation (35). Therefore, tamoxifen has the potential to function as a radiation and chemotherapeutic sensitizer. In fact, other complex I inhibitors, such as metformin, have shown anticancer effects through both their metabolic effects and activation of AMPK. Although our results are reminiscent of the metabolic effects of metformin treatment, both the inhibition of oxygen consumption and activation of AMPK occur at a far more rapid rate (5–10 minutes) than that seen with metformin (12–24 hours). Also important is the distinction that although in vitro metformin exerts its OXPHOS-inhibitory effects at doses ranging from 1 to 50 mmol/L, tamoxifen does so at low micromolar concentrations.

We also show that like metformin, tamoxifen inhibited OXPHOS and increased dependence on glycolysis (Fig. 7). Intriguingly, soon after breast cancer patients begin treatment with tamoxifen, a “metabolic flare” is often seen in their FDG-PET scans (36). Although the cause for this radiologic abnormality has not been clearly delineated, it is reasonable to hypothesize that this could be due to increased uptake of 2[18F]fluoro-2-deoxy-D-glucose (FDG) in response to metabolic reprogramming and AMPK activation, as GLUT 1 transporters, the major transporter system of glucose and FDG uptake, are AMPK targets (37).

Our results are also in agreement with other metabolic effects of tamoxifen in patient populations. Women on tamoxifen therapy have an increased risk for developing nonalcoholic fatty liver disease (38). However, the effects of tamoxifen on tumor fatty acid metabolism have not been investigated in detail. Accumulation of malonyl-CoA following tamoxifen treatment has been reported previously (39), and we also observed this in breast cancer cells. The synthesis of malonyl-CoA is the rate-limiting step in fatty acid synthesis; thus, its accumulation leads us to suspect that there is a decrease in the activity of FASN, which is consistent with AMPK activation. It is also possible that increased upstream flux through glycolysis contributes to the accumulation of malonyl-CoA with tamoxifen treatment. It is worth noting that AMPK is also known to inhibit fatty acid synthesis and promote β-oxidation, primarily through inhibiting ACC (21).

The antitumorigenic effects of metformin in breast cancer have been shown to be dependent on its activation of AMPK (40). We asked whether AMPK had a similar role in beast cancer survival of tamoxifen treatment. We used the CRISPR-Cas9 system to delete AMPKα1 from MDA-MB-231 cells and observed a compensatory increase in AMPKα2 expression. This would lead us to hypothesize that the isoforms have redundant effects, yet we observed opposing effects on cell survival. Other work examining the isoforms of AMPKα has identified some differences, including the suppression of α2 expression in primary breast cancer (41). Knockout of AMPKα1 alone rescued cells from tamoxifen treatment, whereas knockdown of AMPKα2 increased cell death. Immunoblot analysis and [³⁵S] labeling of proteins indicated that the α1 subunit is responsible for downregulation of the MTOR pathway and translation. However, both isoforms are needed to inhibit ACC, and either isoform can activate autophagy. We suspect that it is the loss of robust autophagy induction in the dual AMPKα1α2 knockdown that results in increased tamoxifen cytotoxicity under those conditions. The ability of tamoxifen to activate autophagy is well established. In fact, the autophagic response to tamoxifen has been implicated in breast cancer resistance to tamoxifen treatment (42, 43). In addition, the ability of AMPK to activate autophagy via ULK1 is also well described (44). The data reported in this manuscript are the first to directly identify AMPK activation as a mechanism of autophagy induction in tamoxifen treatment. In summary, we have identified AMPK as a mediator of tamoxifen toxicity through isoform-specific effects.

We have also demonstrated the potential for increased therapeutic benefit through drug combinations with tamoxifen that have a synergistic effect via blockade of upregulated metabolic pathways. Because of the fact that inhibition of oxygen consumption by tamoxifen results in upregulated glycolysis, effective strategies for metabolic interventions may require the inhibition of both OXPHOS and glycolysis simultaneously. Glycolytic inhibitors have long been attractive clinical agents due to the Warburg effect, yet systemic toxicity has limited their success in the clinic. 2DG has been tested in clinical trials, and 3BP has shown significant preclinical potential (45). Our data suggest that clinical use of these inhibitors in combination with tamoxifen may result in increased efficacy and use of reduced doses that may limit toxicity.

In conclusion, we have characterized the ER-independent effects of tamoxifen on cancer metabolism and described how these effects can be exploited for therapeutic benefit through novel drug combinations with the potential to expand the clinical use of tamoxifen.

C. Koumenis is a consultant/advisory board member for Ruga Pharmaceuticals. No potential conflicts of interest were disclosed by the other authors.

Conception and design: N.A. Daurio, S.W. Tuttle, E.Y. Song, N.W. Snyder, I.A. Blair, C. Koumenis

Development of methodology: N.A. Daurio, S.W. Tuttle, A.J. Worth, N.W. Snyder, I.A. Blair

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): N.A. Daurio, S.W. Tuttle, A.J. Worth, E.Y. Song, J.M. Davis, N.W. Snyder

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): N.A. Daurio, S.W. Tuttle, A.J. Worth, E.Y. Song, N.W. Snyder, I.A. Blair, C. Koumenis

Writing, review, and/or revision of the manuscript: N.A. Daurio, S.W. Tuttle, A.J. Worth, J.M. Davis, N.W. Snyder, I.A. Blair, C. Koumenis

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C. Koumenis

Study supervision: I.A. Blair, C. Koumenis

This study was supported by NIH grants 2R25CA140116-06, P30CA016520, P30ES013508, T32ES019851, and R01CA182747.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.