The role of cancer cell bioenergetics in dormancy and drug resistance

Abstract

While anti-cancer drug treatments are often effective for the management of cancer, these treatments frequently leave behind drug-tolerant persister cancer cells that can ultimately give rise to recurrent disease. Such persistent cancer cells can lie dormant for extended periods of time, going undetected by conventional clinical means. Understanding the mechanisms that such dormant cancer cells use to survive, and the mechanisms that drive emergence from dormancy, are critical to the development of improved therapeutic strategies to prevent and manage disease recurrence. Cancer cells often exhibit metabolic alterations compared to their non-transformed counterparts. An emerging body of evidence support the notion that dormant cancer cells also have unique metabolic adaptations that may offer therapeutically tractable vulnerabilities. Herein, we review mechanisms through which cancer cells metabolically adapt to persist during drug treatments and develop drug resistance. We also highlight emerging therapeutic strategies to target dormant cancer cells via their metabolic features.

Introduction

The concept of dormancy in solid tumors stems from the clinical observation that between the resection of the primary tumor and the detection of disease relapse, there is a gap of time in which tumor burden is clinically undetectable. This suggests that a subpopulation of tumor cells survives the period of latency and eventually gives rise to recurrent disease. Preclinical studies have termed this cell population dormant, quiescent, tolerant, or persistent, collectively representing the state between drug sensitivity and drug resistance [1]. Herein, we term this population “dormant cells,” which can be non-cycling or slow-cycling, and can exhibit stem-like properties.

The time when dormancy occurs during the metastatic cascade, between the time of leaving the primary site and the time of colonization at a distant site, remains an open question. Nonetheless, dormant cells have been detected at distant sites, and there are shared mechanisms in the survival of cells leaving the primary site and dormant cells such as autophagy and the unfolded protein response stemming from adaptations to cellular stress. Persistence is broadly described as a general mechanism of survival to cytotoxic therapy. However, it remains to be determined whether the class of therapeutic (or therapeutic target) dictates a cancer cell persistence phenotype. Evidence supports a general mechanism whereby treatment of BRAF^V600E-driven melanoma cells with either cisplatin or the BRAF^V600E-selective inhibitor vemurafenib elicits the emergence of a slow-cycling JARID1B^high dormant cell population [2].

Metabolic adaptation is a well-established hallmark of cancer, with studies showing that tumor cells rewire metabolic pathways to promote the uptake and anabolism of nutrients to drive cell proliferation. The ever-changing dynamics of cancer progression select for cells that exhibit phenotypic plasticity to adapt to new microenvironments and levels of nutrient availability. For example, brain-metastatic breast cancer cells exhibit increased lipid synthesis compared to non-metastatic cells [3]. A key adaptation that cancer cells undergo in response to therapeutic intervention is a rewiring of metabolism to persist under treatment. Herein, we review the metabolic adaptations of dormant cancer cells, therapeutic strategies to target metabolic vulnerabilities, and biomarkers to track these metabolic changes.

Dormant cancer cells share similarities to terminally differentiated, non-cancer cells

Metabolic pathway utilization and nutrient uptake in mammalian cells is commanded by proliferation status. When stimulated by growth factors, proliferating cells internalize glucose and convert it to biomass, while quiescent differentiated cells primarily utilize glucose for energy storage in the form of ATP synthesis. Cancer cells metabolically differ from their non-transformed counterparts by usurping growth factor signaling pathways to drive unchecked growth. The role of mitochondria differs in proliferating vs. quiescent cells: mitochondrial metabolism serves to produce intermediates such as citrate and aspartate to support the biosynthesis of lipids, amino acids, and nucleotides in proliferating cells; in contrast, ATP is the main output of mitochondrial metabolism in quiescent cells [4]. Cancer cells also commonly utilize aerobic glycolysis (Warburg effect) to enable shunting of metabolic intermediates towards biosynthesis of macromolecules that promote cell growth and proliferation [5]. Cells can exhibit preferences for types of fuel (e.g., glucose, fatty acids), depending in part on their phenotype and microenvironment [6], and it was recently shown that nutrient uptake was driven more by nutrient preference than availability [7].

Similarities in metabolic adaptations exist between (i) dormant cancer cells and (ii) non-transformed cells that lack growth factor stimulation or are quiescent: even in the presence of ample nutrients, cells without growth factors arrest cycling and deplete their energy sources by downregulation of key ATP-generating pathways such as glycolysis and OXPHOS [8]. Such effects are reversed upon re-introduction of growth factors. Under most therapeutic interventions, cancer cells similarly arrest growth and alter their demand for biosynthetic intermediates and energy. This shift from (i) a state of necessitating intermediates for anabolism and cell cycle progression to (ii) a state of survival induces alterations in metabolic pathways. These pathways activated by clinically approved cancer therapeutics can be compared to that of metabolic stress. Such examples include the induction of fatty acid oxidation from glucose deprivation in a p53-dependent manner [9, 10], suggesting that a common evolutionary strategy directs the response to insult.

AMPK drives cancer cell persistence

Dormant cells are often defined by their slowed cycling kinetics. Due to the links between cell cycle progression and metabolic pathways, it can be inferred that many of the metabolic adaptations of dormant cells are caused by cell cycle disruption. For example, cyclin D1 levels inversely correlate with mitochondrial size and activity in a cyclin-dependent kinase (CDK)-dependent manner [11]. More recently, CDK4 has been found to act directly on AMP-activated protein kinase (AMPK) alpha 2 and to repress fatty acid oxidation [12]; CDK4 canonically drives G1/S progression in a cyclin D1/D3-dependent manner. Taken together, these results suggest that alterations in metabolism that enable cancer cell persistence are directly tied to changes in cell cycle signaling.

In a dormant model of estrogen receptor-positive (ER+) breast cancer, AMPK was found to be upregulated, subsequently increasing fatty acid oxidation and mitochondrial respiration [13]. AMPK is generally regarded as an energy sensor, activated by a low-energy state that increases the AMP to ATP ratio. Thus, AMPK is engaged in times of energy depletion. AMPK has also emerged as a master regulator of mitochondrial homeostasis, promoting both mitochondrial biogenesis and fission [14]. The role of AMPK in cancer seems paradoxical: AMPK has been described as a tumor suppressor, with the observation that AMPK activation by metformin decreases cancer incidence in diabetic patients [15]; however, AMPK activation has been linked to cancer cell survival during metabolic stress [16-19]. Indeed, activation of AMPK with the AMP mimetic AICAR or the small molecule A-769662 decreased viability in proliferating ovarian cancer cells whilst having no effect in dormant ovarian cancer spheroids. Furthermore, knockdown of LKB1, which in turn induced AMPK inhibition, reduced viability and increased sensitivity to carboplatin in the dormant spheroids [20]. Downstream targets of AMPK also promote cell survival under stress. For example, carnitine palmitoyltransferase 1C (CPT1C) is rate-limiting for fatty acid import into mitochondria. CPT1C upregulation promotes cell survival during metabolic stress by promoting fatty acid oxidation, buffering against depletion of alternative fuels such as glucose [10]. AMPK activation has been shown to drive macropinocytosis of necrotic cells as a source of nutrients in prostate cancer cells [21]. In addition, AMPK-mTORC1 pathway activation driving autophagy is necessary for the formation of dormant polyploid giant cancer cells in nasopharyngeal carcinoma [22]. Such findings support the notion that AMPK activation in dormant cancer cells buffers against stress by intrinsically rewiring metabolism.

Cancer cell dormancy and lipid metabolism

Fatty acids serve as building blocks for phospholipids to sustain membrane integrity as well as substrates for oxidation. Fatty acid oxidation drives the production of ATP and acetyl-CoA that feeds into the TCA cycle, and produces NADPH to buffer oxidative stress. Many cancer types have been shown to be dependent on fatty acid oxidation, including MYC-driven triple-negative breast cancer [23]. In addition, free fatty acids in plasma have been identified as a component of obesity-related risk breast cancer incidence [24].

Following the establishment of tumors induced by the MYC and KRAS^G12D oncogenes in transgenic mice, Havas et al. withdrew oncogene expression that yielded a persistent tumor cell subpopulation exhibiting alterations in the oxidative phosphorylation (OXPHOS) and fatty acid oxidation pathways [25]. A similar approach of withdrawing the ERBB2 (HER2) oncogene in mammary tumors in mice elicited a dormant tumor cell subpopulation dependent upon fatty acid oxidation [26]. Fatty acid oxidation is known to occur in two organellar compartments: peroxisomes where this process is rate-limited by the ACOX flavoenzymes; mitochondria where this process is rate-limited by fatty acid transport across the mitochondrial membranes by CPT1/2. Dormant melanoma cells treated with pharmacologic inhibitors of the oncoproteins BRAF^V600E and MEK1/2 were found to upregulate peroxisomal but not mitochondrial fatty acid oxidation [27]. Inhibition of peroxisomal fatty acid oxidation by knockdown of PPARA (PPARα) or ACOX1, or by treatment with the small molecule thioridazine, selectively killed the dormant melanoma cells by steatosis-mediated cell death and repressed emergence from dormancy.

Upstream of fatty acid oxidation lies fatty acid import into cells, facilitated in part by the fatty acid transporter CD36 [28]. CD36 is upregulated in a variety of tumor types [29]. If fatty acid synthesis is pharmacologically inhibited, cells upregulate CD36 to compensate [30], and inhibition of CD36 improves the efficacy of fatty acid synthase inhibition [31]. Moreover, CD36⁺ cells have been identified in a slow-cycling stem-like population of oral carcinoma cells with metastatic potential [32]. CD36⁺ chronic myelogenous leukemia (CML) cells were more resistant to the BCR-ABL inhibitor imatinib compared to CD36⁻ cells, implicating CD36 in drug tolerance/resistance [33]. A polyclonal anti-CD36 antibody was developed to target these refractory CD36⁺ CML cells, leading to antibody-dependent cellular cytotoxicity while sparing normal bone marrow. Strategies to target CD36 and other metabolic interventions with clinical potential are illustrated in Figure 1. Although CD36 positivity is predictive of fatty acid oxidation in dormant melanoma tumors, it is not required for fatty acid oxidation in dormant cells, suggesting that cells have access to multiple routes of acquiring or synthesizing fatty acids [34].

Figure 1. Summary of key metabolic pathways in dormant cells and strategies for therapeutic intervention.

Dormant cancer cells activate AMPK that signals to increase utilization of fatty acid oxidation, feeding acetyl-CoA into the TCA cycle and driving oxidative phosphorylation (OXPHOS). Cells acquire fatty acids either by import via CD36 or from fatty acid synthesis. Therapeutic strategies depicted here involve targeting fatty acid synthesis and import, the TCA cycle, mitochondrial biology, OXPHOS, and dietary restriction. ETC- electron transport chain

Fatty acid oxidation may be an evolutionarily conserved mechanism to buffer cellular oxidative stress. In metastatic breast cancer cells, upregulation of aldo-keto reductase family 1 member B10 (AKR1B10) resulted in downregulation of glycolysis and upregulation of fatty acid oxidation, which in turn decreased the levels of reactive oxygen species (ROS) in low glucose conditions [35]. Further supporting the notion that alterations in the microenvironment can change fuel preference is evidenced by bile acids in lymph nodes. Bile acids canonically promote dietary fat digestion, but have also been found to activate YAP through regulation of the Hippo pathway [36]. Uptake of bile acids has been proposed to mediate metabolic reprogramming towards fatty acid oxidation through YAP activation for melanoma cells to adapt to the lymph node microenvironment [37].

Perhaps the same mechanisms that drive tumor cell persistence can contribute to eventual relapse. Oren et al. transiently expressed mCherry-tagged histone 2A as a reporter of lung cancer cell proliferation, whereby dilution of mCherry signal reflects proliferation [38]. Following treatment with the EGFR inhibitor osimertinib, drug-tolerant cells were non-cycling or cycling, the latter of which were proposed as the subpopulation that escapes dormancy to cause relapse. Interestingly, fatty acid oxidation was associated with the cycling population that showed a concomitant decrease in ROS, suggesting an anti-oxidative benefit of fatty acid oxidation perhaps by regeneration of NADPH [39, 40].

While the impetus for increased fatty acid oxidation can originate from mitochondrial demand, ER+ breast cancer cells showed increased respiration upon treatment with exogenous oleic acid [13]. Thus, increased availability of fatty acid can promote its utilization by cancer cells, and fuel availability can dictate fuel preference. Similarly, mitochondria derived from triple-negative breast cancer are able to upregulate fatty acid oxidation to support respiration and ATP generation [41].

Dormant cancer cells can rely on mitochondrial OXPHOS, the TCA cycle, and/or glycolysis

In an inducible Kras^G12D transgenic mouse model, Viale et al. generated pancreatic tumors. Upon Kras^G12D withdrawal, pancreatic tumors regressed. Dormant pancreatic cancer cells exhibited increased levels of peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC1α) compared to cells in growing tumors. PGC1α is a master regulator of energy metabolism [42], controlling multiple facets of mitochondrial metabolism including biogenesis, OXPHOS, fatty acid synthesis, and degradation. Further investigation revealed a reliance upon mitochondrial OXPHOS in dormant cells, and an inability to upregulate glycolysis when OXPHOS was inhibited. Kras^G12D is a driver of glucose metabolism, promoting the shunting of metabolic intermediates towards hexosamine biosynthesis and the pentose phosphate pathway [45]. In glucose-depleted conditions, OXPHOS is a major pathway required for survival and proliferation [46], highlighting these glycolysis and OXPHOS as the two central metabolic pathways vital for energy maintenance. Surprisingly, AMPK was not activated in dormant pancreatic cancer cells, suggesting that energy supply balanced demand due to compensation from mitochondria, and/or that energy demand was lower in dormant cells. Inhibition of fatty acid oxidation did not alter respiratory capacity in Kras^G12D-withdrawn cancer cells [45], but decreased respiration in dormant ER+ breast cancer cells [13], highlighting the diversity of fuel sources that can feed into the TCA cycle for OXPHOS and potential differences between cancer types.

Farge et al. found that residual AML cells after cytarabine treatment have increased mitochondrial mass and depend upon CD36 for fatty acid uptake and subsequent OXPHOS [47]. Furthermore, inhibiting either (i) mitochondrial translation with tigecycline (Figure 1) or (ii) fatty acid oxidation with the CPT1 inhibitor etomoxir effectively decreased viability in the residual cell population. Similarly, AML cells tolerant of both cytarabine and the BCL-2 inhibitor venetoclax exhibit transcriptional signatures of OXPHOS and electron transport chain signaling. An OXPHOS transcriptional signature identified which AML patients benefit most from cytarabine and venetoclax, a disruptor of mitochondrial metabolism [48]. Likewise, high mitochondrial membrane potential can predict acute resistance to cytarabine in AML cell lines [49]. Analogous observations in TNBC showed a transition to a more OXPHOS-dominant phenotype in residual cancer cells after chemotherapy [50]. Therein, treatment with the complex I inhibitor IACS-010759 (Figure 1) delayed tumor regrowth, indicating that inhibiting mitochondrial OXPHOS can target dormant cells. In agreement with these findings, mitochondrial OXPHOS can promote resistance to chemotherapy and the expansion of cancer stem cell populations through upregulation of MYC and MCL [53]. Mitochondrial activity has been associated with cancer stemness as evaluated by an upregulation of mitochondrial proteins in breast cancer mammosphere culture [52].

There is evidence that the microenvironment can contribute to a shift towards greater reliance upon OXPHOS by cancer cells. Buschhaus et al. utilized both an in vitro three-dimensional spheroid co-culture model with breast cancer cells and bone marrow stromal cell lines, and an in vivo mouse model with human breast cancer cells implanted in bone marrow [51]. Cancer cell quiescence of co-cultures was demonstrated by G1 arrest, and decreased NADH turnover suggested greater utilization of OXPHOS compared to monocultures, indicating that the bone marrow microenvironment can induce metabolic reprogramming of cancer cells.

In most of the aforementioned models, a rise in the dependency upon mitochondrial OXPHOS was concomitant with a decrease in reliance upon glycolysis. However, not all models of tumor cell persistence exhibit downregulation of glycolysis. CDK4/6 inhibition resulted in an upregulation of glycolysis and the metabolism of glucose and glutamine driven by MTOR pathway activation [54]. A hypermetabolic state where all levels of metabolism are increased might reflect increased demand. Gefinitib-induced DTPs revealed two cell subpopulations: a CD133^high stem-like phenotype, and a CD133^low senescent cell phenotype [55]. The senescent cells displayed the senescence-associated secretory phenotype (SASP), which has higher metabolic demands, and were sensitive to inhibition of glycolysis. Another model of SASP in lymphoma displayed similar features where glycolysis inhibition was effective as a therapeutic strategy [56].

Human breast cancer cells tolerant of taxanes, which exhibit increased expression of the cancer stem cell marker CD44, engage both increased glycolytic and oxidative pathways [57]. Engagement of glycolysis is attributed to activation of hypoxia-inducible factor 1α (HIF1α) that in turn activates expression of the glucose transporter Glut1. Cell persistence during osimertinib treatment resulted in upregulation of miR-147b in non-small cell lung cancer [58]. This resulted in suppression of mitochondrial complex II (succinate dehydrogenase) due to Von Hippel-Lindau tumor suppressor (VHL) loss, leading to a pseudo-hypoxia gene expression signature. In addition, imatinib treatment of CML cells forced a switch from OXPHOS to glycolysis with an observed upregulation of HIF1α signaling [59]. Cells are evolutionarily programmed to engage in glycolysis when oxygen levels are too low to support OXPHOS. Subsequent activation of HIF1α triggers a broad metabolic adaptation by upregulating glucose transporters, glycolytic enzymes, and inhibition of TCA by PDK1 [60]. Taken together, these results suggest that if dormant cancer cells engage HIF1α and hypoxia-related signaling, then glycolysis would be the dominant ATP-producing metabolic pathway over OXPHOS.

Emergence of resistance to approved anti-cancer drugs via metabolic shifts in dormant cancer cells

Inherent to the transition out of dormancy is the acquisition of drug resistance by a resumption of cell cycling. Whether metabolic reprogramming incurred from dormancy remains following the acquisition of drug resistance is unknown. Studies in murine leukemia cancer cell lines suggest that levels of glycolysis are higher and mitochondrial membrane potential is lower in resistant cells compared to their drug-naïve counterparts [61]. However, fatty acid oxidation was also increased in drug-resistant cells when glucose availability was limited, suggesting that metabolic features acquired in dormancy remain after acquisition of drug resistance.

It remains unclear whether alterations in metabolism can drive cancer cell outgrowth and emergence from the dormant state into a drug-resistant state. Early evidence from cells selected for resistance to doxorubicin revealed that their increased glycolysis sensitized cells to treatment with 2-deoxyglucose (2-DG) [62]. La Belle Flynn et al. utilized the D2-hyperplastic alveolar nodule model of murine breast cancer dormancy and found the glycolysis mediator 6-phospho-fructo-2-kinase/fructose 2,6-bisphosphatase 3 (Pfkfb3) to be upregulated in metastatic tumors compared to their unprimed, dormant counterparts [63]. Further investigation showed that Pfkfb3 expression can drive tumor cell emergence from dormancy, suggesting that activation of glycolysis can drive dormancy exit. In BRAF^V600E-driven melanoma cells driven into dormancy by treatment with the BRAF inhibitor PLX4720, inhibition of fatty acid oxidation with etomoxir increased glycolysis and tumor xenograft growth in mice [34]. In hepatocellular carcinoma, dormancy induced by the kinase inhibitor sorafenib resulted in greater glycolytic flux concomitant with resistance to drug [64]. This increase in glycolysis was facilitated by hexokinase 2 (HK2) activity. Some models have shown how metabolic reprogramming initiates emergence from dormancy. The enzyme glycerol-3-phosphate dehydrogenase (GPD1) was found to be critical for dormant glioma stem cell survival and regulates their metabolism by redirecting glucose away from glycolysis, leading to the generation of glycerol [65]. Interestingly, glycerol accumulation is found in silkworm eggs in the diapause state, which is an evolutionarily conserved dormancy program [66]. Furthermore, glycerol is converted back into glycogen to serve as a reservoir of fuel for glycolysis during diapause exit.

Tumor recurrence and drug resistance can also emerge from OXPHOS upregulation. Transcriptomic profiling of paired primary and recurrent human HPV-related head and neck cancers revealed that recurrent cancers generally upregulated OXPHOS in a nuclear factor, erythroid 2 like 2 (NRF2)-dependent manner [67]. As a master regulator of the anti-oxidant response, NRF2 is thought to buffer against ROS generated from OXPHOS, thereby enabling cancer cells to further increase OXPHOS. NRF2 was also found to drive tumor recurrence in a murine HER2+ breast cancer dormancy model by regulating redox homeostasis and increasing nucleotide synthesis [26]. Taken together, these results support myriad pathways, including glycolysis and mitochondrial OXPHOS, that can support emergence from dormancy and a shift toward a metabolic state for proliferation. There is still a limited understanding of the metabolic profile of awakened cells in comparison to the baseline proliferative and dormant state, and the function of how metabolic pathways support awakened cells.

Therapeutic strategies of targeting metabolic adaptations in dormancy

Inhibitors of OXPHOS

Many compounds that target OXPHOS in preclinical studies such as oligomycin [68], antimycin A, and rotenone are deemed too toxic for clinical use [69]. As such, several inhibitors are currently being evaluated to target mitochondria in cancer cells with less systemic toxicity. Recently, a selective inhibitor of complex I was developed to overcome challenges in OXPHOS inhibition, demonstrating efficacy in multiple cancer types [48, 50, 70, 71]. IACS-010759 has completed Phase I clinical testing, which provided biologically active doses in patients with advanced cancers [72]. As mitochondria have their own system of transcription and translation of proteins, a system to degrade damaged proteins exists to maintain quality control. One such mitochondrial protease, ClpP, has been found to be overexpressed in cancer and promotes OXPHOS function [73]. A series of ClpP agonists were developed as anti-cancer agents that induce degradation of OXPHOS proteins [74], with one agent, ONC-212, displaying efficacy against refractory AML and delaying relapse in mice [48]. Recently a compound screen to target glioblastoma cells resulted in the discovery of Gboxin, an inhibitor of ATP synthase that selectively targets increased proton gradients [75]. Gboxin has displayed cell killing in glioblastoma while sparing astrocytes. One of the concerns of targeting mitochondrial OXPHOS is the ability of cells to transition to alternative ATP-generating pathways such as glycolysis [70, 76]. While such adaptability may confer resistance to OXPHOS inhibitors in some cancers, cancers that are inherently deficient in glycolysis or residing in microenvironments with low glucose levels may be sensitized to OXPHOS inhibition [76].

Strategies to disrupt mitochondrial function

As mitochondria share a common ancestry to bacteria, many antibiotics with a mechanism of action involving inhibition of translation have been repurposed as mitochondrial poisons. Such strategies include the addition of doxycycline or tedizolid to overcome venetoclax resistance in AML [77], in addition to tigecycline to overcome drug resistance in CML [78]. Mitochondria have unique machinery to replicate DNA, and the expression of up to 37 genes encoded by the mitochondrial genome is crucial to support complexes I, III, and IV [79]. Recently, the small molecule IMT1 was discovered as a selective inhibitor of mitochondrial RNA polymerase (POLRMT) [80]. This small molecule was well-tolerated in mice and elicited an anti-tumor response while sparing normal tissue, showing promise as an effective therapy to target dormant cells.

Inhibitors of fatty acid metabolism

As cancer cells are capable of utilizing fatty acid synthesis to feed fatty acid oxidation in a feed-forward fashion [81, 82], it is warranted to consider targeting both fatty acid synthesis and fatty acid import. Early development of small molecules targeting fatty acid synthase (FASN) yielded compounds with poor pharmacokinetic properties and unreasonable adverse effect profiles [83]. Recently, small molecules (TVBs) were developed [84] that resulted in lowered adenine analogs and lowered lipid stores, suggesting ATP depletion due to fatty acid synthesis blockade [85]. This has led to completed Phase I clinical trials demonstrating low-grade adverse events with efficacy against KRAS^mutant lung, ovarian, and breast cancers [86].

Targeting fatty acid import is complicated by the existence of multiple plasma membrane importers. Nonetheless, inhibition of HER2 with the small molecule lapatinib induced increased expression of the CD36 scavenger receptor and lipid droplets in preclinical models of HER2+ breast cancer. Oral carcinoma cells also express high levels of CD36. Antibody-mediated inhibition of CD36 prevented the growth of lapatinib-resistant cells and tumors, and prevented the metastasis or oral carcinoma cells in mice [32, 87]. Although no humanized CD36 antibody currently exists, efforts are underway to develop one [88]. Targeting fatty acid catabolism at the level of mitochondrial import, such as using CPT1-alpha/beta inhibitors like etomoxir, was considered a promising anti-cancer strategy based on preclinical results [89]. However, etomoxir proved systemically toxic upon clinical testing and its development was discontinued [90].

Inhibitors of the TCA cycle

Another option for targeting mitochondrial metabolism is to inhibit enzymes that produce TCA cycle intermediates. Pyruvate dehydrogenase complex (PDC) converts pyruvate into acetyl-CoA, which can feed into the TCA cycle. A series of PDC inhibitors were developed that compete for binding to the essential cofactor lipoate. The PDC inhibitor devimistat (Figure 1) disrupts mitochondrial metabolism by inhibiting pyruvate dehydrogenase and α-ketoglutarate dehydrogenase, has efficacy against cancer cells, and synergizes with chemotherapeutics [48, 91, 92]. Several clinical trials have tested devimistat in combination with chemotherapies for several cancer types. Factors such as age-dependent decline in mitochondrial function were found to predict devimistat efficacy [93, 94]. Recently, the anti-cancer mechanism of action of the copper ionophore elesclomol was shown to be dependent upon mitochondrial respiration by inhibition of lipoylated TCA cycle enzymes [95]. This mechanism results in a selective dependency where cells more reliant upon OXPHOS are more sensitive to elesclomol [95, 96].

Dietary modification

Perhaps the strongest body of evidence in support of dietary modification has been the efficacy of fasting in potentiating chemotherapy while protecting normal tissue, resulting in a handful of clinical trials in several cancer types [97]. Furthermore, a fasting-mimicking diet can revert tumor resistance to endocrine therapy [98]. Thus, it stands to reason that dietary modification can serve as a therapeutic avenue for the treatment of dormant cancer.

An alternative strategy to drugging OXPHOS is to alter the translation of protein subunits within complex I. As with metabolic reprogramming, cancer cells can reprogram mRNA translation. It was recently found that valine is rate-limiting in complex I subunit translation in leukemia; dietary valine restriction reduced leukemic burden in mice [99]. In addition, dietary modification by limiting intake of fatty acids has been considered in controlling breast and prostate cancer progression. Specifically, linoleic acid and other n-6 polyunsaturated fatty acids have been found essential to normal and cancer cell growth. However, more work is needed on the underlying mechanism(s) and the types of fatty acids that promote cancer progression [100].

Conjugated linoleic acid (CLA) is a dietary fatty acid and supplement shown to suppress fatty acid synthesis to inhibit growth of breast cancer cells. Mechanistic studies showed that CLA decreases expression of the genes encoding FASN and SPOT14, a regulator of fatty acid synthesis [101]. Dietary CLA supplementation for ≥10 days prior to surgery decreased SPOT14 protein levels in primary human breast tumors [102], suggesting that dietary modification may be used to inhibit cancer cell lipogenesis.

Perspective

While the treatment of early-stage, site-localized tumors with surgical resection +/− radiotherapy is highly effective and frequently curative, patient outcomes following the treatment of advanced/recurrent/metastatic disease are relatively poor and often palliative. Indeed, metastatic cancers are rarely cured, often ultimately exhibit resistance to all treatments, and are usually fatal. Thus, the prevention of cancer recurrence would drastically improve patient outcomes through reduced morbidity and mortality. Cancer recurrence can be prevented through improved treatment during the (neo)adjuvant phase, where the main goal of systemic therapy is to target disseminated and micrometastatic cancer cells. Improved (neo)adjuvant treatment targeting the cancer cells that are destined to become drug-tolerant persisters, or which already exhibit features of drug tolerance, would decrease residual cancer burden in patients. However, determining whether decreasing residual disease burden translates into reduced risk of recurrence will require prospective clinical testing.

Herein, we reviewed metabolic adaptations that can enable cancer cells to persist during treatment with FDA-approved drugs. If a novel treatment strategy were developed preclinically to target dormant cancer cells, existing clinical testing paradigms are not readily amenable to evaluation of such a treatment. The status quo for the typical cancer drug development paradigm is to demonstrate efficacy in the advanced/metastatic setting, obtain regulatory approval, and then test the drug in earlier disease settings. However, drugs that target vulnerabilities unique to (occult) dormant cancer cells, such as the metabolic phenotypes described herein, may be ineffective against advanced/metastatic (measurable) tumors, and so such drugs would fail early-phase clinical testing in subjects with late-stage disease.

Another confounder in the typical cancer drug development paradigm is the requirement for demonstration of preclinical efficacy against established tumors prior to clinical testing. Once safety in humans has been demonstrated, those same drugs are allowed to be tested in patients in the (neo)adjuvant setting to prevent recurrence without preclinical evidence of efficacy against dormant/disseminated cancer cells. The biology of dormant cancer cells is not well-characterized in most cancer types, and drug effects on such cells are mostly unknown, calling into question the biological basis for such clinical trials. The lack of requirement for preclinical demonstration of efficacy against dormant cancer cells jeopardizes the success of (neo)adjuvant clinical trials in addition to subjecting patients to toxicities from potentially ineffective treatments. Indeed, numerous agents shown to be effective against metastatic tumors have failed to prevent recurrence when tested in the (neo)adjuvant setting [103-109]. The drug development paradigm to target dormant cancer cells, ideally in the (neo)adjuvant setting, will require revision to enable efficient and rational translation from preclinical to clinical testing.