Fasting, Fats, and Physics: Combining Ketogenic and Radiation Therapy against Cancer

Chemotherapie · Differenzielle Stressresistenz · Fasten · Ketogene Diät · Strahlentherapie

Die Strahlentherapie ist eine Hauptstütze in der Behandlung solider Tumoren und wirkt über physikalisch-chemische Reaktionen, die oxidativen Stress in Zellen erzeugen. Die Wirkung auf Tumorzellen wird durch die Toxizität für das mitbestrahlte gesunde Gewebe limitiert. Deshalb ist jede Behandlung, durch die eine selektive Sensibilisierung von Tumorzellen bzw. eine Stärkung normaler Zellen gegenüber ionisierender Strahlung erreicht wird, eine wertvolle Ergänzung der Strahlentherapie. In dieser Übersichtsarbeit fasse ich Daten aus vorklinischen und klinischen Studien zusammen, die für eine solche selektive Wirkung einer ketogenen Therapie durch Fasten und/oder eine ketogene Diät sprechen und eine Verbesserung der Tumorkontrollwahrscheinlichkeit bzw. eine Senkung der Wahrscheinlichkeit für eine Normalgewebeschädigung nahelegen. Erstere Wirkung hängt mit der metabolischen Umstellung von der Glykolyse zum mitochondrialen Stoffwechsel zusammen, was für Tumorzellen eine Erhöhung der freien Sauerstoffradikale und eine verminderte ATP-Generierung zur Folge hat. Der zweite Effekt beruht auf dem Phänomen der differenziellen Stressresistenz, das durch eine Verminderung von Wachstumsfaktoren und Glukose sowie eine Erhöhung der Ketonkörper vermittelt wird und gesunde Zellen, aber nicht Tumorzellen von Wachstum auf Stressresistenz umprogrammiert. Beide Effekte beruhen letztlich auf der metabolischen Verschiedenheit gesunder und maligner Zellen, die durch eine ketogene Therapie als Zusatz zur Strahlentherapie kostengünstig und vergleichsweise nebenwirkungsfrei ausgenutzt werden könnte und daher weitere klinische Erforschung verdient.

The incidence of cancer is increasing worldwide, placing a high burden not only on health care systems but also on the global society as a whole [1]. While much of this increase is attributed to an increasing life expectancy, many lines of evidence suggest that old age is not sufficient for cancer progression to a life-threatening disease; instead, most cancers, in particular the most common ones of epithelial origin, seem to develop under the influence of a perturbed metabolism associated with the modern lifestyle [2]. Obesity, chronic low-grade inflammation, and high blood glucose, insulin and insulin-like growth factor 1 (IGF-1) levels have been associated with a higher risk not only of developing but also of dying from various cancers [2, 3]. We [4] and others [5, 6, 7] have argued that the deviation from our ancestral dietary patterns [8], most notably the strong increase of carbohydrate consumption, plays a significant role in this metabolic transition and tumorigenesis. Consistent with this are reports from the first half of the 20th century of a complete lack of or much fewer cancer cases among native societies compared to civilized white people [5, 9, 10, 11, 12, 13], despite many natives reaching an old age [9, 13]. Notably, all these native societies seemed to rely on a high consumption of animal-based foods until these foods were replaced by mostly refined grains and sugar, upon which the cancer incidence rose sharply [5, 11]. As Urquhart put it in 1935 [13]: ‘Some associate [the absence of cancer among Eskimos and Indians] with the extraordinarily simple diet of the natives … [which] is remarkable for its very high proportion of fat and its almost complete lack of carbohydrates. It consists almost entirely of fat and protein.'

Such diets of mostly fat and almost complete lack of (digestible) carbohydrate are referred to as ketogenic diets (KDs) because they promote ketogenesis, in this way mimicking the metabolic state of fasting. Already 22years prior to Urquhart's statement, Van Alstyne and Beebe [14] had established that feeding a carbohydrate-free diet to rats had a significantly protective effect against transplantation of the Buffalo rat sarcoma. In the 1920's, Otto Warburg and coworkers provided a mechanistic link between carbohydrate metabolism and tumor growth by showing that tumor tissue in general consumes several-fold more glucose than normal tissue and excretes high amounts of lactate even under aerobic conditions, a fundamental difference to fast proliferating normal tissue in which respiration would minimize lactate production [15, 16, 17]. This phenomenon was first confirmed in vivo by Cori and Cori [18], by comparing blood samples from a tumor-bearing wing and a healthy wing in chickens and a forearm sarcoma with the healthy arm in 1 patient. It is nowadays termed the ‘Warburg effect' and is utilized clinically for imaging and staging tumors via 2-deoxy-2-[¹⁸F]fluoro-D-glucose positron emission tomography (FDG-PET) [19]. Based on Warburg's observations, in 1941/42, the Munich physician Wilhelm Brünings tested the antitumor effects of drastic blood glucose reduction via a KD combined with insulin injections in head-and-neck cancer patients [20, 21]. He reported remarkable results, which were, however, not reproduced by a different group [22]. There are 2 other reports of cancer treatment using Brünings's method: The one from 1952, however, made no mention of the diet component of the treatment [23] and the one from 1957 reported some favorable results but utilized a calorie-restricted low-carbohydrate diet of 124 g carbohydrate, 66 g fat, and 56 g protein per day, which did not induce ketosis [24]. Any reports mentioning a KD for cancer patients seemed to have vanished from the literature, until 1995, when Nebeling et al. [25] published 2 cases of malignant brain tumors in children that were successfully treated with a KD and standard therapy. To date, more than 24 reports have been published, in which more than 200 cancer patients were described to have been treated with a KD [26].

The Warburg effect is sometimes confused with the Warburg hypothesis [27], proposed by Otto Warburg in 1956, stating that cancer is a disease of irreversibly damaged respiration [28]. The Warburg hypothesis has gained considerable experimental support, in particular in its re-formulation as the hypothesis that dysfunctional mitochondria are a frequent characteristic of cancer cells, forcing them into a dependence on substrate fermentation, in particular of glucose [29, 30, 31]. Newer data reveal that glycolysis not only serves as a compensatory pathway for adenoside triphosphate (ATP) production but also serves multiple purposes, among which the production of building blocks for cell proliferation and anti-oxidative substrates for protection against reactive oxygen species (ROS) are of high importance [32, 33]. Consistently, experiments depriving cancer cells of glucose have associated cell death with both energy stress [34, 35] and oxidative stress [36, 37, 38, 39, 40]. Targeting this weakness of tumor cell metabolism through nutritional strategies has recently gained interest as a measure to support pro-oxidative therapies such as ionizing radiation (IR) or hyperbaric oxygen (HBO) [41, 42, 43, 44]. The majority of newly diagnosed cancer patients will be confronted with radiotherapy (RT) at some point during their treatment [45]. Recent technical developments have made RT ever more precise, and hence more effective, but also safer with regard to the occurrence of side effects. Still, however, the doses that can be applied in a single RT fraction are limited by their toxicities to normal tissue. Therefore, any intervention that selectively inhibits the defense mechanisms of tumor cells and/or increases the resistance of normal cells to IR would be a highly valuable addition to this standard treatment. In this article, I summarize the evidence for using ketogenic therapy to improve the therapeutic window in RT, with the incorporation of newer findings that have been published since our first review on this topic [42].

I distinguish between dietary restriction, which is a general term describing any form of targeted restriction of either macronutrients or total energy intake and includes carbohydrate and protein restriction, and calorie restriction (CR), which defines diets restricting the total energy intake without inducing malnutrition. Prescribing x% CR means that energy intake is restricted to (100 - x)% of that which would be consumed ad libitum, with x being typically in the range of 20-50. Fasting is the most extreme form of CR (x = 100) and is usually limited to a maximum of 3 days, which I refer to as short-term fasting (STF). KDs are defined as isocaloric high-fat diets, in which fat usually accounts for ≳75% of the energy intake. Because the adaptions to fasting are mainly driven by the absence of carbohydrates [46], KDs are fasting-mimicking diets, mainly by their elevation of the ketone bodies, β-hydroxybutyrate (BHB) and acetoacetate (AcAc) [47, 48, 49]. Ketogenic therapy is an umbrella term describing the application of nutritional strategies (CR, KDs, fasting, or exogenous ketone bodies) with the goal to induce systemic ketosis for therapeutic purposes [50]. Unfortunately, studies on CR rarely measure ketone body levels, but results from those that have done so reveal significant elevations of ketone body levels, at ≳30% CR in mice [51, 52, 53]. Mahoney et al. [51] have shown that an average of 40% CR over 3 weeks resulted in a 367% elevation of the BHB concentrations and a 41% drop in the glucose levels in mice, results that are comparable to very-low-calorie diets or STF in humans. Due to this translation, this review focusses on STF and KDs that have been shown to be feasible in their application to cancer patients in first pilot studies.

Modern-day RT is mainly delivered percutaneously using polyenergetic X-rays produced in 4-18-MV linear accelerators [54, 55]. Upon entering the body, the photons are attenuated through the 3 main interactions of the photoelectric effect, Compton scattering, and - starting at energies greater than 1,022 keV - pair production. These processes ionize atoms and molecules, setting free electrons, which in turn can induce ionization and excitation events until they are completely absorbed. The initial ionization events occur within 10^-16-10^-13 s, which is called the physical phase [54, 56]. In the subsequent physicochemical phase, the absorbed energy is distributed within and between molecules through thermodynamical energy transfer, producing free radicals (highly reactive molecules with an unpaired valence electron) and other reactive particles and molecules within about 10^-13-10^-2 s after radiation exposure. The most common event is the radiolysis of cell water, in which different interactions of water with a photon or free electron yield free radicals (H^·, ^·OH), hydrated electrons (electrons surrounded by a shell of 5-7 water molecules), and other reactive compounds (H₂, H₂O₂, OH^-) [54, 57]. The action of ionizing radiation is classified as direct or indirect, depending on whether the absorption of energy (ionization or excitation) occurs within the same molecule that is damaged or on surrounding molecules that diffuse to and react with the target molecule [58].

Warters and Hofer [59] and Warters et al. [60] provided evidence that the crucial target of RT is the nuclear DNA (nDNA) molecule, not the cytoplasm or the cell membrane. This nucleus-centered view still prevails today in textbooks [54, 55, 56]. Biophysical analyses reveal that initial ionization events from X-rays are not homogeneously distributed but occur along distinct tracks, with a potential to cluster within a few nanometers. In this case, the nDNA damage is associated with multiple damage sites within a few base pairs from both direct energy absorption and indirect modifications through ROS (mainly ^·OH) produced within the immediate surroundings that were able to reach the nDNA before being scavenged [61, 62]. The indirect nDNA damage therefore depends on the amount of oxygen and anti-oxidants such as glutathione in the microenvironment, as discussed further below. In an aerated diploid cell, 1 Gy of X-irradiation produces ≳10⁵ ionizations, ≳1000 nDNA base modifications, ∼1000 single-strand breaks, and 25-40 double-strand breaks [56, 63]. nDNA damage triggers a highly complex DNA damage response involving the recruitment of certain proteins to damaged sites and the activation of checkpoint, DNA repair, and cell death pathways [56].

Ionization clusters are also predicted to occur within mitochondria [64], which seems to have much higher relevance for RT outcomes than realized in the past [65, 66]. Mitochondrial DNA (mtDNA) is more vulnerable to ionizing radiation than nDNA, due to a lack of histone protection and less efficient repair mechanisms [67, 68]. Damage to mtDNA causes or augments mitochondrial dysfunction, increasing leakage from the electron transport chain, with long-lasting increases in mitochondrial reactive nitrogen species (mtRNS, formed from NO) and mitochondrial ROS (mtROS) formation, in particular of O₂^·- (superoxide). Mitochondrial dysfunction triggers a retrograde stress response altering nuclear gene transcription [69]. According to Seyfried and Shelton [29], this mechanism is able to account for all the hallmarks of cancer and can be therapeutically targeted using CR and/or KDs. Leach et al. [70] proposed a mechanism by which IR-induced mtROS and mtRNS trigger mitochondrial Ca²⁺ release with subsequent uptake by adjacent mitochondria, which in turn undergo a transient permeability transition with mtROS and mtRNS production and Ca²⁺ release, in this way propagating the signal. A recent experiment using carbon ion and proton microbeams demonstrated that energy absorption within a mitochondrial cluster caused a near-instant (< 1 s) and simultaneous depolarization of all the mitochondria belonging to the cluster and of connected mitochondria as far as 18 µm away from the irradiated site [71]. The authors proposed changes in the membrane structure such as lipid peroxidation caused through a direct or indirect (ROS-mediated) action as possible causes of the depolarization. Along these lines, Kam and Banati [65] proposed the diffusion of excessively produced superoxide from damaged mitochondria as another damage propagation pathway that could ultimately reach the nucleus and induce ‘mitochondrial superoxide-mediated nuclear damage'. All these mechanisms are consistent with data implying that the mitochondrial location of anti-oxidative enzymes is more cell protective than the cytosolic one, in particular in case of manganese superoxide dismutase (MnSOD) which dismutates superoxide to the less toxic hydrogen peroxide (H₂O₂) [65]. According to Richardson and Harper [66], the mtROS and mtRNS produced through IR could fully account for the majority of nDNA damage and most of the observed RT effects.

Collectively, it seems that, while nDNA damage is the most important aspect of RT, this could be mainly a secondary effect of mtDNA damage and mtROS/mtRNS production. On the cellular level, both the mtDNA and the nDNA damage response can lead to a transient or permanent stop in the cell cycle, or induce programmed cell death, usually after a few or more cell divisions. On the tumor level, the outcome of RT will depend on 5 factors, classically known as the 5 R's of radiobiology, which determine whether long-term tumor control will be achieved or not (fig. 1). Evidence that each of these factors can be modified through ketogenic therapy has been reviewed by us recently [42, 44].

As already indicated above, the amount of radical scavengers, additional radicals from mitochondrial metabolism and oxygen within the microenvironment of the DNA, all have an influence on the resulting DNA damage. In 1942, Wilhelm Brünings noted a sensitization of head-and-neck tumors to radium therapy when patients underwent his ‘Entzuckerungsmethode' (de-glycation method) consisting of a KD combined with insulin injections in order to maximally lower the blood glucose levels [21]. On the other hand, modern PET imaging studies have confirmed that tumor areas with a high glycolytic metabolism are more radioresistant than those with a more oxidative metabolic phenotype [72, 73].

As already mentioned, a large body of evidence supports the notion that tumor cells frequently differ from their normal cell counterparts by dysfunctional mitochondria. The alterations include morphological abnormalities, lipid composition changes, and nDNA and mtDNA mutations encoding components of the respiratory chain complexes, resulting in respiratory insufficiency and increased production of mtROS/mtRNS (reviewed in [29, 74, 75, 76]). The Warburg effect seems to help such cells in neutralizing the high intrinsic levels of ROS [43]. Accordingly, glucose withdrawal was shown to lead to mtROS-mediated cell death in tumor cells, but not in normal cells with intact mitochondria [36, 37, 38, 39, 40]. In contrast, high glucose concentrations in the tumor microenvironment help to scavenge ROS through increased production of anti-oxidative substrates such as lactate and glutathione through glycolytic pathways, which would also aid against IR or chemotherapy-associated ROS [33, 77]. Nicotinamide adenine dinucleotide phosphate (NADPH) produced during the oxidation of glucose-6-phosphate in the pentose phosphate pathway maintains glutathione, the most important scavenger of H₂O₂ and other peroxides, in the reduced state [78]. It has been shown that KDs and STF are able to downregulate this anti-oxidative defense mechanism in tumor cells and sensitize them to additional therapy-induced oxidative stress (reviewed in [42, 43]). In a case study of cancer patients, those who kept a strict KD showed reduced levels of the tumor marker TKTL-1 (transketolase-like 1), which is supposed to be associated with the pentose phosphate pathway [79]. Studies using FDG-PET imaging [25, 80] or microdialysis measurements [81] have shown that a KD can downregulate the glycolytic tumor metabolism in some patients. A decrease in tumor lactate production has also been observed in mice kept on a KD [82, 83] and in cultured cancer cells treated with ketone bodies [84, 85] or fasting-mimicking conditions [86, 87]. The decreased tumor lactate concentrations after a few days on a KD in head-and-neck cancer patients measured by Schroeder et al. [81] are of particular interest, as tumor lactate concentrations have been directly linked to radioresistance in a variety of xenografted head-and-neck tumors that were irradiated using a clinically relevant schedule of 30 fractions over 6 weeks [88, 89].

Besides glycolysis, another adaption to high mtROS production frequently occurring in cancer cells is uncoupling of oxidative phosphorylation and ATP production through overexpression of uncoupling protein 2 (UCP2), which is also associated with increased resistance against chemotherapy and RT. Uncoupling allows protons to leak from the intermembrane space back into the matrix, decreases the mitochondrial membrane potential and thus reduces the emission of mtROS [90]. However, this comes at the expense of inefficient ATP generation. Fine et al. [91] have targeted this protective mechanism through administration of AcAc, which led to ATP depletion and growth inhibition. In their explanation, increased uncoupling or mitochondrial dysfunction in general would prohibit mitochondrial production of sufficient ATP to compensate for the reduced glycolytic ATP production that follows from the Randle cycle-like inhibitory effects of free fatty acids and ketone bodies on glycolysis.

Under non-hypoxic conditions, a shift from glycolysis towards (inefficient) mitochondrial metabolism through KDs and fasting would selectively increase the mtROS levels in tumor cells by increasing electron leakage through the electron transport chain. In a murine CT26 colon cancer model, STF decreased both glycolysis and glutaminolysis while shifting metabolism towards the mitochondria, which resulted in enhanced mtROS production and ATP depletion [86]. Marini et al. [87] confirmed these findings, showing that STF in vitro enhanced the oxygen consumption rate and increased the complex I and IV activity in murine CT26 colon and 4T1 breast cancer cells, leading to a subsequent boost in mtROS production. Similarly, Lee et al. [92] measured increased oxidative stress under in vitro fasting conditions in 4T1 cells, which they associated with the increased superoxide levels and the synergistic responses they observed when allografted tumors were treated with combined fasting and cyclophosphamide. Synergistic effects between fasting and chemotherapy were also observed in murine B16 melanoma and GL26 glioma allografts, human breast cancer and ovarian cancer xenografts (all treated with doxorubicin [92]), pancreatic cancer xenografts treated with gemcitabine [93], CT26 colorectal allografts treated with oxaliplatin [86], and murine fibrosarcoma allografts treated with mitoxantrone or oxaliplatin [94] (table 1). Furthermore, Morscher et al. [95] revealed synergistic effects between a KD and metronomic chemotherapy with cyclophosphamide in 2 neuroblastoma xenograft models. However, some other studies found no synergistic cytotoxic effects between fasting and chemotherapy (table 1). It is possible that ATP production in these tumors could be maintained at sufficiently high levels to enable efficient DNA repair [96]. Collectively, considering that all these chemotherapy regimens provoke oxidative stress in the tumors to varying degrees [97], these data support the hypothesis that KDs and STF are able to impair the anti-oxidative defense or further increase the ROS levels in several tumor cell lines, which sensitizes them to treatment by chemotherapy and RT.

In contrast to some of the chemotherapy studies, thus far, all studies adding fasting or a KD to RT have indicated synergistic effects (table 2). In a murine GL26 glioma model, fasting for 48 h prior to each of 2 fractions delivering 5 and 2.5 Gy IR retarded tumor growth and prolonged survival of the treated mice more than IR or fasting alone [98]. Using the same murine glioma cells, Abdelwahab et al. [99] described more than additive effects when a KD was combined with 2 × 2 Gy RT, apparently curing most of the irradiated mice completely. Paradoxically, however, the ROS levels were decreased in tumors of mice fed with the KD.

Allen et al. [100] investigated the combination of a KD with IR against lung cancer xenografts using both conventionally fractionated (34 × 1.8 Gy) and hypofractionated (2 × 6 Gy) schedules. In both cases, adding the KD to RT achieved the greatest efficacy regarding tumor growth delay and mouse survival. In an additional experiment of irradiating with 6 × 2 Gy for 3 times per week, the efficacy could be further enhanced by combining the KD and IR with carboplatin administration. Tumor samples harvested from mice on the KD at the end of the hypofractionation experiment exhibited significantly higher levels of protein modifications with the lipid peroxidation marker 4-hydroxy-2-nonenal (4HNE), consistent with increased KD-mediated oxidative stress in tumor cells [100]. Similar results were obtained using a pancreatic cancer xenograft model treated with 6 × 2 Gy [101]. In 2 clinical trials, the same working group had measured increases in plasma protein carbonyl content in lung and pancreatic cancer patients who adhered to a strict KD while receiving chemo-RT [101]. However, the small sample size (n = 4) and the lack of a control group did not allow any conclusion as to whether the KD contributed to this increase of oxidative protein damage or whether it was solely due to the chemo-RT.

Finally, 2 studies investigating the combination of 30% CR and IR in triple-negative breast cancer models have revealed synergistic anti-tumor effects that were related to a downregulation of the IGF-1 receptor (IGF-1R) and its downstream targets Akt and phosphoinositide 3-kinase (PI3K) in both primary tumors and metastases [102, 103]. Although ketone bodies were not measured, these findings suggest an intriguing role for ketogenic therapy as a targeted therapy against the IGF-1R pathway in order to sensitize tumor cells against IR. IGF-1R overexpression in tumor cells is associated with high radioresistance due to the IGF-1R being involved in ATM-mediated DNA double-strand break repair (ATM = ataxia-telangiectasia mutated) [104].

The oxygen effect describes the enhancement of RT efficacy with increasing oxygen concentrations and is one of the main reasons why RT is applied in a fractionated scheme in order to utilize re-oxygenation of hypoxic tumor areas between RT fractions. For example, there is evidence that single-fraction stereotactic RT lacks a dose-response relationship and achieves lower tumor control rates than 3 or more fractions even if the same ‘biologically effective doses' are applied [105, 106], which is consistent with a detrimental effect of missing re-oxygenation [107]. Oxygen is required for the production of superoxide from cell water radiolysis products, which greatly enhances the toxicity of IR [57]. Therefore, HBO before a RT session can be used to sensitize hypoxic tumor cells to IR, resulting in improved local control rates and overall survival [108, 109]. While HBO alone is of limited clinical efficacy, the D'Agostino lab has shown that HBO therapy increased ROS production and inhibited the growth of highly aggressive VM-M3 mouse tumor cells. Feeding mice a KD and/or exogenous ketones thereby enhanced its efficacy in vivo [41, 110]. This would support the use of both ketogenic therapy and HBO prior to an RT session as a complementary treatment approach. A recent case report of a poly-metastasized breast cancer patient in which a combination of HBO, KD, STF, glucose deprivation, hyperthermia and chemotherapy was used over 6 months reported a complete clinical, radiological, and pathological response and provides a proof-of-principle example for such integrative treatment concepts [111].

Recently, Richardson and Harper [66] proposed a 2-component model for the oxygen effect. In their explanation, as oxygen levels rise, damage from increasing levels of mtROS constitutes the main component which is more and more counteracted by more efficient DNA repair due to increasing ATP production. This model, however, assumes efficient oxygen-dependent ATP production and would not apply to those tumor cells that are unable to compensate for a loss of glycolytic ATP if forced to use mitochondrial metabolism. The failure to compensate for glycolytic ATP production would have severe consequences in terms of impaired DNA repair capacity. In this case, the overall oxygen enhancement of RT-induced damage against these tumor cells would be higher than predicted. Consistent with this, Bhatt et al. [112] have shown that glycolysis inhibition in cells with inefficient respiratory function significantly decreased the DNA double-strand break repair kinetics and radioresistance compared to those cells that were able to compensate for inefficient oxidative phosphorylation by maintaining high levels of glycolytic ATP production.

A frequent cause of inefficient oxidative phosphorylation and enhanced glycolysis in tumors cells are loss-of-function mutations of the tumor suppressor p53 [113]. A study utilizing a MYCN-driven neuroblastoma model revealed that p53 loss-of-function mutations confer high resistance against IR by reprogramming glutathione-associated metabolism and increasing the glutathione pool [114]. It is noteworthy that another MYCN-amplified neuroblastoma model derived from SK-N-BE(2) cells was found to be extremely sensitive to metronomic cyclophosphamide, which had anti-angiogenic effects especially when combined with a KD [95]. It is also noteworthy that, in autophagy-competent cells, various forms of mutant but not wild-type p53 were degradable through deacetylation-induced autophagy triggered by glucose restriction in vitro and a low carbohydrate diet in vivo; depletion of mutant p53 subsequently induced autophagy-mediated cell death [115]. These findings support the application of ketogenic therapy along with RT against p53-mutated tumor cells.

In 2008, Raffaghello et al. [116] introduced the term ‘differential stress resistance' to describe the protective effects of STF on normal but not tumor cells against high-dose chemotherapy. Mechanistically this was shown to occur, in part, through downregulation of the Ras-Raf-MAP kinase (MAPK) und Akt-mammalian target of rapamycin (mTOR) signaling pathways [116, 117]. The fasting-induced reduction of growth factors such as glucose, insulin, and IGF-1 inhibits Akt-mTOR signaling, promotes adipose tissue free fatty acid release and hepatic ketogenesis and globally activates a stress resistance program involving activation of adenosine monophosphate-activated protein kinase (AMPK), peroxisome proliferator-activated receptors (PPARs) and forkhead box class O (FOXO) transcription factors (reviewed in [118, 119, 120]). PPARs are ‘the nuclear transcription factors of fat and fasting' [119] which are activated by a number of fatty acids and certain eicosanoids [121]; in target tissues, they promote numerous metabolic actions, but also anti-inflammatory effects [119, 121]. FOXOs could be termed ‘survival transcription factors' because they, among others, promote the transcription of a number of cell cycle arrest and anti-oxidant genes [122]. FOXOs are negatively regulated by phosphorylation through Akt downstream of the insulin receptor and IGF-1R, which prevents their translocation into the nucleus or translocates them from the nucleus into the cytosol, respectively, where they can get degraded by MDM2-induced polyubiquitylation (MDM2 = murine double minute 2). In contrast, phosphorylation by AMPK activates FOXOs without directly regulating their localization [122]. It was recently shown in vitro that BHB increases AMPK phosphorylation and FOXO3a-mediated expression of catalase and MnSOD [123]. Other data has revealed that BHB increases FOXO3a transcription by altering the chromatin structure at the FOXO3a promoter on nDNA through class I histone deacetylase (HDAC) inhibition [124], providing another mechanism for the upregulation of MnSOD and catalase (fig. 2). Accordingly, incubation of cardiomyocytes with BHB increased FOXO3a, MnSOD, and catalase expression and protected against H₂O₂ toxicity [125]. This supports the hypothesis that ketosis could be used in addition to modern RT techniques such as deep inspiration breath hold [126] to further reduce the risk of late cardiac toxicities from RT involving exposure of the heart. Opposite effects on tumor cells were recently described, showing that BHB added to mouse and human glioma cells in vitro inhibited HDAC activity in a dose-dependent manner, which impaired the DNA damage repair in the tumor cells [127].

Another anti-oxidative property of ketone bodies derives from their catabolism in the tricarboxylic acid (TCA) cycle, which yields NADPH produced during the conversion of isocitrate to α-ketoglutarate catalyzed by the action of the NADP⁺-dependent isocitrate dehydrogenases 1 (in the cytosol) and 2 (in the mitochondria). This NADPH donates electrons for reduction of glutathione in the same way as does NADPH stemming from the pentose phosphate pathway (reviewed in [128]). Finally, BHB in physiological concentrations achieved during STF or a KD has been shown to exert an anti-inflammatory action, in part through inhibiting the NOD-like receptor protein 3 (NLRP3) inflammasome (reviewed in [129]).

To summarize, the combined action of low insulin signaling and elevated BHB levels achieved by STF or a KD exerts anti-inflammatory and anti-oxidative effects in normal cells. The same would not apply to tumor cells with constitutive activation of oncogenes or loss of function of tumor suppressors that are involved in the stress resistance program [118]. The result is a differential stress resistance between normal and tumor cells that could be exploited by using ketogenic interventions during chemotherapy and/or RT (fig. 2). All of the preclinical studies summarized in table 1 have confirmed a reduction of side effects from a variety of chemotherapeutic drugs without interfering with, or even boosting, their anti-tumor effects. To date, 3 small studies in humans have also found evidence of a protective effect of STF against chemotherapy-related toxicity [130, 131, 132]. These studies and their relevance for patients receiving chemotherapy have been discussed in detail elsewhere [133, 134, 135, 136] and will only briefly be reviewed here. The first study, by Safdie et al. [130], was a case study of 10 patients who voluntarily fasted for 48-140 h prior to and 5-56 h after each of an average of 4 chemotherapy cycles. In 6 patients who partly had fasted and partly had not, side effects were reduced during those cycles in which chemotherapy had been combined with fasting, in particular fatigue and gastrointestinal problems. The second study, by deGroot et al. [131], was a randomized controlled trial in 13 human epidermal growth factor receptor 2 (HER2)-negative stage II and III breast cancer patients who were treated with 6 cycles of combined docetaxel, doxorubicin, and cyclophosphamide. The intervention group who fasted for 24 h prior up to 24 h after chemotherapy administration experienced a significant reduction in IGF-1 levels and lower insulin levels than the control group and exhibited signs of less DNA damage or more efficient DNA double-strand break repair in peripheral blood mononuclear cells [134]. The third study was a dose escalation study in a heterogeneous sample of 20 cancer patients who underwent 2 cycles of platinum-based chemotherapy in combination with fasting [132]. There was no control group, but the results indicated an inverse dose-response relationship between the fasting duration and the amount of DNA damage in peripheral blood mononuclear cells as well as the chemotherapy-induced myelosuppression [136]. Importantly, all 3 studies have shown that STF was feasible and resulted in only minor (grade 1 and 2) side effects.

An important role for ketone body-mediated protection was obtained from the study by Dorff et al. [132] in which longer fasting duration prior to chemotherapy appeared more protective, yet only BHB, but not insulin, IGF-1 or glucose, were significantly different between 24 and 48 h of fasting prior to chemotherapy. This supports the hypothesis that KDs might be used as fasting-mimicking diets to reduce side effects from oxidative stress during prolonged treatment situations such as during several weeks of RT. Alternatively (or additionally) STF could be considered for patients receiving weekly chemotherapy during RT or for patients undergoing a few fractions of high-dose stereotactic RT. These and other possibilities of combining ketogenic therapy with different RT schedules should be tested in future preclinical and clinical studies.

Finally, it is important to note that the human STF studies have failed to reproduce the reductions in insulin and glucose levels seen in the mouse studies because corticosteroids were routinely administrated during chemotherapy [134, 136]. This could have prevented some further reductions in side effects or blunted the effect sizes. More alarming are findings that corticosteroid use prior to RT was associated with significantly shorter overall and progression-free survival in 3 large cohorts of glioblastoma patients [137]. This was linked to a corticosteroid-induced redistribution of tumor cells from the relatively radiosensitive G2/M phase to the relatively radioresistant G1 phase, with maintenance of cell viability. Additionally, the corticosteroid-induced elevations of blood glucose levels would provide a radioprotective environment. We have argued that this could be countered with ketogenic therapy [138]. The proof of principle was provided by Champ et al. [139] who showed that consumption of a KD during RT lowered the blood glucose levels in glioblastoma patients even under concurrent corticosteroid treatment.

The preclinical data and mechanistic insights reviewed here support the notion that ketogenic therapy could improve the outcome after RT, both in terms of higher tumor control and lower normal-tissue complication probability. The first effect relates to the metabolic shift from glycolysis towards mitochondrial metabolism, which selectively increases ROS production and impairs ATP production in tumor cells. The second effect is based on the differential stress resistance phenomenon that is achieved when glucose and growth factors are reduced and ketone bodies are elevated, which causes normal cells to switch to a cellular maintenance and stress resistance program. Underlying both effects are the metabolic differences between normal and tumor cells that have been reviewed here and elsewhere [42, 43, 118, 140]. First clinical studies indicate that a differential stress resistance can be induced by STF in humans; however, it has not been studied yet whether such as resistance could be mimicked by KDs and in this way be utilized for a longer course of RT. The large variety of RT and chemotherapy schedules employed in modern cancer treatment opens up a large number of possible ketogenic interventions, ranging from STF over KDs to the administration of exogenous ketone bodies. Ultimately, patient preferences and patient-related factors such as body composition will determine which form of ketogenic intervention can be considered.

I thank Prof. Harald Walach for the kind invitation to write this review.

The author declares that he has no conflicts of interest to disclose. No funding was received for writing this article.