Generally, cancer cells adapt to hypoxia and glucose deficiency without relying on lipid metabolism, primarily through the hypoxia-inducible factor (HIF) pathway (11,12) and the unfolded protein response (UPR) (13) (Fig. 2). The UPR involves endoplasmic reticulum (ER)-associated degradation of incorrectly folded proteins produced under the above stress conditions, preventing accumulation of toxic misfolded proteins followed by ER stress. However, expression of HIF1α, required for adaptation to hypoxia, is impaired in cancer cells cultured under low-glucose conditions (14). The combination of hypoxia and glucose deficiency (HGD) is cytotoxic to cancer cells. Indeed, recent studies have shown that HGD induces cancer cell death in association with CRE-binding protein (15) and overproduction of poly (ADP-ribose) polymer (16). However, cancer cells can synergistically respond to HGD and activate genes required for adaptation to this harsh condition (17).

The common adaptive response mechanism in cancer cells exposed to HGD may involve both HIF pathway activation (18–20) and the UPR (13,15,16,19,20). Hypoxia induces the UPR through activation of the ER stress sensor molecules activating transcription factor 6, inositol-requiring protein 1 (IRE1), and PKR-like ER kinase (PERK) (1 in Fig. 2) (13,19), which helps cancer cells tolerate low-oxygen conditions. These mechanisms can be induced under hypoxia and/or glucose deficiency. The HIF pathway is mediated by hypoxia-inducible transcription factors HIF1α and HIF2α, which are mainly induced in mildly to moderately hypoxic tumor regions (Figs. 1 and 2), along with the constitutively expressed aryl hydrocarbon receptor nuclear translocator (ARNT) (2 in Fig. 2) (12).

Mammalian target of rapamycin (mTOR) plays a role in adaptation to HGD. mTOR complex 1 (mTORC1) promotes the translation of proteins, including HIFα proteins, to maintain energy homeostasis (3 in Fig. 2) (19,20). However, suppression of mTORC1 signaling can also be important for adapting to hypoxia by enabling appropriate mRNA translation and activating the UPR (19,20). mTOR can be inhibited through the HIF1α-REDD1 and HIF1α-BNIP3 pathways under hypoxia, suggesting a negative feedback loop between HIFα and mTOR (route 4 in Fig. 2) (19). Thus, HIFα, the UPR, and mTOR are interconnected under HGD (13,19).

Several studies have revealed the effects of HGD on cancer cell survival. Resistance to apoptosis induced under HGD is mediated by the serine/threonine kinase proviral integration site for Moloney murine leukemia virus 1 (PIM-1) in some cancer cells (21). Meanwhile, proline oxidase (POX) expression is increased under HGD in an AMP-activated protein kinase (AMPK)-dependent manner (22). Expression of both PIM-1 and POX is HIF-independent (5 in Fig. 2) (21,22). Survival of cancer cells exposed to HGD may be dependent on POX because proline oxidation results in ATP production under HGD (22). Thus, POX is a potential target of cancer therapy. HIF1α and the UPR cooperate to enhance the stemness of breast cancer cells via HIF1α binding to XBP1 under HGD (6 in Fig. 2) (23). HIF1α-driven expression of LIMS1 not only facilitates glucose uptake but also enhances HIF1α translation via the AKT-mTOR pathway in pancreatic cancer cells exposed to HGD (route 7 in Fig. 2), resulting in a cell survival advantage (24).

Recent studies have shown that GD generates an HGD-mimicking condition because even under normoxia, expression of HIF1α can be induced by GD to augment cancer cell survival (25–27). This induction of HIF1α expression occurs through EZH2-dependent suppression of PHD3 expression (8 in Fig. 2) (25). This simple GD-driven pseudo-HGD condition with HIF1α expression through inhibition of its degradation augments the aggressiveness of lung adenocarcinoma cells (25). GD under normoxia also increases HIF1α expression via ER stress-inducible molecular chaperone GRP78 in pancreatic cancer cells to augment chemoresistance (9 in Fig. 2) (26). The GRP78-HIF1α complex binds to the regulatory region of the HIF1A gene to promote transcription. Thus, GRP78 can induce HIF1α expression at the mRNA level (26). In summary, GD can enhance tumor hypoxia by upregulating HIF1α expression at both the protein and mRNA levels. Meanwhile, lysophosphatidic acid receptors were shown to contribute to chemoresistance in pancreatic cancer cell line PANC-1 under HGD (28). Expression of immune checkpoint receptor PD-1 and TIGIT can be synergistically increased in esophageal cancer cells under HGD and is responsible for immune tolerance (29). The roles of HIFs and the UPR were not examined in these two studies (28,29).

Glucose uptake and subsequent catabolism are activated in cancer cells. Recent studies have shown that GD influences lipid metabolism in cancer cells to support their survival, although its relationship to hypoxia has not yet been established. Indeed, some glioma cells have these characteristics and become susceptible to GD in culture (30). GD in cancer cells is associated with multiple lipid metabolism pathways, as summarized in Table I and Fig. 2. GD together with serum deficiency was also reported to drive the use of extracellular glutamine and lactate in glycerophospholipid synthesis for biomembrane generation via oxaloacetate-phosphoenolpyruvate conversion (Fig. 3A) in lung cancer cells (31). Mitochondrial phosphoenolpyruvate carboxykinase (PEPCK-M) plays key roles in this process (Fig. 3A) (31). Pharmacological activation of AMPK, a cellular energy sensor molecule, activates fatty acid oxidation (FAO) to acquire ATP in Akt-transformed cells under GD (30) (1 in Fig. 3B). Collectively, these studies have revealed novel mechanisms for cancer cell adaptation to severe nutrient deficiency.

Membrane phospholipids can be substrates for phospholipase D1-mediated autophagy in multiple cancer cell types under GD (32). FAs generated during this process (2 in Fig. 3B) can be used for FAO to sustain cell survival (32). In hepatocellular carcinoma and leukemia cells, GD activates MEK-ERK5 signaling to increase FA uptake (3 in Fig. 3B), followed by ATP generation through existing FAO activity (33). The importance of FAO under GD was also demonstrated in drug-resistant slow-cycling glioblastoma subpopulations (34). The same study showed that FA transport (4 in Fig. 3B) by fatty acid-binding protein (FABP)-7 is crucial for survival of mitochondria (oxidative phosphorylation)-active glioma cells (34). LD catabolism by lipophagy, followed by existing FAO activity (route 5 in Fig. 3B), also contributes to the survival of glucose-starved glioblastoma cells (35,36). This involves hyperactivation of mTOR (35) (Fig. 3B). Meanwhile, choline kinase 2 (CHKα2) is responsible for phosphorylation of LD-associated perilipins, finally resulting in chaperone-mediated autophagy-driven degradation of LDs to generate FAs (35) (route 5 in Fig. 3B). FAO under GD is also important for prostate cancer cells, in which LD accumulation via the PIM1-GSK3β-PPARα axis can be activated under nutrient stress conditions for FA generation (37) (route 6 in Fig. 3B).

Glycogenolysis serves as an alternative pathway for cellular glucose supply under GD. Phosphoglucomutase PGM1, a key enzyme for glycogenolysis followed by glycolysis, contributes to the viability of gastric cancer cells exposed to GD by suppressing lipogenesis (38) (7 in Fig. 3B). FAS can compensate for the reduced cell viability caused by PGM1 inhibition under GD conditions (38). Thus, combined inhibition of interconversion of the phosphate position within glucose molecules and FAS may be promising for gastric cancer treatment (38). A study using colon cancer cell line Caco-2 revealed that the composition of phospholipids, cholesterol, and unsaturated FAs in the cell membrane is considerably altered under GD, leading to metabolic adaptations that augment cell survival (39). GD also increases prostaglandin E2 synthesis at the nuclear membrane (40) (8 in Fig. 3B) by repressing 15-hydroxyprostaglandin dehydrogenase expression in colon cancer cells, resulting in a survival advantage for these cells (41). The stemness of lung cancer cells exposed to GD may be enhanced by oleic acid supplied by surrounding cancer-associated fibroblasts (CAFs), followed by stearoyl-CoA desaturase expression (9 in Fig. 3B), nuclear translocation of polymerized actin, and yes-associated protein (42). This process can be inhibited by treatment with an FAO inhibitor (42).

In summary, this section has revealed that various cancer cells generally circumvent GD through FAO (Table I). Additionally, the biosynthesis of glycerophospholipids and prostaglandins can contribute to their adaption to GD.

Glutamine is another primary nutrient for cancer cells in addition to glucose. The levels of non-essential AAs, including glutamine, in pancreatic cancer tissues are considerably lower than those in adjacent benign tissues (43). In cells with an inactivated von Hippel Lindau (VHL) gene under hypoxia, glutamine is used to generate α-ketoglutarate, which is then used to produce the citrate required for de novo LCFA synthesis through reductive carboxylation (44,45) (route 1 in Fig. 4A) and the glutathione (GSH) synthesis (2 in Fig. 4A) necessary for suppression of lipid peroxide accumulation, which is responsible for ferroptosis (46). Thus, clear cell renal cell carcinoma (ccRCC) cells lacking functional VHL are susceptible to glutamine deficiency (GlnD) caused by glutaminase inhibition (44) or glutamine withdrawal from the culture medium (45) (Table II).

Cancer cells can adapt to stressful GlnD conditions through multiple metabolic mechanisms. A recent study using lung cancer cell lines showed that differential regulation of mTORC1 and mTORC2 via Sestrin2, followed by FAO, contributes to cell survival under GlnD (47) (Table II). Under both GD and GlnD, mitochondrial protein coiled-coil helix tumor and metabolism 1 can facilitate LCFA synthesis and FAO via LD formation in multiple cancer cells (48) (routes 3 and 4 in Fig. 4A and Table II). GlnD likely activates the SREBF1 gene, which encodes the lipogenic enzyme SREBP1 (49). SREBF1 activation, mediated by O-linked N-acetylglucosaminylated transcription factor specificity protein 1 (Sp1) and followed by acetyl-CoA carboxylase (ACC) expression, contributes to LD biosynthesis in various cancer cells (49) (route 3 in Fig. 4A). Similar to the previously described glioblastoma cell response to GD (35), CHKα2 plays a key role in LD lipolysis in lung cancer cells exposed to GlnD (50) (route 4 in Fig. 4A and Table II). In this context, neutral lipolysis and lipophagy likely contribute to LD catabolism (route 4 in Fig. 4A). Notably, the same study demonstrated that phosphorylation of CHKα2 at S279, which is involved in neutral lipolysis, is associated with a worse prognosis in patients with non-small-cell lung cancer (50).

HMG-CoA reductase degradation protein 1 (HRD1) was found to suppress the proliferation of breast cancer cell line MDA-MB-231 through inhibition of FAO (51). HRD1 can interact and ubiquitinate FA transporter protein carnitine palmitoyltransferase 2 (CPT2) (5 in Fig. 4A), resulting in blockade of mitochondrial LCFA transport through its degradation (51). Under GlnD, this mechanism can be abolished to activate FAO, resulting in resistance to glutaminase inhibitors. Thus, simultaneous inhibition of FAO and glutamate production may be therapeutically promising, and high HRD1 expression potentially predicts better efficacy of glutaminase inhibitors (51).

PI3K-C2γ expression is inactivated to activate mTOR in cancer cells from pancreatic tumor tissues. These cells depend on exogenous glutamine to activate lipogenesis (52) (6 in Fig. 4A). Thus, treatment with mTOR and glutaminase inhibitors may be therapeutically beneficial. GlnD induces lipophagy-driven LD degradation to increase cellular cholesterol levels in hepatocellular carcinoma cells (53) (7 in Fig. 4A). Cholesterol inhibits SREBP2 maturation, leading to decreased cholesterol synthesis, which helps maintain redox balance and provides a survival advantage to cancer cells under GlnD conditions (53). Interaction between p53 and TP53-regulated inhibitor of apoptosis 1 (TRIAP1) may also be important for survival of colon cancer cells under GlnD (54) (Table II). TRIAP1 affects lipid homeostasis through multiple lipids, including glycerolipids, sphingolipids, and cholesterols, in HCT116 cells (Table II). The TRIAP1-p53 interaction is important for glutamine metabolism. Indeed, under GlnD, p53 can compensate for TRIAP1 function to overcome metabolic stress (54). Collectively, GlnD can couple to multiple lipid metabolism pathways (Table II). Cancer cells likely acquire ATP through the activation of FAO via LD metabolism under this nutrient stress (Fig. 4A). Maintaining cellular lipid homeostasis is also important (Fig. 4A).

GlnD associated with hypoxia is known to trigger expression of UPR-executor transcription factor ATF4 (19,20) in glioblastoma cells (55). This UPR-mediated stress response is responsible for resistance to temozolomide treatment (55). Hypoxia with GlnD can also promote cancer progression through epigenetic mechanisms. In melanoma tissue, the glutamine concentration in the hypoxic tumor core region is considerably lower than that in the tumor periphery (56). Hypoxia with GlnD can cause hypermethylation of histones through inhibition of demethylation and promote cancer cell dedifferentiation, resulting in resistance to BRAF inhibitor treatment (56). The implications for lipid metabolism involvement in these hypoxia- and GlnD-associated features observed in tumor tissues remain to be determined.

Serine and glycine can be enzymatically interconverted through a single reaction mediated by serine hydroxymethyltransferase, involving the conversion of tetrahydrofolate to its methylated form (1 in Fig. 4B), and they contribute to common metabolic pathways such as GSH synthesis (57,58) (2 in Fig. 4B). Thus, studies have tended to consider the effects of these AAs together (Fig. 4B, Table II).

Studies using colon cancer cell line HCT116 showed that serine/glycine deficiency (SGD) can suppress FAO through impaired mitochondrial function (58). Specifically, the ceramide level is decreased in serine-depleted cells (3 in Fig. 4B), leading to mitochondrial dysfunction (58). Thus, restriction of the serine supply is promising in treatment of p53-deficient cancers. However, SGD causes metabolic reprogramming with increased oxidative phosphorylation activity (59). p53-dependent synthesis of GSH with suppression of purine synthesis (4 in Fig. 4B), followed by scavenging of reactive oxygen species, can contribute to resistance of cancer cells against this metabolic stress (59). This process is accompanied by an increase in pyruvate transfer via 3-phosphoglycerate (3-PG) to the mitochondria, which activates FAO (route 5 in Fig. 4B). Furthermore, SGD facilitates the biosynthesis of toxic deoxysphingolipid, in which serine residues are substituted with alanine residues (6 in Fig. 4B), in cancer cells (60). Thus, dietary restriction and pharmacological targeting of the serine supply pathway can lead to tumor regression. This effect was prominent for growth of alanine-rich spheroid cancer cells (60). A more recent study further demonstrated that deoxysphingolipid (deoxysphinganine) generation is non-toxic and important for adaptation of cancer cells to SGD (61). This response mechanism involves accumulation of sphingosine in cancer cells due to the proteasomal degradation of sphingosine kinase 1 (7 in Fig. 4B), leading to increased expression of phosphoglycerate dehydrogenase (PHGDH) and subsequent de novo serine synthesis (route 8 in Fig. 4B) to overcome SGD (61). In summary, cancer cells can adapt to serine deficiency through inhibition of lipid peroxidation, activation of FAO, and modulation of sphingolipid biosynthesis.

Regarding the relationship between SGD and hypoxia, glioblastoma cells were reported to activate the AMPK-HIF1α pathway and induce a pseudo-hypoxia condition under SGD to bypass this metabolic stress (62). The effects of lipid metabolism involvement in this adaptation mechanism remain to be elucidated.

The major source of cellular cysteine is dietary cystine, which is metabolically associated with glutamine (63). Cysteine is also produced by de novo synthesis from methionine (63). The major effect of cysteine deficiency (CD) on cancer cells is impaired GSH synthesis, resulting in ferroptosis through accumulation of toxic peroxidized phospholipids (Fig. 4C, Table II) (46,64). Thus, CD may be therapeutically beneficial (46,64–66).

The mechanisms underlying how cancer cells circumvent CD fall into a single category (Table II): ferroptosis resistance through suppression of peroxidized lipid accumulation. Multiple pathways may be involved in this process in various cancer cells (Table II). For example, acute myeloid leukemia cells are auxotrophic for cysteine and can acquire CD resistance through overexpression of glutathione peroxidase 4 (1 in Fig. 4C) and microsomal glutathione-S-transferase 1, which may be associated with activation of FA metabolism (64) (Table II). In VHL-defective ccRCC cells, generation of peroxidized lipids is suppressed by FAO inhibition and GSH synthesis (46). In ovarian cancer cells, transcription factor NRF2 enhances cystathionine β-synthase (CBS) gene expression and activates autonomous synthesis of cysteine (route 2 in Fig. 4C), resulting in resistance to CD-induced ferroptosis through GSH synthesis (route 3 in Fig. 4C) (67). A recent study showed that albumin can be an extracellular source of cysteine under CD (68). Albumin incorporated into cancer cells via macro-pinocytosis in association with mTOR inhibition undergoes lysosomal degradation (68). Cysteine released into the cytoplasm also contributes to GSH synthesis (4 and route 3 in Fig. 4C), thereby preventing ferroptosis of cancer cells (68). This mechanism is especially important under spheroid culture conditions associated with tumor-like stress conditions (68).

Deprivation of cellular free iron ions is an alternative mechanism of ferroptosis resistance. Expression of genes involved in cellular iron storage can be increased via the ATM-MTF1 axis under CD. As a result, the balance between iron storage and release is expected to shift toward storage (5 in Fig. 4C), leading to ferroptosis resistance in multiple cancer cells (69). Expression of iron-sulfur cluster protein CISD3 is also increased in cancer cells and contributes to iron storage under CD, followed by ferroptosis resistance (70).

Tryptophane metabolites contribute to ferroptosis resistance under CD through a detoxification mechanism. Specifically, serotonin and 3-hydroxy-anthranilic acid produced by kynureninase act as radical trapping agents and reduce peroxidized lipids to non-toxic lipid alcohols (71) (6 in Fig. 4C). Overall, cancer cells utilize multiple molecular defense mechanisms against CD-driven ferroptosis (Table II).

In the context of CD and hypoxia, CD-induced death of MDA-MB-231 cells can be alleviated by hypoxia through inhibition of ATF4 expression (72). Thus, hypoxia may reduce CD-mediated cytotoxicity in certain cancer cells.

Cancer cells are also sensitive to arginine deficiency (AD), and this vulnerability is often associated with lipid metabolism (Table II). Expression of argininosuccinate synthetase 1 (ASS1), an enzyme in the urea cycle (Fig. 4D) that also catalyzes arginine biosynthesis, is low in >60% of clinical breast cancer samples (73). Thus, many cancer cells are arginine auxotrophs (Table II). AD induces cytotoxic autophagy in breast cancer cells, leading to cell death due to impaired mitochondrial function (Table II). In such cases, FAO suppression is followed by a metabolic shift to increase LDs (73) (route 1 in Fig. 4D). Meanwhile, ASS1-high non-small-cell lung cancer cells can confer ferroptosis resistance through monounsaturated FA synthesis. To facilitate this process, ASS1-driven synthesis of arginine activates the mTOR-SREBP1-SCD5 pathway to cause unsaturation of autonomously synthesized FAs, leading to inhibition of peroxidized lipid generation (74). This mechanism provides additional insight into why arginine auxotroph lung cancer cells are susceptible to AD-induced cell death. These observations suggest that arginine-deficient diets may be a therapeutic strategy through induction of ferroptosis. However, cancer cells can circumvent this stress condition. For example, AD can activate the MEK-ERK signaling pathway (route 2 in Fig. 4D) to induce c-Myc-Max transcription factor-driven ASS1 gene activation (3 in Fig. 4D) in ASS1-negative vulvar leiomyosarcoma (SKLMS-1) cells (75) (Table II). This cellular response causes adaptive metabolic reprogramming, including reduced ACLY expression, followed by suppression of de novo FA synthesis (75) (4 in Fig. 4D). Moreover, reports have described resistance to pharmacological AD in melanoma cells with ASS1 overexpression. In these cells, de novo lipogenesis is also inhibited through suppression of ACLY (Fig. 4D), acetyl-CoA carboxylase 1 (ACC1 in Fig. 4D), and FAS (Fig. 4D) to promote the Warburg effect (76). In this scenario, c-Myc contributes to AD resistance not only through ASS1 expression but also by enhancing glycolysis and glutaminolysis (5 in Fig. 4D) via the expression of glucose transporter 1 and glutaminase, respectively. Collectively, cancer cells can evade the effects of AD by autonomously synthesizing arginine, with the Warburg effect, rather than FAO, likely playing a key role in supporting cell survival under AD.

Under hypoxia, pharmacological AD inhibits expression of HIFs, inducible nitric oxide synthase, and ASS1 in HCT116 cells to inhibit tumor growth (77). This effect is associated with UPR induction. ASS1-deficient bladder cancer cell line UMUC3 shows greater sensitivity to hypoxia under AD (77). Thus, hypoxic cancer cells are vulnerable to AD.

Deprivation of methionine, an essential AA, from culture medium likely affects lipid metabolism in cancer cells (Fig. 4C, Table II). Methionine deprivation (MD) is linked to GSH synthesis through the methionine cycle and cysteine synthesis pathway (routes 2 and 3 in Fig. 4C). Consequently, MD can impair GSH synthesis (65), leading to ferroptosis, similar to the effect of CD. However, acute myeloid leukemia cells are cysteine auxotrophs as indicated by the fact that methionine supplementation does not rescue them from CD-driven ferroptosis (64). MD also affects cellular lipids by increasing phosphorylethanolamine and decreasing choline (Table II), thereby contributing to a synergistic therapeutic effect with chloroethylnitrosourea treatment (78). Meanwhile, methionine is a precursor of S-adenosylmethionine (7 in Fig. 4C), a key methyl donor for biomolecules. Thus, MD affects methylation-driven lipid metabolism pathways. For example, MD causes S-adenosylmethionine deficiency in HepG2 cells and enhances glycerophosphocholine synthesis through demethylation of CpGs within the PNPLA7 promoter region (79) (8 in Fig. 4C). This adaptation mechanism potentially promotes cancer progression via activation of mitochondrial oxidative phosphorylation (79). MD may also be therapeutically promising for glioma-initiating cells because it impairs cholesterol synthesis and increases cholesterol excretion, thereby suppressing key glioma-initiating cell functions (80).

In general, tumor tissues are poorly vascularized. Thus, many blood components, including lipids, are poorly supplied. Impairment of lipogenesis may also enhance lipid insufficiency in cancer cells exposed to hypoxia with limited exogenous lipid supply. The effect of simultaneous deprivation of lipids and molecular oxygen on cancer cells was first demonstrated by experiments examining how serum starvation and hypoxia (SSH) affects their phenotype in relation to mTOR activity (81). The same study using mouse embryonic fibroblast cells and various cancer cells, including kidney cancer cells, showed that mTOR activation under SSH augments ER stress, resulting in UPR-mediated cell death (Table III) (81). This is likely due to abnormal ER expansion induced by increased protein synthesis and insufficient unsaturated LCFA supply because cell death was alleviated by supplementation of albumin-conjugated oleic acid (81).

Hypoxia and delipidated serum cell culture conditions, characterized by poor supply of molecular oxygen as an electron acceptor and poor supply of extracellular lipids, was also shown to reduce regeneration of NAD⁺, a cofactor required for lipogenic citrate production in cancer cells, followed by cell proliferation (82). Thus, cancer cells exposed to SSH are likely to become lipid auxotrophs (Table III). However, NAD⁺-independent metabolism of exogenous acetate to acetyl-CoA followed by lipogenesis in cancer cells can rescue this auxotrophy (82).

ccRCC is a histological subtype of most kidney cancers (70–80%). Most of these cancer cells lack VHL function, resulting in constitutive HIF expression. Consequently, these cancer cells exhibit hypoxia-mimicking phenotypes (83,84). The phenotypes of ccRCC cells, such as ER homeostasis (83), motility (85), invasiveness (85), and anti-apoptosis (84), are highly dependent on LD biogenesis rather than LD catabolism (83–85). The kidney is a well-perfused organ (85), and studies on this cancer type under true hypoxic conditions are limited because ccRCC cells express HIF even in normoxic environments. However, hypoxia is a general tumor condition. Indeed, hypoxic regions exist within the normal renal medulla (86) and renal tumors (87), and adaptation of cancer cells to SSH may contribute to kidney cancer progression (Table III).

The HIF2α-perilipin 2 pathway contributes to ER homeostasis-mediated survival of ccRCC cells under SSH (1 in Fig. 5 and Table III) (83). LD catabolism also contributes to ccRCC cell phenotypes under LCFA starvation and hypoxia. LDs in cancer cells undergo hormone-sensitive lipase-driven degradation under SSH to maintain the cellular unsaturated FA (oleic acid) level, thereby suppressing the effect of toxic saturated FAs synthesized under hypoxia (2 in Fig. 5 and Table III) (88). Thus, the roles of LDs in malignancy are context-dependent.

Ovarian clear cell carcinoma (OCCC) has morphological and biological similarities to ccRCC (89). As described above, ccRCC cells exhibit hypoxia-mimicking phenotypes. This characteristic may also be true for OCCC cells because the HIF pathway is more active in this cancer subtype than in other histological subtypes of epithelial ovarian cancer (90–92).

The effects of SSH on gene expression and phenotypes of OCCC cells have been examined, revealing that multiple genes can be synergistically activated in certain OCCC cells exposed to SSH stress. We first discovered this phenomenon for the FVII gene (93) (3 in Fig. 5), which is responsible for initiation of the physiological blood coagulation cascade and a potential contributor to cancer-associated thromboembolism (Fig. 5 and Table III) (94). Unlike typical hypoxia response genes, such as VEGF, this transcriptional activation involves an Sp1-HIF2α interaction on the FVII gene promoter (Fig. 5) (93). Subsequent studies revealed that multiple genes, including ICAM1 (95) and CD69 (96), exhibit the same SSH-driven transcriptional activation (Fig. 5) and have much higher synergism than the FVII gene (95,96). We found that unlike hypoxia alone and serum starvation alone, SSH has UPR involvement (Table III) (93). However, the UPR does not contribute to the synergistic expression of FVII and ICAM1 (93,95).

Albumin serves as a major LCFA transporter in the blood and is important for LCFA uptake by cells (97). Addition of albumin-LCFA complex was found to abolish the SSH-driven ICAM1 and CD69 expression, suggesting that LCFA starvation is responsible for the synergistic transcriptional activation under hypoxia (95,96) (4 in Fig. 5). Indeed, addition of low-density lipoproteins, including LCFAs and cholesterol as their esterified form, also abolished the SSH-driven ICAM1 expression (4 in Fig. 5), whereas addition of cholesterol alone did not (98). mTOR is involved in this ICAM1 expression via NFκB (5 in Fig. 5) (95). Further studies revealed that lipophagy is induced under SSH in OCCC cells and is responsible for the synergistic ICAM1 and CD69 expression (6 in Fig. 5) (96,98). Lipophagy enhances binding of pro-inflammatory transcription factor NFκB to the ICAM1 promoter region via Sam68 and hTERT (7 in Fig. 5) (98). Currently, ICAM-1 is considered to suppress the lipotoxicity-mediated apoptosis induced by lipophagy-driven LD degradation (8 and 9 in Fig. 5 and Table III) (98). Meanwhile, CD69 causes epithelial-mesenchymal transition in a fibronectin-dependent manner (Fig. 5 and Table III), leading to OCCC cell survival in vitro and in vivo (96).

The effect of SSH on synergistic transcriptional activation in OCCC cells is mediated not only by LCFA starvation but also by cholesterol deficiency. The synergistic FVII gene expression under SSH is mediated through activation of the mTOR-SREBP1 axis (10 in Fig. 5) under cholesterol deprivation (11 in Fig. 5) (99). SREBP1 and HIF1α-ARNT complex indirectly promotes FVII expression through transcriptional activation of glucocorticoid-induced leucine zipper (GILZ) protein (12 in Fig. 5) to generate procoagulant extracellular vesicles, which are potentially responsible for cancer-associated thromboembolism (Fig. 5 and Table III) (94,99). It is noteworthy that GILZ suppresses SSH-driven ICAM1 expression because this anti-inflammatory transcriptional regulator binds and inhibits NFkB (13 in Fig. 5) (99).

Glioblastoma cells have been found to exhibit synergistic gene expression mediated by SREBP under both hypoxia and lipoprotein-deficient medium conditions (100). These SSH conditions synergistically enhance transcriptional activation of the lipogenic stearoyl-CoA desaturase, FABP-3, and FABP-4 genes (14 in Fig. 5) to suppress apoptosis and promote spheroid growth (Fig. 5 and Table III). An SREBP-dependent gene signature involving these genes predicts a poor survival rate for patients with glioblastoma (100). Lipogenesis via acetate also supports the survival of cancer cells under SSH, similar to the effect observed with exogenous acetate supply in HeLa cells (82,101). In this process, SREBP2 activates ACSS2 gene expression to enhance acetyl-CoA production (15 in Fig. 5 and Table III) (101). Additionally, HIFs may contribute to this mechanism because ARNT can upregulate ACSS expression under SSH (Fig. 5 and Table III) (101).