Lipid metabolism alterations affect not only tumor cell proliferation, but also dissemination and resistance to chemotherapeutic drugs (46). Most of adult tissues obtain FAs, cholesterol, and lipids from diet; meanwhile, de novo synthesis of FAs and cholesterol is restricted to the liver and adipocytes. Tumors frequently present the capability to activate the de novo synthesis of cholesterol and FAs (47) making them more independent from externally provided lipids (48, 49). Importantly, targeting enzymes associated with de novo lipogenesis and/or the mevalonate pathway has been shown to inhibit tumor growth (6, 50).

FAs are synthesized from cytoplasmic acetyl-CoA (AcCoA), generated from citrate produced from glucose, glutamine, or acetate (48). ATP-citrate lyase (ACLY) generates AcCoA and oxaloacetate (OAA) from citrate (48, 51). AcCoA carboxylases (ACC1/2) carboxylase AcCoA to form malonyl-CoA. Subsequent condensation steps, catalyzed by FA synthase (FASN), forms the 16-carbon saturated FA palmitate. Palmitate is then elongated by FA elongases (ELOVL) and desaturated by stearoyl-CoA desaturase (SCD1) or FA desaturases (FADS) to form other nonessential FAs, such as the 18-carbon monounsaturated FA (MUFA) oleate (C18:1) (Figure 2).

Many enzymes implicated in de novo synthesis of FAs and cholesterol have been proposed as biomarkers for prognosis in specific types of cancer. FASN is found upregulated in prostate and breast cancer (47, 52), and ACLY has been shown to support tumor formation and transformation (51). Inhibition of several enzymes of de novo lipogenesis, such as FASN, and ACC1 and ACC2, has been tested in different cancer models showing their relevance on tumor growth inhibition (53).

Similarly, inhibition of hydroxymethylglutaryl-CoA (HMGCoA) reductase (HMGCR), by statins, leads to inhibition of cell proliferation of breast cancer cells (54) and tumor regression in several preclinical mouse models, and it is being tested in clinical trials (43). The overexpression of enzymes of the mevalonate pathway has been proposed as biomarkers of poor prognosis in breast cancer (55). Cholesterol is generated by the mevalonate pathway, by condensation of two AcCoA molecules to form 3-HMGCoA, which is then reduced to form mevalonate, and then isoprenoid farnesyl-pyrophosphate. Several studies have shown that targeting the synthesis of cholesterol inhibits cancer cell proliferation and transformation (56).

De novo synthesis of FAs and cholesterogenesis are transcriptionally regulated by SREBPs, which are downstream oncogenic pathways including PI3K/Akt (57) and c-Myc (47) (Figure 2).

The SREBP family includes three transcription factors: SREBP1a and SREBP1c, which are derived from SREBF1 gene by alternative splicing (58), and SREBP2, which is encoded by SREBF2 gene. SREBPs are bound to the endoplasmic reticulum (ER) as inactive precursors (59). When the intracellular levels of cholesterol are high, insulin-induced genes interact with SREBP-cleavage–activating proteins (SCAPs) to retain SREBP inactive precursors attached to the ER. When cholesterol levels are low, SCAPs facilitate the translocation SREBPs to the Golgi apparatus to be further processed releasing the active forms (56). SREBP1 promotes the expression of lipogenic genes; meanwhile, SREBP2 regulates the expression of genes involved in the synthesis, uptake, and efflux of cholesterol. Nevertheless, SREBP1 and SREBP2 have overlapping activities. Both SREBP1 and SREBP2 are found overexpressed in several cancers. Regulation of the intracellular content of cholesterol has also been shown crucial for cancer cell survival. The ATP-binding cassette transporter (ABCA1) controls the efflux of cholesterol to ApoA-coated lipoproteins (57). Recently, it has been demonstrated that activation of p53 increases the retrograde transport of cholesterol from the plasma membrane to the ER, to prevent SREBP2 maturation (60). In addition, cholesterol levels are fine tune regulated by microRNA33—encoded by an intron within the SREBF2 gene (51)—which targets ABCA1. In addition, the esterification of cholesterol for storage in LDs, by sterol O-acyltransferase 1 (ACAT1), has been shown to augment the survival in prostate cancer (61).

In addition to de novo synthesis of FAs and cholesterol, the mobilization of intracellular FAs for FAO at mitochondria is crucial for cancer survival and dissemination. It is well-known that tumor cells present higher levels of reactive oxygen species (ROS) than not tumor cells, which allow them to activate prosurvival and epithelial-to-mesenchymal transition programs to support cancer progression and dissemination. Nevertheless, excessive ROS may promote apoptotic cell death. It has been demonstrated that enzymes implicated in the mobilization of intracellular neutral lipids provide metabolic flexibility to increase the levels of FAs for oxidation at mitochondria. In the FAO pathway, acyl-CoAs are cyclically dehydrogenated, hydrated, and decarboxylated, resulting in the progressive shortening of the FA, together with the production of NADH and FADH₂ and AcCoA. NADH and FADH₂ will be used for ATP production in the electron transport chain, and AcCoA can enter the Krebs cycle. AcCoA together with OAA gives rise to citrate, which after being exported to cytoplasm, can enter two metabolic pathways to produce cytosolic NADPH (62).

Enhanced mitochondrial β-oxidation of FAs has been described in pancreatic cancer (63, 64) and in metastatic breast cancer (65). FAO not only provides energy when glucose becomes limiting, but it also contributes to a better control of the oxidative stress, by augmenting the intracellular levels of NADPH (66). Increased FAO augments survival in leukemia and gliomas by counteracting the metabolic oxidative stress. Moreover, FAO has been shown crucial for the survival of cells from solid tumors when undergoing loss of attachment, which triggers anoikis or cell death due to oxidative stress (67, 68).

In addition, FAO is also influenced by the tumor microenvironment such as in the case of ovarian cancers, which preferentially metastasizes to the omentum enriched in adipocytes, which provides lipids for ATP and NADPH production to control metabolic stress during metastasis.

De novo synthesis of FAs in cancer cells is coupled to additional processes to accommodate the increase in the intracellular lipid content, to preserve the homeostasis between lipid storage and lipid mobilization (69). FAs from de novo lipogenesis are accumulated into neutral lipids (stored in LDs) and phospholipids (in membranes). LDs are complex and dynamic organelles consisting of a neutral lipid core surrounded by a phospholipid monolayer and a complex proteome associated. LDs itself have been proposed as novel diagnostic biomarkers for glioblastoma. It has been demonstrated that while they are not detectable in low-grade gliomas or normal brain tissues, they are common in glioblastoma, the most lethal brain tumor (70). Among the LD-associated proteins, there are enzymes of the sterol biosynthetic pathway, the acyl-CoA metabolism (ACSLs), and triacylglycerol (TAG) biosynthesis. Structural proteins, such as perilipins (PLINs) or caveolins, are critical for the integrity of LDs to avoid collapse and to protect them from lipolysis (Figure 2). Cancer cells present higher amounts of LDs than normal cells (71). Increased expression of PLIN2 has been shown to favor the accumulation of LDs (72), contributing to a better control of the ER stress, to increase the protection against ROS, and to augment the resistance to therapeutic drugs in cancer cells. On the contrary, PLIN2 depletion significantly attenuated the proliferation of colon cancer cells (73), supporting the LD-associated proteins as potential druggable targets for cancer treatment (11).

The increase in de novo synthesis of FAs in cancer cells requires efficient and complementary lipolytic mechanisms to accommodate the intracellular lipid content. Thus, lipolysis allows the stored lipids to be available for the synthesis of phospholipids and lipid signaling mediators and/or to increase the levels of ATP or NADPH when required. Several enzymes involved in lipolysis—adipose TAG lipase (ATGL), hormone-sensitive lipase (HSL), monoacylglycerol lipase (MAGL)—have been described to promote tumorigenesis (74). In this sense, ATGL knockdown in HCT116 CRC cells reduced cell proliferation (75). Increased levels of MAGL are associated with aggressive cancer types such as melanoma and ovarian and breast cancer (74), and inhibition of MAGL suppresses cancer cell migration, invasion, and survival (76). Recently, it has been demonstrated that glioblastomas, which acquire large amounts of free FAs, upregulate diacylglycerol-acyltransferase 1 (DGAT1) to store the excess FAs into triglycerides and LDs (77). Inhibition of DGAT1 disrupted lipid homeostasis, resulting in increased levels of ROS leading to apoptotic cell death.

In addition, a specific function of autophagy associated with the regulation of the intracellular lipid content—lipophagy—has been described to augment resistance to cell death in cancer (78).

In addition, similar to normal cells, cancer cells can uptake exogenous lipids when de novo lipogenesis is inhibited. Upregulation of cell surface receptors, such as cluster of differentiation 36 (CD36) (Figure 2), has been found to augment metastasis (79, 80). CD36 inhibition diminished tumor growth and metastasis in preclinical models of prostate cancer (80). Moreover, the expression of low-density lipoprotein receptor (LDLR) for the internalization of low-density lipoproteins (LDLs) has been found upregulated in renal cell carcinoma (RCC) cells (81). FA-binding proteins (FABPs) contribute to augment the lipid uptake, as well as the intracellular lipid trafficking in cancer cells (82). In breast cancer and glioblastoma cell lines, it has been shown that the uptake of extracellular FAs during hypoxia is sustained by the upregulation of FABP3 and FABP7; meanwhile, FABP5 increases cell proliferation and growth in prostate cancer (83).

Depending on the source of FAs, de novo lipogenesis or extracellular lipid uptake, the levels of saturated FAs incorporated in the phospholipids of cell membranes are different. The lipogenic pathway increases the saturation level of cell membranes with saturated and MUFAs (84), which are less sensitive to suffer lipid peroxidation compared to polyunsaturated acyl chains (PUFAs) mainly obtained from diet. This way, de novo lipogenesis contributes to augment the resistance to oxidative stress and chemotherapy in cancer cells (85).

Nevertheless, excessive accumulation of saturated FAs in the cell membranes can lead to lipotoxicity. In this regard, SCD1 inhibition induces ER stress and apoptosis in cancer cells and diminishes the tumor growth in xenografts models of colon and lung cancers (86). During tumor growth, inner parts of the tumors are faced to hypoxia and reduced nutrient availability. Tumors have developed different strategies to balance the levels of saturated vs. unsaturated FAs. Thus, tumors anticipate lipotoxicity by augmenting the uptake of MUFAs/PUFAs from plasma, which are further stored into LDs or incorporated into phospholipids at the cell membranes. As SCD1 activity requires oxygen, during hypoxia some tumors rely on the activity of DGATs to incorporate MUFAs into TG, which are further accumulated into LDs (Figure 2). In addition, tumors balance, via the Lands cycle, the levels of saturated vs. unsaturated FAs in the phospholipids at the cell membranes. Recently, a process known as ferroptosis has been described associated with high levels of MUFA/PUFAs in the phospholipids of cell membranes, which induce cell death by means of their oxidation through the Fenton pathway. Long-chain FA acyl CoA synthetases (ACSLs)—implicated in the long chain FA activation—may control ferroptosis, as distinct isoforms use distinct substrates. Meanwhile, ACSL4 has PUFAS as main substrates such as AA, ACSL3 can activate both MUFAs and PUFAs, allowing a better control of the excessive accumulation of PUFAs in phospholipids (87). In addition, ACSL3 allows a better control of FA distribution between LD storage or β-FAO, providing a better control of the oxidative stress (42).

De novo FA biosynthesis occurs in the cellular cytoplasm. FAs originate from acetyl-coenzyme A, which is mostly provided by citrate produced by the TCA cycle. Switch of citrate into AcCoA is catalyzed by ATP citrate lyase (ACLY) (Figure 2). Consequently, ACLY is a key enzyme connecting carbohydrate to lipid metabolism by producing AcCoA from citrate for both FA and cholesterol synthesis (61). Several studies have associated ACLY expression in tumor tissues with worse prognosis. ACLY overexpression correlated with stage, differentiation grade, and a poorer prognosis in non–small cell lung cancer (NSCLC) (61). Besides, in combination with the glucose transporter GLUT1, ACLY was also an independent prognostic factor for overall survival (OS) in node-negative patients with NSCLC (102). However, one study reports that young NSCLC patients overexpressing ACLY had longer OS, in contrast to older patients where overexpression of ACLY appears to predict the opposite prognosis (103). ACLY also facilitates colon cancer cell metastasis, and high expression levels of ACLY and Catenin β1 (CTNNB1) protein were positively correlated with metastasis of colon cancer (104). Data from TCGA showed ACLY as a putative unfavorable marker of cervical and liver cancer (Table 1).

At the genomic level, single nucleotide polymorphisms (SNPs) in ACLY gene have been described as independent cancer prognostic markers in Asiatic populations. SNP rs9912300 in ACLY gene was significantly associated with OS in lung cancer patients (165). rs9912300 and rs2304497, both functional ACLY SNPs, exhibited a significant association with risks of death and recurrence in patients with advanced stages of colon cancer (166).

The following step of FA biosynthesis involves the activation of AcCoA to malonyl-CoA, which is catalyzed by AcCoA carboxylase (ACC) (Figure 2). ACC is a complex multifunctional enzyme system. There are two ACC forms, α (ACACA) and β (ACACB), encoded by two different genes. High phospho-acetylCoA carboxylase (pACC) was an independent marker for prediction of better survival in lung adenocarcinoma patients (105), and low pACC levels detected by immunohistochemistry were associated both with worse OS and progression-free survival in advanced stage CRC (167). In the same line, gene expression analysis reported that patients with upregulation of six of these hub genes (genes with high correlation in candidate modules) (ACACB, acyl-CoA dehydrogenase medium chain, adiponectin, C1Q and collagen domain containing, acyl-CoA synthetase short-chain family member 2, phosphoenolpyruvate carboxykinase 1 and PLIN1) displayed improved breast cancer prognosis (106). In TCGA dataset, ACACA gene expression is an unfavorable risk factor for liver cancer, whereas ACACB is a favorable prognostic factor for both renal and pancreas tumors (Table 1). Finally, it has been described in prostate cancer that genetic alterations of ACACA, FASN, and SREBF1 predicted worse overall patient survival (168).

Malonyl-CoA is coupled to the multifunctional enzyme FASN. Repeated cycles of acetyl group's condensation produce the primary FA palmitate that can suffer separate elongation and/or unsaturation cycles to yield other FA molecules (169) (Figure 2). FASN is the key enzyme necessary for the de novo synthesis of long-chain FAs. FASN has been found overexpressed in nearly all of cancer tissues, and its expression is associated with a poorer prognosis.

One study reported that FASN gene expression was higher in the adjacent non-cancer tissue than in the NSCLC tissue, but authors concluded that it was a weaker predictor of shorter patient survival (170). However, a correlation analysis between expression levels of CD276 (B7-H3) and FASN exhibited a positive correlation with poor prognosis in clinical lung cancer tissues (107).

FASN levels were clearly upregulated in CRC tissues with high expression of FASN significantly associated with lymph node metastasis (108), liver metastasis (109), TNM (tumor, node, metastasis) stage, and poor prognosis (36). Moreover, a significant association was shown between FASN and VEGF expression, suggesting the involvement of FAS in tumor angiogenesis (110). Interestingly, one study reported that, among non-obese patients with colon cancer, tumoral FASN overexpression is associated with better survival, while among moderately overweight or obese patients, FASN overexpression may predict a poorer outcome (111). Furthermore, a panel of five genes including FASN (ACOT8/ACSL5/FASN/HMGBCS2/SCD1) has been reported to display a improved prognostic performance than validated clinical risk scales, and it is applicable for early discovery of CRC and tumor recurrence (112). Finally, FASN levels in serum were also examined in CRC patients, where it was associated with tumor stage (171), and high FASN levels are considered as a promising independent predictor of CRC with advanced phases, late clinical stages, and shorter survival (172).

FASN is associated with poor prognosis in breast and prostate cancer, and its inhibition is selectively cytotoxic to human cancer cells (113). FASN was found overexpressed in most of the triple-negative breast cancer (TNBC) patients but not always correlated with OS or disease-free survival. High FASN was significantly associated with positive node status (114). A greater part of clinically HER2-positive tumors was achieved as FASN overexpressors. Reclassification of HER2-positive breast tumors based on FASN gene expression predicted a significantly inferior relapse-free and distant metastasis-free survival in HER2⁺/FASN⁺ patients (115).

A substantial subset of prostatic cancers displays clearly elevated expression of immunohistochemically detectable FASN, a feature that has been associated with poorer prognosis (116–119). Furthermore, high expression level of FASN resulted in a significantly poor prognosis of pancreatic cancer (173), and data from TCGA study suggest that FASN expression could be a marker of bad outcome in cervical and renal cancer (Table 1).

In addition, several genetic changes in FASN gene have been associated with cancer prognosis. Two SNPs rs4246444 and rs4485435 were significantly associated with the recurrence of NSCLC (165). Finally, as it has been previously mentioned in prostate cancer that genetic alterations of FASN together with ACACA and SREBF1 predicted worse prognosis (168).

Then, FAs are activated with CoA by fatty acyl-CoA synthetases (ACSLs) (Figure 2), which is essential for phospholipid and triglyceride synthesis and lipid modification of proteins in addition to for FA β-oxidation (169).

Family of long-chain acyl-CoA synthetases has been extensively proposed as putative prognostic biomarkers of cancer. ACSL3 is up-regulated in lung cancer compared to the healthy lung tissue (174), and recently, an association with ACSL3 expression, NSCLC prognosis, and the efficacy of statins treatment has been discovered (L. P. Fernandez et al., unpublished results). ACSL3 was also found to be overexpressed in estrogen receptor–negative breast cancer (175) and prostate cancer (176). ACSL1 and ACSL4 overexpression was associated with a poor clinical outcome in stage II CRC patients (7–9, 120, 121). In addition, ACSL4 is considered a biomarker for liver and breast cancers (122, 177). By contrast, downregulation of ACSL5 in breast cancer was associated with a poorer prognosis (121, 123). There have not been reported associations between ACSL6 and cancer survival (178).

An in silico study (121) also suggested that high ACSL1 expression was associated with worse outcome in lung cancer patients, and ACSL3 overexpression was associated with worse survival in patients with melanoma. In contrast, high ACSL3 expression predicted a better prognosis in ovarian cancer. In the same study, ACSL4 overexpression predicted bad prognosis in CRC, but good prognosis in breast, brain, and lung cancers. High expression of ACSL5 predicted good prognosis in breast, ovarian, and lung cancers. Finally, low ACSL6 predicted a worse prognosis in acute myeloid leukemia. In silico analysis of TCGA data (Table 1) suggested that ACSL1, ACSL4, and ACSL5 are associated with favorable outcome in renal, urothelial, and endometrial cancers, respectively, whereas ACSL3 expression predicts poor survival in lung and liver tumors.

Genetically, a 3′-UTR polymorphism in ACSL1 is associated with ACSL1 expression levels and poor clinical outcome in CRC patients (14, 120). Patients carrying the ACSL1 rs8086 T/T genotype had significantly reduced disease-free survival compared with patients carrying the C/T or C/C genotype, with 3-fold higher risk of recurrences. T/T genotype for rs8086 is correlated with worse clinical outcome and simultaneously associates with high ACSL1 mRNA levels (14, 120).

Stearoyl CoA desaturase 1 (SCD1) catalyzes the rate-limiting step in the synthesis of MUFAs that are the main components of tissue lipids. SCD1 has been associated with tumor development, late stage, and reduced survival in lung adenocarcinoma (124). Together with other three lipid metabolism–related genes (ABCA1, ACSL1, and AGPAT1), SCD1 expression separated stage II colon cancer patients with a 5-fold higher risk of relapse (7). Moreover, positive associations between SCD1 expression and CRC patient clinical status and the expression of cancer stem cell–related genes (WNT and NOTCH signaling) were found based on TCGA data analysis (125). In the same line, high SCD1 expression is associated with shorter survival in breast cancer patients (126). Table 1 shows that SCD1 is an unfavorable marker of survival in renal and urothelial cancer in TCGA tumors. Other desaturases have also been analyzed as prognostic markers, and, for example, reduced expression of FADS1 suggests pessimistic prognosis for NSCLC patients (127).

Glycerol-3-phosphate acyltransferase (GPAT) catalyzes the first step in the production of almost all membrane phospholipids. GPAT transfers an acyl group from acyl-CoA or acyl-ACP at the sn-1 or-2 position of glycerol 3-phosphate originating lysophosphatidic acids (LPAs) (179). LPA is a substrate for synthesis of numerous important glycerolipid intermediates, such as storage lipids, extracellular lipid polyesters, and membrane lipids (Figure 2). Four GPATs have been discovered; nevertheless, only GPAT1 (GPAM) has been related to cancer outcome. High GPAT1 expression has been associated with reduced OS in ovarian cancer (180). Data from TCGA suggested that GPAT1 could be a favorable prognostic marker in renal cancer, while GPAT3 is a putative biomarker of good prognosis in renal cancer in contrast to urothelial cancer. Finally, GPAT4 expression could have a risk effect in ovarian and endometrial cancers, and a protective one in prostate and urothelial cancer (Table 1).

LPA is further metabolized to phosphatidic acid (PA) by AGPATs (1-acylglycerol-3-phosphate O-acyltransferases) (Figure 2). AGPAT1 belongs to previously mentioned transcriptional signature where combined analysis of four genes, ABCA1, ACSL1, AGPAT1, and SCD1, is associated with higher risk of relapse in stage II CRC patients (7). Furthermore, individuals with upregulation of AGPAT1 expression have an increased risk of CRC recurrence, independently of tumor stage (8). Expression of AGPAT2 was significantly related to decreased OS as well as to shorter progression-free survival in ovarian cancer patients younger than 60 years (181). When we consider tumors from TCGA study, several associations were found (Table 1). AGPAT3 is a marker of good prognosis in renal cancer and predicts bad outcome in cervical cancer. High expression levels of AGPAT4 may be associated with poor prognosis in cervical and renal cancers, whereas AGPAT5 is an unfavorable prognostic marker in liver cancer and a favorable one in CRC.

Then PA is converted to diacylglycerol (DAG) by LPIN, a PA phosphatase. Three LPIN isoforms have been described. LPIN1 is upregulated in lung adenocarcinoma tumor tissues, and high LPIN1 expression was correlated with poor prognosis of patients with lung adenocarcinoma (134). In breast cancer, previous results seem to indicate that the high LPIN expression is related to a good prognosis (135). However, in basal-like TNBC, high LPIN1 expression correlates with the poor prognosis of these patients (136). In TCGA dataset analysis, LPIN2 appears as a favorable prognostic marker in head and neck cancers, while LPIN3 could be an unfavorable biomarker of endometrial, ovarian, and renal tumors (Table 1).

The final step in triacylglycerols synthesis is catalyzed by DGAT, which esterifies the DAG with a FA. Two human DGAT isoforms have been described (182). The expression of DGAT2 in HER2-positive breast cancer was decreased and was closely related to patient prognosis (140). However, data from TCGA reported DGAT2 as an unfavorable prognostic factor for endometrial cancer (Table 1).

Subsequently, TAGs could be stored in LDs, and PLINs, an LD surface family of proteins, are necessary for optimal lipid storage and FA release. There are multiple PLIN proteins encoded by mRNA splice variants of a single PLIN gene. PLIN1 expression in lung adenocarcinoma is associated with apocrine-like features and poor clinical prognosis (137). In contrast, PLIN1 mRNA expression is significantly downregulated in human breast cancer. The reduced expression of PLIN1 is an independent predictor of OS in estrogen receptor–positive and luminal A-subtype breast cancer patients (138). Also in breast cancer, low expression of PLIN2 was associated with favorable prognosis (139). The prognostic effects of PLINs in several types of cancer from TCGA analysis are multiple and very diverse (Table 1).

Eicosanoids are biologically active metabolites of AA and are produced by cyclooxygenases 1 and 2 (COX1 and COX2) [also known as prostaglandin-endoperoxide synthase 1 and 2 (PTGS1 and PTGS2)]. They are overexpressed in a variety of malignant tumors. It has been reported that the mRNA levels of COX-1 and COX-2 in lung cancer patients were significantly higher than in normal patients (183). However, another study reports that in tumor cells COX-2 rather than COX-1 expression may account for the variable prostanoid production seen in NSCLC (128). It is clear that multivariate analysis showed that tumoral COX-2 mRNA expression and lymph node status were the most important independent prognostic predictors for NSCLC survival and disease relapse (129). Elevated COX-2 expression in tumors was significantly associated with lower survival in NSCLC and might be useful in identifying patients who would benefit from additional therapies for managing their disease (130).

The same tendency was observed in CRC, where elevated COX-2 expression, but not that of COX-1, was significantly associated with reduced survival and recognized as an independent prognostic factor (131). However, it has been reported that COX-1 and COX-2 expression is highly variable in Dukes' C tumors, and changes in COX-1 expression may be of importance in CRC (184).

COX-2 expression level and its prognostic value are also a matter of debate in breast cancer (185). Nevertheless, at least eight immunohistochemical reports have explored expression of COX-2 in a total of 2,392 primary breast carcinomas, of which 40% were found to be COX-2 positive (132). At least, four studies have detected that overexpression of COX-2 is linked to poor prognosis in breast cancer. These studies provide the basis for further estimation of a possible therapeutic effect of COX inhibitors in therapy of breast cancer.

In prostate cancer, a subset of Chinese patients with high-COX-2 expression showed minor disease-free and OS rates than those with low COX-2 expression. In this work, univariate and multivariate analyses suggested that the status of COX-2 protein expression was an independent prognostic factor for patients' survival (133).

Data from TCGA showed COX-1 and COX-2 as unfavorable markers of renal cancer, whereas only COX-1 was a risk biomarker of urothelial cancer (Table 1).

Chronic inflammation is a recognized risk factor for CRC, and polymorphisms in genes regulating inflammatory processes appear to modify the risk of neoplasia and the efficacy of non-steroidal anti-inflammatory drugs in CRC chemoprevention. COX-1 polymorphism G213G was significantly associated with an increased CRC (186). Finally, another study reports four COX-1 variants that were associated with CRC survival. rs1213266 was associated with approximately 50% lower CRC mortality. Three other variants, including L237M, resulted in significantly elevated CRC mortality risk (187).

Proteins related to FAs transportation are also relevant as cancer biomarkers. Carnitine palmitoyltransferase, CPT1A, is a protein that is responsible for the translocation of FAs from the cytosol to the mitochondrial matrix, where FA oxidation occurs. Associations of shorter disease-free survival with CPT1A positivity in invasive lobular carcinoma of the breast have been found (142).

Another study recognized a gene expression signature composed of 19 genes associated with FAO that was significantly associated with breast cancer patient survival. These 19 genes are referred to as the “fatty acid oxidation (FAO)” signature. Included in this signature were genes that have previously been identified as the core components of the FA β-oxidation pathway, such as CPT1A. Moreover, the expression of CPT1A was elevated in estrogen receptor–positive, compared to estrogen receptor–negative tumors and cell lines (143). Data from TCGA clearly confirm a CPT1A association with poor prognosis in breast cancer, whereas CPT1A is a marker of good prognosis in renal cancer and CPT1C in pancreas (Table 1).

Other relevant FA transporter is CD36. CD36, a scavenger receptor expressed in multiple cell types, mediates lipid uptake, immunological recognition, inflammation, molecular adhesion, and apoptosis. CD36 has been continually proposed as a prognostic marker in diverse cancers, mostly of epithelial origin (breast, prostate, ovary, and colon) and also for hepatic carcinoma and gliomas (141). Through systematic analysis of the multiple omics data from TCGA, it has been found that the most widely altered lipid metabolism pathways in pan-cancer are FA metabolism, AA metabolism, cholesterol metabolism, and peroxisome proliferator-activated receptor (PPAR) signaling. Genes related to lipid metabolism and immune response that were associated with poor prognosis were discovered including CD36 (188).

First step of cholesterol or mevalonate pathway is catalyzed by acetyl-coenzyme A ACAT1 (Figure 2). ACAT1 is a mitochondrial enzyme that catalyzes the reversible formation of acetoacetyl-CoA from two molecules of AcCoA. An increased expression of ACAT1 in intratumor cholesteryl ester–rich breast tumors was reported (189). Also it has been proposed that ACAT1 expression could serve as a potential prognostic marker in prostate cancer, specifically in differentiating indolent and aggressive forms of cancer (144, 145). Data from TCGA suggest that ACAT1 is a marker of good prognosis in liver and renal tumors. Interestingly, isoform 2 (ACAT2) is a marker of good prognosis in CRCs, whereas in endometrial and renal tumors, ACAT2 has the opposite effect (Table 1).

Next step in cholesterol synthesis is mediated by 3-hydroxy-3-methylglutaryl-CoA synthase (HMGCS). This enzyme, with two isoforms, condenses AcCoA with acetoacetyl-CoA to form HMG-CoA, which is the substrate for HMG-CoA reductase. HMGCS2 expression is associated with reduced clinical prognosis and outcomes in patients with CRC and oral squamous cell carcinoma. It has been suggested that HMGCS2 may act as a helpful prognostic marker and essential target for potential therapeutic strategies against advanced cancer (146). Also, it has been described that HMGCS2 works as a tumor suppressor and has a prognostic impact in prostate cancer, capable of predicting the risk of biochemical recurrence (147). However, in TCGA population, both isoforms are favorable makers of renal cancer. Besides, HMGCS2 determines good prognosis in ovarian and liver cancer (Table 1).

HMGCR is the rate-limiting enzyme of the mevalonate pathway (Figure 2). HMG-CoA reductase expression in CRC and breast cancer correlates with favorable clinicopathological characteristics and an improved clinical outcome (148–150). Besides, HMCGR expression is a predictor of response to tamoxifen in breast cancer (190) and also may predict patient response to radiotherapy in ductal carcinoma in situ (191). In TCGA subset, HMGCR also is a good prognosis marker of renal tumors (Table 1). Statins, lipid-lowering compounds commonly used in cardiovascular disease, are competitive inhibitors of HMGCR. The value of HMGCR as a predictor of response to neoadjuvant or adjuvant statin treatment in cancer was also studied (192).

Once that cholesterol is synthesized, there are several cholesterol transporter proteins that play key roles in cholesterol and phospholipids homeostasis. The ATP-binding cassette transporter ABCA1 is a transmembrane protein responsible for the reverse cholesterol transport from the inner cell to circulatory system. ABCA1 is significantly overexpressed in patients of all stages of CRC, and its overexpression gives proliferative advantages together with caveolin-1–dependent increased migratory and invasive capacities (151). Individuals with upregulation of ABCA1 expression have an improved risk of CRC recurrence and OS independently of tumor stage (8). ABCA1 also forms part of the metabolic-signature ColoLipidGene able to precisely stratify stage II CRC with 5-fold higher risk of relapse (7). Moreover, the presence of tumoral genetic variants located in ABCA1 coding region seems to be associated with CRC risk of death (8). In other tumor types, ABCA1 expression was related to positive lymph nodes, but not significantly associated with tumor recurrence or breast cancer–specific survival (193).

Together with ABCA1, ATP-binding cassette G1 (ABCG1) also initiates and propagates cellular cholesterol efflux. Several genetic variants in ABCG1 have been associated with survival of NSCLC patients (194). Moreover, ABCG1 expression seems to be a favorable prognostic marker of renal cancer in data from TCGA (Table 1).

Other members of the family are the ATP-binding cassettes G4, G5, and G8. High ABCG4 expression has been associated with poor prognosis in NSCLC patients treated with cisplatin-based chemotherapy (152). ABCG5 positivity in tumor buds have been proposed as an indicator of poor prognosis in node-negative CRC patients (153), whereas in TCGA tumors, ABCG5 seems to have a favorable effect in liver prognosis (Table 1).

While cellular cholesterol efflux is mainly performed via ABCA1, cholesterol uptake is principally executed via the LDLR. The prognostic value of LDLR expression was analyzed in CRC where authors found that the absence of LDLR predicts a shorter survival (154). In the same line, lower LDLR expression was an independent prognostic factor associated with longer survival in patients with small cell lung cancer (195). By contrast, TCGA data suggest that LDLR could be a bad prognostic marker of pancreatic, renal, and urothelial cancers (Table 1).

Five are the main transcription factors that regulate the expression of mediators of lipid metabolism: SREBP1, SREBP2, PPARγ, NR1H3, and NR1H2. Sterol regulatory element-binding protein 1 (SREBP1) is a known transcription factor of lipogenic genes, which plays important roles in regulating de novo lipogenesis. SREBP1 is overexpressed and strongly associated with worse clinical outcomes in breast cancer (155). Moreover, SREBP1 also seems to have an essential role in pancreatic cancer, regulating tumorigenesis and being associated with bad prognosis (196). However, data from TCGA propose SREBP1 as a favorable prognostic marker in pancreatic and endometrial cancers (Table 1).

The combined expression of sterol regulatory element-binding protein 2 (SREBP2) together with HMGCR, NR1H3, and NR1H2 genes was associated with poor CRC clinical outcome independent of lymph node metastasis, distant metastasis, and advanced stage (156). Besides, expression of SREBP-2 was elevated in advanced pathologic grade and metastatic prostate cancer and significantly associated with poor clinical outcomes (157).

The PPARγ is a nuclear receptor that controls expression of mediators of lipid metabolism but also the inflammatory response. Additionally, it has been demonstrated that PPAR b/d and a isotypes also have important roles in FAO, FA storage, and cholesterogenesis (197).

Decreased expression of PPARγ has been observed in many tumor types. In this sense, reduced PPARγ expression within the tumor is associated with poor prognosis in lung cancer patients (158, 159). In the same line, tumor expression of PPARγ is independently associated with increased survival of CRC patients (160). Also in patients with breast and prostate cancer, PPARγ is a marker of better prognosis and is associated with better survival (161–163). Importantly, one study reports that cytoplasmic PPARγ expression appeared as an independent marker of poor prognosis in primary breast cancers (198). TCGA analysis proposed PPARγ as a favorable prognostic marker for renal and urothelial cancers (Table 1).

Finally, several studies have also evaluated the association between PPARγ genetic variants and the risk of CRC (199). In patients with stages II/III CRC, polymorphism rs1801282 in PPARγ was significantly associated with tumor recurrence (200).

NR1H3 and NR1H2 encode for liver X receptor (LXR) α and LXR β, respectively. They are intimately related nuclear receptors that react to elevated levels of intracellular cholesterol by enhancing transcription of genes that control cholesterol efflux and FA biosynthesis. NR1H3 expression was significantly correlated to better survival in completely resected stages II and III NSCLC patients (164). Moreover, one study reports that NR1H3 and NR1H2 belong to a transcription signature associated with poor CRC clinical outcome independent of lymph node metastasis, distant metastasis, and advanced stage (156). This result is validated in TCGA dataset (Table 1) where NR1H2 was also associated with CRC poor prognosis.