Metabolic Signaling in Cancer Metastasis

Metastasis formation is a multistep process that requires cancer cells to dynamically change their phenotype. The early metastatic steps include invasion into the surrounding tissue and dissemination via the blood and lymphatic circulation to distant organs. Once the cancer cells have settled outside the primary tumor, they have to colonize and outgrow in the new organ. In recent years, it has emerged that alterations in metabolism are a prerequisite for the early and later steps of metastasis formation because metabolism responds to but also enables the plasticity and heterogeneity of metastasizing cancer cells (1, 2). Thus, there is a rising interest in exploiting the metabolic vulnerabilities of metastasizing cancer cells for metastases prevention and treatment.

Classic activities of metabolism provide energy, an antioxidant defense, and biomass building blocks for the survival, proliferation, and motility of cancer cells. Research on these more classic functions of the metabolic network in cancer cells was initiated by Otto Warburg 100 years ago when he observed that malignant tumors produce significantly more lactate than non-transformed tissues (3). Thus, we nowadays can rely on a wealth of information regarding the structure and players in the metabolic network and how its activity fuels cancer cells by converting metabolites in enzyme-catalyzed reactions into ATP (energy), amino acids, nucleotides, and fatty acids (precursors for proteins, DNA/RNA and lipids) as well as antioxidants such as glutathione. Yet, in more recent years, it has also become evident that metabolites are not only intermediates of biochemical reactions but that they are also sensed to activate receptor and protein kinase signaling and that they are substrates for posttranslational modifications (PTM) that change protein localization, activity, interaction, and stability (Fig. 1). In addition, for more and more enzymes moonlighting functions have been detected. These nonclassical functions of enzymes include altered catalytic activities, often in the nucleus, where the catalyzed reaction does not involve a metabolite but a protein substrate. Furthermore, also noncatalytic moonlighting functions of enzymes have been observed in which enzyme-protein interactions change the activity, localization, or stability of the partner protein; or enzyme-receptor interactions where the enzyme acts as an agonist or antagonist of the associated receptor. These nonclassical functions of metabolism thus trigger metabolic signaling, here defined as the involvement of metabolites or metabolic enzymes in any noncanonical cellular process. Thus, metabolic signaling constitutes one important intersection of metabolism, with the cellular processes that enable plasticity and heterogeneity of (cancer) cells.

Here, we provide an overview of how metabolic signaling contributes to metastasis formation and where promising strategies may be explored to leverage metabolic signaling for metastasis prevention and treatment.

Protein kinases and receptors allow cells to respond to their internal cell state, neighboring cells, and their environment. classic triggers for the signaling cascades downstream of protein kinases and receptors are cytokines, growth factors, and hormones, but accumulating evidence also highlights the importance of metabolites in triggering and regulating receptor and protein kinase–mediated signaling cascades during metastasis formation (Fig. 2).

The mammalian target of rapamycin complex 1 (mTORC1) signaling, which is induced by metabolites and growth factors, has been studied for decades in cancer proliferation. Although amino acids are not the sole metabolites involved, they play a pivotal role as the primary metabolites activating mTORC1 signaling. Recent studies showed that amino acid activation of mTORC1 signaling is similar between primary tumors and metastases, but differences may exist for activation of mTORC1 by other metabolites such as α-ketoglutarate (Fig. 2).

mTOR is activated by intracellular amino acids. Therefore, amino acid transporters become very relevant to mTORC1 activation. In the context of mTORC1 signaling in metastasis mainly the role of L-type amino acid transporters has been studied and similar effects on primary tumors and metastasis were observed. Genetic LAT1 (SLC7A5) or LAT3 (SLC7A6) loss suppressed primary tumor growth and spontaneous metastasis formation in prostate cancer xenograft (4) and colorectal cancer (5) mouse models. Moreover, it was recently shown that LAT3 expression depends on sufficient iron levels (6), which could explain the correlation of high-circulating iron levels with poor prognosis in some cancers (7). Similarly, deletion of the general amino acid transporter Slc6a14, which can transport 18 out of 20 amino acids, decreased primary tumor growth as well as metastasis formation and consequently improved overall survival in pancreatic ductal adenocarcinoma KPC mouse models (8). All the aforementioned studies proposed a direct or indirect involvement of mTORC1 activation in metastasis formation via the SLC-transported amino acids. Yet, the observation that simultaneous loss of Lat1 and targeting of mTORC1 activity with rapamycin in murine Kras-mutant colorectal cancer significantly extended lifespan compared with only Lat1 loss suggests roles of LAT1 in tumor growth and metastasis that go beyond direct activation of mTORC1 (5).

There is also evidence that mTORC1 activation may differ between primary tumors and the corresponding metastases. In patients with breast cancer, a discordance between mTORC1 activity in primary tumors and matched metastases assessed by phosphorylation of eukaryotic translation initiation factor 4E-binding protein 1 (p-4E-BP1) was observed (9). In line, a higher mTORC1 activity in lymph node metastases and general metastases compared with primary tumors was seen for medullary thyroid carcinoma (assessment of p-S6) and renal cell carcinoma (assessment of p-mTOR) samples from patients, respectively (10, 11). Consistently, increased mTORC1 activity based on p-S6 and p-4-EBP1 was observed in lung metastases compared with primary breast cancer of mice (12). Mechanistically, it was found that mTORC1 was activated in lung metastases by α-ketoglutarate derived from an increased de novo serine biosynthesis (12, 13). In contrast, serine biosynthesis was not needed for mTOR activation in primary breast tumors (12). Accordingly, lung metastases only responded to rapamycin when the tumors expressed phosphoglycerate dehydrogenase (Phgdh), which is the first enzyme of the serine biosynthesis pathway (12).

mTORC1 is negatively regulated by AMP-activated protein kinase (AMPK) activity, which is a master regulator of energy homeostasis and accordingly has implications for cancer progression (Fig. 2). AMPK activation is triggered by elevated AMP/ATP and ADP/ATP ratios. Thus, the loss of Ampk catalytic subunit alpha 1 and 2 (AMPKα1 and 2, i.e., PRKAA1 and 2) has been shown to promote lung and brain metastasis from breast cancer (14) and melanoma mouse models (15), respectively. Moreover, decreased expression of the fatty acid transport protein 5 (FATP5/SLC27A5) enhanced both, hepatocellular carcinoma growth and metastasis, by suppressing AMPK activation (16). Mechanistically, loss of FATP5 increased glucose uptake and consequently glucose-derived ATP production leading to AMPK inhibition and eventually mTORC1 activation in HCC cells (16). In line, the reactivation of AMPK using metformin impaired the metastatic capacity of FATP5-depleted hepatocellular carcinoma (HCC) cells (16). In triple-negative breast cancer, however, the disruption of mitochondrial ATP production and the subsequent activation of AMPK signaling, achieved by chelating copper with tetramolybdate, only hindered lung metastasis formation without affecting the growth of the primary tumor (17). This suggests that the same metabolic effect of impaired ATP production may have different effects depending on the tumor context and the trigger inducing the ATP shortage (Fig. 2).

In summary, although mTORC1 and AMPK signaling commonly hold significance in both primary tumors and metastases, there are instances and cancer types where the metabolic activation of the AMPK-mTORC1 axis diverts between metastases and primary tumors.

G-protein–coupled receptors (GPR or GPCR) constitute a cell-surface receptor superfamily of over 800 members, for many of them odors, chemokines, neurotransmitters, and hormones act as agonists. So far only a small fraction of GPRs have been identified that use small molecules as activators and some of them have been implicated in the cancer cell–stroma interaction during metastasis formation (Fig. 2).

Free fatty acids are known agonists of GPR40 and 120. In patients, little is known about the concentrations and origin of the free fatty acids that activate these GPRs. Yet, GPR40 expression was found to be correlated with the occurrence of metastasis in colorectal cancer (18) and melanoma (19) patients. Moreover, GPR120 expression correlated with lymph node metastasis and poor prognosis in PDAC patients (20). Multiple fatty acids including elaidic acid can activate GPR120 and GPR40. Interestingly, blood concentrations of palmitoleic acid and elaidic acid have been associated with colorectal cancer in patients (21). In line, it was found that treatment of colorectal cancer cells with elaidic acid promoted in vitro sphere formation, which was abrogated when both GPR40 and 120 were silenced (22). Moreover, oral administration of elaidic acid to mice promoted colorectal cancer xenograft and subsequent metastasis growth (22). A prometastatic influence of GPR40 and 120 signaling observed in colorectal cancer and papillary renal cell carcinoma (pRCC) cells converged into the activation of the epidermal growth factor receptor (EGFR) signaling pathway (22, 23).

Succinate has been identified as an agonist of GPR91, which was later renamed to succinate receptor 1 (SUCNR1). Plasma succinate levels are not only a biomarker for cancers with succinate dehydrogenase (SDH) mutation but can also distinguish patients with non–small cell lung cancer (NSCLC) from healthy individuals, likely because some cancer cells secrete succinate due to reduced SDH activity (24). In this respect, it was recently shown that besides the canonical succinate transporter SLC13A3, the monocarboxylate transporter 1 (MCT1) can shuffle succinate across the plasma membrane (25). Moreover, tumor reoccurrence in patients with head and squamous cell carcinoma correlated with high SUCNR1 expression (26). Activation of Sucnr1 by succinate mainly occurred in stromal cells such as macrophages, CD8⁺ T cells, and endothelial cells of the tumor microenvironment (27–29), but also autocrine activation of the receptor in cancer cells has been observed (24). In stromal cells, the antitumor activity of CD8⁺ T cells required autocrine Sucnr1 activation (28), whereas protumor M2-polarized macrophages and angiogenesis were stimulated upon Sucnr1 activation by cancer cell–derived succinate (27, 29). Treatment of murine lung cancers with succinate was shown to increase metastasis formation in multiple organs including lung and liver, whereas primary tumor growth was not affected (24). Moreover, exosomes from human umbilical cord mesenchymal stem cells reduced SUCNR1 protein expression via the miR-1827 in cancer cells, which decreased primary tumor growth and metastasis formation of colorectal cancer xenografts. The effect of this exosome treatment was rescued by overexpression of SUCNR1 in cancer cells (30). Mechanistically, it was shown that autocrine activation of SUCNR1 by succinate activated an epithelial-to-mesenchymal transition via phosphoinositide 3-kinase (PI3K) and hypoxia-inducible factor 1 subunit alpha (HIF1α) cascade in NSCLC cells (24).

GPR81 and GPR132 have been reported as lactate receptors. Cancer cell–derived lactate promoted autocrine or, in the tumor microenvironment, paracrine receptor activation that has been linked in breast cancer to metastasis formation. Accordingly, lower GPR132 expression correlated with increased metastasis-free survival in patients with breast cancer (31). Moreover, injection of murine breast cancer cells to Gpr132 knockout mice and crossing of the genetic PyMT breast cancer mouse model into Gpr81-null mice resulted in reduced lung metastasis and, in the case of the PyMT/Gpr81^−/− mice, also delayed tumor onset (31, 32).

Mechanistically, the observed outcomes in both cases were associated with paracrine signaling where cancer cell–derived lactate interacted with Gpr132 on macrophages or Gpr81 on dendritic cells, ultimately contributing to protumor microenvironments. These microenvironments were characterized by a reduced presentation of tumor-specific antigens by dendritic cells (Gpr81-null mice) and the polarization of macrophages toward the M2 phenotype (Gpr132-null mice). The latter observation may be breast cancer-specific because, in lung cancers growing in Gpr132-null mice, no differential macrophage polarization was observed (33). Notably, Gpr81 activation during sensory neuron excitation has been linked to bone pain induced by metastatic breast cancer hinting at a potential link between this metabolic signaling and tumor innervation (34).

Also, cancer cell–intrinsic effects of GPR81 silencing have been observed. In multiple cell lines including breast cancer cells, GPR81 silencing reduced in vitro cancer cell survival and motility (35, 36). Moreover, the knockdown of GPR81 in human breast cancer cells implanted in the bone showed reduced growth suggesting a cancer cell–intrinsic role of lactate signaling in metastasis formation (37).

In summary, GPRs activated by metabolites are understudied in cancer, and current evidence shows that cancer cell–intrinsic but also paracrine GPR signaling contributes to metastasis formation. Preclinical antagonists against SUCNR1 have been developed (38, 39), yet it was recently discovered that extracellular succinate-SUCNR1 signaling was important for inducing transcriptional responses for muscle remodeling upon exercise (40) and adipose tissue thermogenesis (41). This suggests that further studies are needed to evaluate GPRs as drug targets in cancer metastasis.

Presynaptic neurons, in the brain parenchyma, secrete vast amounts of glutamate, which acts as an excitatory neurotransmitter in the central nervous system by activating the ionotropic N-methyl-D-aspartate (NMDA) receptor. Cancer cells can hijack this neuron-specific signaling cascade.

Specifically, it was shown that a human brain tropic triple-negative breast cancer cell line highly expressed the NMDA receptor subunit GRIN2B compared with the parental cell line and that the corresponding brain metastases had an elevated GRIN2B expression compared with lung metastases and the primary breast tumor (42). Moreover, genetically engineered mouse models of pancreatic neuroendocrine tumor (PNET), pancreatic adenocarcinoma, and luminal breast cancer showed GRIN2B expression predominantly in the invasive front of the tumors (43). Supporting these data, increased GRIN2B expression was observed in PDAC, breast cancer, ovarian cancer, and glioma patient samples (43, 44). Underscoring the importance of this finding, it was shown that silencing GRIN2B reduced breast cancer–derived brain metastases in mice (42). The NMDA receptor inhibitor MK801 prevented and regressed PNET tumors in mice, whereas the weaker clinically approved inhibitor memantine only regressed them (43). Both inhibitors prolonged the survival of mice harboring PDAC tumors (45) and MK801 treatment arrested breast cancer xenograft growth (46) as well as lung metastasis (47). Accordingly, in patients with breast cancer, elevated GRIN2B expression was associated with poor distant-relapse-free survival and a gene-expression signature indicative of low NMDA receptor activity correlated with improved survival in PDAC, brain cancer, kidney cancer, and uveal melanoma (42, 45).

Glutamate is an agonist of the NMDA receptor. It was found that the glutamate activating the NMDA receptor in breast cancer–derived brain metastases from mice was provided through the formation of pseudotripartite synapses between cancer cells and glutamatergic neurons (42). Yet, also an activation of the NMDA receptor by glutamate released from the cancer cells themselves was observed in PNET cell lines. This glutamate release was triggered by low interstitial fluid pressure, which is found at the invasive front of a tumor (43). Moreover, increased secretion of glutamate was also triggered by 17-β estradiol (E2)-mediated activation of G protein–coupled estrogen receptor (GPER) in triple-negative breast cancer cells (47). Once glutamate activated the NMDA receptor, the associated Ca²⁺ influx promoted cAMP response element-binding protein (CREB) phosphorylation and the downstream expression of a proinvasive and prometastatic gene signatures through a multistep signaling cascade in breast and PNET cancer cells (43, 47). Moreover, there are indications that activation of the NMDA receptor also promoted the translation of proinvasive genes via activation of the heat shock transcription factor 1 (HSF1) and Fragile X Messenger Ribonucleoprotein 1 (FMR1) axis in PNET cancer cells (45). Finally, the activity of the matrix metalloproteinase 2 (MMP2), which can promote invasion, was correlated with NMDA receptor activity in glioma cell lines (48).

Thus, cancer cells exploit the activation of the NMDA receptor by autocrine- or paracrine-released glutamate to gain invasive capacities promoting metastasis formation. Clinically, it is interesting to target the NMDA receptor in cancer metastasis because several approved inhibitors exist. Esketamine and dextromethorphan are approved for major depressive disorder, whereas memantine is approved for dementia and amantadine for the treatment of Parkinson's disease. Epidemiologic studies analyzing the metastasis risk of patients taking NMDA receptor antagonists, and especially memantine, for which antitumor effects have been reported in mouse models (43, 49, 50), are currently missing.

Metabolites are substrates for PTMs of proteins. Enzymes known as transferases facilitate the covalent attachment of metabolites to proteins, whereas erasers undo these modifications. Furthermore, the extent of protein modification is intricately linked to the concentration of available metabolite substrates. Metabolite substrate concentrations and transferase/eraser expression modulate PTMs in metastasis formation (Figs. 3 and 4). Although transferases and the corresponding erasers are currently explored as drug targets, approaches to modulate metabolite concentrations in metastasizing cancer cells are mostly indirect. Moreover, how diet influences meta­bolite concentrations and consequently PTMs in metastasis formation is only emerging.

Glycosylation is the attachment of a carbohydrate to a protein (or lipid). Various modifications exist such as glucuronidation, O-GlcNAcylation, and sialylation, which use UDP-glucuronic acid, UDP-acetylglucosamine (O-GlcNAc), and CMP-sialic acid as metabolite substrate, respectively. It has been consistently observed across various cancers that increased glycosylation predominantly enhances the metastasis initiation capacity of cancer cells, with only a few exceptions (Fig. 3). In this respect, mainly the expression of glycosylating enzymes has been studied rather than the effect of metabolite substrate availability for driving this posttranslational modification. Yet, there is some evidence that increased production of the metabolites needed for glycosylation promotes metastasis initiation (51).

N-acetylgalactosaminyltransferase (GALNT) transfers GalNAc to the serine and threonine residues of proteins resulting in O-linked protein glycosylation in the Golgi. Mostly invasiveness-promoting consequences of O-glycosylation have been recorded in cancer cells (Fig. 3A).

Several proteins, crucial for the initiation of metastasis by facilitating invasive properties and extracellular matrix (ECM) degradation in cancer cells, undergo O-linked glycosylation. Calnexin, integrin α₅ and β₁, EGFR, AXL receptor tyrosine kinase (AXL), and matrix metalloproteinase 14 (MMP14) were observed to be O-glycosylated in human and mouse liver tumors (calnexin) as well as human liver (calnexin, integrin β₁, MMP14), oral (EGFR), colorectal cancer (AXL), skin (MMP14), and NSCLC (integrin α₅) cancer cell lines supporting invasiveness (52–58). MMP14 and AXL/integrin α₅ O-glycosylation were dependent on GALNT1 and GALNT2, respectively (53, 55, 58). Moreover, GALNT2 expression indicated poor prognosis in patients with colorectal cancer and GALNT14 promoted breast cancer–derived lung metastasis (53, 59), while conversely, GALNT8 inhibited the migratory capacity of breast cancer cells because the corresponding glycosylation impaired EGFR activity (60). Interestingly, shifting the localization of GALNT1 from the Golgi to the endoplasmic reticulum in a mouse liver tumor model resulted in MMP14 glycosylation and advanced tumor development as well as metastasis in multiple organs compared with no metastases in control tumors. Moreover, overexpression of Golgi-localized GALNT1 only moderately promoted metastases (58). This implies that the subcellular localization of O-glycosylating enzymes plays a significant role in determining the specificity of their protein targets.

Sialylation is the attachment of CMP-sialic acid via β-Galacto­side sialyltransferases (STGAL) to N-glycans. In general, increa­sing sialylation either through increasing CMP-sialic acid availability and increased sialyltransferase or decreased sialidase activity promotes metastasis initiation in various cancer types (Fig. 3B).

In patients with breast cancer, it was observed that the gly­cosylation patterns of metastases were mostly organ dependent but that some N-glycans increased with metastatic progression (61). Moreover, ST3GAL6, ST3GAL1 gene, and ST3GAL1 protein expression were indicative of poor prognosis in urinary bladder cancer (62), melanoma progression (63), and decreased disease-free survival in clear cell renal cell carcinoma (ccRCC) (64), respectively. Consistent with this, altered expression of six sialylated N-glycans was found to be correlated with lymph node metastasis in patients with oral squamous cell carcinoma (OSCC) (65). Mechanistically, elevated expression of ST3GAL1 and ST3GAL3 induced the activation of the receptor tyrosine kinase AXL in melanoma cells (63) and integrin β₁ and MMP2/9 in breast cancer cells (66), respectively, promoting invasion in vitro. Moreover, extracellular vesicles from B16 melanoma cells expressing high levels of ST3GAL5 induced premetastatic niche formation and subsequent peritoneal dissemination in mice (67). Conversely, ST6GAL1 was downregulated in HCC tissues and overexpression of ST6GAL1 inhibited HCC metastasis through impaired melanoma cell adhesion molecule (MCAM) expression at the cell surface (68).

CMP-sialic acid is the crucial substrate for sialylation, and it is produced by a side pathway of glycolysis with cytidine monophosphate N-acetylneuraminic acid synthetase (CMAS) being the last enzyme of the pathway. Cmas deletion reduced lung metastasis formation in a breast cancer mouse model (69). Moreover, diverging carbon flux toward CMP-sialic acid production increased metastasis initiation in different breast cancer mouse models and invasion in triple-negative breast cancer cells via integrin β₃ sialylation, which was abrogated by Cmas silencing or blockage of integrin α_vβ₃ activity, respectively (51).

The erasers of sialylation are less studied than the transferases. So far only four sialidases (NEU) have been identified. Decreased NEU1 expression correlated with bladder cancer progression (70). NEU4 was found to be downregulated in HCC and inhibited cell motility at least in part through cleavage of sialic acid residuals from the cell-surface adhesion receptor CD44 (71). Moreover, NEU1 expression impaired a proinvasive fibronectin-integrin α₅β₁ interaction in bladder cancer cells (70). In contrast, NEU1 and NEU3 expression was increased in melanoma and bladder cancer tissues from patients, respectively (72, 73). Accordingly, NEU3 knockdown reduced bladder cancer cell invasion in vitro suggesting a protumoral function (73). Hence, sialidases exhibit both anti- and prometastatic effects, with potential specificity to certain cancer types.

UDP-GlcNAc is the substrate for the enzyme O-linked N-acetylglucosaminyltransferase (OGT) which catalyzes O-GlcNAcylation of proteins. Moreover, UDP-GlcNAc is required for N-linked glycosylation by N-acetylglucosaminyltransferases (MGAT). Current evidence suggests that O-GlcNAcylation in the tumor stroma and cancer cells promotes early metastatic steps by changing cell–cell and cell–matrix interactions (Fig. 3C). Moreover, based on the type of N-glycosylation, observed pro- or antimetastatic phenotypes are promoted in cancer cells (Fig. 3C).

mRNA expression of OGT was positively correlated with metastasis status in patients with lung adenocarcinoma (74) and colon cancer (75), whereas increased O-GlcNAcylation levels were correlated with poor prognosis in patients with breast cancer (76). Moreover, the metabolic pathways leading to UDP-GlcNAc production increased during PDAC progression in the KPC mouse model (77). This effect was based on glutamine–fructose aminotransferase 1 (GFAT1) expression, which is the first step of UDP-GlcNAc production using glutamine and fructose-6-phosphate as substrates (77). Interestingly, 6-diazo-5-oxo-l-norleucine (DON), which inhibits all glutamine-catalyzing reactions, decreased UDP-GlcNAc synthesis and impaired PDAC xenograft growth and lymph node metastasis (78). Moreover, UDP-GlcNAc levels can also be affected by the hexosamine salvage pathway. This was shown to support primary tumor growth in PDAC (79); however, no link to metastasis has been established yet.

Mechanistically, it was found that increased O-GlcNAcyla­tion of cathepsin B, a lysosomal protease of the papain family from macrophages, promoted its secretion into the tumor microenvironment, leading to ECM degradation and increased melanoma-derived lung metastasis (80). In breast cancer, O-GlcNAcylation of the transcription factor Forkhead box protein A1 (FOXA1) was shown to shape its interactome, recruiting the transcriptional repressor methyl-CpG–binding protein 2 (MECP2) (76). This resulted in the suppression of adhesion-related genes, thereby promoting invasion in vitro (76). In lung adenocarcinoma patient samples, tumor-adjacent fibroblasts from the invasive edge displayed an altered glycosylation pattern compared with tumor core fibroblasts. Although the exact cross-talk is unclear, glycoproteomics revealed that glycosylation altered the cyclin-dependent kinase 4 (CDK4)—phosphorylated retinoblastoma protein (pRB) axis in the stroma, indirectly promoting an epithelial-to-mesenchymal transition (EMT) in lung cancer cells (81). Thus, there is evidence that O-GlcNAcylation is promoting metastasis formation.

The expression of the N-glycosyltransferase MGAT1, 2, and 4A was found to be increased in glioblastoma versus adjacent brain tissue (82). Moreover, silencing of MGAT1 decreased the proliferation and migration of glioblastoma cells through decreased N-glycosylation and expression of GLUT1 (82). In line, mutations in MGAT5 impaired N-glycan branching and migration of glioblastoma stem-like cells (83). In contrast, it was observed that MGAT3 protein or mRNA was downregulated in metastatic ovarian and/or breast cancers compared with primary tumors and nonmetastatic breast cancer tissues (84, 85). Moreover, MGAT3-mediated N-glycosylation of CD82 blocked integrin α₅β₁-mediated cellular adhesion, migration, and metastasis of ovarian cancers (84). In line, overexpression of MGAT3 suppressed cervical cancer cell migration and invasion and induced a more epithelial phenotype (86). A similar block of an EMT by MGAT3 activity was observed in OSCC cell lines (87). Thus, MGAT3-mediated N-glycosylation has antimetastatic effects, whereas the expression of the other MGAT enzymes leads to prometastatic N-glycosylation patterns on proteins.

In summary, the majority of studies underscore the critical role of glycosylation in the early steps of metastasis and underscore the importance of glycosylation for the cross-talk of cancer cells with the ECM and stromal cells. At the same time, fewer investigations have demonstrated its significance in the later stage of colonization (88, 89). Although the later stages of metastasis are clinically easier to target, several inhibitors have been developed, for example, against ST3GAL3 and ST3GAL6 (90) Moreover, not only the lack of protein glycosylation but also the accumulation of the required metabolite substrates such as UDP-glucuronic acid may be a therapeutic vulnerability of cancers (91). In addition, in studies regarding congenital disorder of glycosylation, it has been shown that providing nutrients that feed into the substrate production for glycosylation alleviates some symptoms in patients (92). Thus, studies on the role of diet in cancer cell glycosylation may provide novel therapeutic strategies.

Acetyl coenzyme A (acetyl-CoA) is involved in energy production and lipid synthesis, yet it can also serve as a substrate for the acetylation of histone and nonhistone proteins. There is ample evidence that acetylation promotes an EMT in cancer cells as reviewed by Kong and colleagues (93). Yet, also additional mechanisms exist in how acetylation, including histone acetylation and deacetylation, fosters metastasis formation within the same tumor type. Here, we summarize these additional mechanisms for the example of breast cancer (Fig. 4A).

Histone deacetylase (HDAC) increases the global acetylation of histones. The two FDA-approved pan-HDAC inhibitors SAHA and LBH589 were observed to promote breast cancer metastasis in a mouse model (94). In line, genetic and pharmacologic targeting of HDAC11 in mice increased breast cancer cell migration and distant metastasis although lymph node metastasis was reduced (95). Molecularly, HDAC4 inhibition enhanced H3K9 acetylation at the promoter of metastasis marker NEDD9 (neural precursor cell expressed developmentally downregulated 9), increasing its expression and consequently activating invasion via focal adhesion kinase (FAK) phosphorylation (94).

Also, non-histone proteins are frequently acetylated. Zinc finger MYND-type containing 8 (ZMYND8), which affects transcriptional regulation and chromatin remodeling, and the superoxide dismutase SOD2 were observed to be acetylated in breast cancer cells promoting migration and invasion as well as stemness, respectively (96, 97). Mutation of the p300 (histone acetyltransferase P300) acetylation site in ZMYND8 was sufficient to impair primary tumor growth and metastasis (97). Moreover, acetylation of the signal transduction protein SMAD family member 3 (SMAD3) by the acetyltransferase KAT6A promoted breast cancer cell stemness and metastasis formation (98, 99). In patients with breast cancer, KAT2A expression was elevated in metastatic lesions growing in palmitate-enriched organs, whose oxidation provides the acetylation substrate acetyl-CoA, compared with organs with low palmitate content (99). Mechanistically, palmitate oxidation to acetyl-CoA requiring carnitine palmitoyltransferase 1A (Cpt1a) activity promoted the acetylation of the NFκB protein p65 via KAT2A (99). Targeting CPT1A abrogated high-fat diet-induced lung metastasis growth in breast cancer mouse models (99). This is consistent with the observation that inhibiting palmitate uptake using a blocking antibody against CD36 impaired diet-induced metastasis in various mouse models (100, 101).

Notably, not only acetylation changes in cancer cells promote metastasis. In line, it was found that senescent lung fibroblasts that recruit immune cells during premetastatic niche formation show low expression of the enzyme acetyl-CoA carboxylase α (ACACA), which consumes acetyl-CoA and that overexpression of this enzyme reduced metastasis formation (102).

Although numerous studies underscore the importance of acetylation in promoting metastasis, the deacetylation of certain proteins can also be prometastatic in breast cancer. Acetylation of the methyltransferase METTL3 via acetyl-CoA acetyltransferase 1 (ACAT1) expression or downregulation of sirtuin (SIRT) 1 activity, impaired migration and invasion of breast cancer cells (103, 104). Yet, in tissue samples from patients with breast cancer, it was found that the deacetylases SIRT1 and SIRT3 decreased in invasive tumors and lymph node metastasis (compared with primary tumors), respectively (96, 105). Thus, these observations require further investigation to link the cell line data to clinical observations.

Currently, four histone deacetylase inhibitors are clinically approved, and others are passing through clinical trials (106). Yet, inhibitors blocking protein acetylation are less advanced with a drug targeting CREB-binding protein (CBP) and p300, which are highly homologous acetyltransferases, currently in clinical phase I in patients with metastatic prostate cancer (107). Despite these promising advancements in translation, data from preclinical models raise concerns regarding potential adverse effects, given the dual role of acetylation and deacetylation in cancer metastasis. However, a generally positive effect may be observed through dietary interventions or targeting the uptake of the acetyl-CoA precursor palmitate. Accordingly, clinical trials blocking CD36 are about to start (108). Notably, targeting palmitate uptake will have various effects beyond blocking acetylation as reviewed by (109).

Lipids play a crucial role as signaling molecules by posttranslationally modifying proteins and influencing their functions. One common protein modification is palmitoylation, which involves the covalent addition of a palmitate moiety. In humans 23 and in mice 24 palmitoyl acyl transferases (DHHCs) have been annotated, whereas 7 depalmitoylating enzymes have been identified. Strikingly, the gene-expression patterns of these enzymes are cancer-type dependent when analyzed in tumor samples from patients and may correlate with both overall and progression-free survival (110). The resulting changes in palmitoylation can influence the localization, interaction, and activity of the substrate proteins with implications in metastasis formation (Fig. 4B).

Glioblastomas are highly invasive cancers. DHHC5 and 9 expression has been associated with an EMT and colony-forming capacity in glioblastoma (111, 112). Mechanistically, DHHC5 promoted FAK palmitoylation sustaining its membrane localization needed for interaction with the ECM (111), whereas DHHC9 palmitoylated the glucose transporter GLUT1 which subsequently promoted glycolysis (112). In HCC, DHHC7 was important for STAT3 palmitoylation, and the DHHC7 inhibitor S-(2-acetamidoethyl) 2-bromohexadecanethioate (MY-D-4; ref. 113) impaired colony growth in HCC cell lines (114). In contrast, DHHC7-mediated palmitoylation of junctional adhesion molecule 3 (JAM3) increased tight junctions and impaired lung cancer cell migration (115). Cervical cancer cell migration, invasion, and EMT were dependent on a dynamic palmitoylation and depalmitoylation of flotillin 1 (FLOT1) by DHHC19 and LYPLA1, which resulted in sustained insulin-like growth factor-1 receptor (IGF-1R) tyrosine kinase activation (116, 117).

Furthermore, it was observed that palmitate levels regulate protein palmitoylation in colorectal cancer and prostate cancer cells (118, 119). Acyl-CoA oxidase 1 (ACOX1), the first enzyme of the peroxisomal fatty acid oxidation pathway, was found to be progressively downregulated in tissue samples from patients with colorectal cancer comparing normal tissue, primary tumors, and metastases (118). In cell lines, the loss of ACOX1 resulted in elevated palmitate levels which fostered β-catenin stability because the associated palmitoylation prevented degradation via ubiquitination (118). Similarly, the loss of fatty acid synthase (FASN) decreased Ras homolog family member U (RhoU) palmitoylation and cell migration in prostate cancer cells (119). Moreover, it was shown that palmitoylation of CD36 by DHHC4/5 led to monounsaturated fatty acid uptake balancing lipotoxicity in high-fat diet-induced breast cancer–derived lung metastasis (120, 121).

Palmitoylation is a widespread posttranslational modification that is only starting to be explored in cancer and metastasis formation. Besides the general inhibitor 2-bromopalmitate (2-BP), several more specific inhibitors have been developed for use in cell lines including inhibitors against DHHC3, 7, and 20 (113). Moreover, molecular docking studies predict favorable binding activity of the polo-like kinase 1 inhibitor BI-2536 to all DHHC enzymes except DHHC3, which could be further explored to develop selective inhibitors against the DHHC family members (110).

Lactylation is a recently discovered protein modification of lysine residues that is regulated by the lactate concentration in cells (122) and has implications for the early steps of metastasis formation (Fig. 4C).

A global lactylome analysis in HCC cells identified almost 10,000 lactylated proteins including many enzymes such as adenylate kinase 2 (AK2), which was inhibited by lactylation (123). In line, the lactylation of AK2, but not AK2 levels itself, was identified as an unfavorable prognostic marker in HCC patients (123). In cervical cancer cell lines, it was shown that lactylation increased the stability of discoidin, CUB, and LCCL domain containing 1 (DCBLD1), which in turn activated the pentose phosphate pathway resulting in elevated migration, invasion, and growth (124). In addition, nucleolar and spindle-associated protein 1 (NUSAP1) was observed to be lactylated promoting its protein stability, and elevated NUSAP1 levels were associated with a worse prognosis in PDAC patients (125). In line, NUSAP1 deletion decreased migration and metastasis of PDAC cells (125). Thus, lactylation emerges as an additional regulator of protein abundance in cancer cells. Moreover, the direct regulation of lactylation by lactate concentrations raises the question of whether the metastatic potential of tumor cells gained through the expression of the lactate transporter MCT1 (126) is also in part mediated through lactylation.

S-Adenosyl methionine (SAM) is the substrate for methylation reactions, and it is produced in the methionine cycle that in turn intersects with the folate cycle and NAD+ metabolism. Thereby, methionine and serine can both increase SAM availability, whereas nicotinamide N-methyltransferase (NNMT), which converts SAM and NAM (nicotinamide) to SAH (S-adenosyl-L-homocysteine) and 1NMA (N1-methylnicotinamide) decreases it. Whether elevated SAM levels and methylation promote or inhibit metastasis formation may be dependent on the tumor context, the biochemical alteration, and the specific methylation mark that is affected (Fig. 4D and E).

Tumoral and blood SAM levels were found to be correlated with metastatic reoccurrence in patients with colorectal cancer (127). It was also shown that efficiently metastasizing melanoma PDX models show elevated levels of methylation-related metabolites and H3K9 and H3K27 trimethylation levels (128), indicating a role of increased methylation in metastasis formation. Moreover, methionine restriction suppressed the invasiveness of triple-negative breast cancer and gastric cancer cells, which led to impaired lung metastasis in orthotopic triple-negative breast cancer (129) and a gastric cancer metastasis mouse model (130), respectively. This is in line with patient data showing that the levels of the methionine adenosyltransferase 2A (MAT2A), which catalyzes the synthesis of SAM from methionine and ATP, correlated with higher-grade tumors and metastases in patients with lung cancer (131). Moreover, palmitate and a palm-oil-rich diet increased—through an unknown mechanism—methyltransferase Set1A–mediated H3K4 trimethylation and metastasis initiation in oral cancers and melanoma (100). Thus, some diets may have metastasis-promoting effects by changing protein methylation.

Most of the evidence that methylation and elevated SAM levels are antimetastatic stems from the study of NNMT, whose expression decreases SAM levels. A meta-analysis including over 2,500 patients with cancer found that high NNMT expression was correlated with worse tumor differentiation, earlier lymph node, and distant metastasis (132). Accordingly, NNMT expression was observed to be high in cancer stem cells of different tumor types and indicated poor outcomes in patients with breast cancer and patients with triple-negative breast cancer undergoing chemotherapy (133–135). NNMT loss inhibited metastasis formation in breast cancer mouse models (134, 135). Mechanistically, the loss of NNMT promoted histone H3K9 trimethylation (H3K9me3) and DNA methylation at multiple promoters leading to decreased expression of collagen and ECM genes in breast cancer cells (135). Moreover, the methylation of protein phosphatase 2 (PP2A), which enhances activity, increased upon NNMT silencing in triple-negative breast cancer cells leading to decreased phosphorylation and activity of the MEK (mitogen-activated protein kinase kinase)-ERK (extracellular signal-regulated kinase) pathway blocking a downstream EMT (134). Metastases in intrahepatic cholangiocarcinoma mouse models were promoted by NNMT activity because of decreased H3K27 and H3K9 trimethylation leading to elevated EGFR expression (136). Thus, NNMT is emerging as an interesting drug target, and several inhibitors for the use in cell lines have been developed (137–140) although currently no clinical trials have been conducted yet.

Many enzymes have besides their canonical catalytic functions in metabolism also moonlighting functions of catalytic or noncatalytic nature. These moonlighting functions have been extensively studied in primary cancers as reviewed by (141), yet their role in metastasis formation is only emerging with most of them suggesting roles in the early steps of the metastatic cascade (Fig. 5).

Metastasis-related noncanonical functions of glycolysis-related enzymes have been mainly linked to protein phosphorylation in the nucleus resulting in the activation of signaling pathways leading to cancer cell migration and invasion (Fig. 5A).

Fructose is besides glucose a major carbohydrate in the Western diet. Dietary and polyol pathway (glucose to fructose conversion)–derived fructose is converted by ketohexokinase (KHK) to fructose-1-phosphate. In breast and gastric cancer cells, it was found that nuclear KHK-A suppressed the mRNA expression of the repressive EMT marker CDH1 (E-cadherin) by phosphorylating YWHAH (tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein eta), which recruits snail family transcriptional repressor 2 (SLUG) to the promotor of CDH1 (142, 143). In mice, Khk-A promoted fructose-induced breast cancer metastasis to the lung and a phospho-deficient mutant of the Khk-A site in Ywhah (S25A) abolished the metastasis-promoting effect of fructose (142).

6-Phosphofructo-2-kinase/fructose-2,6-biphosphatase 4 (Pfkfb4) interconverts fructose-6-phosphate and fructose-2,6-phosphate, which is an allosteric activator of the glycolysis enzyme phosphofructokinase (PFK). Two studies in breast cancer showed that PFKFB4 also acted in the nucleus by phosphorylating nuclear receptor coactivator 3 (Src-3) which activated collagen remodeling and purine biosynthesis (144, 145). In lung adenocarcinoma cells, PFKFB4 could phosphorylate SRC-2 (146). In breast cancer mouse models, Pfkfb4 silencing decreased systemic metastasis but, dependent on the mouse model, not primary tumor growth. A phospho-deficient mutant of the PFKFB4 site in SRC-3 (S857A) did not induce lung metastasis (144). In addition, PFKFB4 expression was higher in lung, liver, and brain metastases compared with primary tumors of triple-negative breast cancer patients (145), and PFKFB4 expression and a PFKFB4-SRC-3 proteomics signature correlated with poor prognosis in basal and triple-negative breast cancer whereby patients with intense nuclear staining showed a worse prognosis (144, 145).

Pyruvate kinase M2 (PKM2) converts phosphoenolpyruvate (PEP) to pyruvate. The functional role of PKM2 in phosphorylating proteins is debated (147). Yet, two studies suggest that PKM2 has a nuclear function in HCC and colorectal cancer metastasis (148, 149). Thereby, a PKM2 mutant that cannot enter the nucleus (R399/400A) but can still form a tetramer does not induce metastasis (149). Molecularly, it was suggested that PKM2 phosphorylates STAT3 inducing EMT needed for migration and invasion in HCC cells (148).

Metastasis-initiating capacity in cancer cells can be promoted by protein–protein interactions of enzymes. In most described cases, an increase in this interaction promoted metastasis but also the opposite, namely, the loss of an interaction, has been linked to metastasis formation (Fig. 5B).

Several glycolytic enzymes have noncatalytic functions in early metastasis formation. Hexokinase 2 (HK2), the enzyme converting glucose to glucose-6-phosphate (G6P), was recently shown to have a noncatalytic function in promoting breast cancer metastasis in two mouse models (150). Mechanistically, it was observed that HK2 interacts with glycogen synthase kinase 3 (GSK3) and the regulatory subunit of protein kinase A (R1A), which enabled PKA to phosphorylate GSK3. This led to the degradation of GSK3 and subsequent snail family transcriptional repressor 1 (SNAIL) expression promoting an EMT in cancer cells (150). Further, it was shown that HK2 interacts with proteins of chromatin organization in the nucleus increasing the transcription of stemness genes and DNA repair in AML cells (151). Meanwhile, HK2 interaction with AIMP2 (proapoptotic protein aminoacyl tRNA synthetase complex interacting multifunctional protein 2) enhanced its degradation in HCC cells (152). These additional noncatalytic functions of HK2 may potentially foster metastasis-initiating capacities in cancer cells by increasing stemness and decreasing cell death. In line, it was observed that elevated HK2 expression was associated with decreased reoccurrence-free survival in HCC patients (152).

Fructose-1,6-bisphosphate (FBP) is converted to dihydroxyacetone phosphate (DHAP) and glyceraldehyde-3-phosphate (G3P) by aldolase (ALDO), whereas triosephosphate isomerase 1 (TPI1) interconverts DHAP and G3P. In HCC cells, it was seen that ALDOA interacted with insulin-like growth factor 2 mRNA-binding protein 1 (IGF2BP1) enhancing eukaryotic translation initiation factor 4G (eIF4G) translation and subsequently global mRNA translation (153). A catalytically inactive mutant of ALDO and TPI1 induced, like the wildtypes, colony formation and migration in HCC and lung cancer cells, respectively (153, 154). ALDOA and TPI1 expression were associated with poor outcomes in HCC and lung adenocarcinoma patients, respectively, suggesting a role of the noncatalytic functions of these enzymes in metastasis initiation (153, 154).

3-Phosphoglycerate (3PG) is converted to 2-phosphoglycerate (2PG) which then is converted to PEP by phosphoglycerate mutase 1 (PGAM1) and enolase (ENO), respectively. It was found that PGAM1 in exosomes from prostate cancer cells was taken up by endothelial cells where it induced angiogenesis, migration, and proliferation through an interaction with actin gamma 1 (ACTG1; ref. 155). Catalytically inactive ENO1 promoted, like wild-type, lung cancer cell invasion presumably through an interaction of ENO1 with hepatocyte growth factor receptor (HGFR) which can promote an EMT via decreased GSK3 activity (156). Moreover, treatment of mice with PGAM1-containing exosomes promoted prostate cancer cell–derived metastasis formation compared with treatment with PGAM1-deficient exosomes. Moreover, patients with metastatic prostate cancer displayed more PGAM1 expression and endothelial cells compared with nonmetastatic patients (155).

Finally, it was found that in lung cancer cells, autocrine secreted, and in breast cancer cells, macrophage secreted, PKM2 binds to integrin β₁, consequently triggering an EMT through SRC signaling (157, 158). In line, breast and lung cancer cells treated with recombinant PKM2 and challenged with a wound closure assay or intravenously injected into mice resulted in more migratory cells and enhanced lung and liver metastasis (157, 158).

The serine biosynthesis pathway branches off glycolysis converting 3PG to serine in a three-step enzymatic reaction by PHGDH, phosphoserine aminotransferase 1 (PSAT1), and phosphoserine phosphatase (PSPH). The acquisition of metastatic initiation capability in a triple-negative breast cancer mouse model and lung cancer cells was driven by the loss of a noncatalytic function of PHGDH (51) or an increase in the noncatalytic function of PSAT1 (159), respectively. These alterations induced migration and invasion, underlining their pivotal roles in promoting metastasis. Mechanistically, the loss of PHGDH disrupted a protein–protein interaction with the glycolytic enzyme Pfk, which decreased its activity redirecting carbon flux to CMP-sialic acid driving sialylation of integrin α_vβ₃ in triple-negative breast cancer cells (51). In line, overexpression of a catalytically inactive mutant of PHGDH decreased metastasis to the same extent as the wild-type in Phgdh knockdown cells. In lung cancer cells PSAT1 was found to interact with IQ motif containing GTPase activating protein 1 (IQGAP1), which recruited STAT3 and increased its phosphorylation and function in inducing MMP9, C–X–C motif chemokine ligand 8 (CXCL8), and serum amyloid A1 (SAA1) expression (159). Interestingly, heterogeneous and/or low expression of PHGDH correlated with decreased metastasis-free survival in patients with triple-negative breast cancer indicating a role not only for the level of expression but also for heterogeneous expression within a tumor (51).

Nicotinamide adenine dinucleotide kinase (NADK) phosphorylates NAD⁺ to NADP⁺, and its catalytic function has been shown to promote metastasis formation in breast cancer (160). Also, a noncatalytic function of NADK in lymph node metastasis was observed in an NSCLC mouse model (161). Thereby, NADK interacted with the ubiquitination regulating SMAD-specific E3 ubiquitin protein ligase 1 (SMURF1). Subsequently, the ubiquitination and degradation of bone morphogenetic protein receptor type 1A (BMPR1A) by SMURF1 was impaired leading to BMP signaling-dependent migration and invasion. In patients with NSCLC, NADK expression correlated with lymph node metastasis and poor prognosis (161).

Moonlighting functions of enzymes are more and more frequently detected. However, their targeting may be more difficult especially when catalytic and noncatalytic functions or metabolic and nuclear activities have opposite effects on metastasizing cancer cells. For example, targeting only the catalytic function but not protein stability would be important in the case of leveraging PHGDH as a drug target against metastasis (51, 162).

Metastasis formation is the leading cause of death in patients with cancer and innovative approaches to prevent and target metastases are urgently needed. Functions of metabolites and enzymes outside their canonical role in the metabolic network are emerging as vulnerabilities of metastasizing cancer cells. In this respect, most of our knowledge is currently limited to the early metastatic steps where cancer cells gain metastasis-initiating capacities through increased motility and ECM remodeling capabilities leading to invasiveness. Much less is currently known about metabolic signaling in metastases outgrowth and persistence in distant organs.

Whether and how intercepting the early steps of metastases is beneficial in clinical practice depends on the tumor type (Fig. 6). In glioblastoma, a neoadjuvant treatment reducing the invasiveness of the primary tumor may lead to patient benefit because clear resection margins can be easier achieved, and less tissue will need to be resected from the brain to remove the tumor potentially sparing cognitive functions. In breast cancer, metabolic gene and protein signatures indicating metastasis-initiating capacity (51) and thus potential early dissemination of cancer cells from the primary tumor could be further developed to improve the prediction of patients with high metastasis risk. This may benefit the patients by receiving follow-up treatments and checkups according to their predicted metastasis risk. Moreover, also in pancreatic cancer, it was shown that neoadjuvant treatment before surgery extends median survival by about 1.8-fold (163). Recently, two studies in preclinical mouse models of PDAC were published showing that targeting multiple enzymes that use glutamine as a substrate, rather than only targeting glutaminase, decreased primary tumor growth, metastasis incidence, and metastasis size (164, 165). This increased efficacy may be due to the inhibition of noncanonical functions of glutamine metabolism because an earlier study suggested that the beneficial effect of using an inhibitor that targets multiple glutamine-using enzymes in PDAC mouse models is mainly driven by its inhibitory effect on glycosylation (77).

Currently, most patients will benefit from treatments that can target already disseminated cancer cells growing in distant organs (Fig. 6). An arising opportunity may be to target the metabolic signaling that is induced in metastases growing in palmitate-rich organs (99). Yet, more knowledge is needed on the canonical and particular noncanonical functions of metabolites and metabolic enzymes in distant metastases to find innovative treatment strategies. Moreover, a higher acceptance of therapeutic strategies only targeting metastases and their formation is needed at all levels of translation ranging from drug-developing companies to regulatory bodies to enable nonclassical approaches to tackle metastases. A first step in this direction was the 2018 FDA acceptance of two clini­cal trials with metastasis-free survival as the primary clinical endpoint (166).

S.-M. Fendt reports grants from Gilead, other support from Auron Tx, Blackbelt Tx, and Alesta Tx outside the submitted work.

We apologize to all colleagues whose excellent work we could not cite due to text and reference limits. S. Krieg is supported by an EMBO Postdoctoral Fellowship (ALTF 164-2023). S.I. Fernandes acknowledges funding from Stichting tegen Kanker. C. Kolliopoulos acknowledges funding from the Swedish Research Council (VR, Vetenskapsrådet). M. Liu acknowledges funding from the China Scholarship Council (CSC). S.-M. Fendt acknowledges funding from FWO and iBOF (INTERCEPt) Projects, the Beug Foundation, Fonds Baillet Latour, KU Leuven, Francqui Stichting, Foundation ARC, Wereld Kanker Onderzoek Fonds (WKOF) as part of the World Cancer Research Fund International grant programs and Stichting tegen Kanker.