Involvement of redox signalling in tumour cell dormancy and metastasis

During the initial phase of the metastasis cascade, namely tumour cell invasion and dissemination, a fraction of cancer cells requires undergoing morphological changes that allow them to invade the surrounding environment [119] (Fig. 1). These changes are thought to occur via a biological process which implies a loss of epithelial traits towards a gain of mesenchymal traits, named EMT [22]. Importantly, it has been described that EMT in tumour cells leads to distinct intermediary states, which represent a biological continuum between the two absolute phenotypes, epithelial and mesenchymal. These intermediary states are interconvertible and/or reversible given the epigenetic nature of the molecular mechanism that modulates the phenotypic transitions [120]. This genetic flexibility may be behind the higher adaptability to different microenvironments observed in DTCs. ROS are involved in several biological processes in cancer cells as cellular secondary messengers, playing an important role in EMT. Through different redox alterations, ROS can modify the biological function of proteins, which are sensitive to redox status of the cell. In addition to induce molecular and functional changes in proteins that are involved in the remodelling of the ECM, other proteins implicated in cytoskeleton remodelling, the establishment of cellular junctions and cell motility, may undergo variations which have a profound impact on metastatic invasion and dissemination [60, 121]. For instance, it has been observed that the knockdown of NADP-dependent malic enzyme (ME1), a major source of reducing equivalents, significantly decreased NADPH production and generated ROS, which decreased migration and invasion as well as altered EMT biomarkers expression. These data suggest that the promotion of EMT mediated by ME1 is driven in an ROS-dependent manner [121]. Furthermore, ME1 knockdown in human colon and lung cancer cells repressed cell growth under conditions of glucose starvation and induced senescence. In other words, ME1 knockdown diminished NADPH and induced high levels of ROS and apoptosis when subjected to stress conditions, for example, glucose starvation and anoikis [59].

Anoikis is a biological process defined as a caspase-dependent cell death mechanism which is induced by loss of integrin-mediated attachment to the ECM [122]. ROS modulation during anoikis has been studied through different experimental approaches. During matrix detachment, it has been observed a deficiency in glucose uptake that reduce pentose phosphate pathway (PPP) flux, diminishing NADPH and subsequently increasing ROS intracellular levels. However, the expression of the oncogene receptor tyrosine-protein kinase erbB-2 (ERBB2) has been shown to restore defective glucose uptake, rescuing PPP flux through the increase of glucose 6-phosphate. These changes lead to the generation of NADPH which fortified the antioxidant capacity of the cell [123]. In addition, cellular detachment from the ECM activated AMP-activated protein kinase (AMPK), which in turn limited NADPH consumption in fatty acid synthesis (FAS) and increased NADPH levels, through the production of fatty acid oxidation-induced NADPH. Therefore, tumour cells are capable to buffer the increase of intracellular ROS through the modulation of lipid metabolism and the maintenance of the intracellular levels of reducing equivalents, such as NADPH [124]. Subsequent studies have corroborated the direct effect of ECM detachment on AMPK activation. Interestingly, [125] found that AMPK was rapidly phosphorylated and upregulated in several cancer cell lines following matrix deprivation. Further investigations revealed that this early AMPK activation upon ECM detachment was triggered by an intracellular increase of ROS and Ca²⁺ levels and operated through the AMPK upstream activators serine/threonine-protein kinase STK11 (STK11) and calcium/calmodulin-dependent protein kinase kinase 2 (CaMKK2). Therefore, a Ca²⁺-ROS intracellular increase triggers the activation of the STK11/CaMKK-AMPK signalling cascade which promotes cell survival following ECM detachment. On the other hand, the activity of other antioxidant enzymes, such as catalase copper zinc superoxide dismutase (CuZnSOD) and MnSOD, in mitigating intracellular ROS levels to increase ECM-detached cell survival have been explored. Catalase and MnSOD, when antagonized, compromise the survival of detached mammary epithelial cells in a specific manner [126]. These observations were supported by another study indicating that overexpression and activity of MnSOD facilitate the survival of ECM-detached breast cancer cells and metastasis. Furthermore, the same study found a positive correlation between histologic tumour grade and expression of MnSOD in human breast cancer samples [106].

Additionally, the role of mitochondria and amino acid metabolism in tumour cell invasion and dissemination has been addressed through different experimental approaches. Lung carcinoma cells detached from the ECM-used isocitrate dehydrogenase (NADP) cytoplasmic (IDH1) to decarboxylate glutamine to citrate, which enters the mitochondria and participates in oxidative metabolism to provide energy to the cells. Concomitantly, but in the mitochondrial compartment, isocitrate dehydrogenase (NADP), mitochondrial (IDH2), synthesized NADPH, neutralizing mitochondrial ROS and stimulating cell survival [127]. Proline metabolism has also been shown to impact metastasis development. Proline has been found to support mitochondrial ATP production via proline dehydrogenase 1, mitochondrial (Prodh), which produces FADH2, a redox cofactor, and provides electrons to the electron transport chain, balancing redox homeostasis and enhancing metastasis formation [128].

As mentioned before, there are other redox cellular changes that can lead to cell death due to ECM-detachment. A novel type of programmed cell death, known as ferroptosis, has been related to cell death induction in ECM-detached cells [129]. Ferroptosis is characterized by the accumulation of lipid hydroperoxides to lethal levels and in an iron-dependent manner [130]. From a biochemical perspective, lipid peroxides and other intracellular ROS such as superoxide (O2⁻), H₂O₂, and peroxyl radicals (ROO^•) are generated through a cascade of reactions which involve different forms of ROS as intermediaries. O2⁻ is generated extracellularly or intracellularly by NADPH oxidase or the mitochondrial electron transport chain, respectively. In the mitochondria, O2⁻ produces ferrous iron (Fe²⁺) through its release from iron-sulfur (FE-S) or the reduction of ferric iron (Fe³⁺). O2⁻ can also be dismutated to H₂O₂ by superoxide dismutases in the mitochondria. H₂O₂ penetrate through membranes reacting with proteins and DNA, ultimately being detoxified by cellular peroxidases. ROO^•, which originated from both the decomposition of ONOO⁻ and/or the reaction of H₂O₂ with Fe²⁺, initiates the lipid peroxidation cascade. ROO^• forms lipid radicals (L^•) by reacting with lipids, which subsequently generate lipid peroxyl radicals through its reaction with oxygen. In a propagation reaction, lipid peroxyl radicals form lipid peroxides by reacting with polyunsaturated fatty acids (PUFAs) [131].

Sensitivity of the cells to ferroptosis has been associated with several biological mechanisms, such as the metabolism of amino acids, iron, and PUFAs, as well as glutathione, phospholipids, NADPH, and coenzyme Q10 biosynthesis. It has been shown that the induction of ferroptosis is driven by erastin, which is a small molecule that selectively inhibits the cystine/glutamate transport receptor (Xc⁻ system). Cell treatment with erastin inhibits the import of cystine and leads to the depletion of glutathione, finally causing the deactivation of the phospholipid hydroperoxide glutathione peroxidase (PHGPx). PHGPx acts converting lipid peroxides that might be toxic, into non-toxic lipid alcohols. Therefore, suppression of lipid peroxidation have been described to prevent ferroptosis in mammalian cells [130]. Furthermore, lipid peroxidation induced by ROS have been considered to play an important role in different types of cell death, based on chemical and structural alterations of the cell membranes, secondary to lipid peroxidation chain reactions [132]. In this line, the toxic products generated from lipid peroxidation have been shown to trigger apoptosis through distinct pathways, such as NF-kB [133], MAPK [134], and protein kinase C (PKC) [135], as well as autophagy via AMPK/mammalian target of rapamycin complex (mTORC) [136] and c-Jun-amino-terminal kinase (JNK)-apoptosis regulator Bcl-2 (Bcl-2)/Beclin-1 (BECN1) [134, 137]. Another study revealed the important protective role of Nrf-2 against lipid peroxidation and, consequently, ferroptosis. Nrf-2 is a key regulator of the cellular antioxidant response and controls the expression of genes that code proteins which catalyze reactions leading to oxidative and electrophilic stress neutralization [138]. In particular, the expression of the Xc⁻ system and PHGPx, which play a prominent role in ferroptosis inhibition, is under the control of Nrf-2.

Taken together, these studies exemplify a well-known fact in tumour biology, which is that the net effect of intracellular ROS can either promote tumour cell endurance through the activation of pro-survival signalling pathways or induce cell death, if the intracellular level increases beyond a lethal level that cannot be quenched by the several antioxidant mechanisms. Therefore, in order to survive, tumour cells might reach a situation of equilibrium through which they favor the intracellular raise of ROS, at sub-lethal levels, to activate signalling pathways that promote several biological processes involved in tumour progression. At the same time, the intracellular raise of ROS might be maintained at non-toxic levels through the action of the antioxidant system. This non-physiological redox balance might confer survival advantages to tumour cells and be at the base of the metastatic behavior [139, 140].

In determining cell fate when entering the secondary sites, the re-adhesion of tumour cells to the ECM at the metastatic niche is known to be rate-limiting step [143]. The biophysical properties of the ECM have been related to tumour cell antioxidant homeostasis. A recent study [144] has shown that breast cancer cells cultured on a soft ECM increase the production of mitochondrial ROS (mtROS) due to changes in mitochondrial dynamics, driven by an increase in peri-mitochondrial F-actin, which was found to control dynamin-1-like protein (DRP1) activity and mitochondrial fission. These changes conferred enhanced tolerance to oxidative stress through Nrf-2 antioxidant transcriptional response to finally provide breast tumour cells with a survival advantage. Interestingly, the activation of the F-actin/DRP1/Nrf-2 axis was shown to be upregulated in breast tumour dormancy models where inhibition of DRP1 and Nrf-2 prevented the dormant-to-proliferative switch, showing the deep impact of mitochondrial dynamics and redox signalling in DTC fate and metastasis.

Not only the biophysical properties of potential secondary sites have a profound impact on DTCs fate, but biochemical and metabolism-derived products released to the tissue stroma determine the adoption of different survival strategies from DTCs. In this line, an enrichment of pyruvate in the lung stroma has been shown to enforce the adaptation of breast DTCs through the conversion of pyruvate to α-ketoglutarate catalyzed by mitochondrial alanine aminotransferase 2 (ALT2). The raise in the extracellular concentration of α-ketoglutarate potentiated mTORC1 signalling and supporting the remodelling and deposition of collagen [145], a known inducer of the dormant-to-proliferative switch and metastasis formation [146]. Indeed, metabolic rewiring has been shown to discern between micro- and macrometastasis. Using patient-derived-xenograft (PDX) models of breast cancer, Davis et al. [147] showed that the expression of genes involved in mitochondrial oxidative phosphorylation (OXPHOS) was upregulated in micrometastasis as compared to primary tumour cells, where expression of glycolytic enzymes were highest. The authors corroborated these findings through metabolomic data and pharmacologic inhibition of OXPHOS, which further demonstrated its functional importance in the context of metastasis since OXPHOS blockade significantly reduced metastatic burden in the lungs of their PDX model. Finally, Viale et al. [148] showed that oncogene ablation, using an inducible model of KRAS-mutant pancreatic ductal adenocarcinoma (PDAC), selected a subpopulation of dormant PDAC cells which displayed stem cell features and were highly dependent on OXPHOS for survival. In addition, transcriptomic and metabolic analyses performed in surviving cells revealed major expression of genes regulating autophagy and mitochondrial and lysosome activity. Considering that mitochondrial OXPHOS is a major source of intracellular ROS [149] and that OXPHOS may be a commonly activated pathway in quiescent or dormant cancer cells, the raise in intracellular ROS and its consequences in dormant DTC signalling and survival warrants further investigation.

Hypoxia, defined as a biophysical condition of the cellular microenvironment by which a tissue is not able to provide sufficient oxygen to sustain homeostasis [150], has been linked to DTC fate, as it is known to significantly impact tumour progression, resistance to the treatment, and metastasis. Indeed, hypoxia is a poor-prognosis feature of solid tumours [151]. Hypoxia signalling has been shown to regulate diverse steps of the metastatic cascade: EMT, invasion, intravasation, survival of CTCs, extravasation, establishment of the premetastatic niche, and proliferation from micrometastasis to macrometastasis [152]. Under low oxygen, HIF is stabilized though the inhibition of its degradation by the von Hippel-Lindau disease tumour suppressor (pVHL), consequently being translocated to the nucleus where it promotes the expression of genes that lead to the adaptation to the low levels of oxygen [153]. Fluegen et al. [154] showed that a hypoxic microenvironment induce the reversible upregulation of hypoxia (glucose transporter 1 (GLUT-1), HIF1α) and dormancy (COUP transcription factor 1 (COUP-TF1), differentiated embryonic chondrocyte gene (DEC2), p27) genes in primary breast and HNSCC. Remarkably, post-hypoxic DTCs were characterized by a dormant phenotype and high expression of COUP-TF1, DEC2, p27, and TGFβ2. Upon hypoxia, COUP-TF1 and HIF1α were shown to induce the expression of p27 and quiescence in tumour cells. However, dormant DTCs did not show a high-level expression of other hypoxia markers, such as GLUT-1. These data suggested that the preservation of the dormant phenotype in post-hypoxic DTCs was independent of GLUT-1 expression or that additional hypoxia-responsive pathways may contribute to the maintenance of dormancy in DTCs. In addition, following hypoxia exposure, T-Hep3 cells in the lung significantly overexpressed TGFβ2 as compared to proliferating metastases, where its signalling was silenced. In this line, the authors suggested that post-hypoxic DTCs home to TGFβ2 expressing metastatic niches that may contribute to dormancy induction. Indeed, Bragado et al. [155] showed that TGFβ2 signalling in the bone marrow activates MAPK p38α/β, lowering the ERK/p38 ratio, and thus, inducing dormancy in T-Hep3 DTCs in vivo. The same signalling axis has been shown to induce tumour dormancy in a model of prostate cancer [156].

In line with these findings, Erler et al. [157, 158] found that Lysine Oxidase (LOX), a protein of the ECM, is controlled by hypoxia and HIF. Under hypoxic conditions, breast cancer cells in the primary tumour secrete LOX, which accumulates in the premetastatic niche and modifies the secondary site ECM. LOX crosslinks collagen IV at the basement membrane, which is essential for the recruitment of CD11b + myeloid cells and secretion of MMP2. In turn, CD11b + myeloid cell-derived MMP2 cleaves collagen at the pre-metastatic niche, which promotes the invasion of bone marrow-derived cells (BMDCs) and metastatic cells. Indeed, LOX inhibition significantly decreased metastatic growth. In addition, hypoxia has also been shown to induce a quiescent state in a model of patient-derived primary lung cancer harboring activating EGFR mutation [159]. Mechanistically, hypoxia upregulated the expression of mitogen-inducible gene 6 protein (MIG-6) a negative regulator of ERBB signalling, which prevented heterodimer formation of ERBB family receptor tyrosine kinases (RTKs) and downstream signalling inducing tumour cell quiescence and resistance to EGFR tyrosine kinase inhibitor (TKI) treatment. Interestingly, when MIG-6 was downregulated in mutant EGFR lung cancer cells under hypoxic conditions, EGFR signalling was restored, promoting the phosphorylation of ERK and AKT through increased EGFR-HER3 binding, as well as their sensitivity to EGFR-TKI and radiation. In accordance with this results, analyses of tumour samples from lung cancer patients with EGFR mutations revealed a significant inverse correlation between MIG-6 expression their survival rates after the treatment with EGFR-TKI [159].

Autophagy is an evolutionarily conserved physiological mechanism involved in the maintenance of cell homeostasis under metabolic stress [164]. This physiological mechanism can be hijacked by tumour cells to survive the numerous stresses encountered along the metastatic cascade. In this regard, Vera-Ramirez et al. [58] reported that breast DTCs activate autophagy via BECN1-independent non-canonical pathway to survive upon the establishment of a cell dormancy program both in vitro and in vivo. Pharmacological inhibition (through administration of different autophagy inhibitors, such as bafilomycin, 3-methyladenine, and hydroxychloroquine) or genetic (through knockdown of autophagy related 7 (ATG7) but not BECN1) means dramatically impaired the survival of dormant breast cancer cells. Mechanistically, autophagy inhibition resulted in the intracellular accumulation of depolarized mitochondria and toxic oxidative by-products which drove apoptotic cell death. Interestingly, quenching of mitochondria-derived ROS rescued cell viability, further showing the crucial importance of ROS management in the survival of dormant DTCs. Additionally, autophagy has been shown to be activated under hypoxic conditions and decrease intestinal inflammation through the downregulation of the mTOR/NOD-like receptor, pyrin domain-containing 3 (NLRP3) pathway [165]. Moreover, whereas the inhibition of the PI3K/AKT/mTOR and MAPK signalling pathways resulted on autophagic cell death [166], lower levels of PI3K in association with microenvironmental factors were correlated with the promotion of autophagy-induced dormancy in ovarian cancer [167]. These data highlight the hormetic and environment-dependent net effect of the activation of the molecular networks related to redox signalling and autophagy. On the other hand, oxidation of ATG4, ATG7, and ATG3 have also been found to decrease autophagy. A study revealed that ATG4B activity was reversibly modulated by the formation of intramolecular disulfide bonds in response to oxidative stress. Modifications were described in Cys292 and Cys361 residues of the ATG4B protein [168]. Furthermore, inactive ATG3 and ATG7 are covalently bond to microtubule-associated protein 1A/1B-light chain 3 (LC3) and protected from oxidation. When activated, LC3 is transferred to phosphatidylehanolamine, the covalent interaction is lost, and the catalytic thiols of both ATG3 and ATG7 are exposed to inhibitory oxidation, preventing LC3 transfer and autophagy [169].

The accumulation of unfolded or misfolded proteins in the ER lumen is a salient feature of specialized secretory cells leading to a cellular condition known as endoplasmic reticulum (ER) stress. ER stress is counteracted by the activation of the unfolded protein response (UPR), which is a homeostatic signalling network that either promote the recovery of ER function or potentiates apoptosis, depending on the duration and intensity of the stress stimuli. Chronic UPR has been shown to be involved in the pathogenesis of a wide variety of human diseases, from diabetes to cancer, highlighting the role of this highly interconnected signalling network as a stress “dimmer” in the cell, beyond ER stress and protein folding [170‐172]. Indeed, some types of cellular stress and extracellular stress signals, such as hypoxia, activate p38 MAPK which can in turn inhibit the expression of Forkhead Box protein M1 (FoxM1), Proto-Oncogene C-Jun (c-Jun), and the uPAR transcripts, supressing the activation of ERK and triggering downstream UPR signalling [173]. An analysis revealed the upregulation of ER-stress response genes in HEp3 cells under p38 regulation: binding immunoglobulin protein (BiP, also known as GRP-78), protein disulfide-isomerase (PDI)/protein disulfide-isomerase A3 (PDIA3, also known as ER60), serpin H1 (HSP47), and cyclophilin B [174]. The data showed that under normal conditions, BiP is bound to domains located in the ER-luminal kinase receptors protein kinase R-like endoplasmic reticulum kinase (PERK), serine/threonine-protein kinase/endoribonuclease IRE1 (IRE1α), and activating transcription factor 6 (ATF6), among others. However, in the presence of misfolded proteins, BiP dissociated from the protein complexes and translocated to the ER lumen activating PERK, IRE1α, and ATF6. When active, PERK phosphorylated eukaryotic translation initiation factor 2A (eIF-2A) which inhibited eIF-2A-dependent protein synthesis and induced ATF4 to upregulate the expression of genes associated with protein synthesis and antioxidant response. The activation of this molecular cascade promoted cell survival and growth arrest. PERK has also been considered to induce the activation of the Nrf-2 transcription factor, to inactivate C/EBP-homologous protein (CHOP) and block apoptosis [175]. Activated IRE1α promoted the spliced form of X-box-binding protein 1 (XBP-1) (XBP-1 s), which is a transcription factor that binds to chaperones and endoplasmic-reticulum-associated protein degradation (ERAD) gene promoters to modify or degrade misfolded proteins. IRE1α can stimulate Grp78/BiP to block apoptosis regulator BAX (Bax)-elicited proapoptotic signals. Last, ATF6 has also been shown to promote cell survival in HEp3 cells through the activation of mTOR signalling, controlled by GTP-binding protein Rheb (Rheb) [176‐178].

Therefore, the UPR-mediated transcriptional induction of genes that increase protein folding capacity and antioxidant machinery is an important mechanism to re-establish cellular homeostasis and to alleviate protein folding stress [176, 177]. In addition, protein post-translational modifications, which are key for appropriate folding, are frequently highly sensitive to the redox status of the cell. The formation of disulfide bridges is catalyzed by protein disulfide isomerases and other oxidoreductases which, in turn, contribute to ROS generation and redox unbalance in the ER [172, 178]. Evidence suggests that ROS production and oxidative stress are not only coincidental to ER stress. Rather, ROS are an integral part of the UPR signalling network, supporting either proapoptotic or proadaptive UPR signalling [170‐172].

Regarding to the onset of proliferation following quiescence, several lines of evidence support the role of the microenvironment in triggering the dormant-to-proliferative switch of DTCs in the metastatic niche, as reviewed in previous sections of this manuscript. On the contrary, our knowledge about the intracellular molecular mechanisms contributing to the dormant-to-proliferative switch is much more limited, although redox signalling is thought to play an important role in the reactivation of dormant DTCs. For instance, Nrf-2 has been shown to be upregulated upon downregulation of ERBB2 in breast cancer cells and secondary to an increase in intracellular ROS due to altered cellular metabolism. It has also been reported that Nrf-2 was activated during dormancy and its signalling has been found to induce the re-establishment of redox homeostasis in recurrent tumours and upregulation of de novo nucleotide metabolism, ultimately promoting the reactivation of dormant breast cancer cells [179].

It has also been suggested that lipid metabolism is relevant to this step of the metastatic cascade. Pascual et al. [180] showed that a sub-population of metastasis-initiating oral carcinoma cells (defined by overexpression of CD44) expressed high levels of the fatty acid receptor CD36 and lipid metabolism genes. Supplementation with palmitic acids or a high-fat diet fuelled tumour proliferation and metastatic outgrowth in vitro and in vivo, respectively. Blocking CD36 dramatically inhibited the formation of metastasis in vivo. These data showed that metastatic cells rely on lipid metabolism in early steps of the metastatic colonization and growth.

Remarkably, a series of experiments using organoid cultures and mouse models revealed that tumour recurrent cells showed increased ROS intracellular levels and modified lipid metabolism. When the synthesis or intramitochondrial transport of fatty acids was inhibited, the levels of cellular ROS and DNA injury decreased, promoting tumour cell proliferation. Moreover, a significant reduction on the rate of in vivo recurrence was achieved through NAC-dependent ROS scavenging and proliferation inhibition through treatment with progesterone antagonist [181]. Interestingly, Carracedo et al. [182] unveiled the role of promyelocytic leukemia (PML) gene as an inhibitor of peroxisome peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC-1-α) acetylation and as a potent activator of peroxisome proliferator-activated receptors (PPAR) signalling and fatty acid oxidation in breast cancer. PML expression promoted tumour cell invasion in vitro and was correlated with poor prognosis in breast cancer patients. Therefore, fatty acid oxidation and fatty acid metabolism have been shown recurrently to critically modulate metastasis development. Lipid metabolism is intimately ligated to redox signalling, both as a ROS source through their catabolism [183] and as a ROS intracellular scavenger when stored as cytoplasmatic lipid droplets (LDs) [184]. In the context of metastasis, the antioxidant role of intracellular lipids stored in the form of LDs has been shown to protect DTCs against paclitaxel-induced ROS production and promote breast cancer metastasis. Mechanistically, metastasis competent cells overexpressed FoxM1, whose transcriptional activity stimulated the expression of phospholipase D1 (PLD1), which in turn, promoted LD accumulation [168]. These innovative approaches and data underscoring the antioxidant and protective role of lipids, beyond their traditional metabolic and signalling roles, open new research avenues in our understanding of dormant DTC biology and metastatic relapse.

As mentioned before, the tissue microenvironment is known to play a pivotal role in cancer metastasis. A widespread characteristic of tumours is their inability to develop adequate blood vessels with the consequence being relative hypoxia. Hypoxic conditions stimulate blood vessel development, whereby the blood flow in these new vessels is often chaotic, causing oxidative stress [77]. Consequently, within a growing tumour mass, cancer cells repeatedly face cycles of hypoxia and reoxygenation [85]. H₂O₂ secreted by cancer cells has been proposed to mimic the effects of hypoxia under aerobic conditions in adjacent fibroblasts and other stromal cells, resulting in production of ROS, which has been shown to stimulate angiogenesis. For instance, it has been shown that the administration of Mn(III) ortho-tetrakis-N-ethylpyridylporphyrin, a potent scavenger of ROS and RNS, attenuates angiogenesis in vivo, modifies the density of microvessels, and decreases the proliferation rate of endothelial cells in a mouse model of breast cancer [185].

Moreover, glucose deficiency is one of the main factors leading to oxidative stress in tumours, either in the stage of tumour initiation or progression [186]. In relation to metastasis, it is important to note that detaching from ECM leads to ATP deficiency owing to a reduction of glucose transport [123], which would lead to a strong induction of ROS production. VEGF expression can be regulated by nutrient deprivation and hypoxia, which both increase intracellular levels of ROS [187]. Increased generation of H₂O₂ also led to accumulation of HIF-1α [188]. MnSOD suppresses the induction of HIF-1α in human breast carcinoma cells [188]. Likewise, suppression of endogenous ROS by mitochondrial inhibitors or GPx decreased HIF-1 induction and VEGF expression in cancer cells [189]. Moreover, it has been reported in three different tumour mouse models that antioxidants, as NAC and vitamin C, exert their antitumoural effect mainly acting on HIF-1α level [190]. These results suggest that both, superoxide and H₂O₂, may contribute to HIF-1α accumulation. At moderate levels, ROS induce transcription of HIF1A gene and stabilize the encoded protein by inhibiting the activity of the iron-dependent prolyl 4-hydroxylase (4-PH) as a consequence of catalytic Fe²⁺ oxidation of Prolyl hydroxylase domain-containing protein 2 (PHD2) [191]. This mechanism seemed involved in physiological stabilization of HIF under mild (1–3% O₂), but not deep, hypoxia, a condition reportedly accompanied by mitochondrial production of ROS [192, 193]. This is probably mediated by an increase in basal levels of H₂O₂ and superoxide, due to decreased expression of several antioxidant enzymes such as peroxiredoxins (Prxs) and CuZnSOD [194]. Moreover, S-nitrosylation of HIF at Cys-800 has been shown to increase factor stability and transcriptional activity, likely by promoting HIF binding to p300 [195]. Interestingly, PH inhibition by ROS may contribute to normoxic accumulation of HIF in cancer cells [196, 197].

As a consequence of hypoxia-induced HIF-1α activation, HIF-1 and its cofactor p300 induced the expression of pro-angiogenic genes [198], such as VEGF and VEGF receptors [199, 200]. HIF-1 stabilization is important for cancer metastasis due to several reasons. Hypoxia, through HIF activation, drives cancer cell invasion and metastatic progression in various cancer types. In epithelial tumours, hypoxia induces the transition to amoeboid cancer cell dissemination. Accordingly, hypoxia triggers EMT in several human tumour cells through the generation of mitochondrial ROS which correlated with elevated intracellular levels of HIF [201]. Invasion and metastasis in cells where metastasis was promoted by the generation of mitochondrial ROS as consequence of mitochondrial DNA mutations have been also associated with elevated intracellular levels of HIF, suggesting a direct link between ROS and the pro-invasive program triggered by this transcription factor [93].

Moreover, HIF-1 regulates glycolysis-related genes and inhibits mitochondrial respiration resulting in metabolic adaption of tumour cells to hypoxia [202, 203]. Likewise, HIF-1 prevents intracellular acidification, which leads to an increased formation of lactate and CO₂ [202], both compounds favoring ECM degradation and cell invasion [204]. In addition, HIF-1α-induced promotion of metastasis and EMT in osteosarcoma cells has been shown to depend on a reduction of intracellular ROS production. In osteosarcoma (OS) cells under hypoxic conditions, HIF-1α induces the overexpression of the mitochondrial NADH dehydrogenase (ubiquinone) 1 alpha subcomplex 4-like 2 (NDUFA4L2) protein, which in turn promoted OS cell migration, invasion, proliferation, and EMT. Drug-and genetically-induced inhibition of the NDUFA4L2 protein negatively impacted OS cell survival, which could be in part compensated through a reduction in intracellular ROS levels through autophagy activation [205].

Along with oxygen availability, interstitial fluid hydrostatic pressure that generally increases with the rapid growth of malignant tumours also has been shown to influence HIF activity. Higher protein levels of HIF‑1α have been found in Lewis lung cancer (LLC) cells that were exposed to 50 mmHg hydrostatic pressure for 24 h. This condition was associated with upregulation of numerous metastasis‑promoting genes (hepatocyte growth factor (HGF), cadherin 11 (CDH11), and ephrin type-B receptor 2 (EPHB2)) and the downregulation of metastasis suppressing genes (KiSS-1 metastasis suppressor (KISS1), spleen associated tyrosine kinase (SYK), and HIV-1 Tat interactive protein 2 (HTATIP2)). Thus, elevated interstitial fluid hydrostatic pressure in malignant tumours may promote the onset of the metastatic cascade by stabilizing HIF‑1α [206]. Moreover, it was found that the exposure of the LLC cells to high hydrostatic pressure, prior to intravenous injection into healthy mice, resulted in higher cell survival/retention in the lungs 24 h later and resulted in more metastatic tumour lesions 4 weeks later [206].

In the previous model, elevated HIF‑1α levels also correlated with the levels of the antioxidant enzymes, CuZnSOD and MnSOD, improving tolerance to oxidative stress in LLC cells exposed to high hydrostatic pressure. These data suggested that metastasis promotion in cancer cells exposed to high hydrostatic pressure would be driven by oxidative damage prevention through HIF transcriptional activity [206]. In addition, under hypoxic conditions, HIF-dependent upregulation of the one-carbon metabolism enzyme serine hydroxymethyltransferase, mitochondrial (SHMT2), promotes mitochondrial serine catabolism and NADPH production [207], to raise the antioxidant capacity and prevent ROS-mediated tumour cell death [207].

The Rac1 GTPase has important functions in ROS-mediated actin reorganization of migrating tumour cells [208, 209]. Intracellular ROS increases are involved in MMP-3–induced EMT in murine breast epithelial cancer cell lines which, concomitantly, correlated with increased expression of a Rac1b splicing variant. It has been suggested that the treatment with MMP-3 stimulates the expression of Rac1b that would be responsible for increasing ROS levels [73]. Importantly, a cell shape change is required for EMT induction by MMP3 and Rac1b [210], in keeping with the notion that Rac-dependent ROS transduce mechanical perturbations into a pro-invasive gene expression program [76]. Secretion and activation of MMP-2 has been reported to be dependent on Rac1 and ROS generation in pancreatic carcinoma PANC-1 cells exposed to EGF. Moreover, it has been suggested that signalling events downstream of EGF receptor involved PI3K- and Src-dependent activation of Rac1 [211]. ROS are in turn responsible for the activation of the SNAI1 factor, a key transcriptional inducer of the EMT program [212‐214]. In addition, synergistic signalling between growth factors and integrins leads to an intracellular oxidative burst through a Rac1-dependent increase in mitochondrial ROS production [215, 216]. In this sense, Rac1-mediated ROS generation was involved in prometastatic signalling through c-Met, one of its upstream regulators, in a murine model of invasive melanoma [217]. The redox cascade triggered by overexpression of c-Met involved generation of H₂O₂ via CuZnSOD-mediated dismutation of superoxide. Importantly, tumour cell capacity to form experimental lung metastases in vivo was inhibited after blocking the abovementioned cascade [217].

Some studies also have demonstrated the involvement of Rac signalling in cytoskeletal rearrangement and in mediating integrin signalling. ROS generated by Rac1-induced NOX have been shown to activate the cofilin pathway and, thus, to contribute to increased cell migration [90, 91]. Conversely, a density-dependent decrease of NOX-Rac1 activity and ROS production in response to receptor tyrosine kinases engagement mediates growth inhibition by cell–cell contact in fibroblasts [218]. Interestingly, it has been reported that mesenchymal and amoeboid motility styles are interconvertible through the reciprocal regulation of Rac and Rho GTPases in melanoma cells [219]. Moreover, activation of Rac and subsequent generation of ROS leads to NF-κB activation and MMP-1 production in response to integrin-mediated cell shape changes [216]. Rac1-mediated changes in cellular ROS levels also increase the migratory potential of MCF-7 and T47D breast cancer cells probably through NF-κB [220]. Lastly, increased activity of Rac1 in primary endothelial cells has been shown to mediate the loss of cell–cell adhesions and loosens the integrity of the endothelium, which allows the intravasation of cancer cells [221].

The transcription factor NF-κB is considered as a redox-sensitive transcription factor. ROS can activate NF-κB through several mechanisms and, at the same time, ROS production is also regulated by NF-κB. Several NF-κB target genes are involved in the detoxification of ROS but some of them, such as NOS, have been shown to exert a pro-oxidant function [278]. These interactions highlight a complex interplay between ROS and NF-κB. By regulating gene expression, NF-kB can promote metastasis. NF-κB can modulate EMT by inducing the expression of MMPs and key cellular adhesion molecules [85, 253, 254, 279, 280]. NF-κB upregulates the expression of MMPs, urokinase-type plasminogen activator, and cytokines in highly metastatic breast cancer cell lines. Rac-induced in response to integrin-mediated cell shape changes leads to NF-κB activation and MMP-1 production [216]. NF-κB was also suggested to be involved in the increase in the migratory potential of the MCF-7 and T47D breast cancer cells mediated by Rac [220]. In addition, ROS regulate the expression of interleukin-8 (IL-8) and the cell surface protein intercellular adhesion molecule 1 (ICAM-1) through NF-κB. Both ICAM-1 and IL-8 can regulate the trans-endothelial migration of immune cells [281, 282]. On the other hand, VEGF and its receptors are known to be regulated by NF-κB, in the promotion of angiogenesis [283, 284]. In a recent study, it was demonstrated that Kras-derived mitochondrial ROS activated NF-κB through polycystin-1 (PC1) to upregulate EGFR pro-proliferative signalling [198]. Overexpression of the inhibitor of nuclear factor kappa-B kinase (IKK) in breast cancer cells with constitutive NF-κB activity resulted in reduced expression of the receptor CXCR4 in vitro and the corresponding loss of migration mediated by its ligand, the SDF-1α. Introduction of CXCR4 cDNA into IκB-expressing cells restored SDF-1α-mediated migration. Thus, NF-κB regulates the motility of breast cancer cells by directly upregulating the expression of CXCR4, a receptor of the SDF-1α. NF-κB subunits p65 and p50 bind directly to CXCR4 promoter and activate its transcription. Cell surface expression of CXCR4 and the SDF-1α-mediated migration were enhanced in breast cancer cells isolated from mammary fat pad xenografts compared with parental cells grown in culture. A further increase in CXCR4 cell surface expression and SDF-1α-mediated migration was observed with cancer cells that metastasized to the lungs [285].

There are numerous interactions, links, and cooperativities between the NF-κB pathway and other signalling pathways. More recently, crosstalk of NF-κB with another transcription factor involved in certain types of cancer, that is, the transcriptional regulator ERG (ERG), has been identified. Interestingly, an increased NF-κB activity was detected in ERG fusion-positive prostate cancer patients and cell lines. It was shown that increased NF-κB activity is associated with phosphorylation of p65 on Ser536 involving signalling through Toll-like receptor 4 (TLR4) [286]. ERG also appears to stimulate the SDF-1/CXCR4 axis, which contributes to metastasis [287]. Furthermore, the redox-sensitive transcription factors NF-κB and FOXO3a have been described as regulators of MMP expression [85, 253, 254].

Nrf-2 can be activated by cigarette smoke, infection, oxidative stress, or inflammation. High amount of ROS activates tyrosine kinases to dissociate Nrf-2:Keap1 complex allowing the nuclear import of Nrf-2 [288]. Moreover, the modifications of cysteine residues in Keap1 apparently alter the interaction of Keap1 with Nrf-2 and lead to its relocation to the cytoplasm, where it is subsequently degraded by the ubiquitin proteasome [289]. Therefore, under physiological conditions, Keap1 and Nrf-2 act as a cellular sensor of damage caused by free oxygen radicals, through the constant shuttling of Keap1 between the nucleus and the cytoplasm [290]. Consequently, certain ROS levels can induce the expression of multiple antioxidants and cytoprotective genes via Nrf-2 transcriptional activity.

Because it regulates a wide spectrum of antioxidants and detoxification genes, Nrf-2 is the main inducible defense against oxidative stress [291, 292] but, whether it is more readily activated by ROS than other redox-responsive transcription factors, is unclear. Nevertheless, assuming that Nrf-2-induced gene expression provides an initial means to adapt to oxidative stress, it may offer a type of “floodgate” protection in which only once the antioxidant protection elicited by the proteins whose expression is induced by Nrf-2 is overwhelmed (and the ROS concentration threshold that induces Nrf-2 transcriptional activity can vary for different tissues and experimental systems), additional antioxidant transcription factors within the network are activated. A modification of the floodgate model would include induction by Nrf-2 of Krueppel-like factor 9 (KLF9), which is a DNA-binding transcriptional regulator that downregulates the antioxidant genes TXNRD2 and PRDX6 [293, 294]. Therefore, induction of KLF9 would shut down antioxidant defenses when ROS levels exceed a certain threshold or duration. In this scenario, other members of the network would only be activated when the antioxidant capacity of Nrf-2 target genes is exceeded or when KLF9 is induced. Once Nrf-2-directed floodgate defenses have been breached, the question of whether individual antioxidant transcription factors are activated in a stratified or coordinated manner is uncertain. In this context, it should be noted that Nrf-2 regulates the expression of HSF-1 [295] whose increased levels have been reported to be associated with metastasis [296‐298] and promote infiltration, metastasis and neoangigoenesis in different cancer cell lines [299, 300]. In addition, Nrf-2 activation also led to overexpression of the NF-κB p65/RelA subunit, which antagonizes Nrf-2 by inhibiting the recruitment of CBP and activating histone deacetylase 3 (HD3) [301] and that, when oxidative stress is sufficient to cause DNA damage, ensues the activation of TP53 which in turn antagonizes Nrf-2 activity, thereby heightening oxidative stress intracellular levels and facilitating apoptosis [302]; together, these findings suggest that Nrf-2 could be downregulated by oxidative stress to an extent to which it induces inflammation and pro-apoptotic signalling.

In animal models and breast cancer patients with poor prognosis, it has been shown that Nrf-2 is activated during dormancy and in recurrent tumours. Constitutive activation of Nrf-2 accelerates recurrence, while suppression of Nrf-2 impairs it [179]. On the other hand, knockdown of Nrf-2 greatly impairs migration and invasion of a variety of cell lines including malignant and non-malignant cells [303, 304]. In a study comparing normal and tumour tissues of colorectal cancer patients, Nrf-2 and Keap1 protein levels and their tumour to normal tissue ratios were correlated with the lymph node/distant metastasis status. The authors found that the Keap1 tumour to normal tissue ratios was predictive of lympho-vascular invasion, which in turn was a significant predictor of metastasis in these patients [305].

Impairment of Nrf-2/antioxidant responsive element (ARE) pathway leads to oxidative stress, inflammation, and mitochondrial dysfunction [306]. Nrf-2 has been traditionally considered as tumour suppressor because of its cytoprotective functions against oxidative stress. However, hyperactivation of the Nrf-2 pathway creates an environment that favors the survival of normal as well as malignant cells, protecting them against oxidative stress, chemotherapeutic agents, and radiotherapy and providing a survival advantage that might favor tumour progression [307‐310]. In addition, Nrf-2 also controls free Fe²⁺ homeostasis since it upregulates the expression of HO-1, which generates free Fe²⁺ via the breakdown of heme molecules. Since Fe²⁺ induces a Fenton reaction to produce the free radical ^•OH from hydrogen peroxide, the upregulation of HO-1 leads to a paradox. In addition, Nrf-2 boosts the expression of the genes encoding several components of the ferritin complex that detoxifies Fe²⁺ by converting it to Fe³⁺ and storing it [311]. Notably, high serum concentrations of ferritin have been described in several cancers with a poor prognosis [312, 313]. As Nrf-2 induces HO-1, the enzyme catabolizing heme, its accumulation in lung cancer tissue causes the stabilization of transcription regulator protein BACH1 (BACH1), a pro-metastatic transcription factor [314‐316] whose degradation is triggered by heme [317]. BACH1 pro-metastatic effects are a consequence of its interaction with the ubiquitin ligase Fbxo22 [318]. Human metastatic lung cancer displays high levels of HO-1 and BACH1 supporting the pro-metastatic role of both proteins. Moreover, BACH1 transcriptional signature is associated with poor survival and metastasis in lung cancer patients [318].

Despite Nrf-2 inhibits EMT in non-transformed cell lines [212, 319, 320], overexpression of Nrf-2 in cancer cells can enhance metastasis through the process of EMT. Namely, expression of Nrf-2 has been reported to be important for the migration of normal and malignant cells since it is needed for MMP upregulation. In this sense, it has been reported that Nrf-2 downregulation correlates with reduced expression or gelatinase activity of MMP2 and MMP9 [303, 321, 322]. Moreover, it has also been reported that Nrf-2 promotes EMT by downregulation of CDH1 expression in cancer cell lines [323, 324]. Likewise, Nrf-2 silencing reduces CDH2 expression, a process which was proposed to be mediated by the downregulation of the Nrf-2 target gene NOTCH1 [321, 325]. Inhibition of ROS production suppressed MCUR1-induced EMT of HCC cells which would also depend on Nrf-2 and Notch1 activity [326]. Thus, ROS may act as second messengers for Nrf-2 activation leading to EMT promotion. However, the mechanism by which Nrf-2 regulates these enzymes and proteins needs to be clarified. ROS/Nrf-2/Notch1 pathway was also activated by mitochondrial Ca²⁺ [326], as discussed before in the “6.3.3” section. Interestingly, cancer cells that exhibit constitutively high levels of Nrf-2 can grow in an anchorage-independent manner and have a higher metastatic capacity [327]. Cell detachment generates ROS [328] and activates autophagy [264], which could induce Nrf-2-dependent gene expression. A possibility is that this anchorage-independent growth might be regulated by Nrf-2-dependent induction of secreted phosphoprotein 1 (SPP-1, also known as osteopontin) [329]. In cells that lose ECM attachment by adduction of integrins with the toxic metabolite methylglyoxal, Nrf-2 activation induces the expression of glyoxalase I (Glx I), which metabolizes methylglyoxal and prevents anoikis [330, 331]. Moreover, in recurrent tumours, Nrf-2 signalling induces a transcriptional metabolic reprogramming to re-establish redox homeostasis and upregulate de novo nucleotide synthesis. The Nrf-2-driven metabolic state renders recurrent tumour cells sensitive to glutaminase inhibition, which prevents reactivation of dormant tumour cells in vitro [179]. In addition, mechanistic studies showed that Nrf-2 binds to the promoter region of steroid hormone receptor (ERR1) and may function as a silencer in breast cancer cells. This may enhance RhoA protein stability and lead to RhoA overexpression that promotes migration and metastasis by the RhoA/Rho-associated protein kinase (ROCK) pathway [332]. Likewise, Nrf-2 silencing suppressed stress fiber and focal adhesion formation leading to decreased cell migration and invasion of breast cancer cells by downregulating RhoA [332]. Moreover, restoration of RhoA in MCF7 and MDA-MB-231 cells induced Nrf-2 knockdown-suppressed cell growth and metastasis in vitro, and Nrf-2 silencing suppressed stress fiber and focal adhesion formation leading to decreased cell migration and invasion [332].

On the other hand, Nrf-2 expression in the metastatic microenvironment has also been reported to exert an antimetastatic role. In xenograft mouse models of metastasis, whole-body and myeloid-specific Nrf-2 deletion increased susceptibility to lung metastases [333, 334]. This effect could be attributed to a persistent inflammation featured by a high number of myeloid-derived suppressor cells (MDSCs) [333] combined with redox alterations in immune cells. Redox alterations were especially relevant in MDSCs since they suppressed CD8 + T cell proliferation, depending on their intracellular ROS levels [335, 336]. Intracellular ROS levels were higher in MDSCs from Nrf-2-deficient mice as compared to those from wild-type mice [333]. In addition, MDSCs derived from Nrf-2-deficient mice produced RNS and ROS that prevented CD8 + T cell antigen recognition, a tolerance mechanism known as anergy [335, 336]. Therefore, Nrf2 deletion would lead to a metastasis-conducive microenvironment in this xenograft model of lung cancer. In contrast, in Keap1 knockdown mice (Keap1-kd or Keap1f/f) or in wild-type mice treated with the Nrf-2 inducer bardoxolone, the number of lung metastases were reduced [291, 337, 338]. Nrf-2 activation in Keap1f/f mice limits metastasis, probably due partly to decreased ROS levels in MDSCs [333]. Consistently, deletion of Nrf2 or Trsp, the gene that encodes for selenocysteine tRNA necessary for the translation of the selenocysteine-containing antioxidant proteins GPx and Thioredoxin Reductase 1 (TR1), in the myeloid lineage confirmed that the anti-metastatic activity of Nrf-2 relates to its regulation of ROS in MDSCs [334]. Furthermore, Nrf-2-dependent downregulation of IL-6 could also prevent recruitment of myeloid precursor cells to tumours [339, 340] independently of redox regulation.

Additional proteins acting upstream Nrf-2 also have been related to metastasis. These include carbonyl reductase (NADPH) 1 (CBR1), hepatitis B virus X-interacting protein (HBXIP), nestin, and karyopherin subunit alpha-6 (KPNA6). CBR1 is another important enzyme that regulates the expression of Nrf-2 during oxidative stress and helps to detoxify ROS [341]. The levels of HBXIP that can compete with Nrf-2 for binding with Keap1 protein, via its highly conserved GLNLG motif, are positively correlated with Nrf-2 expression in breast cancer cells and clinical breast cancer tissue samples [342]. KPNA6 is a protein which facilitates nuclear import and attenuates Nrf-2 signalling, clearance of Nrf-2 protein from the nucleus, and restoration of the Nrf-2 protein to basal levels [343, 344]. Knockdown of nestin, a protein that binds to Keap1, resulted in downregulation of Nrf-2 and repressed the in vivo development of gastric cancer metastasis arising from the injection of SGC-7901 and MKN-45 cell lines in mice. The restoration of Nrf-2 expression, or treatment with the Nrf-2 activator sulforaphane, counteracted the inhibitory effect of nestin knockdown on the proliferation, migration, invasion, and antioxidant enzyme production [345].

Wnt signalling modulates major developmental processes and is a dominant mediator of stem cell self-renewal, cell fate, and cancer stem cell biology. β-catenin resulted upregulated in HNSCC after inhibition of human CBR1, whose expression is lower in HNSCC patients with lymph node metastasis compared to those without lymph node metastasis. In addition, CBR1 inhibition was shown to increase intracellular levels of ROS. Consistently, CBR1 inhibition resulted in increased in vitro invasion ability of several HNSCC cell lines and in the activation of several EMT markers, such as vimentin, CDH1, or the zinc finger protein SNAI2 (SNAI2) [346], indicating that the Wnt signalling pathway could have a pro-metastatic role. On the other hand, β-catenin expression was suppressed after induction of glutathione S-transferase omega-2 (GSTO-2), an enzyme that exhibits thioltransferase activity, in a lung squamous cell carcinoma model. The overexpression of GSTO-2 correlated with a lower oxygen consumption rate and mitochondrial membrane potential in the lung squamous carcinoma cells. Moreover, GSTO-2-overexpressing cells formed smaller tumours and the incidence of liver metastasis was lower as compared to control cells in a subcutaneous xenograft lung cancer model using nude mice. Interestingly, when cells transfected with GSTO2 were treated with a p38 inhibitor, β-catenin expression and mitochondrial membrane potential were recovered, suggesting that GSTO-2 loss contributes to lung cancer progression, modulating tumour cells metabolism via the p38/β-catenin signalling pathway [347]. In line with these observations, pharmacological inactivation of GSK-3β, a well-known negative regulator of the Wnt/β-catenin signalling pathway, promoted the migratory activity of 4T1 murine breast cancer cells, which correlated with higher levels of ROS and functional abnormalities in the mitochondrial respiratory chain complex I/III. In addition, NOX3 and NOX4 expression were upregulated in the 4T1 cells, which would further affect the generation of ROS. Furthermore, the authors found that the expression of pSer535-eIF-2B promoted the expression of NKG2-D type II integral membrane protein (NKG2-D) ligands, Mult-1 and Rae1, followed by phosphorylation of Ser9 in GSK-3β and increased generation of ROS [255]. In the same line, the treatment with SSTC3, novel small-molecule activator of CK1α that inhibits the Wnt signalling pathway, has been proved to inhibit the growth of patient-derived metastatic colorectal cancer xenografts in mice [348, 349]. Activation of Wnt/β-catenin signalling also correlated with the activation of EMT in cervical cancer cells in vitro [350]. Surprisingly, GPx2 was highly expressed in cervical cancer tissues compared to normal individuals and it was showed to reduce apoptotic damage by reducing hydroperoxides and promoting proliferation and metastasis [350]. On the other hand, the modulation of the Wnt/β-catenin pathway in neighboring cells, in particular immune cells, has been relevant to study its pro-metastatic role. Indeed, pharmacological inactivation of GSK-3β downregulated NKG2-D ligands H60a and Rae1 in NK cells and suppressed their cytotoxicity [255].

Importantly, several interactions with other signalling pathways have been put of manifest. For instance, PP2A has been shown to both positively and negatively regulate Wnt pathway at multiple levels [351]. Additionally, AKT could also activate β-catenin. Through a bioinformatics analysis and Western blot assays, Ma et al. [352] showed that Zi Shen Decoction (ZSD) decreased the enzymatic activity of PI3K and AKT in vivo and in vitro. The authors also found that the AKT/GSK-3β/β-catenin pathway mediated anticancer effect of ZSD in lung cancer cells. Indeed, they showed that oral administration of ZSD suppressed the LLC growth in a subcutaneous allograft model and promoted necrosis and inflammatory cell infiltration in the tumour tissues. Pharmacological attenuation of p-GSK-3β formation by inhibiting the PI3K/AKT pathway, reversed the abovementioned effects [255]. Moreover, EGFR has also been suggested to form a complex with β-catenin contributing to invasion and the promotion of metastasis [353, 354]. Since many of the abovementioned pathways are in part related to cell redox state, Wnt/beta-catenin signalling could indirectly be conditioned by or modulate it.

Aberrant activation of Sonic Hedgehog (SHH) signalling pathway by mutations within its components drives the growth of several malignant tumours [355, 356]. Different SHH ligands have been evidenced to have some effect on the metastatic potential of lung cancer cells. In this sense, it has been reported that the canonical SHH signalling pathway is activated in lung stroma by SHH ligands secreted from transformed lung epithelia [357]. Likewise, SHH signalling dysregulation has been reported to contribute to metastasis and angiogenesis in colon cancer [358], and termination of SHH signalling through glioma-associated oncogene (GLI1) inhibition resulted in the inhibition of proliferation in colon cancer [359, 360]. Moreover, smoothened homolog (SMO) and GLI1 were highly expressed in triple negative breast cancer, and their increased expression was correlated with metastasis, poor prognosis, and recurrence of triple negative breast cancer [361]. In addition, a GLI1 isoform, called truncated GLI1, has also been reported to induce the activation of genes related to proliferation, migration, and angiogenesis of breast cancer [362].

Similarly, inhibiting SHH pathway by interfering with the interaction between the already mentioned ligands and their receptors has been shown to have an antimetastatic potential. Early abrogation of the SHH pathway using an anti-SHH/Indian hedgehog protein (IHH) antibody 5E1 against Shh, the primary Hh ligand expressed in the lung of Kras^G12D/+;Trp53^fl/fl autochthonous murine lung adenocarcinoma, led to significantly worse survival with increased tumour and metastatic burden, while genetic deletion of Shh has no effect on survival [357]. Likewise, loss of IHH, another SHH ligand, by in vivo CRISPR led to more aggressive tumour growth suggesting that IHH, rather than SHH, activates the pathway in stroma to drive its tumour-suppressive effects—a novel role for IHH in the lung. Interestingly, the authors found that genetically engineered mice treated with the 5E1 antibody against SHH/IHH presented decreased blood vessel density and increased DNA damage, all of which indicated a raise in ROS and subsequent tissue damage. Indeed, Kras^G12D/+;Trp53^fl/fl mice treated with the 5E1 antibody and the antioxidant NAC showed inhibited tumour growth and prolonged mouse survival, suggesting that IHH may supress tumour growth through Hh signalling pathway by quenching or detoxifying ROS through a yet to explore mechanism [357].

Hh signalling has also been shown to promote tumour metastasis by being actively involved in EMT. SHH exerts its effects on EMT via the upregulation of the SNAI1 and downregulation of CDH1 [363, 364]. Namely, ectopic expression of GLI led to increased invasiveness of pancreas cancer cell lines [365]. Likewise, metastasis was induced by overexpression of GLI1 in the rarely metastasizing clone of prostate cancer AT2.1 [364]. The role of SHH pathway in EMT was supported by the reduction in the invasive properties of pancreatic cancer cells, following SNAI1 expression downregulation [365]. Moreover, in breast cancer, SHH signalling has been shown to induce angiogenesis independently of VEGF activation [366].