Mechanisms of action of NME metastasis suppressors – a family affair

Two original discoveries established NME1 as a metastasis suppressor. First, up-regulation of NME1 in several metastatic tumor cell lines from multiple histological types, including melanoma, breast, colon, lung, liver, ovary, prostate, and oral carcinoma cell lines, reduced their metastatic potential both in experimental and spontaneous mouse models of metastasis [34‐42]. Second, the incidence of lung metastases increased significantly in NME1 knockout mice prone to develop hepatocellular carcinoma [43]. In addition, numerous studies analyzing human patients with melanoma or epithelial tumors of the breast, liver, colon or ovary, found that loss of NME1 expression correlated with a greater risk of metastasis and a poorer clinical prognosis [10, 37, 44‐49]. By contrast, in neuroblastoma and in hematological malignancies, NME1 overexpression correlated with metastasis dissemination and unfavorable outcome [50‐52]. These conflicting reports of both positive and negative correlation of NME1 expression with metastasis depending on the tumor type, are highly suggestive of context-specific mechanisms.

Overexpression of NME1 in invasive tumor cell lines that have low levels of endogenous NME1 expression reduces the migratory and invasive potential of the cells. This is observed in cell lines derived from melanoma and from breast, colon, lung, liver, ovarian, prostate and oral carcinoma [42, 53‐59]. By contrast, silencing NME1 in non-invasive tumor cell lines that express high NME1 levels, results in a migratory and invasive phenotype in melanoma, glioblastoma, and carcinoma cells [60‐64]. Silencing NME1 in epithelial tumor cells results in an intermediate phenotype with both epithelial and mesenchymal features, refered to as a ‘partial’ EMT [65]. Indeed, NME1-depleted cells express both epithelial and mesenchymal markers, indicating that they do not transition fully to a mesenchymal state. Nevertheless, this ‘partial’ EMT phenotype is more metastatic than the fully mesenchymal state [66, 67], providing an explanation for the high metastatic potential of NME1-deficient tumor cells.

In human cancer samples, there is an inverse correlation between NME1 levels and the expression of well-known EMT markers [65]. Consistent with this, NME1 is downregulated or absent at the invasive front of clinical samples of hepatocellular carcinoma and colon cancers whereas it is robustly present in the center of the tumors [60]. In addition, the ability of NME1 to influence tumor invasion in vivo has been demonstrated in a xenograft model in which human breast tumor cells were injected into the primary duct of the mammary gland in immunodeficient mice [68, 69]. In this model, tumor cells form intraductal tumors that recapitulate the transition from carcinoma in situ to invasive breast carcinoma [68, 70]. Gene-editing inactivation of NME1 in the injected cells accelerated the development of invasive lesions [61].

The anti-metastatic activity of NME2 has been demonstrated mainly in breast, lung and oral squamous carcinoma, and in melanoma [15, 40, 71‐73]. Moreover, analysis of gene expression datasets from breast, colon, lung and ovarian tumors revealed that there is significantly less NME2 expression in metastatic tumors than in non-metastatic tumors [74].

NME2 has been reported to be an inhibitor of cell motility and invasion [18, 55, 57, 75, 76], suggesting that it might have a similar function to NME1 during tumor progression as an inhibitor of metastasis. In contrast to these earlier studies, however, our work rather suggests that NME2 is not a suppressor of EMT and tumor invasion [60, 61, 65, 77]. Indeed, loss of NME2, unlike NME1, does not impair the invasive capacity of tumor cells or the transition from in situ to invasive breast carcinoma in the intraductal xenograft model [61]. Importantly, inactivation of either NME1 or NME2 expression did not alter the protein level of the other isoform [61], suggesting that there is no compensatory effect when one isoform is deficient.

NME1 and NME2 isoforms are 88% identical in their amino acid sequence, however, sixteen out of eighteen non-identical amino acids are located at the peripheral surface of the hexamer [78]. Thus, NME1 and NME2 homohexamers may interact with different partners and may have distinct cellular functions. Indeed, there is evidence that NME1 and NME2 have more specific partners than partners in common [78]. Furthermore, the relative abundance of NME1 and NME2 proteins and the stoichiometry of NME1–NME2 heterohexamers may provide additional specific regulation and function.

In addition to specific binding partners and regulation of function of NME1 and NME2, the two proteins may function at distinct times in tumorigenesis. We hypothesise that that NME1 and NME2 function as NDPKs at early stages of tumorigenesis, providing nucleoside triphosphates to sustain the high proliferation of tumor cells. At later stages, however, NME1 might become the only isoform controlling EMT, invasion and metastasis. Consistent with this hypothesis, cells induced to proliferate have high levels of NME1 [60, 79‐83] and cells overexpressing NME2 proliferate faster than control cells in culture [73]. Accordingly, NME1 and NME2 are overexpressed early in the primary tumor and only NME1 expression is reduced or repressed at later stages in colorectal and hepatic tumors that have undergone metastasis [43, 60, 84]. Additionally, in breast cancer, we found a similar up-regulation of NME1 in ductal carcinoma in situ when compared to the surrounding non-malignant tissues, whereas NME1 levels were significantly reduced in invasive tumor foci and in microinvasive carcinoma buds extending beyond the ruptured basement membrane [61]. By contrast, NME2 was upregulated in ductal carcinoma in situ, but remained highly expressed in invasive tumors supporting the conclusion that NME2 is not a repressor of invasion.

The NME4 gene encodes nucleoside diphosphate kinase D (NDPK-D). Mutations in NME4 that inactivate either the enzymatic activity of NDPK-D or its ability to bind cardiolipin in the mitochondrial inner membrane both induce a strong metastatic phenotype in the cervical carcinoma cell line HeLa and in the breast adenocarcinoma cell line MDA-MB-231 [19], including pronounced cell scattering, loss of intercellular adhesion, increased cell migration in 2D and 3D assays, and increased invasion through a type I collagen matrix. Overexpression of wild-type NDPK-D had the opposite, anti-metastatic effect. Conversely, silencing NME4 in the breast epithelial carcinoma cell line ZR75-1 reduced cell–cell adhesion and increased migration [19]. The metastasis suppressor activity of NME4 was most clearly demonstrated in a metastasis assay in mice in which HeLa cells overexpressing wild-type NDPK-D were injected intravenously [19]. Significantly fewer lung metastases were seen in these mice than in mice injected with HeLa cells overexpressing kinase-dead NDPK-D or expressing a low level of wild-type NDPK-D. These effects were specific to the altered function/expression of mitochondrial NME4 and not due to modified expression of NME1 and NME2 [19].

Low levels of NME4 expression also correlated with high levels of metastasis in various types of cancer in humans. NME4 expression is lower in hepatocarcinoma-derived cell lines with high metastatic potential than it is in those with low metastatic potential [85]. In human cancer, NME4 expression correlates negatively with markers of EMT and tumor aggressiveness [19]. In several cohorts of breast cancer patients, expression of NME4 is negatively associated with markers of mesenchymal cells, the EMT, and tumor invasion, but is positively associated with epithelial markers. In oral cancer, a miRNA that promotes cell migration, invasion and metastasis by inhibiting NME4 expression, miR-196, is highly expressed and correlates with lymph node metastasis [86]. Consistent with the role of NME4 as a metastasis suppressor in human subjects, low expression of NME4 is associated with shorter overall survival (i.e. poor prognosis) in patients with breast tumors or with several other tumor types [19].

Among the other members of the NME gene family, there is sparse evidence for their functions as suppressors of metastasis. Human NME3 and NME5 might modulate tumor cell motility depending on the cell context [55, 87], but the mechanism(s) are unclear. Also, in one study, overexpression of NME3 inhibited the metastatic potential of colorectal tumor cells [87]. Further work will be required to pursue the characterization of these and other NME family genes roles in cancer.

The first evidence for a functional link between NME proteins and dynamin came from a study in Drosophila [91], that showed that Awd, the counterpart of mammalian NME1 and NME2 facilitated dynamin-mediated neurotransmitter uptake at neuromuscular junctions in the fly. Further studies reported a key implication of Awd function in endocytosis during cell migration [96, 97]. In cooperation with the Drosophila homolog of dynamin, Shibire, Awd inhibits cell migration by promoting endocytosis of chemotactic receptors including the receptors for FGF and PDGF/VEGF, from the surface of migrating tracheal cells during tracheogenesis and of migrating border cells during oogenesis [96‐101]. Loss of awd in these two cell types decreases endocytosis, leading to up-regulation of the receptors on the cell surface and increasing migration. By contrast, overexpression of awd, increases the endocytosis rate of receptors from the cell surface, so decreasing cell migration. The severity of the awd phenotype is exacerbated in a shibire mutant background whereas overexpression of awd can revert the phenotype associated with a dominant-negative shibire mutation. Likewise in mammalian cells, NME1 mediates endocytosis of the FGF receptor, FGFR1, induced by expression of the von Hippel-Lindau (VHL) protein and prevents cell migration [102]. Interestingly, a loss-of-function mutant in VHL [103] resembles the tracheal phenotype in the awd mutants [96], suggesting that the functional relationship between VHL and NME is evolutionary conserved and is important during development. In addition, mammalian tumor cell lines overexpressing NME1 have increased endocytosis of the EGF receptor and also migrate less than the control cells, and both the increased endocytosis and suppression of migration are blocked by inhibitors of dynamin [57]. In the nematode Caenorhabditis elegans, the NDPK homolog of NME1 and NME2, NDK-1, also influences migration of distal tip cells [104]. Although the underlying mechanism is unknown, the genes encoding C. elegans NDK-1 and dynamin, DYN-1, interact genetically [104]. Additionally, in a genome-wide RNAi screen for genes involved in membrane trafficking, knockdown of NDK-1 caused failure of receptor-mediated endocytosis [105]. Thus, it is possible that NDK-1 regulates the amount of a chemotactic receptor on the surface of the distal tip cells, in the same way as it does in Drosophila and in mammals.

Together, the evidence discussed above indicates that NME proteins facilitate endocytosis of surface receptors and possibly other proteins, altering their availability to transduce migration signals, which, in turn, can suppress cell migration and chemotactism.

Regulation of endocytosis rate by NME proteins may also influence the ability of tumor cells to invade surrounding tissues by clearing membrane-bound proteases from the cell surface. Transmembrane membrane type 1 metalloproteinase (MT1-MMP) is instrumental during cancer progression by mediating proteolytic breaching of tissue barriers, basement membrane and interstitial stromal type I collagen. MT1-MMP is cleared from the cell surface by dynamin-dependent clathrin-mediated endocytosis and is found in clathrin-coated pits associated with dynamin [106‐108]. Overexpression of NME1 in breast tumor cells was found to increase MT1-MMP endocytosis resulting in removal of the protease from the cell surface, whereas silencing of NME1 decreases MT1-MMP uptake [61]. Collectively, these data indicate that NME1 controls the endocytic clearance and surface exposure of MT1-MMP in human breast cancer. As a consequence, loss of NME1 function enhances both matrix degradation and the invasive potential of breast tumor cells in vitro. At the mechanistic level, MT1-MMP, NME1 and dynamin, interact in clathrin-coated vesicles at the plasma membrane. Loss of NME2, by contrast, has no such effect. Consistent with a role for NME1 in endocytosis of MT1-MMP, in human hepatoma and colorectal tumor cell lines in which NME1 is silenced, expression of a mutant MT1-MMP deleted of its catalytic domain inhibits invasion [60]. Conversely, overexpression of proteolytically active invasion-promoting MT1-MMP increases invasion by cells in which NME1 is silenced. Thus, NME1 enhances endocytosis of MT1-MMP, so suppressing cell invasion.

Another way by which the function of NME proteins in dynamin-mediated endocytosis may suppress metastasis is by promoting cell adhesion. Evidence for this mechanism, again, came first from studies of the awd mutation in the Drosophila homolog of NME1 and NME2. During oogenesis in mutant awd larvae, adherens junction components, including E-cadherin, β-catenin and α-spectrin, in follicle epithelial cells are mislocalized in the oocyte, disrupting the integrity of the epithelium structure, whereas awd overexpression promotes the turnover of these components by controlling endocytosis [109]. Consistent with this, awd mutations cause developmental defects in the imaginal discs, the sac-like epithelial structures in Drosophila larvae from which legs, antennae and wings develop in the adult fly [110, 111]. A kinase-dead awd mutation failed to rescue the awd mutant phenotype in contrast to the wild-type awd [112], indicating that Awd kinase activity is essential for cell adhesion during Drosophila development.

Consistent with a role for Awd in cell adhesion, dynamin is also necessary to maintain epithelial integrity [113, 114]. In epithelial tissues with a shibire mutation, E-cadherin accumulates in the cytoplasm and adherens junction stability is disrupted, indicating that E-cadherin endocytosis is regulated in epithelial tissues and necessary to maintain epithelium integrity [113‐116]. Similarly, in C. elegans embryos, dynamin-mediated endocytosis is crucial to maintain cell polarity [117]. Moreover, in non-invasive hepatoma and colon tumor epithelial cell lines, silencing NME1 reduces the amount of E-cadherin on the cell surface, correlating with reduced cell–cell adhesion [60]. Also in mammalian epithelial cells, NME1 promotes dynamin-mediated endocytosis of E-cadherin [118].

NME proteins may also regulate cell adhesion to the substratum by modifying the endocytosis of integrin receptors. ICAP-1, an adaptor protein for clathrin-dependent endocytosis of integrins, recruits NME1 and NME2 close to integrins, the clathrin adaptor protein complex, AP2, and dynamin at clathrin-coated pits to ensure integrin turnover at focal adhesions and regulate integrin signaling and cell adhesion [119‐121].

This abundant evidence that NME proteins are involved in clearing from the cell surface various receptors with key functions in cell–cell adhesion and cell adhesion to the matrix by promoting their dynamin-mediated endocytosis, so inhibiting cell migration and invasion, likely explains the strong metastasis suppressor activity of NME1.

Additional members of the dynamin superfamily act elsewhere in eukaryotic cells to mediate membrane fission and fusion [122, 123]. Three human mitochondrial dynamin-like proteins, dynamin-related protein 1 (DRP1), optic atrophy protein 1 (OPA1) and mitofusin (MFN), could be implicated in the metastasis suppressor function of mitochondrial NME4 and possibly NME3.

Changes in mitochondria structure and function are potent determinants of EMT and metastasis [124‐126]. In particular, fragmentation (or fission) of the mitochondrial network facilitates invasion and migration of tumor cells, whereas mitochondrial fusion is rather inhibitory [127]. Generally, metastatic tumor cells express low levels of the fusogenic protein, MFN, as compared to non-metastatic cells, while they express higher levels of the pro-fission protein, DRP1 [128‐131]. In addition, activation of DRP1 [132] or MFN silencing [128] increase the metastatic potential, whereas silencing or pharmacological inhibition of DRP1 or MFN overexpression reduce cell migration and metastasis [128, 129, 133, 134].

The two isoforms of mitofusin, MFN1 and MFN2, are integral membrane proteins of the mitochondrial outer membrane that mediate fusion [122, 123]. NME3 is similarly localized at the outer membrane depending on its N-terminal region [93], and it is known to interact with MFN1 and MFN2 [135]. Silencing of NME3 increases fragmentation of mitochondria [136], suggesting that NME3 enhances outer membrane fusion mediated by mitofusins. Strikingly, expression of NME3 was shown to rescue mitochondrial fusion and elongation in NME3-silenced cells irrespective of its NDPK activity [135]. Further studies will be necessary to elucidate exactly how NME3 functions in mitochondrial fusion and whether this mechanism can contribute to suppress metastasis. In addition, it should be noticed that the NDPK ortholog DYNAMO1 (dynamin-based ring motive-force organizer 1) of mammalian NME3 locally generates GTP for the optimal activity of DRP1 during the division of mitochondria in the red alga, Cyanidioschyson merolae [137]. However, the relevance of these data regarding the role of NME3 in metastasis dissemination is currently unknown.

In addition, the dynamin-like protein, OPA1, mediates the fusion between the inner membranes of mitochondria [122, 123]. NME4, is also located in the intermembrane space and is bound to the mitochondrial inner membrane through cardiolipin, an abundant phospholipid in the inner membrane [138]. Silencing of NME4 alters mitochondrial morphology by producing fragmented, swollen and ‘blebby’ mitochondria reminiscent of those produced upon defective mitochondrial fusion [93]. Depletion of NME4 phenocopies the effect of OPA1 loss-of-function on mitochondria morphology. Moreover, NME4 forms a complex with OPA1 on the mitochondrial inner membrane, facilitated by the binding of both proteins to cardiolipin [138‐140]. The effect of NME4 binding to OPA1 is not fully understood, however NME4 increases GTP-loading onto OPA1 [93], suggesting that NME4 promotes OPA1-mediated fusion activity. The functions of NME3 and MFN at the outer membrane and NME4 and OPA1 at the inner membrane, respectively, likely contribute to maintaining a fused mitochondrial network important for preventing the EMT and metastasis.

Members of the dynamin superfamily are evolutionarily conserved membrane-remodeling GTPases involved in both membrane fission and fusion reactions. However, unlike myosin and kinesin motor proteins that use ATP to produce forces, dynamins hydrolyse GTP. In addition, dynamins are very sensitive to intracellular GTP concentration due to their remarkably low affinity for GTP and high intrinsic GTPase activity, resulting in rapid GTP hydrolysis and GDP–GTP exchange, which is enhanced by dynamin oligomerization [90, 93, 141]. Thus, vigourous replenishment of GTP is necessary to sustain cellular activity of dynamin. Unlike for myosins and kinesins, which are fueled by high cytosolic ATP concentration, the lower concentrations of GTP are not sufficient to maintain high rate of GTP loading and GTP hydrolysis by dynamins [142]. Thus, a mechanism of GTP channeling achieved by enzymes that synthesise GTP in close proximity to dynamins is required to secure a high GTP/GDP ratio and favour GTP hydrolysis. The strong NDPK activity of NME proteins [143], together with their high affinity for GDP [144], are ideal to maintain a high local concentration of GTP required for dynamin function. Indeed, several evidence support a model in which NDPKs physically interact with members of the dynamin superfamily to maintain high local GTP concentration for optimal dynamin function in membrane remodeling [94]. NME1 and NME2 fuel cytoplasmic endocytic dynamins at plasma membrane clathrin-coated pits to drive endocytosis (Fig. 1A). In addition, NME1 and NME2 may facilitate the oligomerization of dynamin, which is necessary for membrane fission [57]. Thus, local production and channeling of GTP to endocytic dynamins and stimulation of dynamin oligomerization by NME1 and NME2 contribute to stimulate dynamin’s function in endocytosis. NME4 has the same function on OPA1 by fueling GTP on it at the mitochondria inner membrane to drive inner membrane fusion (Fig. 1B). The molecular mechanism underlying NME3 function on mitofusins must be different to this GTP channeling mechanism as a kinase-dead mutant of NME3 can rescue mitochondrial fusion in NME3-silenced cells. One possibility is that NME3 recruits cytosolic NDPKs, NME1 and/or NME2, to the mitochondrial surface by forming hetero-oligomers. In this scenario, NME1/NME2 recruited by NME3 would channel GTP to mitofusins to promote outer membrane fusion (Fig. 1C). If the molecular mechanisms of action of NME proteins on dynamin superfamily proteins remain to be defined precisely, convergence of subcellular localisations of NME proteins and their dynamin superfamily counterparts, acting as a team is emerging as a new, interesting perspective to explain the antimetastatic activity of NME proteins (Table 1).