The acquisition of a migratory and invasive phenotype by carcinoma cells is a crucial step in cancer progression, as it facilitates metastasis that accounts for 90% of human cancer death [39]. AQP1 plays a significant role in tumour cell migration and invasion. Hu and Verkman observed that AQP1 expression accelerated migration of mouse melanoma B16F10 and breast cancer 4T1 cell lines in vitro, with polarised expression of AQP1 observed at the leading edge of the migrating cells [40]. In mice, they observed that AQP1 expression increased cancer cell extravasation and lung metastases. These findings were supported by observations of Jiang who showed alteration of osmotic water permeability induced by over- or under-expression of AQP1 channels influenced the migratory property of HT20 human colon cancer cells in vitro; AQP1 over-expressing HT20 cells demonstrated increased extravasation in nude mice [41].

The importance of water fluxes across the plasma membrane in the facilitation of lamellipodia formation has been acknowledged in earlier reports on cell migration [42,43]. One proposed mechanism for AQP1-modulated tumour cell migration is induction of osmotic water flow across the plasma membrane by AQP1 in response to an osmotic gradient created by actin depolymerisation and active solute influx at the leading edge of migrating cells [44]. Water influx through AQP1 can increase hydrostatic pressure, causing local expansion of plasma membrane, followed by actin re-polymerisation to stabilise cell membrane protrusion. Chen and colleagues demonstrated infection of 603B cholangiocyte cells with Cryptosporidium parvum, an intracellular parasite, led to recruitment of host-cell AQP1 and SGLT1, a Na⁺/glucose co-transporter, to the attachment site with resultant localised water influx, and inhibition of AQP1 resulted in reduction in the parasite cellular invasion, which was dependent on host-cell membrane protrusion [45]. Alternative mechanism to explain APQ1 modulated tumour cell migration is changes in cell shape and volume of migrating cells during passage through tight environments and generation of hydrostatic forces induced by AQP1. Actin polymerization/depolymerisation and transmembrane ionic fluxes may facilitate osmotic water flow through AQP1 [46]. Stroka and colleagues also proposed an “Osmotic Engine Model”, involving an actin- and myosin-independent cell migration mechanism based on water permeation and active and passive ion transport through AQPs and Na⁺/H⁺ pumps for cell migration in confined microenvironments [47].

Migration of cells requires coordinated regulation of [Ca²⁺]_i, intra- and extra-cellular pH, cellular membrane potential and volume, making ion channels and transporters integral components of cellular migration mechanism [48]. Cells establish polarity during directional migration and several ion channels and transporters have been suggested to polarise at the leading edge of migrating cells where they may play a role in triggering osmotic water flow across plasma membrane [49,50]. Ion channel and transporters such as AQPs, K⁺ channels, Na⁺/H⁺ exchanger and Na⁺/glucose co-transporter and calcium-regulatory transporter proteins are suggested to be involved in the key steps of tumour metastasis cascade, namely loss of cell-cell contacts, invasion of surrounding stroma and intra- and extra-vasation of vasculature [51,52]. Kourghi and colleagues showed several AQP1 ion channel blockers with no effect on water channel activity inhibited migration of HT29 cells and the magnitude of inhibition was dependent on the potency for AQP1 ion channel block [53]. Ion channel property of AQP1 alone may be sufficient to facilitate tumour cell migration in some cancers.

More recently, association of AQP1 with tumour microenvironment-driven tumour growth has been suggested. Pelagalli and colleagues demonstrated that treatment with bone marrow-derived mesenchymal stem cells (BM-MSCs) conditioned medium increased AQP1 expression with resultant enhanced migration and invasion of U2OS osteosarcoma and SNU-398 hepatocellular carcinoma cells, which were hampered by AQP1 inhibitor, tetraethylammonium chloride [54]. BM-MSCs have been shown to differentiate into cancer-associated fibroblasts which promote tumour progression and metastasis [55,56]. AQP1 may play an important role in the cross-talk between tumour microenvironment and tumour cells together with various growth factors, cytokines and chemokines.

Tumour angiogenesis—the formation of new blood vessels within the tumour—is another fundamental property of cancer and is vital for tumour metastasis [57]. These vessels provide essential nutrients to sustain tumour growth, and a route by which the cancer cells can exit the tumour and enter the circulation. There has been an increasing recognition that ion channels and transports are concerned with neovascularization through mediating activation, proliferation, migration and differentiation of endothelial cells. Ion channels and transporters may serve as enzymes, chemical and mechanical sensors, receptors and scaffolding proteins [58]. Up-regulation of AQP1 was shown in the perivascular area of astrocytoma where infiltration of tumour cells occur, in contrast to the scarce expression in the necrotic centre, suggesting a link between AQP1 and tumour angiogenesis [22].

Saadoun and colleagues observed reduced density of tumour microvessels in AQP1 null mice subcutaneously implanted with melanoma cells, which delayed tumour growth, and prolonged the survival of the mice [59]. Abnormal tumour microvascular anatomy with reduced density resulting from inhibition of AQP1 expression was also reported by Nicchia and colleagues who tested RNA interference knockdown using AQP1 siRNA on mice subcutaneously implanted with B16F10 mouse melanoma cells, and by Esteva-Font and colleagues who examined the impact of AQP1 deficiency in mouse mammary tumour virus-driven polyoma virus middle T oncogene (MMTV-PyVT) mice which spontaneously developed epithelial cancer [60,61]. These findings were accompanied by reduced staining for vascular endothelial growth factor receptor 2 (VEGFR2), an indicator of angiogenic sprouting, and reduced expression of the endothelial marker, factor VIII [60,61].

Using endothelial cells isolated from AQP1 null mice, Saadoun and colleagues showed impaired migration of AQP1-deficient endothelial cells associated with abnormal vessel formation in vitro [59]. Silencing of AQP1 using siRNA in HMEC-1 human endothelial cells resulted in a lack of F-actin polarisation at the leading edge of the plasma membrane and failure of these cells to organise a cord-like network in vitro, the finding also demonstrated in AQP1-silenced WM115 human melanoma cells [62]. This implied that AQP1 facilitated the migration not only of tumour cells but also endothelial cells, enabling tumour angiogenesis. As endothelial cell migration is linked to vascular permeability, Clapp and Escalera proposed that enhanced vessel permeability facilitated by AQP1, which increases cellular water transport, initiates angiogenic cascade by inducing extravasation of plasma proteins needed as a scaffold for migrating endothelial cells [63].

In comparison to tumour cell migration and tumour angiogenesis, the evidence for AQP1-facilitated tumour cell proliferation is less substantial. Inhibition of AQP1 activity in colon cancer cell line HT29 showed no effect on proliferation, while in another colon cancer cell line HCT-116, proliferation was reduced by 17% [64]. Mouse cancer cell lines, B16F10 and 4T1 with induced over-expression of AQP1 failed to show increased cell proliferation despite exhibiting increased extravasation and the formation of distant metastases [40]. Contrary to this, Hoque and colleagues observed enhanced cell proliferation in NIH-3T3 mouse embryo fibroblast cell line with forced AQP1 expression in vitro [20]. Similar finding was reported with rat pheochromocytoma cell line PC12 [65]. Down-regulation of AQP1 expression with shRNA in two osteosarcoma cell lines U2OS and MG63 was accompanied by inhibition of proliferation [66]. AQP1 inhibition by an inhibitor AqB050 or siRNA knockdown was reported to result in a reduction in cell proliferation in primary malignant mesothelioma cells harvested from pleural effusions [67].

Resistance to apoptosis has been proposed as a part of the mechanism underlying enhanced cell proliferation of AQP1 expressing cells [20,66]. In addition to reduced induction of apoptosis on nocodazole treatment to synchronise cell cycle, transfection of PC12 cells with AQP1 cDNA increased the proportion of cells in S and G2/M phases and this was associated with enhanced expression of cyclin D1 and E1, key proteins needed for cell cycle progression through G1 phase and G1/S transition, respectively [68]. Progression through the cell cycle is accompanied by increase in cell volume, while apoptosis involves reduction in cell volume [69,70]. With AQP1 over-expressing PC12 cells exhibiting a larger cell size and higher intracellular complexity compared to the wild-type control, Galan-Cobo and colleagues postulated that the cell morphology change induced by AQP1 over-expression facilitates tumour cell progression through the cell cycle and antagonizes apoptotic process [68].

Meng and colleagues demonstrated co-immunoprecipitation of AQP1 and β-catenin with up-regulation of β-catenin when mesenchymal stem cells (MSC) were made to over-express AQP1 [91]. β-catenin functions as an intracellular signal transducer in canonical Wnt signalling pathway and regulates gene transcription (Figure 3) [92,93]. In the absence of Wnt ligand, cytoplasmic β-catenin forms a complex with Axin, APC, GSK3β and CK1 and becomes phosphorylated. This allows β-catenin to be recognized by E3 ubiquitin ligase β-Trcp for proteasomal degradation. In the presence of Wnt ligand, a receptor complex forms between Frizzled and LRP5/6 resulting in recruitment of Dishevelled, which phosphorylates LRP5/6 with subsequent Axin recruitment. This disrupts Axin-mediated phosphorylation/degradation of β-catenin, allowing β-catenin to transport to the nucleus where it serves as a co-activator for TCF to activate Wnt responsive genes including those responsible for tumorigenesis such as c-Myc, cyclin D1, c-Jun and FRA1.

The pathway is concerned with establishment and maintenance of stem cells and selective destruction of a cytosolic β-catenin pool resulted in a loss of tumorigenic potential in the mouse model of a colon cancer cell line DLD1 [94]. c-Myc, cyclin D1 and the components of AP-1 transcription factor complex, c-Jun and FRA1 have been proposed to be the transcriptional targets of the Wnt signalling pathway, and are relevant in human tumorigenesis in that they are known to control cell cycle progression (G1/S transition), as well as proliferation, differentiation and apoptosis [95,96,97]. Because treatment of AQP1-silenced WM115 human melanoma cells with the proteasome inhibitor MG132 induced the recovery of β-catenin, it has been suggested that AQP1 negates proteasomal degradation of β-catenin to augment Wnt signalling pathway [62]. Yun and colleagues demonstrated AQP1-mediated enhanced β-catenin expression in pulmonary arterial myocytes required AQP1 C-terminal tail, which contains protein binding sites [98]. Interaction of AQP1 with β-catenin via its C-terminus may hinder association of β-catenin with APC/Axin destruction complex for proteasomal degradation [93].

β-catenin also plays a role in cell-cell adhesion. Reorganisation of cadherin-catenin complexes is implicated in epithelial-mesenchymal transition (EMT), which enables epithelial tumours to become invasive [99]. Silencing of AQP1 using siRNA in WM115 human melanoma and HMEC-1 human endothelial cells resulted in a lack of polarisation of F-actin at the leading edge of plasma membrane and failure of these cells to organise a cord-like network in vitro [62]. Lin-7, another component of cadherin-catenin complex which interacts with PDZ domain of β-catenin, was shown to co-immunoprecipitate with AQP1 in WM115 human melanoma and HMEC-1 human endothelial cells and its protein expression was reduced by siRNA down-regulation of AQP1. Researchers have speculated AQP1 stabilises the cadherin/β-catenin/Lin-7/F-actin complex to enhance the migratory and invasive capacity of tumour cells [62].

Cadherin is currently considered to negatively regulate Wnt signalling by sequestering cytoplasmic β-catenin as it shares the binding site on β-catenin with adenomatosis polyposis coli protein of Wnt signalling; however, the exact interplay between the cytoplasmic pool of β-catenin for Wnt signalling pathway and cadherin-bound β-catenin at plasma membrane is poorly understood [100]. Further research into interaction of AQP1 with each pool of β-catenin is needed.

FAK is a cytoplasmic tyrosine kinase that induces growth factor receptor- and integrin-mediated signal transduction. It is implicated in tumour progression; activated and/or over-expressed FAK is found in a variety of human cancers. FAK contributes to tumour progression by enhancing tumour cell motility and EMT and by promoting tumour vascularisation (Figure 4) [101]. Vascular endothelial growth factor-A (VEGF-A)/VEGF or angiopoietin-1 signalling activates PI3K/AKT through the action of FAK to promote migration, sprouting and angiogenesis of endothelial cells [102]. It is postulated AQP1 partly exhibits its tumorigenic effect through activation of FAK signalling. FAK co-immunoprecipitated with AQP1 as well as β-catenin in AQP1 over-expressing MSC and depletion of FAK abolished AQP1-driven MSC migration (Figure 3) [91]. As over-expression of AQP1 did not alter its expression at mRNA level, the authors concluded that up-regulation of FAK seen in AQP1 over-expressing MSCs again occurred at post-translation level.

Expression levels of MMP2 and MMP9 were found to be down-regulated by AQP1 siRNA in LTEP-A2 and LLC lung cancer cell lines [103]. These enzymes degrade type IV collagen, the major structural component of basement membrane and activate TGF-β to promote EMT [104]. Up-regulation of MMPs are common in many tumour cells and has been shown to facilitate tumour cell migration in vitro and metastasis in vivo [105,106,107,108]. FAK was reported to induce secretion of MMP2 and MMP9 in mouse fibroblasts in a Concanavalin-dependent manner [109]. Wnt signalling has been reported to induce MMP2 and MMP9 expression in effector T cells [110]. It is speculated AQP1 enhances the activity of MMP2 and MMP9 through FAK and Wnt signalling pathways, augmenting migration of tumour cells and endothelial cells.

Rho GTPases are involved in cell cytoskeleton rearrangements, migration and proliferation, and transformation. Aberrant regulation of Rho GTPases has been implicated in tumour progression via several mechanisms; loss of cellular polarity, stimulation of dedifferentiation, alteration of cell-cell and cell-matrix adhesion which allows tumour cells to become invasive and metastasise to distant sites via vascularization of tumours and intravasation of tumour cells [111]. RhoA and Rac1, in particular, have been shown to regulate mesenchymal and amoeboid movements, respectively, where RhoA drives actomyosin contraction, while Rac1 mediates cell polarisation and lamellipodia formation [112]. These two modes of cell motility play an important role in tumour cell invasion in three-dimensional extracellular matrix.

The association of AQP1 with RhoA and Rac has been previously suggested by Jiang who observed increase in the activation of these small G proteins in migrating HT20 colon cancer cells over-expressing AQP1, with more frequent polarised expression of actin at the cells’ leading edges. Impaired expression of RhoA was also seen in U2OS and MG63 cells following AQP1 down-regulation by shRNA [66]. Both RhoA and Rac1 are known to be involved in FAK-mediated tumour invasion and metastasis via Rho guanine exchange factor and the PI3K signalling pathway (Figure 4), and this may be one of the mechanisms whereby AQP1 influences the activities of Rho GTPases [101].

As described earlier, hypoxia-induced glycolysis may enhance AQP1 expression in tumour cells to maintain their viability. This is thought to cause acidification of the extracellular compartment with increase in secretion of a lysosomal cysteine protease cathepsin B (Figure 2) [113]. Up-regulation of cathepsin B together with AQP1 and LDH in 9L cells under glycolytic conditions was previously demonstrated [72]. High levels of expression of cathepsin B have been observed in epithelial tumours and glioblastoma, and are implicated in tumour development and progression via initiation, proliferation, angiogenesis, invasion, inflammation, apoptosis and metastasis [114]. Cathepsin B has been suggested to be a part of proteolytic network that involves MMPs and the plasminogen activator cascade, which may contribute to enhanced motility, invasion and angiogenesis [115,116,117]. Cathepsin B has additionally been thought to cleave anti-apoptotic proteins, such as Bcl-2, Bcl-xl and Bak, resulting in deregulation of tumour cell apoptosis [118].

Whether TGF-β plays any role in AQP1-facilitated tumour progression remains to be further evaluated. There have been two articles published in which expression of TGF-β was examined in AQP1 down-regulated tumour cells. Gene set enrichment analysis performed by Wu and colleagues indicated a positive correlation between AQP1 over-expression and TGF-β signalling pathway in osteosarcoma and expression of TGF-β1 and TGF-β2 was reduced in U2OP and MG63 cell lines when their AQP1 expression was down-regulated by shRNA [66]. Conversely, no alteration in the level of TGF-β was observed in AQP1 siRNA-treated LTEP-A2 and LLC cells [103]. The conflicting results reported may be secondary to the paradoxical action of TGF-β during tumour progression; while it initially supresses tumour development by regulation of cell proliferation, differentiation, adhesion and tumour microenvironment, TGF-β can manifest its pro-tumorigenicity by inducing uncontrolled cell proliferation, loss of apoptosis, EMT, sustained angiogenesis and evasion of immune surveillance [119].