AQP1 Is Not Only a Water Channel: It Contributes to Cell Migration through Lin7/Beta-Catenin

Background AQP1 belongs to aquaporins family, water-specific, membrane-channel proteins expressed in diverse tissues. Recent papers showed that during angiogenesis, AQP1 is expressed preferentially by microvessels, favoring angiogenesis via the increase of permeability In particular, in AQP1 null mice, endothelial cell migration is impaired without altering their proliferation or adhesion. Therefore, AQP1 has been proposed as a novel promoter of tumor angiogenesis. Methods/Findings Using targeted silencing of AQP1 gene expression, an impairment in the organization of F-actin and a reduced migration capacity was demonstrated in human endothelial and melanoma cell lines. Interestingly, we showed, for the first time, that AQP1 co-immunoprecipitated with Lin-7. Lin7-GFP experiments confirmed co-immunoprecipitation. In addition, the knock down of AQP1 decreased the level of expression of Lin-7 and β-catenin and the inhibition of proteasome contrasted partially such a decrease. Conclusions/Significance All together, our findings show that AQP1 plays a role inside the cells through Lin-7/β-catenin interaction. Such a role of AQP1 is the same in human melanoma and endothelial cells, suggesting that AQP1 plays a global physiological role. A model is presented.


Introduction
Aquaporins are a family of water-specific, membrane-channel proteins expressed in diverse tissues. Two functional groups of mammalian aquaporins are now recognised: aquaporins (AQP1, AQP2, AQP4, AQP5 and AQP8) which are primarily water selective and aquaglyceroporins (AQP3, AQP7, AQP9 and AQP10) which are permeable to small uncharged solutes such as lactate, glycerol and urea in addition to water [1]. The characterization of the organization of aquaporin genes and identification of their position within the human and mouse genomes have established a primary role for some aquaporins in clinical disorders such as congenital cataracts and nephrogenic diabetes insipidus. More recently, in the control of fat accumulation, aquaporins were demonstrated to play an important role [2][3][4][5]. A characterization of AQPs was carried out in neuronal stem cells [6].
Herein, we focused on AQP1 which is specifically and strongly expressed in most microvasculature endothelial cells outside the brain [7]. Recently, in AQP1 null mice, an impairment of endothelial cell migration, without altering their proliferation or adhesion, was shown [8]. In this connection, AQP1 has been proposed as a novel promoter of tumor angiogenesis [9].
Cell migration involves transient formation of membrane protrusion (lamellipodia and membrane ruffles) at the leading edge of the cell. AQP1 is localized at the leading edge and co-immunoprecipitated with several transporters involved in migration including the Na/H + and Cl 2 /HCO3 2 exchangers [10]. Verkman proposed, recently, that water entry might then increase local hydrostatic pressure to cause membrane protrusion that, in turn, creates space for actin polymerization [1].
The organization of cytoskeleton requires the presence of scaffolding proteins, the plasma membrane-associated proteins containing one or several PDZ domains [11]. One scaffold complex common for epithelial and neuronal cells is the heterotrimeric complex consisting of the CASK/Lin-2, Lin-7 and Lin-10 PDZ proteins [12][13][14][15]. In mammals, Lin-7 can recruit cell adhesion molecules, receptors and signaling proteins [12][13][14]. CASK/Lin-2 contains bindings sites for interactions with cytoskeletal proteins [15]. On the other hand, transporters, receptors and ion channels such as the epithelial c-aminobutyric acid (GABA) transporter (BGT-1), the N-methyl-D-aspartate (NMDA) receptor and N-type Ca 2+ channel, interact with the PDZ domains of mammalian Lin-7 [16][17][18][19]. The heterotrimeric PDZ complex plays a role in regulating the localization of interacting proteins but it is still uncertain how these components are targeted to the junction surface [17,18].
Using targeted silencing of AQP1 gene expression, herein, we demonstrated, for the first time, that AQP1 knock down dramatically affects the actin cytoskeleton organization through Lin-7/b-catenin interaction. As cellular model we used a melanoma cell lines which was extensively studied and character-ized by our group and that express many angiogenic/lymphoangiogenic factors (WM115) [20]. Finally, in order to assess if their effects were restricted to a specific melanoma cell line or AQP1 plays a global physiological role, we also analyzed a human endothelial cell line.

GFP-Lin7
Full-length mouse Lin7A protein was fused to the C-terminus of the enhanced Green fluorescence (GFP-Lin7) by subcloning the BamH1 restriction fragment of the pGEX1-mLin7A plasmid (kindly provided by S. Kim, Standford Medical School, CA) in the mammalian expression vector pEGFP-C1 (Promega) as previously described.

MG132 treatment
The cells were treated with MG132 (10 mmol/L, Sigma) for 1 h under the standard growth medium.

Immunofluorescence
Cells were fixed at 4uC in 4% paraformaldehyde (methanol for co-staining with anti-AQP1 and anti-Na+/K+ ATPase antibodies) and then permeabilised with 0.01% Triton X-100 for 30 min at room temperature. Then, the cells were blocked with normal goat serum (1:20) in a solution of 1% glycine, 2% bovine serum albumine in saline for 20 min and incubated with rabbit anti-AQP1 (1:100, Alpha Diagnostic International), rabbit anti-Lin7 For F-actin visualisation, the cells were fixed in 4% paraformaldehyde, permeabilised with 0.1% Triton X-100 for 30 min and incubated with 50 mg/ml FITC-conjugated phalloidin (Sigma) or Rhodamine-conjugated phalloidin (Molecular Probes) for 40 min at room temperature.

Cell migration and invasion
2610 5 cells were plated on the upper chamber of transwell inserts and maintained in serum-free medium. The lower chamber contained medium plus 1% serum or 10% serum (chemoattractant) or no serum (negative control). Cells were incubated for 6 h at 37uC in 5% CO 2 . After incubation, the membranes were washed briefly with saline and the non-migrated cells were scraped from the upper surface of the membrane with a cotton swab. Migrated cells remaining on the bottom surface were stained for AQP1 for wild type cells and counted 6hr later (see immunofluorescence for AQP1). For AQP1 silenced cells, the number of cells migrated were counted after haematoxylin staining. The migrated cells were counted in 20 random high-power microscope fields under 6200 magnification.

RT-PCR analysis
Total cellular RNA was extracted using the RNeasy Mini Kit (Qiagen) according to the manufacturer's protocol. Complementary DNA (cDNA) was synthesized using 3-5 mg of total RNA, 2.5 mM of random nonamers, 500 mM of each of the four deoxynucleotides, and water up to 10 ml. After incubation for 10 min at 70uC and 5 minutes on ice, 10 ml of RT reaction mixture containing 1 U of enhanced avian myeloblastosis virus reverse transcriptase enzyme (eAMV-RT, Sigma), as well as 1 U of RNase inhibitor, 16buffer for AMV-RT, and water, was added for a total volume of 20 ml. The reaction proceeded at 25uC for 15 min. and 45uC for 50 min. PCR reactions were performed in 50 ml using 200 mM dNTPs, 1.5 U of REDTaq polymerase (Sigma), 16 RED buffer, 0.3 mM of specific oligonucleotide primers, and 2 ml of cDNA. The sequence of the primers, the lengths of PCR products and the annealing temperatures were as follows: AQP1, sense 59-TGGGGACCAAGATTTACCAA-39 and antisense 59-CGCAGAGTGTGGGCCACATCA, 221 bp 68uC; AQP4, sense 59-GAATCCTCTATCTGGTCACA-39, antisense

Results
Knock down of AQP 1 in HMEC-1 and WM115 cell lines induces changes in cell shape and F-actin organization WM115 and HMEC-1 cells express mainly AQP1 (see Supplementary materials, Fig. S1). We knocked down the AQP1 water channel using three different siRNA: AQP1-1, AQP1-2, AQP1-3 in both WM115 and HMEC-1 cells (see Supplementary materials, Fig. S2). A significant suppression of AQP1 (90%) PLoS ONE | www.plosone.org occurred in WM115 tranfected cells with siRNA AQP1-1 and with siRNA AQP1-2 (72%) but not with AQP1-3 (23%) (Supplementary materials, Fig. S2). Furthermore, siRNA CTRL1 and 2 did not modify the level of expression of AQP1 (Supplementary Materials, Fig. S2). Interestingly, in both cell lines, the silencing of AQP1 (siRNA AQP1-1) induced a dramatically change in cell shape. In non-silenced cells, F-actin was preferentially polarized at the leading edge of the plasma membrane and the cells were able to organize a cord-like network in vitro. In contrast, both AQP1-1 silenced cells, were rounded up (Fig. 1, panel B and C). The integrity of the F-actin cytoskeleton is critically dependent on cell volume changes and several volumeregulated transport proteins including water channels exhibit functional interaction with the cytoskeleton playing a pivotal role in cell volume regulation [1]. The ability to establish and maintain a polarized state of the cell, that is the basis for migration, was suggested to be reflected by a polarized distribution of the ion and water channels [24,25] that determine the water influx, providing the force that extend the plasma membrane [26] and that promotes the actin polymerization in the front part of cells formed by lamellipodium [27]. Our data shows an AQP1 dependent organization of cell shape (Fig. 1).
AQP1 affects migration potential of WM115 and HMEC-1 cells. Since AQP1 was demonstrated to be involved in angiogenesis [8], we investigated the effect of this water channel on cell migration in WM115 and HMEC-1 cells. As shown in Figure 2A, AQP1 knock down (siRNA AQP1-1) induced a significant reduction in the migratory capability of both cell lines. In contrast, siRNA AQP1-3 did not induce any significant effects on the migration or on the shape of the cells. Accordingly, siRNA AQP1-2 (which gave an intermediate effect on AQP1 expression) showed a little effect on cell migration (Fig. 2B). On the other hand, CTRL 1 and 2 did not modify significantly the migration capacity of both cell lines (Fig. 2C) Lin-7 co-immunoprecipitates with AQP1 in HMEC-1 and WM115 cells. Lin-7 is a component of the cadherin-catenin complex through the physical interaction with the PDZ target sequence of beta-catenin [28]. Recently, some transportes, receptors and ion channels are demonstrated to interact with the PDZ domains of mammalian Lin-7 [16][17][18]. To verify whether Lin-7 is a component of AQP1-containing complex, we tested the association of these proteins by coimmunoprecipitation (Fig. 3). Furthermore, the polyclonal Lin-7 antibody (but not a pre-immune serum) co-immunoprecipitated AQP1 from both WM115 and HMEC-1 lysates (Fig. 3A). We have also verified that Lin-7 and beta-catenin coimmunoprecipitated in our cells according to previous paper [28] (Fig. 3A). In contrast, Ephrin B2, a protein involved in the maintaining of morphology of the cells and in various biological process such as angiogenesis, did not co-immunoprecipate or coimmunolocalize with AQP1 ( Fig. 3A and B). Since Lin-7 interacting with Lin-2/CASK bind cytoskeletal proteins [15], we investigated whether the two cell line expressed Lin-2/CASK. However, we found no significant expression of Lin-2/CASK by RT-PCR or immunoblot analyses (not shown).
Finally, in figure 4 is shown that AQP1 is retained specifically by the Lin7-GFP, confirming co-immunoprecipitation experiments.
Knock down of AQP1 affects Lin-7/b-catenin expression. Since Lin-7 co-immunoprecipitates with AQP1, we knocked down AQP1 with three different siRNA (AQP1-1, AQP1-2 and AQP1-3) and two CTRLs and investigated what happens to Lin-7, beta-catenin and beta-actin (housekeeping gene) in both cell lines (Fig. 5). Furthermore, we have also analysed the effect of silencing AQP1 on PAR-3, another PDZ-containing protein critical for cell polarity localized at the intracellular junctions of endothelial cells [29]. Negative control (CTRL1 and 2) did not affect significantly, Lin7, beta-catenin or AQP1 (Fig. 5A). Accordingly, siRNA AQP1-3 (which is not able to reduce significantly the level of expression of AQP1 and did not modify the morphology of the cells, Fig. 2B) did not modify beta-catenin or Lin-7 expression (Fig. 5B). siRNA AQP1-2 gave an intermediate effect (Fig. 5B). Importantly, the treatment of AQP1 silenced cells with the proteasome inhibitor MG132 enhanced the recovery of betacatenin (70%, Fig. 6A) and partially of Lin-7 (16%, data not shown). Accordingly, in cells transfected with CTRL1 no effect after MG132 treatment occurred (Fig. 6B).

Discussion
Aquaporins (AQPs) are the integral plasma membrane proteins involved in water transport in many fluid-transporting tissues. Interestingly, in AQP1 null mice endothelial cell migration was impaired, and in vivo angiogenesis reduced [8].
Herein, we show, for the first time, in two different cell line expressing AQP1, a human melanoma cell line WM115 and a human microvascular cell line HMEC-1 that the knock down of AQP1 induced a re-organization of F-actin and affected the cell shape. In non-silenced cells F-actin was preferentially polarized at the leading edge of the plasma membrane and the cells were able to organize a pseudovascular network. In contrast, siRNA AQP1-1 cells have lost the capability to organize such a network. From the functional point of view, AQP1 silenced cells decreased their capability to migrate. In this connection, a remarkably impaired growth of implanted tumors was demonstrated in AQP1 null mice, with reduced tumor vascularity and an extensive necrosis [8]. Interestingly, in AQP-1 null mice cell migration of melanoma cells was greatly impaired with abnormal in vitro vessel formation [8].
The novelties of the present paper are the following. Firstly, AQP1 plays the same role in human melanoma and endothelial cells, suggesting that this water channel has a global physiological role. Secondly, AQP1 interacts at least, with Lin-7/beta catenin signaling. All these aspects are particular intriguing. Actually, the polarization at the leading edge of migrating cells has been demonstrated for several transporters involved in migration, including Na + /H + and Cl 2 /HCO 3 2 exchangers and the Na + / HCO 3 2 co-transporter [30], and cell migration involves the transient formation of membrane protrusions (lamellipodia and plasma membrane ruffles) at the leading edge of the cell, suggesting that a rapid local changes in ion fluxes and cell volume are accompanied by rapid transmembrane water movement [8]. Thereby, the actin cleavage and ion uptake at the tip of lamellipodium might create local osmotic gradients that drive the influx of water across the plasma membrane [10,[31][32][33]. Saadoun et al. postulated that water entry increases local hydrostatic pressure causing the polarization of AQP1 [8]. However, at yet, the intracellular mechanisms triggered by AQP1 are not well known. Herein we show, for the first time, that Lin-7 co-immunoprecipitates with AQP1 and interacts with beta-catenin through Lin-7, thereby affecting the organization of cytoskeleton. Intriguingly, our experiments with the proteosomal inhibitor imply that lack of AQP1 targets the Lin-7/beta-catenin complex to proteolytic degradation.
All together, these data expand the physiological role of AQP1 beyond its water transport function. This protein rather emerges as a critical scaffold for a plasma membrane associated multiprotein complex important for cytoskeleton build-up, adhesion and motility. A model for these functions of AQP1 is presented in Figure 7. In this model we focused the role of AQP1 in the stabilization of cytoskeleton complex through the involvement, at least, of Lin-7/beta-catenin.
All of all, our findings corroborate the analysis of manifold cellular functions of AQPs in normal cells and in diseases and the possibility to consider aquaporins as specific therapeutic targets for various pathophysiological conditions [34]. Figure S1 Expression of AQPs and intracellular localization. Panel A. RT-PCR of AQP1 in WM115 and HMEC-1 cells. Total RNA was purified from subconfluent cells using the RNeasy Mini Kit from Qiagen. One microgram of total RNA was reverse transcribed and amplified by Enhanced avian hs RT-PCR (Sigma) according to the manufacturer's instructions. PCR conditions and primers are reported in the Materials and Methods section. As housekeeping gene, GAPDH was used. In order to check the specificity of the amplified bands they are sequenced. Panel B. RT-PCR of AQP 8, 4 and 9 in WM115 cells for HMEC-1 and WM115 cells as described in panel A. As positive control, human brain tissue expressing AQP4 and AQP8 was used. Panel C. Coimmunofluorescence of AQP1 and Na+/K+ ATPase in WM115 cells. The cells fixed with methanol for 10 min and then permeabilized with 0.01% TRITON X-100 for 20 min, incubated with the primary antibodies (anti-AQP-1, 1:100; mouse anti-Na+/ K+ ATPase (1:250) for 2h at room temperature. The cells were then incubated with secondary antibodies: Alexa Fluor-488 conjugated goat anti-rabbit, (1:700, Molecular Probe), Alexa Fluor-594-conjugated goat anti-mouse, (1:700, Molecular Probes) for 1h and examined on a Leika TCS NT confocal microscope. The localization of both proteins in plasma membrane are shown x-y and z-stack analysis. AQP1 is the main aquaporins expressed in these cells. The protein is mainly expressed at the level of plasma membrane. In order to clearly demonstrate the plasma membrane localization of AQP1, we have co-stained the cells with anti-AQP1 and anti-Na+/K+ ATPase (typical membrane marker) antibodies. As shown in panel C of figure 1, AQP1 shows a plasma membrane localization as demonstrated both by the merge and Z-stack analysis. Found at: doi:10.1371/journal.pone.0006167.s001 (0.14 MB TIF) Figure S2 AQP1 siRNAs in WM115 and HMEC-1 cells. WM115 or HMec-1 cells were tranfected with siRNA AQP1-1, AQP1-2, AQP1-3 or negative control siRNA CTRL1 for 48 h as described in the Materials and Methods section. 50ug protein was submitted to 10% SDS-PAGE and transferred to a PVDF sheet. The latter was incubated with anti-AQP1 (1:2000) or beta-actin (1:5000) as housekeeping protein, for 2h. Then the sheet was incubated with secondary antibody and visualized using the ECL detection system. Densitometric analysis expressed as ratio of each line with respect to beta-actin is shown. The figure shows that