Aquaporin-3 Controls Breast Cancer Cell Migration by Regulating Hydrogen Peroxide Transport and Its Downstream Cell Signaling

Abstract

Most breast cancer mortality is due to clinical relapse associated with metastasis. CXCL12/CXCR4-dependent cell migration is a critical process in breast cancer progression; however, its underlying mechanism remains to be elucidated. Here, we show that the water/glycerol channel protein aquaporin-3 (AQP3) is required for CXCL12/CXCR4-dependent breast cancer cell migration through a mechanism involving its hydrogen peroxide (H₂O₂) transport function. Extracellular H₂O₂, produced by CXCL12-activated membrane NADPH oxidase 2 (Nox2), was transported into breast cancer cells via AQP3. Transient H₂O₂ accumulation was observed around the membrane during CXCL12-induced migration, which may be facilitated by the association of AQP3 with Nox2. Intracellular H₂O₂ then oxidized PTEN and protein tyrosine phosphatase 1B (PTP1B) followed by activation of the Akt pathway. This contributed to directional cell migration. The expression level of AQP3 in breast cancer cells was related to their migration ability both in vitro and in vivo through CXCL12/CXCR4- or H₂O₂-dependent pathways. Coincidentally, spontaneous metastasis of orthotopic xenografts to the lung was reduced upon AQP3 knockdown. These findings underscore the importance of AQP3-transported H₂O₂ in CXCL12/CXCR4-dependent signaling and migration in breast cancer cells and suggest that AQP3 has potential as a therapeutic target for breast cancer.

INTRODUCTION

Breast cancer mortality remains high owing to clinical relapse associated with metastases, primarily to the lungs, brain, and bones (1). Metastases are the result of several sequential processes, including cell migration and invasion (2). Organ-specific metastasis requires chemokine-dependent cancer cell migration toward destination sites (3). In particular, the CXCL12/CXCR4 axis is a key step in breast cancer cell migration toward the lungs (4, 5). The binding of CXCL12 to CXCR4 stimulates downstream G protein signaling, leading to the activation of the phosphatidylinositol 3-kinase (PI3K)/Akt or mitogen-activated protein kinase (MAPK) pathway. These effects regulate a variety of cellular functions, such as cell proliferation and migration, thereby contributing to cancer metastasis and progression (6). However, the underlying mechanism by which the pathways downstream of CXCL12/CXCR4 result in breast cancer cell migration and metastasis remain to be fully elucidated.

Aquaporin-3 (AQP3), a member of the aquaporin water channel family (AQP0 to -12), has the primary function of transporting water and glycerol (7, 8). AQP3 is expressed in various cancer cells derived from diverse types of cancer tissues, including breast, colon, and lung (9, 10). Recent results from clinical studies have suggested the relevance of AQP3 expression in tumor progression and the prognosis of several malignant cancers (11–14). In vitro studies using cancer cell lines have implicated AQP3 expression in cancer cell proliferation and migration (15–17). However, the mechanism by which AQP3 participates as a biological pore channel in cancer progression remains unknown.

AQP3-facilitated cellular uptake of hydrogen peroxide (H₂O₂) was discovered recently (18). We subsequently demonstrated that the uptake of extracellular H₂O₂ by T cells, generated in response to chemokines, including CXCL12, was facilitated by AQP3 and was required for the chemotaxis necessary to produce a sufficient immune response (19). The cellular redox state appears to play a role in the pathology of cancer (20). Many studies have proposed the involvement of reactive oxygen species (ROS), including H₂O₂, in cell growth, survival, motility, and metastasis during cancer progression (21). Furthermore, H₂O₂ is emerging as an important second messenger in cell signaling (22, 23). We hypothesized that H₂O₂, taken up through an AQP3-facilitated process, might act as a second messenger and be involved in a variety of cellular functions, including the focus of the current study—breast cancer cell migration and metastasis.

In this study, we utilized the breast cancer cell lines MDA-MB-231 and DU4475 and their transplantation into immune-deficient mice, since both cells express CXCR4 but do not express epidermal growth factor receptor 2 (EGFR2) (HER2) and are well characterized by their metastatic potentials and properties (4, 5, 24). The results demonstrate that AQP3 is required for CXCL12-induced breast cancer cell signaling and directional migration by a mechanism involving the CXCL12-induced generation of extracellular H₂O₂ and subsequent intracellular transport by AQP3. An in vivo spontaneous-metastasis model using AQP3 knockdown (KD) cells consistently showed markedly reduced breast cancer cell metastasis to the lungs.

MATERIALS AND METHODS

Cell lines.

Breast cancer cell lines, MDA-MB-231 and DU4475, were obtained from the American Type Culture Collection (ATCC). The cells were maintained in Dulbecco's modified Eagle medium (DMEM) or RPMI 1640 containing 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin in a CO₂ incubator (5% CO₂; 37°C).

RNA interference (RNAi) and plasmid DNAs.

For AQP3, NADPH oxidase 2 (Nox2), PTEN, and protein tyrosine phosphatase 1B (PTP1B) knockdown, cells were incubated with ON-TARGETPlus SMARTpool AQP3, Nox2, PTEN, or PTP1B small interfering RNA (siRNA) or nontargeting siRNA (Thermo Scientific). For AQP3 siRNA, three additional independent constructs (Thermo Scientific) were examined. The pCMV6 (empty) vector and human AQP3-expressing pCMV6 vector were obtained from OriGene Technology. The HyPer-Cyto cDNA plasmid was obtained from Evorgen. For the transfection of siRNA and plasmid DNAs, the Lipofectamine transfection system (Invitrogen) was used according to the manufacturer's instructions. For stable knockdown of the AQP3 gene, cells were infected with SMART vector 2.0 lentiviral short hairpin RNA (shRNA) (Thermo Scientific). The expression levels of target genes were examined by real-time PCR or immunoblotting as described below.

Osmotic water permeability assay.

Osmotic water permeability was measured using a stopped-flow spectrometer (Applied Photophysics). MDA-MB-231 cells were collected with Accutase (Invitrogen), and the cells (2.0 × 10⁶ cells/ml) were subjected to a 200 mM inwardly directed mannitol gradient at 37°C for 30 s. The kinetics of the decreasing cell volumes was measured from the time course of 90° scattered light intensity at 450-nm wavelength. The osmotic water permeability coefficient (Pf) was calculated as described previously (25).

H₂O₂ permeability assay.

Cellular H₂O₂ was determined using a fluorescence microplate reader (Molecular Devices) or a fluorescence-activated cell sorter (FACS) Fortessa system (BD Biosciences) with 2′,7′-dicholorodihydrofluorescein diacetate (H₂DCFDA) (Invitrogen) according to the manufacturer's instructions. The data are expressed as percentages of the fluorescent product 2′,7′-dichlorofluorescein (DCF) fluorescence of vehicle-added control cells unless otherwise specified.

Chemotaxis and transendothelial migration assay.

MDA-MB-231 or DU4475 cells (4.0 × 10³ cells) were deposited in the upper chamber of a Transwell containing a fibronectin (10 μg/ml; Sigma)-coated polycarbonate Transwell membrane filter (8-μm pore size; Corning). The lower chamber contained CXCL12 (100 ng/ml; R&D Systems) in 0.1% bovine serum albumin (BSA)-DMEM. After 3 or 6 h of incubation, the recovered cells were analyzed by staining with 0.5% crystal violet. The stained cells were monitored with a microscope (Evos; Invitrogen). The data were expressed as the percentage of control cells undergoing CXCL12-induced migration in three different random areas unless otherwise specified. For transendothelial migration, human umbilical vein endothelial cells (HUVEC) (Japanese Collection of Research Bioresources [JCRB]) were seeded onto a type I collagen (80 μg/ml; BD Bioscience)- and fibronectin-coated 8-μm Transwell filter and grown to confluence. MDA-MB-231 cells (4.0 × 10³ cells) were deposited in the upper chamber, and 0.1% BSA-DMEM containing CXCL12 (100 ng/ml) was added to the lower chamber. After 24 h, transmigrated cells were analyzed by the same method as for the chemotaxis assay.

Immunofluorescence microscopy.

F-actin polymerization was visualized with phalloidin (Alexa Fluor 488; Invitrogen) after cells were fixed with 4% paraformaldehyde, followed by 0.1% Triton X-100 permeabilization. For AQP3 and Nox2 expression, cultured MDA-MB-231 cells were fixed and permeabilized with ice-cold methanol and stained with antibodies against AQP3 (Millipore) and Nox2 (BD Biosciences). Fluorescence images were obtained with an LSM710 confocal microscope (Carl Zeiss).

Time-lapse imaging of chemotaxis and H₂O₂ uptake.

HyPer-Cyto-transfected MDA-MB-231 cells (control siRNA or AQP3 siRNA transfected) were deposited on the μ-slide chemotaxis chamber (Ibidi) according to the manufacturer's instructions. One chamber contained CXCL12 (500 ng/ml) in serum-free DMEM, and another chamber contained serum-free DMEM. Chemotaxis was monitored with an FV10i microscopy system (Olympus). Images were obtained at 3-min intervals, and a total of 500 images were obtained as data unless otherwise specified. Time-lapse imaging of H₂O₂ uptake was monitored at 3-s intervals, and a total of 65 images were obtained with a BZ-X710 microscope (Keyence). The image data obtained were analyzed using ImageJ software (NIH), Chemotaxis and Migration Tool (Ibidi), FV10-Viewer (Olympus), and BZ-X Analyzer (Keyence).

PTEN oxidation.

PTEN oxidation was assessed using the protocol described previously with slight modifications (26). Briefly, cells were scraped in 500 μl of ice-cold 50% trichloroacetic acid (TCA) following CXCL12 and/or H₂O₂ stimulation and centrifuged at 2,000 × g for 5 min. The pellet was washed with ice-cold acetone and solubilized with a Dounce homogenizer in 100 mM Tri-HCl (pH 6.8) buffer containing 2% sodium dodecyl sulfate (SDS) and 40 mM N-ethylmaleimide. The samples were subjected to SDS-PAGE under nonreducing conditions (10% gel; Invitrogen) and transferred to polyvinylidene difluoride (PVDF) membranes for immunoblot analysis. The membranes were analyzed with anti-PTEN (Cell Signaling Technology).

PTP1B oxidation.

PTP1B oxidation was monitored as described previously (27, 28). Stimulated cells were scraped in ice-cold radioimmunoprecipitation assay (RIPA) buffer (Cell Signaling Technology) containing 1% protease inhibitor cocktail (Sigma) with or without 100 mM iodoacetic acid (IAA). The supernatant (10,000 rpm; 10 min; 4°C) was immunoprecipitated with anti-PTPN1 (Novus)-conjugated protein G-Sepharose (GE Healthcare) for 3 h at 4°C. The beads were incubated with 10 mM dithiothreitol (DTT) for 10 min and with 100 μM pervanadate for 1 h at 4°C after each wash with lysis buffer. The samples were analyzed by SDS-PAGE and immunoblotting with antibodies against conserved irreversibly oxidized protein tyrosine phosphatase (PTP) active site (R&D Systems).

Immunoblotting and ELISA.

For chemokine-induced protein phosphorylation, cells were lysed with ice-cold RIPA buffer (Cell Signaling Technology) containing 1% protease inhibitor cocktail (Sigma). The supernatant (10,000 rpm; 10 min; 4°C) was used for immunoblotting with antibodies against phospho-Akt (Ser 473; Cell Signaling Technology). Phosphorylation of Akt (Ser 473) was quantified by enzyme-linked immunosorbent assay (ELISA) (R&D Systems) according to the manufacturer's instructions. For immunoblotting analysis of membrane proteins, cells were lysed with extraction buffer (HEPES, pH 7.4, 250 mM sucrose, 1 mM EDTA, 1 mM EGTA, 1% protease inhibitor cocktail) with a Dounce homogenizer. The supernatant (10,000 rpm; 10 min; 4°C) was used for immunoblotting with antibodies against AQP3 (Abcam), Na⁺/K⁺-ATPase (Millipore), CD98 (Abcam), or Nox2 (Millipore).

Real-time quantitative PCR (qPCR) and RT-PCR.

The mRNA expression level of target genes was analyzed with total RNA using a TaKaRa one-step RNA PCR kit (TaKaRa Bio Inc.). For reverse transcription (RT)-PCR, total RNA was extracted using TRIzol (Invitrogen). The cDNA was reverse transcribed from total RNA using the Prime Script RT reagent kit (TaKaRa Bio). Quantitative RT-PCR was performed using SYBR green I (TaKaRa Bio) and the Light Cycler real-time PCR apparatus (Roche).

Coimmunoprecipitation of AQP3 and Nox2.

Association of AQP3 and Nox2 was examined by coimmunoprecipitation. Cells were homogenized in plasma membrane extraction buffer (HEPES, pH 7.4, 250 mM sucrose, 1 mM EDTA, 1 mM EGTA, 1% protease inhibitor cocktail) with a Dounce homogenizer. Cell lysates were centrifuged twice at 700 × g for 10 min to remove nuclei and unbroken cells. The collected supernatant was subjected to ultracentrifugation (100,000 × g; 1 h; 4°C). The plasma membrane-rich pellet was solubilized with ice-cold RIPA buffer and immunoprecipitated with protein A-Sepharose (GE Healthcare)-conjugated antibodies against AQP3 (Santa Cruz; C-18), Nox2 (Santa Cruz; C-15), or goat IgG (Santa Cruz) for 6 h at 4°C. The protein samples were applied to SDS-PAGE and transferred to a PVDF membrane for immunoblot analysis. The membrane was analyzed with antibodies against AQP3 (Sigma) and Nox2 (Millipore). The lysate of the plasma membrane-rich fractions was used for immunoblotting with Na⁺/K⁺ ATPase (plasma membrane marker; Millipore), EGFR (plasma membrane marker; Santa Cruz), Erk1/2 (cytoplasmic marker; CST), GM130 (Golgi apparatus marker; Abcam), or calnexin (endoplasmic reticulum [ER] marker; Enzo) to confirm the enrichment of plasma membrane in the obtained fraction.

In vivo cell migration assay.

Control RNAi- or vehicle-treated green fluorescent protein (GFP)-expressing cells were stained with CellTracker Blue CMF₂HC (Invitrogen) and mixed with nonstained AQP3 RNAi or AMD3100-treated GFP-expressing MDA-MB-231 cells (1-to-1 ratio), respectively. The cell mixture (8.0 × 10⁵ cells) was injected into the tail veins of female severe combined immunodeficient (SCID) mice (CB-17 ICR/scid; Crea, Japan). For AQP3 overexpression, a one-to-one mixture of empty (CellTracker green CMFDA; Invitrogen) and AQP3 (CellTracker green CMFDA and CellTracker blue CMF₂HC; Invitrogen) vector-transfected cells was injected into SCID mice. To examine the involvement of H₂O₂ in metastasis, the protocol described previously was performed with slight modifications (29). Briefly, water containing N-acetyl-l-cysteine (NAC) (10 mg/ml; Sigma) or regular water was supplied to female SCID mice for 4 days, and on the 5th day, GFP-expressing MDA-MB-231 cells were intravenously injected. For each experiment, after 24 h, the mice were sacrificed and their lungs were collected for analysis. Cell suspensions were subjected to flow cytometry on the FACS Fortessa system (BD Biosciences) to determine the number of MDA-MB-231 cells in the lungs. All animal experiments were approved by the Committee on Animal Research of Kyoto University.

Spontaneous metastasis.

MDA-MB-231 cells (1.0 × 10⁶ cells) infected with lentivirus carrying GFP-shRNA (nontargeting shRNA or AQP3 shRNA) were orthotopically injected into 7-week-old female SCID mice. Primary tumors were allowed to develop, and their sizes were measured every week. Tumor size was calculated by the following formula: volume = (width² × length)/2. To determine the number of MDA-MB-231 cells, tissues were minced and digested with a mixture of dispase (1 mg/ml; Roche) and liberase (50 μg/ml; Roche). The number of GFP-positive cells was analyzed by flow cytometry (FACS Fortessa; BD Biosciences). For AQP3 staining and visualization of macrophage infiltration, frozen or paraffin-embedded sections were stained with antibodies against AQP3 (Millipore), human EpCAM (eFluor 660; eBioscience), turbo-GFP (Origene), or F4/80 (allophycocyanin [APC]; Biolegend). Akt phosphorylation was monitored by LSM710 confocal microscopy (Carl Zeiss) with paraffin-embedded sections using antibodies against phospho-Akt (Ser 473; Cell Signaling Technology).

Statistical analysis.

Statistical analysis was performed using the two-tailed Student t test.

RESULTS

Transport of CXCL12-induced H₂O₂ into breast cancer cells through AQP3.

To investigate its role in breast cancer cells, AQP3 was knocked down by AQP3 RNAi (siRNA or lentiviral shRNA) in two different triple-negative breast cancer cell lines, MDA-MB-231 and DU4475 cells. Introduction of AQP3 RNAi into both breast cancer cell lines (AQP3 KD) consistently reduced AQP3 mRNA and protein expression compared to control RNAi-transfected cells (Fig. 1A and B; see Fig. S1A and B and S8A and H in the supplemental material). Osmotic water permeability was determined using the kinetics of scattered light intensity in response to osmotic challenge, as described previously (25). Osmotic water permeability in response to a 200 mM inwardly directed mannitol gradient was significantly decreased by AQP3 knockdown (see Fig. S1C in the supplemental material).

FIG 1.

Transport of CXCL12-induced H₂O₂ into breast cancer cells through AQP3. (A) The mRNA levels of AQP3 in control (si-control)- or AQP3-siRNA (si-AQP3)-transfected MDA-MB-231 (top) and DU4475 (bottom) cells and in lentiviral control (sh-control)- or AQP3-shRNA (sh-AQP3)-infected MDA-MB-231 cells (middle) were assessed by real-time PCR. The data are expressed as percentages of AQP3 expression relative to β-actin (MDA-MB-231) or glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (DU4475) expression of the control cells. The error bars indicate standard errors (SE) (n = 4 or 5; **, P < 0.01; *, P < 0.05). (B) Immunoblot analysis of plasma membrane-rich fraction with AQP3 and Na⁺/K⁺-ATPase antibodies in si-control or si-AQP3 of MDA-MB-231 (left) and DU4475 (right) cells and in sh-control or sh-AQP3 of MDA-MB-231 cells (middle). (C to F) MDA-MB-231 and DU4475 cells were transfected with control or AQP3 siRNA. (C) H₂O₂ uptake in control- or AQP3 RNAi (AQP3 KD)-transfected cells. The cells were incubated with H₂O₂ for 30 s, and cellular H₂O₂ was detected with H₂DCFDA. The data are expressed as percentages of the DCF fluorescence of vehicle-added control cells (n = 4 to 6; H₂O₂ added versus vehicle, **, P < 0.01, and *, P < 0.05; control versus AQP3 KD, ††, P < 0.01). (D) Intracellular H₂O₂ was monitored by H₂DCFDA following CXCL12 stimulation using a microplate reader (MDA-MB231) or by FACS analysis (DU4475). The cells were stimulated with CXCL12 for 30 s (n = 3 to 6; CXCL12 added versus vehicle, **, P < 0.01, and *, P < 0.05; control versus AQP3 KD, †† P < 0.01, and †, P < 0.05). (E) Intracellular H₂O₂ levels of MDA-MB-231 cells after treatment with DPI (20 μM; 30 min) or extracellular catalase (2,000 U/ml), followed by CXCL12 stimulation (100 ng/ml; 30 s). The data are expressed as percentages of the DCF fluorescence of cells with vehicle added (n = 4 to 6; **, P < 0.01). (F) Effect of Nox2 knockdown (si-Nox2) on cellular H₂O₂ levels. MDA-MB-231 cells were stimulated with CXCL12 (100 ng/ml) or H₂O₂ (100 μM) for 30 s (n = 6; stimulation versus vehicle, **, P < 0.01).

We next investigated whether AQP3 could transport extracellular H₂O₂ into MDA-MB-231 and DU4475 cells. Intracellular H₂O₂ was measured after the extracellular addition of H₂O₂, using the H₂O₂-reactive fluorescent dye H₂DCFDA, by a fluorescent microplate reader for attached MDA-MB-231 cells or by a flow cytometry FACS system for floating DU4475 cells. Intracellular H₂O₂ levels were increased 30 s after supplementation with H₂O₂ and were significantly higher in control cells than in AQP3 knockdown cells (Fig. 1C; see Fig. S1D in the supplemental material). The basal H₂O₂ levels determined by H₂DCFDA were almost identical between control and AQP3 knockdown in both MDA-MB-231 and DU4475 cells. Cellular H₂O₂ levels were also significantly increased in both MDA-MB-231 and DU4475 cells in response to CXCL12 and suppressed by AQP3 knockdown (Fig. 1D; see Fig. S1E in the supplemental material). Intracellular H₂O₂ levels of AQP3 knockdown MDA-MB-231 cells in response to 100 ng/ml CXCL12 were lower than those of control cells in response to any concentration of CXCL12, similar to DU4475 cells. We verified that AQP3 knockdown had no effect on the expression of CXCR4 (see Fig. S1F in the supplemental material). These results indicated the involvement of AQP3 in H₂O₂ transport in breast cancer cells.

NADPH oxidase (Nox1 to -5) is a major source of H₂O₂ in various cancer cells (30). Nox2 is expressed in breast cancer cells (see Fig. S2A in the supplemental material). Pretreatment with diphenyleneiodonium (DPI), a general Nox inhibitor, or incubation with catalase, which removes extracellular H₂O₂, significantly reduced the CXCL12-induced increase in intracellular H₂O₂ levels (Fig. 1E). This suggested that H₂O₂ was produced in the extracellular space via Nox in response to CXCL12. In addition, the knockdown of Nox2 by siRNA (see Fig. S2B to D and S8I in the supplemental material) significantly diminished the CXCL12-induced increase in H₂O₂ levels, whereas the inward permeability of H₂O₂ was unaffected (Fig. 1F). Taken together, our results provide evidence that Nox2-derived extracellular H₂O₂ produced in response to CXCL12 stimulation is rapidly transported into breast cancer cells through AQP3.

CXCL12-induced directional cell migration requires AQP3 and H₂O₂ uptake.

The role of AQP3 in CXCL12-induced cell migration was determined using a Transwell chamber. Migration toward CXCL12 was significantly impaired by AQP3 knockdown compared with control cells (Fig. 2A; see Fig. S3A in the supplemental material). CXCL12-stimulated control cells developed a polarized morphology with F-actin cytoskeletons, visualized with phalloidin staining, at the leading edge. In contrast, AQP3 knockdown cells failed to undergo morphological changes in response to CXCL12 (Fig. 2B, top). Immunofluorescence staining showed that both AQP3 and Nox2 were localized to the leading edge and the plasma membrane (Fig. 2B, bottom).

FIG 2.

CXCL12-induced directional migration requires AQP3 and H₂O₂ uptake. (A to F) MDA-MB-231 and DU4475 cells were transfected with control or AQP3 siRNA. The error bars indicate SE. (A) The chemotaxis efficiency of control or AQP3 KD MDA-MB-231 and DU4475 cells toward CXCL12 (100 ng/ml; 6 h) was examined with an 8-μm-pore-size Transwell chamber (n = 4; control versus AQP3 KD, **, P < 0.01). (B) (Top) Confocal microscopy analysis of visualized F-actin polymerization of control or AQP3 KD MDA-MB-231 cells with phalloidin-Alexa Fluor 488 after CXCL12 stimulation (500 ng/ml; 5 min). The arrows indicate the polymerization of F-actin at the leading edge. (Bottom) Coimmunostaining with anti-AQP3 and anti-Nox2 antibodies after vehicle or CXCL12 stimulation (500 ng/ml; 5 min). Scale bars, 10 μm. The boxed areas are enlarged on the right. The arrows indicate AQP3 and Nox2 localization at the leading edge. (C) Chemotaxis of MDA-MB-231 cells treated with DPI (20 μM; 30 min) or extracellular catalase (2,000 U/ml) toward CXCL12 (100 ng/ml; 6 h) (n = 4; mock versus DPI or catalase, **, P < 0.01, and *, P < 0.05). (D) Transendothelial migration of control or AQP3 KD MDA-MB-231 cells through HUVEC in the presence of CXCL12 (100 ng/ml; 24 h) (n = 3; control versus AQP3 KD, **, P < 0.01, and *, P < 0.05). (E) Two-dimensional (2D) chemotaxis assay of control or AQP3 KD MDA-MB-231 cells toward a CXCL12 gradient (500 ng/ml; 12.5 h; 3-min intervals) using a μ-slide chemotaxis chamber. A total of 60 cells were manually analyzed with NIH ImageJ software. (Left) Chemotaxis in response to a CXCL12 gradient was analyzed with rose plots and Rayleigh's test for vector data (P = 3.43e⁻⁶ for control; P = 0.22 for AQP3 KD). (Right) The chemotaxis index (FMIII, forward migration index parallel to the gradient) was measured from the endpoint of the migration distance parallel to the CXCL12 gradient (n = 60; **, P < 0.01). The horizontal lines indicate the average of the chemotaxis index. (F) (Top) Fluorescent imaging of H₂O₂ uptake at the leading edge in HyPer-Cyto-transfected control and AQP3 KD cells after CXCL12 stimulation (500 ng/ml; 5 min). Scale bars, 50 μm. The arrows indicate H₂O₂ uptake at the leading edge. (Bottom) Time-lapse imaging of cellular H₂O₂ at the leading edge in HyPer-Cyto-transfected control and AQP3 KD cells after CXCL12 stimulation (500 ng/ml; 6-s intervals). Scale bars, 10 μm. The results are representative of more than seven independent experiments performed on at least 6 independent cells.

To elucidate the link between CXCL12-generated H₂O₂ and cell migration, the effects of DPI and catalase on chemotaxis were determined. Treatment with either DPI or catalase significantly abrogated CXCL12-induced migration and F-actin polymerization at the leading edge (Fig. 2C; see Fig. S3B in the supplemental material). This supported the involvement of extracellular H₂O₂ produced by Nox activation in CXCL12-induced breast cancer cell migration. Cancer cell invasion through vascular endothelial cells is another essential step during metastasis (31). The transendothelial migration toward CXCL12 through human umbilical vein endothelial cells was also significantly attenuated by AQP3 knockdown (Fig. 2D).

The involvement of AQP3 in directional migration along a CXCL12 concentration gradient was examined using a μ-slide chemotaxis chamber. Control cells showed directional cell migration toward a higher CXCL12 concentration, whereas AQP3 knockdown cells showed dispersed migration (Fig. 2E, left). The chemotaxis index, which represents the efficiency of forward migration parallel to the CXCL12 gradient, was reduced significantly by AQP3 knockdown (Fig. 2E, right). In contrast, no difference was observed in migration velocities between control and AQP3 knockdown cells (data not shown).

We next tested the hypothesis that intracellular H₂O₂ transiently accumulated around the membrane through AQP3 during CXCL12-induced chemotaxis. A genetically encoded H₂O₂-specific fluorescence sensor, HyPer-Cyto (32), was used for time-lapse imaging of H₂O₂ uptake. Intracellular H₂O₂ dynamically accumulated at the leading edge in control cells during CXCL12-induced chemotaxis (Fig. 2F; see Fig. S3C and Video S1A in the supplemental material). In contrast, AQP3 knockdown cells lost their polarity even under the CXCL12 gradient, and increased intracellular H₂O₂ levels were not observed (Fig. 2F; see Video S1B in the supplemental material). Collectively, these results suggest that AQP3 is required for directional migration during CXCL12-induced chemotaxis and that it acts through H₂O₂ uptake around the leading edge.

AQP3-dependent oxidation of PTEN/PTP1B and phosphorylation of Akt.

We next sought to identify the targets of AQP3-mediated H₂O₂ in response to CXCL12-induced signaling. Because PTEN and PTP1B are sensitive to oxidation by H₂O₂ and subsequently affect the regulation of diverse cell-signaling pathways, such as the PI3K/Akt pathway, we assessed PTEN and/or PTP1B as potential intracellular targets of AQP3-transported H₂O₂ (26, 33). PTEN was oxidized in response to H₂O₂ or CXCL12 stimulation in control cells, whereas in AQP3 knockdown cells, the oxidation of PTEN was impaired (Fig. 3A). PTP1B oxidation was measured using IAA. IAA alkylates all cysteine residues of PTP1B except oxidized residues and inhibits further oxidation. The irreversibly oxidized active site of PTP1B was monitored with antibodies against peptide corresponding to the conserved PTP active site (27, 28). PTP1B was oxidized in response to CXCL12 in control cells but not in AQP3 knockdown cells, where an equal amount of PTP1B was immunoprecipitated (Fig. 3B). We next examined the activation of Akt, which controls cancer cell migration and metastasis (34). Akt phosphorylation at Ser 473 was dose-dependently induced by CXCL12 in control cells, whereas in AQP3 knockdown cells, Akt phosphorylation was suppressed, as assessed by immunoblotting (Fig. 3C; see Fig. S4A and S8B and J in the supplemental material) and ELISA (see Fig. S4B in the supplemental material). These data indicate that AQP3 is required for PTEN and PTP1B oxidation and subsequent Akt phosphorylation in CXCL12-stimulated breast cancer cells.

FIG 3.

AQP3-dependent PTEN/PTP1B oxidation and Akt phosphorylation. (A to E) MDA-MB-231 cells were infected with lentiviral control or AQP3 shRNA, and DU4475 cells were transfected with control or AQP3 siRNA. (A) Immunoblot analysis of PTEN oxidation. Control or AQP3 KD MDA-MB-231 cells were stimulated with CXCL12 (100 ng/ml) or H₂O₂ (100 μM) for 1 min. The cell lysates were analyzed using antibodies against PTEN under nonreducing (top) or reducing (bottom) conditions. (B) The oxidation of PTP1B in control or AQP3 KD MDA-MB-231 cells was analyzed by immunoblotting (IB). After CXCL12 stimulation (100 ng/ml; 1 min), PTP1B (with or without IAA) was immunoprecipitated (IP) and analyzed with antibodies against the oxidized PTP active site. Total immunoprecipitated PTP1B was determined with the antibody against PTP1B. (C) Akt phosphorylation (P-Akt) (Ser 473) in CXCL12-stimulated control or AQP3 KD MDA-MB-231 (0 to 100 ng/ml; 3 min) and DU4475 (0 to 50 ng/ml; 3 min) cells was assessed by immunoblotting. (D) Effect of transfection of PTEN RNAi or cotransfection of AQP3 and PTEN RNAi on Akt phosphorylation of MDA-MB-231 cells in response to CXCL12 (100 ng/ml; 3 min). (E) Effect of transfection of PTP1B RNAi or cotransfection of AQP3 and PTP1B RNAi on Akt phosphorylation of MDA-MB-231 cells in response to CXCL12 (100 ng/ml; 3 min).

RNAi-mediated PTEN and PTP1B knockdown was used to assess whether PTEN and/or PTP1B regulates Akt signaling in response to CXCL12 stimulation in breast cancer cells (see Fig. S5A to D and S8K and L in the supplemental material). Akt phosphorylation was elicited by either PTEN or PTP1B knockdown, even in the absence of CXCL12 stimulation (Fig. 3D and E; see Fig. S8C in the supplemental material). This suggests that the expression of PTEN and PTP1B inversely regulates the activation of Akt. Although AQP3 knockdown impaired CXCL12-induced Akt phosphorylation (Fig. 3C), the double knockdown of AQP3 and either PTEN or PTP1B resulted in constitutive Akt phosphorylation (Fig. 3D and E; see Fig. S8C in the supplemental material). This suggests that both PTEN and PTP1B are downstream of AQP3-mediated cell signaling.

CXCL12-induced H₂O₂ regulates PTP1B/PTEN oxidation and Akt activation.

Previous studies suggested that some stimuli, such as EGF, platelet-derived growth factor (PDGF), tumor necrosis factor alpha (TNF-α), and insulin, induce Nox-mediated H₂O₂, which may subsequently regulate redox signaling (22, 30). Treatment with DPI or catalase suppressed the CXCL12-induced oxidation of PTEN and PTP1B (Fig. 4A and B). Moreover, Akt phosphorylation in response to CXCL12 was also impaired by DPI or catalase treatment (Fig. 4C; see Fig. S8D in the supplemental material). These results suggest that Nox-derived H₂O₂ in response to CXCL12 is involved in the oxidation of PTEN/PTP1B and the phosphorylation of Akt in breast cancer cells.

FIG 4.

CXCL12-induced H₂O₂ regulates PTP1B/PTEN oxidation and Akt activation. (A to C, E, and F) MDA-MB-231 cells were infected with lentiviral control or AQP3 shRNA. (A and B) The effects of DPI (20 μM; 30 min) or catalase (2,000 U/ml) on PTEN (A) or PTP1B (B) oxidation induced by CXCL12 (100 ng/ml; 1 min) in MDA-MB-231 cells were analyzed by immunoblotting. (C) Akt phosphorylation (Ser 473) in DPI (20 μM; 30 min)- or catalase (2,000 U/ml)-treated MDA-MB-231 cells, followed by CXCL12 stimulation (100 ng/ml; 3 min). (D) MDA-MB-231 cells were transfected with control or AQP3 siRNA. Intracellular H₂O₂ levels were detected with H₂DCFDA in control or AQP3 KD MDA-MB-231 cells stimulated with CXCL12 (100 ng/ml) and/or H₂O₂ (100 μM) for 30 s, 1 min, and 3 min (the error bars indicate SE; n = 4 to 6; CXCL12 added, AQP3 KD at 3 min versus CXCL12 added, control or CXCL12 plus H₂O₂ added, AQP3 KD at 3 min, **, P < 0.01). (E) Effects of costimulation with CXCL12 and H₂O₂ on Akt phosphorylation in AQP3 KD MDA-MB-231 cells. The cells were stimulated with H₂O₂ (100 μM) and/or CXCL12 (100 ng/ml) for 3 min. (F) PTEN (top) and PTP1B (bottom) oxidation in AQP3 KD MDA-MB-231 cells stimulated with CXCL12 (100 ng/ml) and/or H₂O₂ (100 μM) for 1 or 3 min.

As further evidence for the involvement of AQP3-mediated intracellular H₂O₂ in CXCL12-induced cell signaling, the effect on cell signaling of cotreating AQP3 knockdown cells with H₂O₂ and CXCL12 was determined. The level of intracellular H₂O₂ in AQP3 knockdown cells induced by CXCL12 did not reach the level in control cells even with a longer incubation time (i.e., 3 min) (Fig. 4D). However, adding exogenous H₂O₂ together with CXCL12 for 3 min, but not for 1 min, caused cellular H₂O₂ levels in AQP3 knockdown cells to reach those of control cells treated with CXCL12 alone (Fig. 4D). This result suggests that exogenous H₂O₂ might slowly diffuse across the plasma membrane. In this context, Akt phosphorylation was restored in AQP3 knockdown cells, whereas adding either H₂O₂ or CXCL12 alone did not activate Akt (Fig. 4E; see Fig. S8E in the supplemental material). Impaired PTEN and PTP1B oxidation in AQP3 knockdown cells was also restored by adding exogenous H₂O₂ together with CXCL12 stimulation for 3 min (Fig. 4F). These results support the conclusion that AQP3-mediated H₂O₂ transport is involved in CXCL12-induced cell signaling.

Overexpressing AQP3 increases H₂O₂ uptake and cell migration upon CXCL12 stimulation.

We next examined whether overexpressing AQP3 could induce H₂O₂ uptake and cell migration in response to CXCL12. Transfection of a plasmid expressing human AQP3 upregulated the expression of AQP3 in mRNA and protein levels, as measured by quantitative real-time PCR and immunoblotting, respectively (Fig. 5A and B; see Fig. S8F in the supplemental material). Immunofluorescence showed that AQP3 was strongly expressed both in cytosol and around the plasma membrane in AQP3 vector-transfected cells (Fig. 5C). Following the addition of exogenous H₂O₂, intracellular H₂O₂ levels were significantly increased in AQP3-overexpressing cells compared to empty-vector-transfected cells (Fig. 5D). Intracellular H₂O₂ was further increased in response to CXCL12 by overexpressing AQP3 (Fig. 5E). These results indicate that increasing the expression of AQP3 enhances H₂O₂ uptake.

FIG 5.

AQP3 overexpression increases H₂O₂ uptake and cell migration upon CXCL12 stimulation. The error bars indicate SE. (A to F) MDA-MB-231 cells were transfected with the vector pCMV6 (empty [Emp]) or human AQP3-expressing pCMV6 (AQP3; 1 to 10 ng for 1.5 × 10⁴ cells). (A) The mRNA levels of AQP3 in empty-vector- or AQP3 vector-transfected cells were assessed by real-time PCR. The data are expressed as percentages of the AQP3 expression level relative to GAPDH expression of the empty-vector-transfected cells (n = 4; empty vector versus AQP3 vector, **, P < 0.01). (B) Immunoblot analysis of naive (NV) or empty-vector- or AQP3 vector-transfected MDA-MB-231 cells with anti-AQP3, anti-Na⁺/K⁺-ATPase, and anti-CD98 antibodies. (C) Confocal images of AQP3 in empty-vector- or AQP3 vector-transfected cells. The cells were stimulated with vehicle or CXCL12 (500 ng/ml) for 5 min. Scale bar, 10 μm. The arrows indicate the localization of AQP3 in the leading edge. (D) H₂O₂ uptake in empty-vector- or AQP3 vector-transfected cells. The cells were incubated with 100 μM H₂O₂ for 30 s, and cellular H₂O₂ was detected with H₂DCFDA by a fluorescence microplate reader. The data are expressed as percentages of DCF fluorescence of empty-vector-transfected cells (n = 6; empty vector versus AQP3 vector, **, P < 0.01). (E) Intracellular H₂O₂ of empty-vector- or AQP3 vector-transfected cells was monitored by H₂DCFDA following CXCL12 stimulation (100 ng/ml; 30 s) (n = 6; empty vector versus AQP3 vector, **, P < 0.01). (F) The chemotaxis efficiency of empty-vector- or AQP3 vector-transfected cells toward CXCL12 (100 ng/ml; 3 h) was examined using an 8-μm-pore-size Transwell chamber (n = 3; empty vector versus AQP3 vector, **, P < 0.01, and *, P < 0.05). (G) Akt phosphorylation (Ser 473) in empty-vector- or AQP3 vector (10 ng)-transfected cells after CXCL12 stimulation (0 to 100 ng/ml; 3 min) was assessed by immunoblotting. (H) (Left) Coimmunoprecipitation assay showing the presence of a complex between AQP3 and Nox2 in MDA-MB-231 cells. The plasma membrane-rich fraction (PM) was solubilized in RIPA buffer and immunoprecipitated with an AQP3 or Nox2 antibody. (Right) Immunoblot analysis was performed with PM- and RIPA-extracted fractions to confirm plasma membrane enrichment.

Chemotaxis in response to CXCL12 was significantly enhanced by overexpressing AQP3 (Fig. 5F). Given the results of both the AQP3 knockdown and overexpression experiments on CXCL12 stimulation, H₂O₂ uptake and cell migration efficacy were positively correlated and were both highly dependent on the expression level of AQP3 (see Fig. S6 in the supplemental material). Furthermore, AQP3 overexpression increased Akt activation at lower concentrations of CXCL12 (Fig. 5G; see Fig. S8G in the supplemental material), thereby supporting a role for AQP3-dependent H₂O₂ uptake in CXCL12-induced cell migration.

Based on the rapid increase in intracellular H₂O₂ upon Nox2 activation, we speculated that AQP3 and Nox2 might form a complex at the plasma membrane in breast cancer cells. Coimmunoprecipitation of AQP3 or Nox2 confirmed the association of AQP3 with Nox2 and vice versa (Fig. 5H), suggesting that the AQP3-Nox2 complex may allow rapid H₂O₂ flux across the plasma membrane.

Involvement of AQP3 in breast cancer cell migration and metastasis in vivo.

The hypothesis that AQP3-mediated H₂O₂ transport is involved in breast cancer cell migration during metastasis via CXCL12-CXCR4 signaling was tested in vivo using an experimental migration assay. Fluorescence-labeled breast cancer cells were intravenously injected into SCID mice as described previously (4). The number of control cells found in the lungs at 24 h postinjection was remarkably higher than the number of AQP3 knockdown cells (Fig. 6A). In contrast, overexpression of AQP3 significantly increased the number of cells that migrated to the lungs (Fig. 6B). The irreversible inhibition of CXCR4 with AMD3100, a highly selective CXCR4 antagonist (35), suppressed the metastatic activity toward the lung (Fig. 6C). Treating SCID mice with the antioxidant NAC markedly decreased the number of control cells that accumulated in the lungs, indicating that oxidants are involved in breast cancer cell migration in vivo (Fig. 6D).

FIG 6.

Involvement of AQP3 expression in breast cancer cell migration in vivo. The error bars indicate SE. (A to D) MDA-MB-231 cells (8.0 × 10⁵ cells) were intravenously injected into SCID mice, and the numbers of cells in the lungs after 24 h were analyzed by flow cytometry. (A) Control and AQP3 KD cells (turbo-GFP-expressing control or AQP3 shRNA) were injected (n = 9; **, P < 0.01). (B) Empty-vector- and AQP3 vector-transfected cells were injected (n = 11; **, P < 0.01). (C) Vehicle- and AMD3100-treated cells (10 μM; 30 min) were injected (n = 8; *, P < 0.05). (D) Effect of the H₂O₂ scavenger NAC on lung metastasis. Control cells (8.0 × 10⁵ cells) were intravenously injected into SCID mice supplemented with NAC (10 mg/ml, 4 days) or given regular water (n = 4 to 6; *, P < 0.05).

Finally, we investigated the potential role of AQP3 in metastasis using an in vivo spontaneous-metastasis model as described previously (4). GFP-expressing control or AQP3 knockdown cells were orthotopically implanted into the mammary fat pads of SCID mice. At 11 weeks posttransplantation, no differences in the tumor size and tumor cell number were found between control and AQP3 knockdown primary tumors (Fig. 7A and B, left). The growth of cultured cells was also unaffected by AQP3 knockdown (see Fig. S7A in the supplemental material). In contrast, FACS analysis and confocal microscopy observations showed that the number of metastatic cells in the lungs was significantly reduced by AQP3 knockdown (Fig. 7B, right, and C, top). Hematoxylin and eosin (H&E) staining showed an accumulation of cancer and noncancer cells in the lungs of mice implanted with control cells, with fewer extrinsic cells in the mice implanted with AQP3 knockdown cells (Fig. 7C, middle and bottom). Immunostaining showed substantial infiltration of F4/80⁺ macrophages in the lungs following implantation of control cells, which is consistent with a previous report (36). In contrast, fewer F4/80⁺ cells were detected in the lungs when AQP3 knockdown cells were transplanted (see Fig. S7B, top, in the supplemental material). We confirmed that there was no difference in macrophage infiltration between control and AQP3 knockdown primary tumors (see Fig. S7B, bottom, in the supplemental material). In addition, we verified that AQP3 knockdown had little effect on the expression of genes related to the epithelial-mesenchymal transition, angiogenesis, and immune cell attraction, which are known to be critical steps in metastasis (3, 37) (see Fig. S7C in the supplemental material).

FIG 7.

AQP3 knockdown abrogates spontaneous metastasis to the lungs. (A to F) Control or AQP3 KD MDA-MB-231 cells (turbo-GFP-expressing control or AQP3 shRNA; 1.0 × 10⁶ cells) were orthotopically injected into SCID mice. Primary tumors and lungs were analyzed at 11 weeks posttransplantation. (A) Tumor sizes were measured weekly for 11 weeks. (The error bars indicate SE; n = 9 or 10.) (B) The numbers of MDA-MB-231 cells in primary tumor (left) or lung (right) tissues were detected by flow cytometry (n = 9 to 11; *, P < 0.05). (C) (Top) GFP-expressing cells (green) were detected in the lungs by confocal microscopy. Scale bar, 50 μm. (Middle and bottom) H&E staining of the lungs. The boxed areas are enlarged on the bottom row. Scale bars, 100 μm (middle) and 20 μm (bottom). The arrows indicate the extrinsic cells in lung. (D and E) Immunofluorescence of AQP3 in primary tumor (D) and lung (E) tissues. Frozen sections were stained with antibodies against AQP3 (red) and human EpCAM (green). Scale bars, 20 μm (D) and 10 μm (E). (F) Immunofluorescence of phospho-Akt (Ser 473) (red) in primary tumor tissues. Scale bar, 50 μm. Enlarged images are shown in the insets.

Immunofluorescence showed that AQP3 was expressed in GFP-positive control cells in the implanted primary tumor (Fig. 7D) and in EpCAM⁺ cancer cells in the lungs (Fig. 7E; see Fig. S7D in the supplemental material). Additionally, phosphorylated Akt was detected by immunostaining in GFP-expressing control cells in the primary tumor, whereas fewer positive cells were observed with AQP3 knockdown (Fig. 7F). Taken together, our results indicate that AQP3 is required for breast cancer cell migration during metastasis in vivo.

DISCUSSION

We reveal a novel role for AQP3 in breast cancer cell migration that involves its H₂O₂ transport function and subsequent H₂O₂-mediated cell signaling. The transport of extracellular H₂O₂, generated in response to CXCL12, into breast cancer cells was dependent on AQP3. Subsequently, AQP3-mediated H₂O₂-dependent PTEN/PTP1B oxidation, followed by activation of the Akt pathway, promoted directional cell migration in response to CXCL12. The efficacy of CXCL12-induced cell migration correlated with AQP3-mediated H₂O₂ uptake in vitro. Studying the in vivo migration of breast cancer cells after intravenous injection revealed that the number of cancer cells that migrated to the lungs was dependent on the expression level of AQP3. This migration was found to occur via CXCL12/CXCR4- and H₂O₂-dependent pathways. The data obtained by orthotopic transplantation of breast cancer cells consistently demonstrated that spontaneous metastasis to the lungs was markedly impaired by AQP3 knockdown. Collectively, our findings provide compelling evidence that AQP3 is required for breast cancer cell migration by a mechanism involving its H₂O₂ transport function and downstream CXCL12/CXCR4-dependent cell signaling, namely, through PTEN, PTP1B, and Akt (see Fig. S9 in the supplemental material).

H₂O₂ plays an important role as a second messenger in biological systems (38). Previous studies have proposed that the transient accumulation of intracellular H₂O₂ around the membrane is essential for it to serve as a signaling molecule (39). Extracellular H₂O₂, which is mainly produced by plasma membrane Nox following its activation in response to diverse cell surface receptors, appears to diffuse readily across the plasma membrane (30, 40). However, passive diffusion of H₂O₂ across the lipid membrane may be limited (41). In addition, H₂O₂ is rapidly degraded in cells by various enzymes (42). Therefore, we surmised that high-capacity, active H₂O₂ influx across the membrane may be required for cell signaling. In the current study, we found that transient H₂O₂ accumulation at the leading edge and in uropods occurred in breast cancer cells during CXCL12-induced chemotaxis, as assessed by real-time cell imaging with the H₂O₂-specific protein HyPer-Cyto. We also found that AQP3 formed a complex with Nox2 in breast cancer cells, as we recently reported in epidermal keratinocytes (43). The localization of the AQP3-Nox2 complex at the leading edge in polarized breast cancer cells may allow efficient H₂O₂ transport to the location where cell signaling occurs during CXCL12 stimulation. AQP3 may thus be required for the rapid accumulation of intracellular H₂O₂ around the membrane and for H₂O₂ to exert its downstream cell-signaling effects.

The H₂O₂-mediated oxidation of PTPs, including PTEN and PTP1B, modulates diverse cell-signaling pathways, such as the PI3K/Akt pathway (23). Active-site cysteine residues of PTPs are deprotonated due to their lowered pK_a values (4.7∼5.4) and are highly reactive with H₂O₂ under physiological conditions (44). Thus, H₂O₂ selectively oxidizes these cysteine residues (Cys-SH) to sulfenic acid (Cys-OH), which simultaneously reacts to form disulfide linkages (33, 45). Disulfide linkages at PTEN or PTP1B active sites disrupt their phosphatase activities, increasing Akt phosphorylation (46). In the case of PTEN, Akt phosphorylation was also regulated via lipid phosphatase activity, which is inhibited by H₂O₂-mediated oxidation of cysteine residues (47, 48). Several stimuli, such as EGF, trigger H₂O₂-mediated Akt phosphorylation (49). However, for the CXCL12-CXCR4 axis, there is no evidence that H₂O₂-mediated oxidation is involved in Akt phosphorylation. Rather, previous studies showed that Akt phosphorylation was controlled by the direct activation of p110γ, a catalytic subunit of PI3K, by activation of the G protein receptor CXCR4 (50, 51). Here, we showed that Nox2-induced, AQP3-transported H₂O₂ was essential for both PTEN/PTP1B oxidation and Akt phosphorylation in response to stimulation by CXCL12. Based on previous studies, as well as the current work, we speculate that AQP3-mediated H₂O₂ accumulation in breast cancer cells stimulated with CXCL12 may boost the G protein activity of CXCR4 and directly activate the PI3K/Akt pathway via oxidizing and inhibiting PTEN and PTP1B. Because Akt phosphorylation regulates a variety of important cell functions, including cell survival, proliferation, and migration (34), the enhanced selectivity and efficiency of cell-signaling regulation through AQP3-mediated H₂O₂ is critical.

Increasing evidence shows that a correlation exists between AQP3 expression and cancer progression and prognosis, particularly in distant or lymphatic metastasis of colorectal, esophageal, hepatocellular, and gastric malignant cancers (11–14). In the current study, we showed that overexpressing AQP3 promoted the migration of breast cancer cells toward CXCL12 in association with higher intracellular H₂O₂ levels and enhanced susceptibility to Akt activation in vitro. The in vivo cell migration into the lungs was also elevated by AQP3 overexpression. Similarly, AQP3 knockdown suppressed breast cancer cell migration both in vitro and in vivo. These results support the potential importance of high AQP3 expression in cell migration during cancer progression. In contrast, we show here that the proliferation of MDA-MB-231 and DU4475 cells was unaffected by AQP3 knockdown. On the other hand, previous reports suggested the involvement of AQP3 expression in cell proliferation in various cancer cells, including adenocarcinoma and squamous cell carcinoma (16, 17). AQP3-mediated cell function may be dependent on the AQP3 expression level or on the expression of genes involved in cell migration or proliferation, such as growth factor receptor or chemokine receptor.

AQP3 is highly expressed in a large number of cancer cells, whereas lower levels of AQP3 expression are often seen in nontumorigenic or premalignant cells (52). However, how evolving tumor cells are reprogrammed to express high levels of AQP3 remains poorly understood. Previous reports have shown that the expression of AQP3 is increased by various stimuli, including EGF, TGF-β1, and phorbol esters (53–55). Cancer progenitor cells or stem cells might acquire higher AQP3 expression levels once various stimuli act as paracrine factors or through oncogenic transformation or mutation, subsequently driving tumor progression. Because cancer cells are exposed to hypoxic microenvironments, low-oxygen conditions or modulated hypoxia-inducible factor levels may serve as plausible mechanisms for AQP3 upregulation during cancer cell evolution. Although further studies are necessary to determine the relevance of AQP3 expression in breast cancer progression and the mechanism by which tumor cells acquire high AQP3 expression levels during evolution or progression, our findings provide new insights into therapies designed to inhibit AQP3 and limit cancer cell metastasis.

In summary, our data indicate a novel role for AQP3 in the migration of breast cancer cells, in which AQP3-mediated intracellular H₂O₂ uptake is required for CXCL12-induced cell signaling and migration. AQPs have been identified as potential targets for cancer therapy (8, 52); however, drug development targeting AQPs has not yet begun. Our findings support a role for AQP3 in regulating cancer progression, in addition to its known biological importance, and a new potential therapeutic target.

Funding Statement

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.