EVIDENCE AGAINST INVOLVEMENT OF AQUAPORIN-4 IN CELL-CELL ADHESION

Summary

Aquaporin-4 (AQP4) water channels are expressed strongly in glial cells, where they play a role in brain water balance, neuroexcitation and glial cell migration. Here, we investigated a proposed new role of AQP4 in facilitating cell-cell adhesion. Measurements were made in differentiated primary glial cell cultures from wildtype vs. AQP4 knockout mice, and in null vs. AQP4-transfected L-cells, a cell type lacking endogenous adhesion molecules, and in null vs. AQP4-transfected CHO-K1 cells and FRT cells. Using established assays of cell-cell adhesion, we found no significant effect of AQP4 expression on adhesion in each of the cell types. As a positive control, transfection with E-cadherin greatly increased cell-cell adhesion. High-level AQP4 expression also did not affect aggregation of plasma membrane vesicles in a sensitive quasi-elastic light scattering assay. Further, we found no specific AQP4 binding of a fluorescently labeled oligopeptide containing the putative adhesion sequence in the second extracellular loop of AQP4. These data provide evidence against involvement of AQP4 in cell-cell adhesion.

Introduction

Aquaporin-4 (AQP4) is a water-selective transport protein expressed in plasma membranes in glial cells throughout the central nervous system. At least three distinct roles of AQP4 in brain physiology have been demonstrated ¹. AQP4 facilitates water movement into and out of the brain in a variety of pathological conditions such as water intoxication, stroke, infection, tumor and hydrocephalus ^2–5, by facilitating osmotically driven water movement across barriers separating brain parenchyma and the blood and cerebrospinal fluid compartments. AQP4 increases the migration of glial cells in culture and in vivo models ^{6, 7}, by a mechanism that likely involves facilitated water influx into lamellipodia at the leading edge of migrating cells ⁸. AQP4 also modulates neuroexcitation phenomena, including seizure activity, cortical spreading depression, and visual, auditory and olfactory signal transduction ^9–13, by a mechanism that may involve altered extracellular space volume and potassium buffering ^{11, 14}.

Recently, an interesting new role of AQP4 in facilitating cell-cell adhesion was proposed, based on structural data from 2-dimensional crystals in which an extracellular helix-helix interaction (at residues 139-142) was predicted ¹⁵. The authors also reported increased clustering of L-cells following AQP4 transfection. Another study showed increased cell adhesion to various substrates when AQP4 was transfected into glioma cell lines ¹⁶. There is also evidence that a related aquaporin, the major intrinsic protein of lens fiber (also called AQP0) facilitates adhesion of lens fiber cells in the eye ^{17, 18}. As reviewed recently ¹⁹, AQP4-dependent cell-cell adhesion could be important in brain development, blood-brain barrier integrity, and osmosensing. However, arguing against this speculation is the absence of such abnormalities in transgenic mice lacking AQP4 ¹.

Here, we tested the potential role of AQP4 in cell-cell adhesion. We used primary glial cell cultures from wildtype vs. AQP4 knockout mice, null vs. AQP4-transfected L-cells, which lack endogenous adhesion molecules, and two other cell types in which transfection produces high levels of AQP4 expression. Utilizing established and novel assays of cell-cell adhesion, we found no evidence for AQP4-facilitated adhesion. Our conclusion that AQP4 residues 139–142 is not involved in adhesion was further supported by binding measurements of a peptide containing the putative adhesion sequence.

Results

Initial adhesion studies were done using L-cells, which lack endogenous adhesion molecules such as cadherin or N-CAM. Some experiments were done on stably transfected L-cells expressing AQP4. As shown in Fig. 1A, AQP4 was expressed in a plasma membrane pattern in the transfected cells with uniform cell-to-cell expression. Experiments were also done in control (non-transfected) L-cells, and in L-cells that were transiently transfected with AQP4 or E-cadherin (positive aggregation control ²⁰). Transient transfection produced high level AQP4 expression in ~60% of cells.

Figure 1. AQP4 expression in L-cells does not increase cell-cell adhesion.

A. Immunoblot (left) and immunofluorescence (right) of non-transfected and AQP4 stably transfected L-cells. B. Cell aggregation activity. Self-adhesion of L-cells in single cell suspension measured by counting the number of particles (isolated cells + aggregates) at 0, 30 and 60 min (S.E. n=6 independent experiments). N_t/N_o is the ratio of number of particles at time t vs. time 0. Differences not significant between non-transfected and AQP4-expressing cells. Transient transfection with E-cadherin induced significant aggregation (* P < 0.01) compared to non-transfected cells. C. L-cell adhesion assessed by phase-contrast microscopy in which cell suspensions were allowed to settle onto coverglass after terminating aggregation by glutaraldehyde. Few non-transfected and AQP4-expressing cells were adherent, whereas many E-cadherin-transfected cells formed large aggregates. Bar, 100 μm.

Cell aggregation was measured using a standard adhesion assay in which a cell suspension was stirred for specific times, and the number of particles (cells + cell aggregates) counted. Increased adhesion is seen as a reduction over time in the total number of particles. Fig. 1B shows the number of L-cell particles at 30 and 60 min. There was a small reduction in particle counts with time, though not different between non-transfected, stably transfected and transiently AQP4 transfected cells. Similar results were found in 3 sets of measurements on separate cell cultures. In the positive control, greatly increased adhesion (reduced particle count) was seen after L-cell transfection with E-cadherin. Fig. 1C shows light micrographs of the L-cells after glutaraldehyde (to freeze aggregation state) and settling onto a coverglass. Little aggregation was seen in non-transfected and AQP4-expressing L-cells, whereas many aggregates were seen in the E-cadherin-expressing L-cells. Similar aggregation studies done twice using control vs. AQP4 stably expressing CHO and FRT cells (generated previously as described ^{21, 22}) also showed no AQP4-dependent aggregation (data not shown).

AQP4 is expressed highly in brain glial cells, the most relevant cell type where AQP4-dependent cell adhesion might occur. Similar cell-cell adhesion measurements were done on cultured, differentiated glial cells having high AQP4 expression (Fig. 2A) and osmotic water permeability ²³. More than 90% of the cultured glial cells were GFAP positive and showed the characteristic stellate morphology. Particles (cells + cell aggregates) were counted following suspension in standard HBSS and Ca²⁺/Mg²⁺-free HBSS containing 10 mM EDTA. Fig. 2B shows greater baseline adhesion of the astrocytes than that seen for L-cells, though no difference in cultures from wildtype vs. AQP4 null mice. Elimination of calcium reduced adhesion, but still no difference was seen in cultures from wildtype vs. AQP4 null mice. Fig. 2C shows micrographs of glial cells after settling onto a coverglass. Though distinct cell aggregates were seen, their numbers and appearance were similar in the cultures from wildtype and AQP4 null mice.

Figure 2. AQP4–independent adhesion in brain glial cells.

A. Immunofluorescence showing ~95% of GFAP-positive cells with membrane AQP4 staining in wildtype glial cells. B. Cell-cell adhesion measured as in Fig. 1 for differentiated primary cultures of glial cells from neonatal brains of wildtype and AQP4 null mice (S.E. n=6 independent experiments). Suspensions contained standard HBSS (+ Ca⁺⁺) or Ca²⁺/Mg²⁺-free HBSS containing 10 mM EDTA (− Ca⁺⁺). Differences not significant between wildtype vs. AQP4-deficient glial cells. C. Phase-contrast micrographs of glial cells after settling on culture plate as in Fig. 1C. Arrows indicate cell aggregates. Bar, 100 μm.

As another sensitive approach to detect adhesion of AQP4-expressing cells, we imaged glial cell aggregates following fluorescent labeling, with cells from wildtype mice labeled red and from AQP4 null mice labeled green. We previously established, for in vivo cell migration studies, fluorescence labeling procedures to stain cells brightly without effect on cell function ⁶. We reasoned that cell aggregates, after mixing equal numbers of red and green cells, would be enriched in red cells if AQP4 enhanced cell-cell adhesion. Fig. 3A shows representative low magnification phase-contrast and fluorescence micrographs of cells/aggregates after a 30 min incubation in stirred suspension, showing comparable numbers of red and green cells, with both colors seen in aggregates. At high magnification the aggregates were seen to contain both red and green cells in an apparently random clustering (Fig. 3B, inset). To quantify the results, the numbers of red and green cells were counted in clusters containing at least 6 distinct cells. As summarized in Fig. 3B by a histogram of red/green cell counts, a statistical distribution as found, as expected, which was centered at a red/green ratio of unity for both cell types. These results indicate random, AQP4-independent glial cell adhesion. Similar results were found at 60 min (not shown).

Figure 3. Two-color fluorescence clustering assay of glial cell adhesion.

Glial cells were fluorescently stained red (wildtype) and green (AQP4 null). Cell suspensions contained equal numbers of wildtype and AQP4 null cells. A. Representative low magnification phase-contrast (left) and fluorescence (right) micrographs of cells after 30 min in stirred suspension. Bar, 50 μm. B. Histogram of the ratio of red-to-green cells in aggregates, with a total of 39 aggregates counted (range 6 to 24 cells per aggregate). Inset shows example of aggregate at high magnification.

We also studied binding to AQP4 of the putative AQP4 adhesion sequence. Differentiated glial cells or stably transfected AQP4-expressing CHO-K1 cells ²² were used. These cell types strongly express AQP4 in orthogonal arrays of particles ^{22, 24}. Binding was measured by quantitative imaging of confluent cell layers following incubation with a fluorescein-labeled 9-mer peptide containing extracellular loop residues of AQP4 proposed to mediate cell-cell adhesion (Fig. 4A, top). Inclusion of an excess of unlabeled peptide was used to detect specific binding. Fig. 4A (bottom) shows fluorescence micrographs of glial cells after incubations with fluorescent peptide in the absence and presence of a 50-fold excess of unlabeled peptide. Cells were then briefly fixed and washed prior to imaging. Though incubation with peptide increased fluorescence above background, fluorescence was similar in the cells from wildtype and AQP4 null mice, and when excess unlabeled peptide was present. Fig. 4C summarizes results from a series of measurements done at different concentrations of the fluorescent peptide in glial and CHO cells. No specific peptide binding was found based on similar fluorescence without vs. AQP4 at each concentration of fluorescent peptide, and by the lack of competition with a large excess of unlabeled peptide. These data show no specific association of the putative AQP4 adhesion sequence with the external-facing surface of AQP4 in cell membranes.

Figure 4. Lack of specific AQP4 binding of a putative AQP4 adhesion oligopeptide.

A. (top) Schematic of AQP4 membrane topology showing putative adhesion sequence (residues 139-142) in extracellular loop. The synthetic fluorescein-labeled 9-mer peptide (NH₂-PPSVVGGLG-COOH) is shown. (bottom) Fluorescence micrographs of confluent astrocytes after incubation with indicated concentration of fluorescein-labeled 9-mer peptide for 30 min at room temperature, followed by brief fixation with 2% glutaraldehyde and washing. C. Summary of relative (background-subtracted) cell fluorescence for incubations done at indicated concentrations of the fluorescein-labeled 9-mer peptide, in the absence or presence of 50-fold excess unlabeled peptide (S.E., n=3–4). Differences not significant comparing AQP4 expressing vs. non-expressing glial and CHO cells. +/+: AQP4-expressing cells, −/− AQP4 null cells.

Last, as potentially the most sensitive test of AQP4-facilitated membrane adhesion, we measured aggregation of small plasma membrane vesicles (diameter ~ 200 nm) that strongly expressed AQP4 vs. control, non-expressing vesicles. Vesicles were prepared by homogenization and density gradient centrifugation of control and AQP4-expressing CHO cells, as done previously ²². After shaking in standard HBSS overnight at high density, ~5 % of particles were aggregates with diameters greater than 2000 nm, for both control and AQP4-expressing vesicles. As positive controls, nearly all vesicles were contained in aggregates when a low-ionic strength sucrose buffer was used in place of HBSS during the incubation.

Discussion

Our data provide evidence against the involvement of AQP4 in cell-cell adhesion and the proposal that AQP4 serves as an ‘adhennel’ ¹⁹ with dual transport and adhesion roles. We found by multiple assays no evidence for increased adhesion/aggregation in cells expressing AQP4 than in the same cell types not expressing AQP4, including the most relevant cell type, brain glial cells. Also, specific AQP4 binding was not seen of a peptide corresponding to the second extracellular domain of AQP4 with its putative adhesion sequence. As a negative study, however, we cannot exclude the possibility that under some conditions extracellular AQP4-AQP4 interactions can occur, though we are doubtful in part because of the absence of any of the predicted adhesion defects in AQP4-deficient mice.

The evidence for possible involvement of AQP4 in cell-cell adhesion came from structural analysis of double-layered 2-dimensional crystals of recombinant AQP4 ¹⁵. Notwithstanding the limited structural resolution (3.2 Angstroms) and the variable orientations and distances between the two layers, it was found that apposing AQP4 monomers interacted at residues 139 and 142, located in their second extracellular loops. However, unlike the precise stacking of AQP0 in crystals ^{17, 18}, the AQP4 crystals were offset such that tetramers in one membrane line overlay the center of tetramers in the apposing membrane. The crystallography data provide no information about the strength of the AQP4-AQP4 interaction or whether it occurs in real hydrated membranes or cells. If such interactions do occur in cells, it was speculated that the interactions might serve to block AQP4 water permeability and that rapid water flow might force interacting membranes apart.

Different cell types utilize different adhesion mechanisms. Well-characterized cell adhesion molecules include cadherins, immunoglobulin (Ig) superfamily molecules, and integrins, which perform a wide range of functions at cell contacts such as embryogenesis, immune response, hemostasis, inflammation, and maintenance of tissue integrity (reviewed in refs. 25 and 26). Cadherins are single-pass transmembrane glycoproteins with calcium-dependent adhesion, whereas adhesion in Ig superfamily molecules is calcium-independent. Cadherins and Ig molecules can form homophilic adhesion, whereas integrins are heterodimeric transmembrane glycoproteins that interact mainly with the extracellular matrix ²⁷. The mouse fibroblast L-cell is a good model to study cell adhesion because it expresses few if any adhesion molecules and shows little aggregation in standard cell aggregation assays ^28–30. We did not find increased aggregation of L-cells upon transient or stable AQP4 transfection, whereas E-cadherin expression increased aggregation.

From the crystal structure data, it was suggested each AQP4 tetramer interacts with four tetramers in the opposing membrane, such that AQP4 orthogonal arrays might enhance AQP4-facilitated adhesion ¹⁹. The various cell types studied here express AQP4 in square, orthogonal arrays ^{22, 31}. Various aggregation assays on these cells, including a highly sensitive vesicle aggregation assay, showed no AQP4 dependence.

Because AQP4 is expressed mainly in brain glial cells, we studied the possibility of AQP4-dependent aggregation in cultured, differentiated glial cells. Glial cells are known to express several adhesion molecules, including the Ca²⁺-dependent adhesion molecule N-cadherin ³², and the Ca²⁺-independent Ig superfamily molecule nectin-1 ²⁵. Our data confirmed Ca²⁺-dependent and independent adhesion of cultured glial cells. However, no differences in aggregation were found, using two different assays, in glial cells cultured from brains of wildtype vs. AQP4 null mice. These data support the conclusion that molecules other than AQP4 are responsible for glial cell-cell adhesion.

Finally, in peptide competition studies we were unable to detect specific AQP4 binding of the putative 3¹⁰ α-helix containing AQP4 residues 139–142. These results suggest that residues in the synthetic peptide have little or no affinity to the extracellular AQP4 surface. In a study of AQP4 membrane diffusion reported recently ³³, we studied a rAQP4M23 construct containing a c-myc between residues 139 and 142, which serendipitously destroys the 3¹⁰ α-helix. The cellular processing, water permeability and ability to form orthogonal arrays of this construct were not different from those of native (untagged) AQP4, providing further support for the conclusions here.

In summary, our data do not support the involvement of AQP4 in cell-cell adhesion, which would have ascribed two completely unrelated roles to AQP4. The well-documented water transporting role of AQP4 probably accounts for its various functions in brain water transport, glial cell migration, and neuroexcitation.

Materials and Methods

Cells and tranfection

Mouse fibroblast L-cells (ATCC CCL-1) were cultured in MEM supplemented with 10% FBS, 100 U/ml penicillin and 100 μg/ml streptomycin under 5% CO₂ at 37 °C. L-cells were transfected with rat AQP4 (rAQP4M23) in plasmid pcDNA3.1 (generated as described previously ²²) using Lipofectamine 2000 (Invitrogen) according to manufacturer’s instructions. Stable transformants were selected with 500 μg/mL G418 (Invitrogen). Resistant colonies were tested for the expression and cloned. For transient transfections, L-cells were plating onto 6-well plates and transfected with Lipofectamine 2000. Experiments were done 48 hr after transfection. Mouse E-cadherin cDNA in pSTEM1 was provided by Dr. Masatoshi Takeichi (Riken Institute, Japan). Stably transfected CHO-K1 cells (ATCC CCL-61) expressing AQP4 ^{22, 31} were grown in Ham’s nutrient mix supplemented with 10% fetal calf serum at 37 °C in 5% CO₂. The clonal cell populations expressed AQP4 protein strongly in their plasma membranes and were highly water permeable. Astroglia were cultured from cortex of wildtype and AQP4 null neonatal mice as described ²³. At confluence (day 10), primary astroglia were shaken at 200 rpm for 18 h at 37 °C to remove microglia ³⁴, and 0.15 mM dibutyryl cAMP was added to induce differentiation.

Immunocytochemistry

Cells were fixed with 4% paraformaldehyde for 15 min and then washed with PBS. Cells were blocked with 1% bovine serum albumin (BSA), and incubated with primary 1:200 rabbit anti-AQP4 antibody (Ab) (Millipore) or 1:500 mouse anti-GFAP (Sigma) for 1 hr at room temperature and detected with Texas Red-conjugated goat anti-rabbit or FITC-conjugated goat anti-mouse secondary antibodies (1:200 for both; Jackson ImmunoResearch Laboratories, Inc.). Nuclei were counterstained blue with 4,6-diamidino-2-phenylindole (DAPI). Sections were visualized on an inverted fluorescence microscope equipped with a high-resolution Spot color cooled CCD camera (Diagnostic Instruments, Sterling Heights, MI).

Cell aggregation assay

An established cell aggregation assay was used as described ^{28, 29}. Briefly, to obtain a single cell suspension, cells were incubated with 0.25% trypsin and 1 mM EDTA at 37 °C for 5 min, and dispersed by gentle pipetting. The cells were then suspended in Hanks balanced salt solution (10⁶ cells/ml), placed in 24-well plates pre-coated with 2% BSA, and rotated on a orbital shaker (80 rpm) at 37 °C for specified periods of time. An equal volume of 2% glutaraldehyde was added to prevent further aggregation or dissociation of aggregates ³⁰. The extent of aggregation was determined from the ratio of the total particle number at time t of incubation (N_t) to the initial particle number (N₀). Particle number was determined by counting representative aliquots on a hemocytometer.

In a related assay, AQP4 wildtype and null astrocytes were separately labeled with CellTracker™ Red CMTPX and CellTracker™ Green CMFDA (Invitrogen). Cells were incubated with dyes (2 μM) in Opti-MEM for 30 min, washed 2 times with Opti-MEM, incubated for 1 h in DMEM, and suspended as above in HBSS. Cells were mixed at a 1:1 ratio and incubated as described above. Green and red fluorescence images were obtained of particles/aggregates after deposition on a coverglass using a Nikon 2000 epifluorescence microscope. In some experiments, aggregation was measured in Ca²⁺/Mg²⁺-free HBSS containing 10 mM EDTA.

Peptide association studies

Astrocytes or CHO cells were seeded on the BD Falcon™ 8-well CultureSlides (Fisher Scientific, Pittsburgh, PA) until confluence. A peptide (NH₂- PPSVVGGLG-COOH) containing amino acid residues 139-142 of AQP4 was synthesized with or without an N-terminus fluorescein label (Sigma Genosys). 100 μL of HBSS containing different concentrations of the fluorescent peptide (1–500 μM) was added in each well and incubated at room temperature for 30 min, 2% glutaraldehyde was then added to prevent dissociation of bound peptide, and cells were washed with PBS. Experiments were also done with a large excess of unlabeled peptide included with the fluorescent peptide. Cell-associated fluorescence was determined by quantitative image analysis.

Vesicle aggregation assay

Plasma membrane vesicles were isolated from control and AQP4-expressing CHO-K1 cells by sucrose gradient centrifugation as described ²². Cells from five 25-cm diameter plastic dishes were grown to confluence (1–3 x 10⁸ cells/dish), released by agitation in PBS containing 10 mM EGTA, and washed twice (100 x g, 10 min) at 4 °C in homogenizing buffer (HB, 250 mM sucrose, 10 mM Tris-HCl, pH 7.4). The pellet was resuspended in HB containing antipain (1 μg/ml), pepstatin (1 μg/ml) and benzamidine (15 μg/ml), and homogenized by 20 strokes of a glass Dounce homogenizer. The homogenate was centrifuged at 500 x g for 10 min at 4 °C and adjusted to 1.4 M sucrose, 10 mM Tris-HCl, 0.2 mM EDTA (pH 7.4). A discontinuous sucrose gradient (2 M sucrose (2.5 ml), 1.6 M (5 ml), 1.4 M (10 ml, containing homogenate), 1.2 M (10 ml), 0.8 M (2.5 ml)) was centrifuged for 2.5 h at 25,000 rpm in an SW 28 rotor, and 5 ml fractions were collected. After centrifugation (25,000 rpm, 30 min), the plasma membrane vesicles were resuspended using a 30-gauge needle in 100 μL HB buffer or HBSS, and shaken overnight at 4 °C. Aggregation was stopped by addition of an equal volume of 2% glutaraldehyde. Vesicle diameter was determined by quasi-elastic light scattering using a Zetasizer (Malvern Instruments, Inc., Southborough, MA). The intensity of the laser light scattered by the samples was detected at an angle of 90°. Vesicle size distributions were determined by autocorrelation analysis using Malvern software.

Immunoblot analysis

L-cells were harvested and homogenized in 250 mM sucrose, 1 mM EDTA, 2 μg/ml aprotinin, 2 μg/ml pepstatin A, 2 μg/ml leupeptin, and 100 μg/ml serine protease inhibitor (Sigma). Homogenates were centrifuged at 500 x g for 5 min and the supernatant was collected and stored at −80 °C. Protein concentration was measured using the Protein Assay Kit II (Bio-Rad Laboratories, Hercules, CA). Proteins were electrotransferred to a polyvinylidene difluoride membrane and incubated with rabbit anti-AQP4 (1:1000, Chemicon) or anti-β-actin antibody (1:2000, GE Healthcare, Piscataway, NJ) followed by horseradish peroxidase-linked anti-rabbit IgG (1:10.000, GE Healthcare), and visualized by enhanced chemiluminescence (Roche Diagnostics, Indianapolis, IN).

Figure 5. Quasi-elastic light scattering assay of plasma membrane vesicle aggregation.

Plasma membrane vesicles isolated from AQP4 expressing and non-expressing CHO-K1 cells were incubated at high density overnight in HBSS. Cell size distributions were measured by quasi-elastic light scattering. Differences not significant in the fraction of aggregates of AQP4-containing vs. control vesicles. As positive control, vesicles were incubated in a low-ionic strength isosmolar sucrose solution to induce aggregation.