POST-GOLGI SUPRAMOLECULAR ASSEMBLY OF AQUAPORIN-4 IN ORTHOGONAL ARRAYS

AQP4 forms OAPs in plasma membrane but not in Golgi or ER

To determine the cellular compartment of OAP formation, the AQP4 oligomeric state was analyzed in membrane fractions purified from homogenates of human brain. Fractions enriched in plasma membranes, ER and Golgi were prepared by sucrose-density gradient centrifugation. By BN-SDS/PAGE and AQP4 immunoblot, AQP4 was detected as high molecular oligomer in the plasma membrane fraction, corresponding to OAPs, whereas only AQP4 tetramers were detected in the ER and Golgi fractions (Figure 1A, left). CX43 was used as a control because of its abundant expression in brain and formation of well-characterized oligomeric structures, with monomers in the ER, hexamers (connexons) in Golgi, and high molecular structures (GAPs) in the plasma membrane (8, 22) (Figure 1A, center). Fraction purity was verified by immunoblot analysis (Figure 1A, right) using as markers Na⁺/K⁺ ATPase (plasma membrane), TGN-46 (Golgi) and calnexin (ER).

Studies were also done in U87MG cells after transient transfection with M23-AQP4 or CX43 (Figure 1B) in which plasma membrane and ER/Golgi fractions were isolated by differential centrifugation. BN/PAGE showed multiple bands at high molecular weight in the plasma membrane fraction, corresponding to AQP4 OAPs, and a single band at ~300 kDa in the Golgi/ER fraction, corresponding to AQP4 tetramers. CX43 native gels showed a single strong band at high molecular weight in the plasma membrane fraction, corresponding to GAPs, and two bands in the Golgi/ER fraction, corresponding to monomers and hexamers. Membrane markers confirmed fraction purity (Figure 1B, right).

FFEM showed OAPs in the plasma membrane fraction of M23-transfected cells but not M1-transfected cells (Figure 1C, top and center). The OAP density in the M23-transfected cells was ~1 per 500 μm². OAPs were never seen in freeze-fracture micrographs of Golgi (Figure 1C, bottom) for eight different specimens in which a total of area of 11500 μm² of Golgi membrane was examined. The FFEM data is in agreement with the conclusion from biochemical analysis that AQP4 aggregates/OAPs are absent in Golgi.

Post-Golgi assembly of AQP4 OAPs was verified in live cells by fluorescence recovery after photobleaching (FRAP). We previously generated and characterized C-terminal fusions of GFP to AQP4 (23). Figure 2A top, left shows the expected molecular sizes of GFP-tagged M1 and M23 by AQP4 immunoblot. BN/PAGE in Figure 2A, top, center shows OAP formation by M23-GFP but not by M1-GFP, as expected, which was confirmed by high-resolution clear native-polyacrylamide gel electrophoresis (Figure 2A, right). Confocal imaging of M1-GFP and M23-GFP showed plasma membrane staining by both isoforms (Figure 2A, bottom, upper panels), with M1-GFP seen in a smooth pattern and M23-GFP in a punctate pattern, reflecting their different oligomeric states on the plasma membrane. TIRFM confirmed OAP formation in the plasma membrane by M23-GFP but not M1-GFP (Figure 2A, bottom, lower panels).

To visualize AQP4 in subcellular compartments in live cells, we used BFA to block protein traffic from the ER, and low temperature to block protein exit from Golgi. Low temperature conditions were optimized to give near perfect M1 and M23 accumulation in Golgi, as confirmed by colocalization with mCherry-labeled Gal-T (Figure 2B, top). BFA treatment resulted in ER retention of M1 and M23, as confirmed by colocalization with mCherry-labeled calnexin (Figure 2B, bottom).

We previously exploited the differential diffusion dynamics of M1 vs. M23-AQP4 in the plasma membrane, as measured by single particle tracking (SPT), to characterize the determinants of OAP formation (24). Here, we used FRAP with confocal image detection to compare diffusion of M1 vs. M23-AQP4 in the Golgi, ER and plasma membrane. Figure 3A, top shows that photobleaching of plasma membrane (white arrows) produced a dark region of fluorescence that fully recovered for M1-GFP but not for M23-GFP, in agreement with OAP formation by M23-GFP but not M1-GFP at the plasma membrane. Figure 3A, bottom shows complete recovery of fluorescence in Golgi for both M1-GFP and M23-GFP. Complete recovery of fluorescence was also found for both M23-GFP and M1-GFP in the ER (not shown). Figure 3B, top shows representative recovery curves as background-subtracted fluorescence in the bleached spot, with percentage recovery summarized in Figure 3B, bottom. Fluorescence recovery was essentially complete over 2 min for M1-GFP and M23-GFP in Golgi and ER, and for M1-GFP but not M23-GFP in the plasma membrane. These results obtained in live cells support the biochemical and freeze-fracture data in Figure 1 indicating OAP assembly by AQP4 in the plasma membrane but not in Golgi.

OAP formation by AQP4 is pH-insensitive

We investigated whether various differences in Golgi vs. plasma membrane could account for OAP formation in the plasma membrane but not in Golgi. Recognizing that Golgi pH is relatively acidic compared to cytoplasm, possible pH-dependent OAP assembly was studied by incubation of plasma membrane vesicles from M23-AQP4 expressing cells in buffers of different pH in the presence of digitonin to permeabilize the membrane. BN/PAGE showed, however, that AQP4 supramolecular assembly was not sensitive to pH in the wide pH range 3–7.4 (Figure 4A).

High membrane curvature disrupts AQP4 OAPs

Proteins in the secretory pathway are transported from the ER to the plasma membrane through the Golgi compartment in vesicles of different sizes. Because AQP4 assembly in OAPs involves the formation of crystalline planar arrays, we postulated that high membrane curvature might disrupt OAPs. To investigate this possibility, M23-containing plasma membrane vesicles were extruded through filters of different pore sizes to generate vesicles of average diameters 57, 113, 223 and 355 nm. Remarkably, BN/PAGE of vesicles of diameters 57 and 113 nm showed a single band at ~300 kDa, corresponding to AQP4 tetramers, whereas multiple bands were seen in vesicles of diameters 223 and 335 nm, corresponding to AQP4 OAPs (Figure 4B, left). In another set of experiments, 405 nm M23-containing plasma membrane vesicles were extruded to yield 61 nm vesicles, some of which were fused together using a high-Ca²⁺ buffer (Figure 4B, right). BN/PAGE showed a single band (AQP4 tetramers) for the extruded vesicles, but multiple bands corresponding to OAPs in the original (non-extruded) vesicles and in the fused, extruded vesicles. The disruption of AQP4 OAPs by vesicle extrusion to small diameter was confirmed by FFEM (Figure 4C). Figure 4D shows a schematic of two adjacent AQP4 tetramers in 350 vs. 50 nm diameter vesicles. In order remain aligned in a parallel orientation, adjacent AQP4 tetramers would need to tilt minimally, by 2.6°, in a 350 nm diameter vesicle, but by 22.6° in a 50 nm vesicle, which is likely to be highly energetically unfavorable.

We studied an interesting consequence of membrane curvature-dependent AQP4 OAP disruption. Autoanti-bodies against AQP4 in the disease NMO bind substantially better to AQP4 in OAPs than as separate tetramers (20). To quantify binding of a recombinant monoclonal NMO autoantibody (rAb-53), M23-AQP4 containing vesicles of diameter 355 nm or 57 nm were immobilized on a cover glass, and stained with rAb-53 antibody (and red fluorescent secondary antibody). Binding was normalized for AQP4 using an antibody against the AQP4 C-terminus and a green fluorescent secondary antibody for ratio imaging. Figure 4E shows greatly reduced red-to-green (R/G) fluorescence in the small vesicles, indicating that a consequence of OAP disruption by high membrane curvature is reduced NMO autoantibody binding.

Membrane cholesterol and SM content do not affect AQP4 OAP assembly

Golgi and plasma membranes have different lipid composition and cholesterol content, which could affect protein oligomerization, as has been found for syndecan-4, aquaporin-0 (AQP0) and glutamate transporter-1 (GLT-1) (25–27). SM-rich domains were reported to be the sites of lysenin oligomerization (28). The effect of depletion of cholesterol and SM on AQP4 OAP assembly was tested on plasma membrane vesicles from M23-expressing cells. Cholesterol was depleted by incubation of vesicles with βMCD, and SM was depleted by conversion to ceramide using SMase. BN/PAGE in Figure 5A, left (cholesterol-depletion) and right (SM-depletion) shows that these maneuvers did not affect AQP4 supramolecular assembly. Cholesterol assay showed lower cholesterol in Golgi vs. plasma membranes, and confirmed near complete cholesterol depletion with βMCD treatment (Figure 5A, right). We and others (29) have shown SM depletion of > 50% at concentrations of SMase lower than used here.

Plasma membranes contain factor(s) promoting AQP4 OAP assembly

Vesicle fusion experiments were done to determine whether Golgi-specific factor(s) are present that prevent OAP assembly in Golgi, and/or whether plasma membrane-factor(s) are present that facilitate OAP formation in the plasma membrane. Our strategy was to fuse M23-containing plasma membrane vesicles with non-AQP4-containing Golgi, and to fuse M23-containing Golgi with non-AQP4-containing plasma membrane vesicles. Vesicle fusion was accomplished by incubation in a high-Ca²⁺ buffer. BN/PAGE showed that fusion of (non-AQP4-containing) Golgi with M23-containing plasma membrane vesicles had little effect on the gel band pattern and hence AQP4 supramolecular assembly, whereas fusion of (non-AQP4-containing) plasma membrane vesicles with M23-containing Golgi resulted in appearance of higher molecular weight bands corresponding to AQP4 OAPs (Figure 5B, left). Fusion of M23-containing plasma membrane vesicles with (non-M23-containing) plasma membrane vesicles did not change the gel band pattern, nor did fusion of M23-Golgi with (non-M23-containing) Golgi (Figure 5B, center). Vesicle size analysis in Figure 5B, right confirmed vesicle fusion by the high-Ca²⁺ buffer method used here. We point out that vesicle sizes before and after fusion are well above those where size-dependent effects on OAP formation were observed. These findings suggest that plasma membrane factor(s) are necessary and sufficient for AQP4 OAP assembly, and that a relatively small amount of such factor(s) can promote AQP4 OAP formation. The incomplete OAP formation after fusion of M23-containing Golgi with plasma membranes may be related to the random nature of vesicle fusion in which Golgi/Golgi, Golgi/plasma membrane and plasma membrane/plasma membrane fusion can occur.

We investigated two candidate plasma membrane factors that may promote AQP4 OAP formation. It has been suggested that AQP4 may interact with α-syntrophin, a scaffold protein, through a C-terminal PDZ binding motif in AQP4 (30–32). The possible involvement of α-syntrophin in OAP formation was investigated by knock-down in human astrocytes. Immunoblot showed ~90 % reduced α-syntrophin in the siRNA-treated astrocytes (Figure 5C, top). BN/PAGE showed, however, that AQP4 supramolecular assembly was not affected by α-syntrophin knock-down (Figure 5C, bottom). To investigate the involvement of the AQP4 C-terminus PDZ-binding domain, measurements were done in U87MG cells transfected with full-length M23-AQP4 or M23-Δ6, a C-terminus truncation mutant of M23-AQP4 lacking its PDZ-binding domain (24). We found, however, that expression of full-length vs. M23-Δ6 AQP4 did not affect α-syntrophin expression (Figure 5D, top) or supramolecular assembly (Figure 5D, bottom). Therefore, neither α-syntrophin nor AQP4 PDZ-domain interactions are required for AQP4 OAP formation.

DNA constructs

cDNAs encoding human AQP4-M1, AQP4-M23 and connexin-43 (CX43) were PCR-amplified using whole-brain cDNA as template. PCR fragments were ligated into the mammalian expression vector pcDNA3.1. To generate AQP4 C-terminal chimeras with enhanced green fluorescent protein (GFP) or mCherry the respective sequences were PCR amplified from their respective plasmids (Clontech, Mountain View, CA) and ligated in-frame into pcDNA3.1. An M23-AQP4 C-terminus truncation mutant was generated as described (24). Chimeras of galactose-1-phosphate uridylyl-transferase (GalT) and calnexin with mCherry were ligated into pcDNA3.1 in-frame with the fluorescent protein. All constructs were verified by sequencing.

Cell culture and transfections

U87MG (human glioblastoma-astrocytoma, ATCC HTB-14) and CHO-K1 (ATCC CCL-61) cell cultures were maintained at 37 °C in 5% CO₂, 95% air in appropriate medium containing 10% fetal bovine serum, 100 units/ml penicillin and 100 μg/ml streptomycin. Cells were transfected with cDNAs in antibiotic-free medium 12–24 h before experiments, using FuGene HD (Roche, Basel, Switzerland) at a 3:1 ratio (μl:μg nucleic acid) and 0.18 μg DNA per cm².

Fluorescence recovery after photobleaching (FRAP)

Photobleaching of AQP4-GFP chimeras was done using a Nikon Eclipse FN1 confocal microscope equipped with Nikon 100x oil immersion lens (numerical aperture 1.49). Transfected cells grown on cover glasses were mounted in a live-cell chamber kept at 37 °C. In some experiments cells were incubated with brefeldin A (BFA) for 8 h at 37 °C to cause retention of AQP4 in the ER. In some experiments cells were incubated at 20 °C for 12–16 h to accumulate AQP4 in Golgi. GFP was excited at 488 nm using a 40 mW argon laser at 2% power, which was increased 30–50-fold for 0.5–1 s for photobleaching of a circular spot of ~10 μm diameter. Images were processed to determine background-subtracted area-integrated intensities using ImageJ (http://rsbweb.nih.gov/ij/).

Freeze-fracture electron microscopy

Pelleted cells or vesicles were fixed with 2% glutaraldehyde/0.1 M cacodylate (Electron Microscopy Sciences, Hatfield, PA) for 4 h, and then rinsed with 0.1 M cacodylate and stored in PBS. Pellets were equilibrated with cryoprotectant, 30% glycerol/0.1 M cacodylate, overnight, placed on copper freeze-fracture supports and frozen in liquid Freon 22 cooled by liquid nitrogen. Unidirectional and rotary shadowed freeze-fracturereplicas were generated as described (30). The replicas were cleaned by immersion for 2 h in concentrated sodium hypochlorite bleach, washed three timesfor 5 min each with distilled water, transferred to copper electronmicroscopy grids, and examined with a JEOL 1011 electron microscope(JEOL, Tokyo, Japan).

Total internal reflection fluorescence microscopy (TIRFM)

TIRFM was done using a Nikon Eclipse TE2000E microscope equipped with a through-objective TIRF attachment and a 100x oil immersion objective (numerical aperture 1.49) mounted on a perfect focus module (Nikon). GFP was excited using an Argon laser, Z488/10x excitation filter and Z488RDC dichroic mirror, and detected through an ET525/50m emission filter (Chroma). Images were acquired using a QuantEM 512SC deep-cooled CCD camera (Photometrics, Tucson, AZ).

Subcellular fractionations

Human brain was homogenized in homogenizing buffer [300 mM sucrose, 1 mM EDTA, protease cocktail inhibitor (Roche, Basel, Switzerland), 10 mM Tris-HCl, pH 7.2] by 20 strokes of a glass Dounce homogenizer. The homogenate was centrifuged at 500g 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 (1 ml), 1.6 M (2 ml), 1.4 M (4 ml, containing homogenate), 1.2 M (4 ml), 0.8 M (1 ml)] was centrifuged for 2.5 h at 25,000 rpm in an SW 27 rotor, and 1 ml fractions were collected. Protein concentration was determined by the Bradford method. Separation of plasma membrane from intracellular vesicles was performed by differential centrifugation as described (42), with minor modifications. Cells were washed three times with PBS, scraped into homogenizing buffer, and homogenized by 20 strokes in a glass Dounce homogenizer. The homogenate was spun at 4,000g for 15 min and the pellet discarded. A plasma membrane-enriched fraction was obtained by centrifugation at 17,000g for 45 min. The resulting supernatant was spun at 200,000g for 1 h to obtain cytosolic (supernatant) and the intracellular vesicle (pellet) fractions.

Electrophoresis

Cells, membrane vesicles or tissues were suspended in ten volumes of Blue Native (BN) lysis buffer (500 mM ε-aminocaproic acid, 50 mM imidazole pH 7.0, 12 mM NaCl, 10% glycerol, 1% Triton X-100, protease inhibitor cocktail). After 30 min incubation on ice, the samples were centrifuged at 22,000 g for 30 min and supernatant protein concentration determined. The supernatants for Blue Native/polyacrylamide gel electrophoresis (BN/PAGE) were supplemented with Coomassie Blue G-250 dye from a 5% suspension in 500 mM 6-aminohexanoic acid. The supernatants for high-resolution clear native/polyacrylamide gel electrophoresis (hrCN/PAGE) were supplemented with Ponceau dye 0.1%. For studies of pH effect, cells were treated with PBS containing 0.005% digitonin (with pH adjusted in the range 3–7.4), briefly washed with PBS, and lysed in BN lysis buffer. Polyacrylamide native gradient gels were prepared as described (43). Samples (20 μg protein) were mixed with 5% Coomassie Blue G-250. Ferritin was used as the molecular mass standard (440 and 880 kDa). The running buffers were: 25 mM imidazole, pH 7 (anode buffer) and 50 mM tricine, 7.5 mM imidazole, 0.02% Coomassie Blue G-250, pH 7 (cathode buffer). Proteins were blotted onto polyvinylidene difluoride (PVDF) membranes (Millipore, Bedford, MA) using a native transfer buffer (50 mM tricine, 7.5 mM imidazole). For the second dimension, lanes from the first dimension were cut into strips and equilibrated in denaturation buffer (1% SDS, 1% β-mercaptoethanol) for 1–2 h at room temperature. A single strip was then placed into a second dimension gel of the same thickness and subjected to tricine SDS-PAGE. Proteins were blotted as above. For one-dimensional Laemmli SDS-PAGE, gels consisted of a 12% running gel and 3% stacking gel. hrCNE/PAGE was performed as described (44). Samples were loaded on polyacrylamide native gradient gels 3–9%. Samples (30 μg protein) were mixed with Ponceau S 0.1% and loaded onto an identical gel to that used for BN/PAGE. The cathode buffer (50 mM tricine, 7.5 mM imidazole, pH 7.0) was supplemented with the anionic detergent sodium deoxycholate (0.05%). A VersaDoc imaging system (Biorad, Hercules, CA) was used to image fluorescent GFP-AQP4 in hrCN/PAGE gels.

Immunoblot

Proteins were blotted onto PVDF membranes at 160 mA for 1.5 h in buffer containing 1.92 M glycine, 0.25 M Tris base, 10 mM EDTA, 1% SDS. Membranes were blocked with 3% BSA and incubated with the following primary antibodies at 4 °C overnight: goat or rabbit anti-AQP4 or rabbit anti-CX43 (Santa Cruz Biotechnology, Santa Cruz, CA), calnexin, trans-Golgi network-46 (TGN-46) and Na⁺/K⁺ ATPase (ABcam, Cambridge, MA). Membranes were then rinsed, incubated for 1 h with horseradish peroxidase-conjugated goat anti-rabbit IgG or donkey anti-goat IgG (Jackson ImmunoResearch Laboratories, West Grove, PA), rinsed extensively, and labeled proteins were detected using the ECL Plus enzymatic chemiluminescence kit (Amersham Biosciences, Pittsburgh, PA).

Cholesterol and sphingomyelin (SM) depletion

Cholesterol was depleted from plasma membrane vesicles by incubation with β-methyl-cyclodextrin (βMCD) (Sigma, St Louis, MO). Vesicles (5 μg protein) in BN buffer containing different concentrations of βMCD in 100 μl total volume were incubated for 2.5 h at room temperature on a rocking platform. The treated vesicles were centrifuged at 20,000 g for 10 min at 4 °C, washed once with 100 μl BN buffer, and the pellets prepared for BN/PAGE. Cholesterol was assayed using the AmplexRed cholesterol assay kit (Invitrogen, Carlsbad, CA). SM depletion was performed on cell pellets of freshly harvested cells by addition of 0.1 U/ml sphingomyelinase (SMase, Sigma, St Louis, MO).

Vesicle fusion and extrusion

Equal amounts (5 μg/ml protein) of Golgi and plasma membrane vesicles were diluted in a final volume of 100 μl BN buffer in the presence or absence of 5 mM Ca²⁺, and incubated for 30 min at room temperature on a rocking platform. Samples were centrifuged at 20,000 g for 10 min at 4 °C and prepared for BN/PAGE. Aliquots (5 μl) of each fusion reaction were diluted in 50 μl BN buffer for size determination by quasi-elastic light scattering (N5 Submicron Particle Size Analyzer, Beckman, Brea, CA). Vesicles of different sizes were prepared by extrusion as described (45) using a Lipofast vesicle extruder and 50, 100 and 200 nm-pore polycarbonate filters (Avestin, Canada).

Quantitative immunofluorescence of vesicles

Vesicles of diameter 355 or 57 nm containing M23-AQP4, prepared as described above, were immobilized on poly-D-lysine-coated coverslips for 1 h, then rinsed with PBS and incubated for 30 min with human recombinant monoclonal antibody rAb-53 (46). Vesicles were then rinsed with PBS, fixed in 4 % paraformaldehyde for 15 min, and permeabilized with 0.1 % Triton X-100. Vesicles were then blocked (PBS containing 6 mM glucose, 1 mM pyruvate, 1 % bovine serum albumin) and incubated for 30 min with 0.4 μg/mL polyclonal, C-terminal specific rabbit anti-AQP4 antibody (Santa Cruz Biotechnology, Santa Cruz, CA), then rinsed with PBS. Finally, vesicles were incubated for 30 min with 4 μg/mL goat anti-human IgG-conjugated Alexa Fluor 555 and goat anti-rabbit IgG-conjugated Alexa Fluor 488 (Invitrogen) in blocking buffer. Quantitative analysis of antibody binding was done by ratio imaging fluorescence microscopy (20) using an Nikon Eclipse FN1 confocal microscope equipped with Nikon 100x oil immersion lens.

siRNA transfections

Human fetal cortical astrocyte cultures were generated as described (47). More than 98% of the cells were positive for glial fibrillary acidic protein (GFAP). Knockdown was achieved using α-syntrophin Stealth siRNA, with a scrambled sequence used as control (Invitrogen, Carlsbad, CA). Astrocytes were seeded the day before transfection using Dulbecco’s modified Eagle’s medium (DMEM) containing 10% fetal bovine serum (FBS). Transient transfection of siRNAs was carried out using siRNA Lipofectamine RNAiMAX (Invitrogen, Carlsbad, CA) according to the manufacturer’s instructions.