Post-Golgi supramolecular assembly of aquaporin-4 in orthogonal arrays

Abstract

The supramolecular assembly of aquaporin-4 (AQP4) in orthogonal arrays of particles (OAPs) involves N-terminus interactions of the M23-AQP4 isoform. We found AQP4 OAPs in cell plasma membranes but not in endoplasmic reticulum (ER) or Golgi, as shown by: (i) native gel electrophoresis of brain and AQP4-transfected cells, (ii) photobleaching recovery of green fluorescent protein-AQP4 chimeras in live cells and (iii) freeze-fracture electron microscopy (FFEM). We found that AQP4 OAP formation in plasma membranes, but not in the Golgi, was not related to AQP4 density, pH, membrane lipid composition, C-terminal PDZ domain interactions or α-syntrophin expression. Remarkably, however, fusion of AQP4-containing Golgi vesicles with (AQP4-free) plasma membrane vesicles produced OAPs, suggesting the involvement of plasma membrane factor(s) in AQP4 OAP formation. In investigating additional possible determinants of OAP assembly we discovered membrane curvature-dependent OAP assembly, in which OAPs were disrupted by extrusion of plasma membrane vesicles to ∼110 nm diameter, but not to ∼220 nm diameter. We conclude that AQP4 supramolecular assembly in OAPs is a post-Golgi phenomenon involving plasma membrane-specific factor(s). Post-Golgi and membrane curvature-dependent OAP assembly may be important for vesicle transport of AQP4 in the secretory pathway and AQP4-facilitated astrocyte migration, and suggests a novel therapeutic approach for neuromyelitis optica.

Figure 1. AQP4 oligomerization in subcellular fractions of brain homogenates and transfected cells

(A) Electrophoresis of human brain plasma membrane, Golgi and ER fractions. The first dimension is BN/PAGE, the second dimension is tricine SDS/PAGE. Immunodetection was carried out against AQP4 or CX43 (left). Immunoblot of Na⁺/K⁺ ATPase, TGN46 and calnexin (right). (B) BN/PAGE of plasma membrane and Golgi/ER fractions of U87MG cells transfected with M23-AQP4 (left) or CX43 (center). Immunoblot of Na⁺/K⁺ ATPase, TGN46 and calnexin (right). (C) Unidirectionally-shadowed freeze-fracture electron micrographs of purified subcellular fractions of M1 and M23-AQP4-transfected U87MG cells. Scale bar, 200 nm.

Figure 2. Oligomerization of AQP4-GFP chimeras and localization in live cells

(A) SDS/PAGE (top left), BN/PAGE (top center) and hrCN/PAGE (top right) of M1-GFP and M23-GFP. Confocal and TIRF microscopy showing plasma membrane targeting of M1-GFP and M23-GFP in transiently transfected U87MG cells (bottom). (B) Live-cell imaging of U87MG cells transiently transfected with M1-GFP and M23-GFP after low-temperature block showing Golgi localization (top), or treatment with BFA showing ER localization (bottom). Cells were co-transfected with compartment-specific marker proteins (GalT for Golgi, calnexin for ER) fused to mCherry.

Figure 3. Diffusion of M1 and M23-AQP4 in live cells

(A) FRAP of M1-GFP or M23-GFP in transiently transfected U87MG cells, with plasma membrane (top) and Golgi (low-temperature treated cells, bottom) localization. Arrowheads indicate bleached area. (B) Representative kinetics of FRAP of M1-GFP and M23-GFP over 120 s (top), with summary of percentage recovery data at 120 s (bottom) (S.E., n > 10 cells, * P < 0.01).

Figure 4. Membrane curvature-dependent AQP4 oligomerization

(A) BN/PAGE of M23-AQP4 in transfected U87MG cells after incubation in buffers of indicated pH. (B) (left) BN/PAGE of M23-AQP4 in plasma membrane vesicles of different sizes generated by extrusion (top). Vesicle size distributions determined by quasi-elastic light scattering (bottom). (right) BN/PAGE of M23-AQP4 in control and extruded plasma membrane vesicles, and in extruded vesicles after fusion. (C) Freeze-fracture electron micrographs of vesicles of average diameters 355 nm (top) and 57 nm (bottom). (D) To-scale model of AQP4 tetramers in membranes of different curvatures correspond to vesicles of diameters 350 or 50 nm. (E) Fluorescence micrographs of coverslip-immobilized 355 and 57 nm diameter vesicles showing binding of NMO monoclonal antibody rAb-53 (red) and anti-AQP4 antibody (green). Red-to-green fluorescence ratios (R/G) (right) (mean ± S.E., n=3).

Figure 5. Determinants of AQP4 oligomerization on the plasma membrane

(A) Effect of cholesterol depletion on M23-AQP4 oligomerization in plasma membrane vesicles from M23-AQP4-expressing U87MG cells. (left) BN/PAGE after treatment with indicated concentration of βMCD. (center) BN/PAGE after treatment with indicated concentrations of SMase to deplete SM (S.E., n=4). (right) Cholesterol content in Golgi and plasma membrane vesicles treated with βMCD (S.E., n=4). (B) BN/PAGE of M23-AQP4 oligomerization in unfused and high Ca²⁺-fused mixtures of plasma membrane and Golgi vesicles. As indicated, vesicles were derived from M23-AQP4-expressing or non-transfected cells. (right) Vesicle size distribution before (pre-fusion) and after incubation in high Ca²⁺ buffer. (C) BN/PAGE of AQP4 after α-syntrophin siRNA knock-down in human astrocyte cultures (bottom). Knock-down (α-syn) and loading (actin) controls are shown (top). (D) BN/PAGE of U87MG cells after transfection with native M23-AQP4 or a C-terminus truncation mutant lacking the PDZ-binding domain (bottom). SDS/PAGE of α-syntrophin expression is shown (top).