AQP0 forms 2D crystals with all tested sphingomyelin/cholesterol mixtures. (A-E) AQP0 purified from sheep lenses was reconstituted with pure sphingomyelin (A), sphingomyelin/cholesterol mixtures at molar ratios of 2:1 (B), 1:2 (C), and 1:4 (D), as well a pure cholesterol (E). AQP0 was reconstituted under all conditions and formed diffracting 2D crystals. The scale bars are 2 μm. (F) Projection map of AQP0 reconstituted with pure cholesterol at 3.2 Å resolution. The 2D crystals show p422 symmetry and have the typical lattice constants for AQP0 crystals of a = b = 65.5 Å, and γ = 90°. The panel shows two-by-two unit cells. See also Figure 1–figure supplements 1 and 2 and Tables 1 and 2.

Electron crystallography provides structures of AQP0 in sphingomyelin/ cholesterol bilayers at 2.5-Å resolution. (A) Electron diffraction pattern of an untilted AQP0 2D crystal reconstituted at a sphingomyelin:cholesterol ratio of 2:1, showing reflections to ∼2 Å resolution. Scale bar indicates (10 Å)-1. (B) Density map at 2.5 Å resolution used to build the AQP02SM:1Chol structure. The 2Fo-Fc map contoured at 1.5σ is shown as gray mesh, the AQP0 model is shown in yellow with oxygen atoms in red and nitrogen atoms in blue. The red sphere represents a water molecule. (C) A diffraction pattern of an untilted AQP0 2D crystal reconstituted at a sphingomyelin:cholesterol ratio of 1:2, showing reflections to better than 1.6-Å resolution. Scale bar indicates (10 Å)-1. (D) Density map at 2.5-Å resolution used to build the AQP01SM:2Chol structure. Color code as in (B). See also Figure 2–figure supplement 1.

The 2:1 sphingomyelin/cholesterol bilayer surrounding AQP0 is similar to bilayers formed by phosphoglycerolipids. (A) The top panel shows the seven sphingomyelins (light green sticks) and one cholesterol (orange sticks) molecules forming the bilayer around an AQP0 subunit (gray surface). The bottom panel shows just the lipid bilayer. (B) The top panel shows DMPC lipids (purple sticks) surrounding an AQP0 subunit (Gonen et al., 2005) and the bottom layer shows an overlay of the DMPC bilayer with the 2:1 sphingomyelin/cholesterol bilayer. (C) The top panel shows an E. coli polar lipids (EPL) bilayer (modeled as PE lipids) (light brown sticks) surrounding an AQP0 subunit (Hite et al., 2010) and the bottom layer shows an overlay of the EPL bilayer with the 2:1 sphingomyelin/cholesterol bilayer. See also Figure 3–figure supplements 1–3 and Table 3.

The 1:2 sphingomyelin/cholesterol bilayer surrounding AQP0 and comparison with the 2:1 sphingomyelin/cholesterol bilayer. (A) The five sphingomyelins (dark green sticks) and four cholesterol (red sticks) molecules surrounding an AQP0 subunit (gray surface). The arrows between the orange and blue lines indicate the average distances between the phosphor atoms of the phosphodiester groups and the nitrogen atoms of the amide groups in the two leaflets, respectively. (B) The AQP02SM:1Chol structure shown for comparison with the AQP01SM:2Chol structure in (A). Arrows as in (A). (C) Overlay of the lipid bilayers in the AQP02SM:1Chol and AQP01SM:2Chol structures. (D) Location of the four cholesterols (red sticks) in the AQP01SM:2Chol structure with respect to AQP0 surface characteristics. Color coding: yellow, aromatic residues; cyan, hydrophobic residues; and light green, polar and charged residues. (E) Position of cholesterol Chol3 (red sticks) in the AQP01SM:2Chol structure and its interaction with residues of two adjacent AQP0 tetramers (brown sticks). See also Figure 4–figure supplements 1–2 and Table 3.

Localization of cholesterol around AQP0 monomers from unbiased MD simulations of individual AQP0 tetramers in SM membranes with low and high cholesterol concentration. Density maps representing the localization of cholesterol around AQP0 over time were computed from simulations starting from unbiased cholesterol positions in membranes at the indicated SM:Chol ratios. After combining the four maps calculated individually for the four subunits of the tetramer, cholesterol densities were projected (blue areas) onto the surface of a single AQP0 monomer (white surface). Projections are shown for the S1 and S2 monomer surfaces, as defined in the representations to the right. Lipids seen in the electron crystallographic structures obtained in membranes at the respective SM:Chol ratios are displayed as sticks and labeled according to the electron crystallographic structures. Densities are contoured at 10σ for the 2:1 SM:Chol membrane and at 9σ for the 1:2 SM:Chol membrane. The density hot spot indicated with an asterisk coincides with the Chol3 position seen at the 1:2 SM:Chol ratio. See also Figure 5– figure supplements 1–3.

Insertion depth and orientation of a cholesterol at the interface between two AQP0 tetramers. (A) Schematic figure illustrating how the insertion depth, d, and orientation angle, θ, of the cholesterol were measured. The cholesterol insertion depth was defined as the distance in z direction of the cholesterol oxygen atom (red stick representation) from the center of mass of the phosphorus atoms of the nearby SM molecules in the extracellular leaflet (top green horizontal line). The cholesterol orientation was defined as the angle between the membrane normal (simulation box z vector) and the vector along the rings of the cholesterol molecule (black dashed line). (B-E) Right panels: The three different systems that were simulated, namely (B) “No AQP0”, a pure lipid membrane without AQP0, (C, D) “AQP0 + EC lipids S2/1”, a membrane with one AQP0 tetramer surrounded by the annular lipids seen in the AQP01SM:2Chol structure, and (E) “2×AQP0”, a membrane containing a pair of AQP0 tetramers together with the lipids in between them from a hybrid AQP02SM:1Chol structure that replaces the two central SM molecules with the EC deep cholesterol molecules found in the AQP01SM:2Chol structure. Left panels: Graphs showing normalized histograms for the insertion depth, d, and orientation angle, θ, for the monitored deep cholesterol in membranes with different SM:Chol ratios (see color code in panel A), except for the 2×AQP0 system, which was only simulated in a pure SM membrane. For the simulations with one AQP0 tetramer, insertion and orientation were computed separately for the deep cholesterol located at surface S2 (C) and S1 (D). The vertical line indicates the most probable cholesterol position in the 2×AQP0 system (E). See also Figure 6–figure supplement 1.

Equilibrium MD simulations of pairs of associated AQP0 tetramers and force-induced separation of two associated AQP0 tetramers. (A) Distance between the centers of mass dCM (top) and minimum distance dtet-tet (bottom) between the pair of AQP0 tetramers during the 2×AQP0 simulations in equilibrium for the interface containing only SM molecules (SM, green) and the interface containing the deep cholesterol (Chol, blue) (n = 10 simulations for each case). (B) Principal component (PC) analysis of the relative movements between the two tetramers. Here, the motion of one of the tetramers (dashed-line rectangle) relative to the other tetramer (solid-line rectangle) was monitored. The three main principal components accounted for 67.9% of the total relative motion between tetramers (PC1: 34.3%, PC2: 24.7%, and PC3: 8.9%). The schematic drawings illustrate the three main modes of motion: bending (depicted in the drawing as viewed from the side of the membrane), lateral rotation (depicted in the drawing as viewed from the top of the membrane), and torsion (depicted in the drawing as viewed from the side of the membrane with one tetramer in front of the other). Lipids at the interface between the two tetramers (first two panels) and lipids surrounding the two tetramers (last panel) are shown. PC1 plus PC2 capture the bending and rotation while PC3 corresponds to torsion. The histograms in the bottom panels show the projections of the MD trajectories onto the three main PC vectors, for the interfaces containing only sphingomyelin (SM, green) or containing the deep cholesterols (Chol, blue). The approximate angular extent for each of the modes, attributed to these projections, is indicated (in degrees). The distributions with and without cholesterol are very similar, except for PC1. Nevertheless, PC1 relates to a small angular variation (∼7.1° bending together with ∼6.5° rotation). (C) Two AQP0 tetramers (white surface) arranged as in 2D crystals and embedded in a pure SM membrane were pulled apart by exerting a harmonic force F on them in the direction that connects their centers of mass (dCOM). Two different interfaces were studied: the “SM interface” consisted solely of SM lipids (green spheres) as seen in the AQP02SM:1Chol structure, whereas the “Chol interface” contained the deep cholesterols seen in the AQP01SM:2Chol structure. The reference position of the harmonic springs used to exert the force F was moved at a constant velocity vpull/2. (D) Force (F) and distance (dCOM) time traces are shown for one of the simulations using a vpull of 0.004 m/s. A Gaussian smoothing function (black continuous lines) was applied to the curves (yellow and purple). The detachment force (black circle) was computed as the highest recorded force when dCOM started to increase and below a cut-off distance dCOM of 7.3 nm (horizontal dashed line indicates the cut-off distance and the vertical dashed line indicated the time when this value was surpassed). Note that using different cut-off distances did not change the overall trend (see Figure 5–figure supplement 1). The inset shows an example of the arrangement of the tetramers at the moment of detachment. The red circle indicates the last contact. (E) Force–distance profiles for the two different interfaces are presented for the three indicated pulling velocities (n = 10 for each case). Dots indicate the point of detachment. (F) Detachment force is presented as a function of the pulling rate for the two interfaces (avg ± s.e., n = 10). A fit of the form Fdetach = A+B*log(Vpull) is shown to guide the eye with lines (Fdetach=[2142+226*log(Vpull)] for the “Chol” interface and Fdetach=[2099+248*log(Vpull)] for the “SM” interface). P-values comparing the two data sets are shown for each pulling velocity (Student’s t-test). See also Figure 7–figure supplement 1.

Interactions formed between adjacent AQP0 tetramers and lipid-AQP0 surface complementarity. (A) Equilibrium MD simulations were performed with pairs of AQP0 tetramers without cholesterol (top row) or with cholesterol (bottom row) at the interface. Density maps were calculated for different components next to the AQP0 surface over time and these maps were projected onto the surface of one of the AQP0 tetramers to represent their localization. For clarity, the second tetramer, which would be in front of the shown tetramer, is not shown. The density displayed at arbitrary density units is color-coded and shows the position of cholesterol (red), protein, i.e., the neighboring AQP0 tetramer (brown), sphingomyelin (green) and water (blue). (B) Tij is defined as the fraction of time during which residues (i, j) from opposite tetramers are in contact. This quantity was extracted from the equilibrium simulations for the pure sphingomyelin interface Tij(SM) and for the interface containing cholesterol Tij(Chol). Tij = 0 means that i and j were never in contact and Tij = 1 means that they were always in contact. (C) The pairwise difference ΔTij = Tij(Chol) - Tij(SM) is shown, discarding insignificant changes (ΔTij < 1 percentage, %, points). Accordingly, a value of ΔTij > 0 (ΔTij < 0) corresponds to protein–protein contacts that were more often observed in the simulations with cholesterol (sphingomyelin). For instance, the residue pair Gln129–Ser106 was observed almost 20 percentage points more time in the simulations with cholesterol. The color of the bars indicates the location of the residues (grey: inner part of the extracellular leaflet, i.e., where deep cholesterol resides; white: the rest of the interfacial protein surface, see inset at lower right). Residues involved in a high ΔTij > 10 percentage points are highlighted in orange in the inset (yellow for the contact Leu217–Phe214 observed in the electron crystallographic structure). (D) The average duration for every established protein–protein contact, <τ>ij, is also displayed for the two different lipid interfaces. In (B) and (D), the horizontal dashed lines indicate the highest value observed for all possible residue pairs. Contacts observed in the electron crystallographic structures are highlighted in bold letters. In B–D, the avg±s.e.m. is presented (n = 20, i.e., 10 independent simulation times with 2 symmetric monomeric interfaces). (E) The schematic drawing depicts a top view of the two associated AQP0 tetramers (squares, “Tet 1” and “Tet 2”) with the two central lipids sandwiched between them (black circles). The other lipids at the interface are not shown for clarity. The respective monomer surfaces S1 and S2 are indicated. The region of the surface of the two tetramers that is in total covered by a lipid, AContact (red line) was normalized by the surface area of the lipid, ALipid (here corresponding to the perimeter of the circles). This ratio gives a measure of the surface complementarity between the tetramers and the sandwiched lipids, i.e., the higher the value of AContact/ALipid the more the two surfaces complement each other. (F) Normalized histograms of AContact/ALipid obtained from the equilibrium MD simulations are shown for the central sphingomyelin (SM) and cholesterol (Chol). See also Figure 8–figure supplements 1–3.

Proposed model for how an increasing cholesterol concentration drives AQP0 2D array formation in the native lens membrane. (A) At a low cholesterol concentration, AQP0 tetramers are mostly surrounded by phospholipids and sphingomyelin. Free cholesterol in the membrane (green ovals) only associates with the highest affinity cholesterol-binding sites. Cholesterols occupying these peripheral binding sites are shown as red ovals and the black double-headed arrow indicates the transient nature of this interaction. The deep cholesterol-binding sites (orange squares) are not occupied. (B) With increasing cholesterol concentration, more cholesterols associate with the AQP0 surface. These cholesterols cause the interacting lipids in the cytoplasmic leaflet to move out from the bilayer center (blue arrow), resulting in the annular lipid shell that has a bigger hydrophobic thickness than the surrounding membrane, creating a hydrophobic mismatch that results in membrane deformation. (C) To minimize hydrophobic mismatch, AQP0 tetramers cluster. Cholesterol in between adjacent tetramers can move into the deep binding sites (yellow arrow) and cholesterol occupying deep binding sites (yellow ovals) act as glue that increases that association of the adjacent tetramers (indicated by the small double-headed black arrow) as compared to adjacent tetramers that do not sandwich a deep-binding cholesterol (indicated by the large double-headed black arrow). Clustering of proteins to minimize hydrophobic mismatch and stabilization by deep cholesterol-mediated protein–protein interactions may be the basis for the formation of transient lipid rafts. (D) Each AQP0 tetramer has four deep cholesterol-binding sites. As a result of the avidity effect, AQP0 can form large and stable 2D arrays.