We determined the x-ray structure of bovine aquaporin 0 (AQP0) to a resolution of 2.2 Å. The structure of this eukaryotic, integral membrane protein suggests that the selectivity of AQP0 for water transport is based on the identity and location of signature amino acid residues that are hallmarks of the water-selective arm of the AQP family of proteins. Furthermore, the channel lumen is narrowed only by two, quasi-2-fold related tyrosine side chains that might account for reduced water conductance relative to other AQPs. The channel is functionally open to the passage of water because there are eight discreet water molecules within the channel. Comparison of this structure with the recent electron-diffraction structure of the junctional form of sheep AQP0 at pH 6.0 that was interpreted as closed shows no global change in the structure of AQP0 and only small changes in side-chain positions. We observed no structural change to the channel or the molecule as a whole at pH 10, which could be interpreted as the postulated pH-gating mechanism of AQP0-mediated water transport at pH >6.5. Contrary to the electron-diffraction structure, the comparison shows no evidence of channel gating induced by association of the extracellular domains of AQP0 at pH 6.0. Our structure aids the analysis of the interaction of the extracellular domains and the possibility of a cell–cell adhesion role for AQP0. In addition, our structure illustrates the basis for formation of certain types of cataracts that are the result of mutations.
Proc Natl Acad Sci U S A. 2004 September 28; 101(39): 14045–14050.
The vertebrate ocular lens is a remarkably transparent and avascular tissue that acts basically as a syncytium of differentiated epithelial cells, called fiber cells. These cells are thin and highly elongated, and they are essentially a plasma membrane-enclosed sack filled with transparent crystallin proteins. The lens is covered on the surface of its anterior hemisphere with a layer of simple squamous epithelial cells and an acellular capsule that encloses the entire lens. The lack of vascular-supply structures and any identifiable active transport systems in the fiber cell mass means that diffusional pathways are of paramount importance to the establishment and maintenance of lens homeostasis and transparency. The transparency of the lens, together with its ability to undergo dynamic shape changes during accommodation, provides for a clear and accurate image of the world to be projected onto the retina. The transparent nature of the lens is contingent on several crucial features that permit light to pass through with a minimum of light scattering. These features are (i) the maintenance of a highly ordered molecular structure of the crystallin proteins; (ii) terminally differentiated fiber cells containing very few organelles; and (iii) intracellular and intercellular spaces being kept smaller than the wavelength of ambient light (1–3).
It is intriguing to understand the cellular and molecular basis for the maintenance of lens transparency, as well as the loss of lens transparency due to pathological and injury-induced conditions. Lens physiology has implicated water as one culprit that is often responsible for the disruption of crystallin molecule transparency; the movement of excess water across the lens fiber cell membrane into the fiber cell induces the hydration of crystallin proteins that disrupts their transparent molecular structure. The major integral membrane protein of the lens fiber cell, aquaporin 0 (AQP0), is thought to be a key player in maintaining a healthy functional lens by regulating water permeation across the fiber cell plasma membrane. Bovine AQP0 (bAQP0) is composed of 263 amino acids and accounts for >60% of the fiber cell plasma membrane protein complement (4–6). The measured functions have been controversial. AQP0 was initially postulated to be a gap-junction protein that forms voltage-dependent, nonspecific channels with the ability to transport substances as large as 1,500 Da, and it was then postulated to be a cell–cell adhesion molecule. The genetic sequence of AQP0 identified it as a member of the AQP, rather than the connexin, family (7) and thereby predicted its role in the establishment and maintenance of lens homeostasis (8). However, measured water transport is 15-fold lower than for AQP1 at pH 6.5 and is reduced to 46-fold lower than AQP1 at pH 7.5 (9).
To address the roles of AQP0 and, in particular, its roles in water transport and cell adhesion in the lens, we determined the 3D structure of bovine AQP0 to a resolution of 2.2 Å. This structure helps to explain the observed water transport and its regulation through these channels. The structure also provides insight into the possible role that AQP0 plays in lens accommodation through its cell-adhesion activity, as well as the effect of cataract-inducing mutations on the structure of AQP0 and the transparency of the lens as a whole.
Since the submission of this manuscript, an electron diffraction-derived structure of sheep AQP0 (sAQP0) at a resolution of 3.0 × 3.5 Å was published by Walz and coworkers (10). Their structure was for a junctional form of doubled membranes at pH 6.0, used 2D crystals of sAQP0, and does not reveal any water molecules. Overall, the x-ray and electron-diffraction structures are remarkably similar to an α-carbon rms deviation of 1.12 Å. Residues 6–239 were common to both structures. The only four residues that are different between bAQP0 and sAQP0 sequences (bAQP0 numbering) are C14F, S20T, M90V, and S240T (S240T does not appear in either structure). The comparisons between the two structures instruct as to differences that may be caused by the junctional molecular contact vs. the isolated tetramers, by the differences in pH, or by any difference between membranebound and detergent-solubilized AQP0, and they could reveal differences related to any of the postulated gating mechanisms that depend on junctional contact or pH effects.
The pH-gating mechanism proposed by Nemeth-Cahalan and Hall (11) and Nemeth-Cahalan et al. (12) by their criteria shows AQP0 maximally conducting at pH ≈6.5 and conducting 3.4-fold less at pH ≈8.5 (however, any pH-dependent gating has been challenged; ref. 13). For Walz and coworkers (10), sAQP0 crystals were formed at pH 6.0, whereas our bAQP0 crystals were formed at pH 10.0. The channel would be expected to be closed if any pH gating were to persist at pH >8.5, and thus, the two structures could illustrate any pH-dependent structural changes associated with closure at high pH (or any pH-dependent structural changes at all). Our structure seems to be functionally open, and thus, there is little evidence in the structure to support blockade in a static sense at pH 10.
In contrast, the “double-layer” structure is thought to be in a closed form (10), even though it is found to be mostly open in a single membrane at pH 6.5. The x-ray structure (P4212) does not have individual AQP0 molecules or tetramers associating through their extracellular surfaces, whereas the electron diffraction double-layered structure (P422) has direct close approximation of the extracellular domains in a conformation that may represent one in vivo form. Therefore, any changes in the channel architecture that are due to extracellular-domain interactions should also be apparent.