Inverting the selectivity of aquaporin 6: Gating versus direct electrostatic interaction

Specialized transmembrane proteins form channels that facilitate the transport of ions and other molecules across the membranes and thus play crucial role in many life processes. The understanding of these systems has increased enormously in recent years because of advances in structural and biochemical studies (e.g., refs. 1 and 2). However, key questions remain with regards to the exact factors that control the selectivity and gating of ion transport across membranes. To understand these factors, it would very useful to quantify the energy contributions that allow some ions to permeate while blocking others, and thus allow channels to fulfill their specific physiological functions. Although the availability of structural information paved the way for detailed scrutiny of different hypotheses and models, there is still a clear need for well defined experimental test cases. In this issue of PNAS, a study of aquaporin 6 (AQP6) by Liu et al. (3) provides a crucial test case in which a mutation of a single residue (a change of Asn-60 to Gly) completely changes the conductance, converting AQP6 from an anion channel to a water-selective channel. This finding provides extremely valuable information whose full “translation” to a functional model probably will require more structural information as well as the use of computer modeling approaches.

This is not the first time that studies of the aqauporin family have provided a challenge and a guide to the understanding of ion selectivity. Most members of the aquaporin family allow water molecules to pass through membranes but block ions, including hydronium ions (2). The reason for the blockage of proton transport through aquaporin was the subject of recent studies (4–8) suggesting that the widely held paradigm that biological proton transfer is controlled by the orientation of water molecules should be replaced by the idea that proton transport is controlled by the electrostatic interaction between the proton and the channel (see, for example, refs. 5–8). However, the detailed nature of the electrostatic barrier remains controversial. Some workers hold that the barrier is caused by specific residues in the so-called NPA motif (which is located in the center of the channel), whereas detailed calculations of the effects of different residues (7) by “computational mutations” indicate that the loss of solvation upon entrance to the channel is sufficient to block the transport of protons and other cations.

Probably the best way to reach unique conclusions about the nature of the electrostatic barriers that govern ion penetration in aquaporins is to have experimental test cases in which a single mutation changes the selectivity and or ion conductivity in a drastically way while still maintaining a functioning channel. In this respect, the report of Liu et al. appears to be particularly important, because it may allow resolving the difference between the two very different limiting cases described in Fig. 1 (or a combination of these limits). In one model, transfer of anions in the native protein is caused by the direct electrostatic interaction between the dipole of Asn-60 and the negatively charged anion. In the other model, Asn-60 facilitates ion conductance indirectly by pushing helix 5 away from helix 2 and thus changing the channel configuration. This alteration may lead to two effects: (i) rearrangement of dipoles and charges that will change the electrostatic barriers for ion transport and (ii) a widening of the pore that will allow the penetration of water molecules whose dipoles can stabilize the anion. This change reduces the loss of solvation energy upon transfer of the ion from the bulk solvent to the channel interior.

Fig. 1.

Illustration of the options of direct electrostatic control (a) and indirect gating control (b). The figure considers, in a completely schematic way, the pore of AQP6 (showing only four of the seven helixes) and illustrates the possible effects of the N60G mutation. In the first case, direct interaction with Asn 60 (designated by a dipole) helps to stabilize the transferred anion. In the second case, Asn 60 (designated as a bulge and a dipole) pushes helix 2 from helix 5 and changes the structure of the channel relative to the mutant structure in a way the pore radius increases, allowing water molecules to solvate the ion and possibly increasing the interaction between dipolar and charged groups of the channel and the anion.

At present, the only structural model for AQP6 and its N60G mutant has been obtained by exploiting the homology to AQP1 (3). Such a tentative model cannot be used to discriminate between two options of Fig. 1, although glycosylation experiments (3) indicate that the mutation leads to significant structural changes. Fortunately, a part of this challenging question can be explored by computational analysis even in the absence of x-ray or NMR structural information. For example, a preliminary semimacroscopic electrostatic calculation that considered the protein relaxation in response to the ion charge (M. Kato, A. Bruyin, and A.W., unpublished data) examined the first possibility (a direct stabilization by Asn-60). It was found that Asn-60 contributes <1 kcal/mol to the reduction of the Cl^– penetration profile (relative to the Gly mutant). Thus, the change in ion selectivity is probably not caused by direct electrostatic interactions, but by an indirect mechanism. Further confirmation of this finding would mean that the N60G mutant offers a rare glimpse on the energy balance in gating mechanisms without the complications involved in analyzing more complex voltage-activated gating (9).

Because the N60G mutation appears to present an excellent model for a gating mechanism it would be very useful to solve the structure of the native and mutant enzyme (a suggestion that is also given in ref. 3). At any rate, once the structures of AQP6 and its N60G mutant are solved, we will have a powerful and relatively simple benchmark for modeling gating in ion channels.

The change in ion selectivity is probably not caused by direct electrostatic interactions.

The finding that the N60G mutation increases the water permeability is also of significant interest. However, this finding may reflect the fact that ions cannot penetrate the channel and thus cannot compete with the transfer of water. It is also important to recognize that simulations of water transfer and evaluations of the corresponding barriers are far less demanding than reliable computational studies of ion transfer. Because water is neutral, its interaction with the channel is quite small and does not involve the enormous challenges associated with long-range electrostatic interactions in proteins (see discussion in refs. 10–13). It is the refinement of electrostatic models for the energetics of ions in ion channels where experiments of the type reported by Liu et al. are so badly needed.

In the absence of a clear structural information about AQP6, it is unclear why this channel allows cations to penetrate. However, preliminary computational studies of AQP0 indicated that even in this water-selective channel, there is a central region where Cl^– is only few kcal/mol less stable than in water. This result is in clear contrast to the finding that protons and other cations are very unstable in this site. This finding does not mean that there is an enormous positive potential at the center of the channel. The effective potential and dielectric in proteins depends on the sign of the charge of the ion and reflects a complicated compensation of different factors. Thus, for example, both positive and negative ions would be destabilized in an equal way in a purely nonpolar environment (rather than having one of them stable and the other unstable). This complication only underscores the importance of combining structural, genetic, and computational analysis.

Finally, in view of the success in using mutational studies to probe the action of AQP6, similar experiments in AQP1 and AQP0 would be very useful. This may allow one to examine whether the water/proton selectivity is due to changes in the generalized solvation energy as concluded by computational mutation analysis (7) or to the NPA motif. It is clear that the most fruitful way to ask this question theoretically is to perform the mutation computationally, rather than trying to calculate the electrostatic contributions from the fixed structure of the native channel.