Structure and selectivity in bestrophin ion channels

See allHide authors and affiliations

Science  17 Oct 2014:
Vol. 346, Issue 6207, pp. 355-359
DOI: 10.1126/science.1259723


Human bestrophin-1 (hBest1) is a calcium-activated chloride channel from the retinal pigment epithelium, where mutations are associated with vitelliform macular degeneration, or Best disease. We describe the structure of a bacterial homolog (KpBest) of hBest1 and functional characterizations of both channels. KpBest is a pentamer that forms a five-helix transmembrane pore, closed by three rings of conserved hydrophobic residues, and has a cytoplasmic cavern with a restricted exit. From electrophysiological analysis of structure-inspired mutations in KpBest and hBest1, we find a sensitive control of ion selectivity in the bestrophins, including reversal of anion/cation selectivity, and dramatic activation by mutations at the cytoplasmic exit. A homology model of hBest1 shows the locations of disease-causing mutations and suggests possible roles in regulation.

Insight into a retinal degeneration disease

Human bestrophin 1 (hBest1) is a membrane protein that forms a chloride channel in the retinal pigment epithelium. Mutations in hBest1 can lead to a retinal degeneration disease known as Best disease. Yang et al. describe the structure of KpBest, a bacterial homolog of hBest1. KpBest forms a pentamer with an ion channel at its center. In contrast to hBest1, KpBest1 is a sodium channel. The structure suggests a mechanism for ion selectivity that was confirmed by mutagenesis of KpBest and hBest1. A model of the hBest1 channel structure based on the KpBest structure reveals how mutations cause disease.

Science, this issue p. 355

The human BEST1 gene encodes a protein [human bestrophin-1 (hBest1)] that is highly expressed in retinal pigment epithelium (14). More than 120 distinct mutations in hBest1 have been identified that result in multiple retinal degeneration disorders (511), notably vitelliform macular degeneration or Best disease. Functionally, hBest1 was identified as a Cl channel that can be activated by Ca2+ (8, 12, 13), and most of the disease-causing mutations in hBest1 are point mutations that cause channel dysfunction (8, 12, 1416). Thus, understanding the structure of the hBest1 channel holds value from both biological and biomedical perspectives.

The bestrophin family identified by hBest1 is distributed widely, with representatives in most metazoan animals, including four in humans, and also in other eukaryotes and in prokaryotes (7, 8). The animal bestrophins are characterized by a highly conserved N-terminal domain that includes four predicted transmembrane helices (TMs) and diverse C-terminal domains that may be involved in protein-protein interactions (8, 12, 15, 17). Bacterial bestrophins lack the variable C-terminal domain and are more divergent in the transmembrane portion. Using a structural genomics approach, we identified a homolog from Klebsiella pneumoniae (KpBest) that could be produced by recombinant expression for structural and functional characterization. The structure-based sequence alignment implies 14% identity between KpBest and hBest1 (fig. S1).

Initial crystals of detergent-solubilized KpBest diffracted poorly; however, constructs from a truncation series did yield suitable crystals. The initial structure was solved from one of these, grown from a solution containing zinc acetate, at 2.9 Å resolution by single-wavelength anomalous diffraction (SAD) at the zinc K-edge resonance. Improved diffraction was obtained after further truncation, removing a total of 11 residues from the C terminus (fig. S1), and the structure was further refined to 2.3 Å resolution (tables S1 and S2). The building and refinement of the structural model were facilitated by five-fold noncrystallographic symmetry. The refined model comprises ordered residues from 22, 23, or 24 through 285 or 289 in different protomers, plus Zn2+ ions and water molecules.

Bestrophins have been predicted by different groups to form dimers, tetramers, or pentamers (12, 18). Here, we found that KpBest forms a stable pentamer (Fig. 1, A and B, and fig. S2) with large intersubunit contacts (26,880 Å2 total buried surface area). The electrostatic potential surface is largely negative on the extracellular surface, neutral in the transmembrane region, and positive at the cytoplasmic membrane surface (Fig. 1, C and D). Consistent with the experimentally determined topology of hBest1 (19), each protomer has four transmembrane helices and the N and C termini both reside on the cytoplasmic side (Fig. 1, E and F). Extracellular interhelix loops TM1-TM2 (12 residues) and TM3-TM4 (3 residues) are short, whereas the intracellular connection between TM2 and TM3 is long (105 residues), comprising five helices (α3 to α7) that form a separate cytoplasmic domain together with the C-terminal helix α10 (red in Fig. 1, E and F). Both the transmembrane and the cytoplasmic helical bundles appear to have novel folds: Dali searches find matches only to fragments of these structures.

Fig. 1 Crystal structure of KpBest.

(A and B) Ribbon diagram of the KpBest pentamer with each protomer colored differently: (A) as viewed from outside the membrane and (B) as viewed from the side (rotated 90° through the x axis). (C and D) Electrostatic potential at the molecular surface viewed as in (A) and (B), respectively. The contour level is at ±5 kT/e; red for negative potential and blue for positive potential. Membrane boundaries in (B) and (D) were calculated by OPM (Orientations of Proteins in Membranes) server. (E) 2D topology of a protomer, colored spectrally from dark blue at its N-terminal segment to red at its C-terminal segment. (F) Ribbon diagram of a protomer. Colored as in (E).

The TM2 helices line the putative ion-conducting pore through the membrane, and they continue intracellularly as long and curved, but uninterrupted, helices α2 (light blue in Fig. 1, E and F). In contrast, TM3/α8 and TM4/α9 are connected to the cytoplasmic domain by extended segments (4 and 12 residues, respectively). The α9-α10 connection is a loop structure that corresponds to a conserved carboxylate-rich segment (EDDDDFE) in eukaryotes, possibly playing a role in Ca2+ regulation. This segment has fewer carboxylate residues in prokaryotic homologs, but it presents an electronegative surface patch in KpBest nevertheless (Fig. 1D). The α7 helices (light yellow in Fig. 1F) and cytoplasmic portions of α2 line a cytoplasmic cavern beneath the transmembrane pore.

An apparent ion conduction pathway is at the center of the KpBest pentamer. A funnel-shaped electronegative vestibule, penetrating midway into the membrane, precedes a hydrophobic five-helix transmembrane pore (Fig. 2A). The pore is followed by a cytoplasmic cavern with a restricted, on-axis exit ~46 Å below the membrane. The pore and upper parts of the cavern are highly conserved, whereas outer surfaces and lower parts of the cytoplasmic domain are not (Fig. 2B). Overall, the ion permeation pathway has a flower-vase shape, with one restriction (radius < 2.0 Å) from three rings of TM2 residues (I62, I66, and F70) at the pore and another (I180, radius = 1.2 Å) at the start of cytoplasmic helix α7 (Fig. 2C). Therefore, the structure of KpBest predicts two distinct permeation restrictions in the ion passageway, providing a vital clue for the functional mechanism of bestrophin channels. Notably, all four residues located at the predicted restrictions are highly conserved and/or disease related in hBest1: I76, F84, and I205 (KpBest I62, F70, and I180, respectively) are identical (fig. S1), whereas point mutation of either F80 or I205 (KpBest I66 and I180, respectively) causes retinal disorders (6, 20, 21).

Fig. 2 Structure of the ion-conducting pathway through KpBest.

(A) Cross section through the pore center. The model is viewed as in Fig. 1D, with the electrostatic potential shown on exposed surfaces of the molecular envelope. (B) Cross section as in (A), but colored by Consurf sequence conservation. Turquoise marks the most variable positions, and maroon marks those most conserved. The calculation used 150 prokaryotic homologs with 95% maximal and 35% minimal sequence identities compared with KpBest. (C) Ribbon diagram of two oppositely facing (144°) protomers of a KpBest pentamer are shown with the extracellular side on the top. The side chains of critical residues are red.

Although eukaryotic bestrophins are known as Ca2+-activated Cl channels, the function of KpBest had not been previously examined. To test its function, purified KpBest was fused into a planar lipid bilayer with 150 mM of NaCl in both the trans (internal) and cis (external) solutions. Applying a range of transmembrane potentials resulted in well-resolved unitary currents with a linear single-channel I-V relationship (Fig. 3, A and B), confirming that KpBest is indeed an ion channel. Ca2+ was not required for KpBest activation, as might have been expected given that KpBest lacks the C-terminal domain that contains putative Ca2+ binding sites in eukaryotic bestrophins (7, 13, 22). Strikingly, with 150 mM of NaCl on the cis side and no NaCl on the trans side, inward single-channel currents were recorded (Fig. 3C) with mean amplitude of –5.3 pA (fig. S3A, left), demonstrating that KpBest is a cation channel that conducts Na+, unlike Cl-conducting eukaryotic bestrophins. To fully assess KpBest ion selectivity, reversal potentials under various bionic conditions were measured. KpBest is permeable to monovalent cations with rank order Na+ > K+ ≈ Cs+ but not to bivalent cations Mg2+, Ca2+, or Ba2+ (Fig. 3D). It is noteworthy that ion channels in the same family can have reversed charge selectivity, as exemplified by TMEM16 Ca2+-activated channels: TMEM16A and 16B are anion channels, whereas TMEM16F may conduct cations (23).

Fig. 3 Ionic conductance measurements of KpBest and hBest1.

(A) Representative families of single KpBest currents recorded from planar lipid bilayers at different voltages (150 mM of NaCl in both trans and cis solutions). No current was recorded at –60 mV. (B) Single KpBest channel current-voltage relationship. (C) Current trace of WT single KpBest channels (150 mM of NaCl in cis and 0 NaCl in trans solutions). The voltage for recordings (C), (E), and (H) was 0 mV. (D) Relative cation permeability; n = 3 recordings in independent bilayers for each point. (E) Current traces of mutant KpBest channels [same condition as (C)]. (F) Critical residues in KpBest and hBest1. (G) Relative permeability (PNa/PCl) of hBest1 WT and mutant variants; n = 12 to 15 cells for each bar. (H) Current trace of I180A single KpBest channels [same condition as (C)]. (I) WT and I180A single KpBest channel open probabilities. (J) Exemplar whole-cell currents of WT (left) and I205A (right) hBest1 in HEK 293 cells. (K) Population steady-state current-voltage relationships; n = 3 to 6 cells for each point. * P < 0.05 when compared with WT using two-tailed unpaired Student’s t test. No Ca2+ was added for hBest1 recordings.

Despite extensive studies on the ion-conducting pores of eukaryotic bestrophins, including hBest1 and mouse bestrophin-2 (mBest2) (15, 2426), the molecular basis for ion selection in these channels is not clear. Our KpBest model predicts three critical residues (I62, I66, and F70) at the first permeation restriction (Fig. 2C) that likely control ion selectivity. To test this hypothesis, we first examined I66, because it is the only residue among the three that is different in KpBest compared with anion-conducting bestrophin channels (where this residue is F) (Fig. 3F and fig. S1) (12, 15, 2427). In bilayer experiments, KpBest I66F showed outward current with 150 mM of NaCl on the cis side and no NaCl on the trans side, indicating that this mutant channel conducts Cl rather than Na+ (Figs. 3E, top, and fig. S4). To further test this premise in the eukaryotic counterpart, wild-type (WT) and F80I mutant hBest1 (corresponding to KpBest I66) were transfected into human embryonic kidney (HEK) 293 cells, and their reversal potentials were determined in whole-cell voltage clamp experiments. Consistent with the KpBest1 I66F results, hBest1 F80I was much less permeable to Cl compared with WT (PNa/PCl = 0.39 for F80I, compared with PNa/PCl = 0.03 for the WT) (Fig. 3G and fig. S5).

Inspired in part by selectivity-flipping changes made at F81 (equivalent to KpBest I66) in Drosophila melanogaster bestrophin-1 (dBest1) (28), we next examined whether KpBest ion selectivity could be altered by individually substituting each of the hydrophobic pore-lining residues with positively charged arginine (R), which in principle might favor negatively charged Cl. When purified KpBest I62R, I66R, and F70R mutants were tested in bilayer experiments, I62R (but not I66R or F70R) conducted Cl rather than Na+ (Fig. 3E and fig. S4). These mutants showed perturbed conductance amplitudes (fig. S4), indicating effects on permeation at all of these sites. Following the same logic as for KpBest, the equivalent residues on hBest1 (I76, F80, and F84, respectively) were individually mutated to negatively charged glutamic acid (E). Consistent with the KpBest results, only I76E flipped the ion selectivity to Na+ (PNa/PCl = 1.54) (Fig. 3G and fig. S5B), although F80E and F84E were also less permeable to Cl than WT (PNa/PCl = 0.33 and 0.46, respectively) (Fig. 3G). None of the hBest1 mutations significantly affected current density (fig. S6). Notably, our results are in accord with previous reports: The rectification of mBest2 could be altered in opposite directions by replacing F80 with either R or E (26), and the corresponding F81E mutation of dBest1 flipped the cation/anion selectivity (28). Taken together, using our KpBest model as a guide, we have identified three residues that sensitively affect ion selectivity in bestrophins.

The dramatic change in ion selectivity from substitutions at the first hydrophobic residue of the pore (I62R in KpBest and I76E in hBest1) suggests a particularly critical role for this position. Interestingly, the expression level of I62R in E. coli was much lower compared with that of WT KpBest (about 1/14) (fig. S7), suggesting that I62 is important for the expression or assembly of the channel. Consistent with this idea, mutating the equivalent mBest2 I76 to C/L/V resulted in no currents (26). Surprisingly, the simple swapping of the central hydrophobic pore residues (KpBest I66 and hBest1 F80) significantly altered ion selectivity without affecting expression levels. The hydrophobic character of the five-helix transmembrane pore of bestrophins is reminiscent of that in the SLAC1 channel (29), where the anion selectivity series, like that observed for mBest2 (24, 25) and also for TMEM16A (30, 31), is inversely related to the hydration energy of monovalent anions (32).

The second restriction in the permeation pathway of KpBest, at the base of the cytoplasmic cavity where I180 residues interact (Fig. 2C), suggests a possible permeation gate. To test this idea, we replaced the bulky isoleucine with alanine and tested the purified I180A mutant channels in bilayer experiments. Indeed, I180A channels conducted unitary currents with the same amplitude as for WT (fig. S3A); however, the open probability was markedly enhanced (Figs. 3, H and I, and fig. S3B). Substituting I180 on KpBest with R, which has a longer side chain than A, yielded mutant channels that still conducted Na+ (Fig. 3E) with a low open probability similar to WT KpBest. To determine whether the equivalent residue in hBest1 also acts as an activation gate, the comparable I205A/E mutants were generated and subjected to a whole-cell voltage clamp. Phenocopying their respective KpBest counterparts, hBest I205A displayed significantly larger currents compared with WT and lost inward rectification, indicating a change in channel gating (Fig. 3, J and K), whereas I205E barely affected channel ion selectivity (PNa/PCl = 0.06) (Fig. 3G). Thus, the gate predicted by our KpBest model (KpBest I180 and hBest1 I205) controls gating of the channels but not selectivity. Notably, hBest1 I205T is a disease-causing mutation with significantly decreased Cl conductance (6), reinforcing the functional importance of I205.

To understand characteristics of hBest1, including the disposition of disease-causing mutations, we generated a homology model for its transmembrane portion based on the KpBest structure (Fig. 4). Consistent with the anion selectivity, the transmembrane pore and the lower restriction of hBest1 are both positively charged (Fig. 4A), contrasting with the negative interior of the KpBest permeation pathway. The conservation pattern among eukaryotic bestrophins (Fig. 4B) is roughly similar to that of prokaryotic relatives (Fig. 2B), although the cytoplasmic exterior is less variable. Disease-causing mutations in hBest1 are abundant ( and have been clustered into hot spots in the sequence (fig. S1) (8, 33). When plotted onto the three-dimensional (3D) model (Fig. 4C and fig. S8B), the hot spots segregate into a cavity-lining set from hot spot 2 along α2 and a juxtamembrane set from hot spots 3 and 4 at the top of α7 and at the α9-α10 carboxylate loop, which is even more negative in hBest1 (fig. S8A) than in KpBest (Fig. 1D). The carboxylate loop could conceivably play a role in regulation by Ca2+. Hot spot 1 is along the N-terminal segment that is disordered in KpBest, but presumably it too would be alongside other hot spots in the juxtamembrane region in hBest1.

Fig. 4 Homology model of hBest1.

(A) (Top) Electrostatic potential at the extracellular surface of the hBest1 homology model. Viewed and drawn as for Fig. 1C. (Bottom) Cross section through the homology model of hBest1. Viewed and drawn as for Fig. 2A, except that the cut surface is plain gray. (B) (Top) Top view as in (A), but colored by surface conservation. Turquoise marks the most variable positions, and maroon marks those most conserved. This calculation used 150 homologs with 95% maximal and 35% minimal sequence identities compared with hBest1. (Bottom) Cross section viewed and drawn as in Fig. 2B but having a plain gray cut surface. (C) Ribbon diagram of a protomer from the hBest1 homology pentamer. The Cα positions of residues that are sites of disease-causing point mutations are marked in red.

In summary, our structure of the bacterial homolog KpBest provides a solid basis for understanding human bestrophin activity and disease-causing mutations, and physiological tests on KpBest and hBest1 provide insights into the control of ion selectivity and activation in bestrophins generally.

Supplementary Materials

Materials and Methods

Figs. S1 to S8

Table S1 and S2

References (3444)

References and Notes

  1. Acknowledgments: We thank O. Clarke for help in the diffraction analysis; A. Marks for use of planar lipid instrumentation; W. Xie for help in planar lipid experiments; Q. Fan, Y. Geng, and members of the Hendrickson laboratory for discussions; and J. Schwanof and R. Abramowitz at National Synchrotron Light Source (NSLS) beamlines X4A and X4C and F. Murphy at the Advanced Photon Source beamline 24-ID-E for their assistance in data collection. This work was supported in part by NIH grants GM095315 and GM107462 to W.A.H. X4 beamlines are supported by the New York Structural Biology Center at the NSLS of Brookhaven National Laboratory, a Department of Energy facility. The data reported in this paper are tabulated in the supplementary materials and deposited to the Protein Data Bank with access codes listed in table S2. Author contributions: T.Y. and W.A.H. designed research, analyzed data, and wrote the paper; T.Y. performed experiments; Q.L. analyzed diffraction data; B.K., R.C.K., and R.B. performed expression tests; Y.G. advised on protein production and analyzed diffraction data; E.K. and B.R. performed bioinformatics analyses; and H.M.C. designed and analyzed electrophysiology experiments. The authors declare no conflicts of interest.

Stay Connected to Science

Navigate This Article