This article reviews the current state of knowledge about the bestrophins, a newly identified family of proteins that can function both as Cl− channels and as regulators of voltage-gated Ca2+ channels. The founding member, human bestrophin-1 (hBest1), was identified as the gene responsible for a dominantly inherited, juvenile-onset form of macular degeneration called Best vitelliform macular dystrophy. Mutations in hBest1 have also been associated with a small fraction of adult-onset macular dystrophies. It is proposed that dysfunction of bestrophin results in abnormal fluid and ion transport by the retinal pigment epithelium, resulting in a weakened interface between the retinal pigment epithelium and photoreceptors. There is compelling evidence that bestrophins are Cl− channels, but bestrophins remain enigmatic because it is not clear that the Cl− channel function can explain Best disease. In addition to functioning as a Cl− channel, hBest1 also is able to regulate voltage-gated Ca2+ channels. Some bestrophins are activated by increases in intracellular Ca2+ concentration, but whether bestrophins are the molecular counterpart of Ca2+-activated Cl− channels remains in doubt. Bestrophins are also regulated by cell volume and may be a member of the volume-regulated anion channel family.
Cl− channels are ubiquitously expressed in both eukaryotic and prokaryotic cells. In eukaryotes, Cl− channels function both at the plasma membrane and in intracellular organelles. Physiologically, Cl− channels can be broadly classified as ligand gated, volume regulated (VRAC), voltage sensitive, or second messenger activated (e.g., cAMP or Ca2+). The molecular counterparts of some of these physiologically described channels are known. Ligand-gated anion channels include the GABAA and glycine receptors, voltage-gated anion channels include some members of the ClC family, and second messenger-activated channels include the cystic fibrosis transmembrane conductance regulator (CFTR) (88). However, many physiologically described anion channels have not yet been unambiguously identified at the molecular level. For example, although a number of molecular candidates have been proposed for VRACs (reviewed in Refs. 29, 88, 147) and Ca2+-activated Cl− (CaC) channels (reviewed in Refs. 44, 74), none of them has yet been universally accepted. The recent discovery of a new family of anion channels, the bestrophins, has therefore been met with considerable interest. Because at least some of the members of this family are activated by increased intracellular Ca2+ (170), bestrophins have been proposed to be the molecular counterpart of at least a subset of CaC channels. Whether this in fact is true remains to be demonstrated convincingly. Bestrophins are also sensitive to cell volume (27, 55), but whether bestrophins are VRACs is also too early to tell.
There is compelling evidence that bestrophins are Cl− channels, even if their mechanisms of activation and regulation remain undefined. Expression of a variety of different bestrophins in HEK cells induces novel Cl− currents (13, 28, 167–171, 196, 199, 200). Because mutation of certain amino acids alters the permeability and conductance of the channel (168, 169, 171) and sulfhydryl reagents alter the properties of cysteine-substituted channels in characteristic ways (169, 171, 199), bestrophin is likely to constitute part of the ion conduction pathway (165). This conclusion is reinforced by the finding that different bestrophin paralogs have different biophysical properties (199). Endogenous Cl− currents can be suppressed by knocking down bestrophin transcripts with RNAi (13, 28).
Despite the evidence that bestrophins constitute a new family of Cl− channels, relatively little is known about their physiological roles, except for the role of hBest1 in the eye (76). Bestrophins were first found by genetic linkage of hBest1 to a juvenile form of macular degeneration called Best vitelliform macular dystrophy (“Best disease”) (124, 159). Subsequently, mutations in hBest1 were found also to be associated with a fraction of other types of adult-onset macular dystrophy. Most, if not all, of the disease-causing mutations in hBest1 produce a dysfunction in the Cl− channel function of hBest1 (116, 196, 214, 215). But, although hBest1 clearly can function as a Cl− channel, it is not yet agreed that Best disease is caused by Cl− channel dysfunction. The uncertainty arises largely from the fact that mBest1 knockout mice have no ocular pathology and apparently have normal Cl− currents in the retinal pigment epithelium (RPE) where Best1 is expressed (123). Furthermore, there is an imperfect correlation between Best1 mutations and the electrophysiological features that define Best disease. Finally, the mechanisms leading from dysfunction of Cl− channels to retinal pathology and macular degeneration are not understood.
The idea that hBest1 is a Cl− channel has been questioned by some investigators (119–121, 123, 177). This challenge is based largely on several observations, the most compelling of which is that CaC currents in RPE cells are not abolished in mBest1 knockout mice (123). This observation, coupled with the observation that hBest1 can regulate voltage-gated Ca2+ channels in an RPE cell line (177), has led to the suggestion that hBest1 is not a Cl− channel but is rather a channel regulator (123). These different viewpoints bring to mind another Cl− channel: CFTR. There are a number of significant parallels between bestrophins and CFTR. CFTR is clearly a Cl− channel (14), but it also regulates other membrane transport proteins, including the epithelial Na+ channel (ENaC) (114, 128, 183, 206). Furthermore, CFTR knockout mice do not exhibit lung pathology as expected, partly because other Cl− channels subserve the function of CFTR in mouse lung (see references in Refs. 58, 114). In both cystic fibrosis and Best disease, multiple mutations (100 in Best disease and 1,525 in CFTR) distributed throughout the protein have been linked to disease. Like CFTR, bestrophins may be Cl− channels that also regulate other ion channels. It remains to be seen whether the macular degeneration caused by mutations in hBest1 can be solely attributed to the loss of Cl− channel activity of this protein or whether, like CFTR, other functions come into play.
In this review, we summarize the state of the bestrophin field at this point in time. There remain many uncertainties, but we will try to develop the idea that bestrophin, like CFTR, is a multifunctional protein that is a Cl− channel and a regulator of other channels.
II. BESTROPHIN MOLECULAR BIOLOGY
A. Discovery of the hBest1 Gene
Identification of the gene causing Best vitelliform macular dystrophy (BVMD) was facilitated by studies in the 1970s in which the inheritance pattern of BVMD was studied in a large Swedish family. More than 250 cases were traced back to a single gene source in the 17th century (150). The locus for BVMD was determined by linkage analysis to be located in a pericentric region on chromosome 11, at 11q13, near the markers D11S956, FCER1B, and UGB (56, 194). The gene for rod outer segment protein-1 (ROM1), which is associated with digenic retinitis pigmentosa, maps within this region. Thus, for a time, ROM1 became a leading candidate for the BVMD-causing gene. However, ROM1 was excluded as the causative gene by several studies which showed that BVMD did not segregate with ROM1 and that mutations in ROM1 were not found in BVMD patients (65, 66, 80, 144). The BVMD locus was further refined (30, 193, 205), and the gene was then cloned by recombination breakpoint analysis by Petrukhin et al. (159) and by Marquardt et al. (124) who systematically analyzed mutations in genes within overlapping YAC (166) and PAC clones (30) of the BVMD locus. Although at one point it was suggested that there was genetic heterogeneity in BVMD, this does not seem to be the case (115). The gene was initially called VMD2 (for vitelliform macular dystrophy type 2).
There are two systems of nomenclature for the mammalian bestrophins: Best-1, -2, -3, and -4 (196, 199) and VMD2, VMD2L1 (VMD2-like protein 1), VMD2L2, and VMD2L3 (98, 124, 192). There has been some confusion about the correspondence between the Best and the VMD2 nomenclatures. However, recently the HUGO and the Mouse Genome Database nomenclature committees have recommended the Best nomenclature. Table 1 shows the characteristics of the four human and three mouse bestrophin genes.
C. Paralogs and Phylogenetic Relationships
Bestrophins have been unambiguously identified in mammals, birds, bony fish, amphibians, echinoderms, insects, nematodes, and flat worms (Figs. 1 and 2). Distant homologs have also been found in fungi, plants, and bacteria (72). Gene orthologs (genes that evolved from a common ancestral gene by speciation) and paralogs (genes related by duplication within a genome) (108) can be inferred from the phylogenetic tree. All of the vertebrate bestrophins fall into one of four paralogous branches (130) (Fig. 1). For example, the human genome contains four bestrophin paralogs (hBest1, hBest2, hBest3, and hBest4) (192, 199). All other mammals have either three or four bestrophin paralogs. Mouse, for example, has three paralogs and one pseudogene (98). Arthropod sequences are sufficiently divergent that they cannot be assigned to one of the four vertebrate bestrophin paralog groups. Because there are relatively few arthropod sequences, it is difficult at this time to be certain how many paralogs exist in arthropods, but it seems that there are four paralogs in insects such as Drosophila, Apis, and Anopheles. Caenorhabditis elegans and Caenorhabditis briggsae have 25 and 21 bestrophin paralogs, respectively. The evolutionary reasons for this expansion of the bestrophin family in Caenorhabditis remain unknown. One bestrophin has been found in each of the primitive chordates Ciona intestinalis and Ciona savignyi and in the echinoderm Stronglyocentrotus purpuratus (sea urchin). Hydropathy analysis and transmembrane domain (TMD) prediction algorithms usually find six potential transmembrane domains (boxes in Fig. 2). Predicted TMD 2, 5, and 6 are very highly conserved from bacteria to human.
Putative fungal, plant, and prokaryotic bestrophin genes are very distantly related to the animal bestrophins, but many key functional residues are conserved, suggesting that these proteins are truly related to the animal bestrophins. However, no physiological evidence yet exists to show that any of these are Cl− channels. Comparing the amino acids that are conserved between mammalian and prokaryotic bestrophins may provide some insights into the key functional domains of these proteins. hBest1 and the bacterial Ralstonia solanacearum bestrophin (NP_521533) are 15% identical and 26% conserved in the first 350 amino acids (Fig. 2). Hydropathy analysis predicts four transmembrane domains that correspond to predicted transmembrane domains 1, 2, 5, and 6 of hBest1. Within these putative transmembrane domains, amino acid conservation is ∼43%. Also, 28 of the 78 positions in hBest1 that are mutated in BVMD are identical in wild-type hBest1 and the bacterial protein. Interestingly, the regions immediately before the first and after the last predicted transmembrane domains are also very highly conserved.
D. VMD2 Promotor
Zack and colleagues (48, 49) have identified the promotor region of hBest1. A fragment consisting of bases −154 to +38 is sufficient to direct RPE-specific expression of a luciferase reporter in transgenic mice. This region contains two E-boxes, 1 and 2, that are essential for promoter activity. The factors that regulate the promotor were identified by yeast one-hybrid screens using bait containing E-box 1 and by chromatin immunoprecipitation with antibodies against these proteins. The factors that were found included MITF, TFE3, and TFEB. Transfection and siRNA studies provided evidence that hBest1 is regulated by the MITF-TFE family through both E-boxes. E-box 1 is required for a direct interaction of MITF-TFE factors, and E-box 2 is required for binding of another unidentified factor. The hBest1 promotor has been used very successfully to drive the specific expression of other proteins in RPE cells (91, 153).
III. BESTROPHIN STRUCTURE-FUNCTION
A. Cl− Channel Function
The first clear evidence that bestrophins are Cl− channels was presented by Sun et al. (196) who showed that expression of hBest1 and several other bestrophins in HEK cells induced Cl− currents. Figure 3, A–C, compares currents from untransfected HEK cells or HEK cells transfected with hBest1. When untransfected HEK cells are patch clamped with an intracellular solution containing micromolar free Ca, the currents are very small, whereas hBest1-transfected cells have large currents. The currents show a small amount of outward rectification but are time independent. hBest1 currents are dependent on intracellular Ca, because when hBest1-transfected cells are patched with nominally zero free Ca2+ (buffered with EGTA), the currents are very small. Although hBest1 has no time dependency and shows little rectification, other bestrophins show pronounced rectification and time dependence (196, 199) (Table 1). This difference in biophysical properties of different bestrophins is a strong indication that the Cl− conductance pathway induced by bestrophin expression is actually formed by bestrophin. The difference would be more difficult to explain by the alternative possibility that bestrophins are Cl− channel regulators.
Another piece of evidence that bestrophins are Cl− channels is provided by the finding that it is possible to alter the gating, the conduction, and the pharmacology of bestrophin channels by mutagenesis (168–171, 199). For example, Figure 3D shows that substitution of tryptophan at position 93 with histidine (W93H) in mBest2 results in a dramatic change in channel gating. In contrast to the wild-type current, which is time independent and has a nearly linear current-voltage (I-V) curve like hBest1 (Fig. 3B), the W93H channel activates slowly with time upon hyperpolarization and strongly inwardly rectifies. The rectification of the channel can be altered in opposite ways by changing the electrostatic charge at position 80 by replacing the native phenylalanine with either a positively charged arginine or a negatively charged glutamate (169) (Fig. 3E). Rectification of the channel can similarly be altered by replacing F80 with cysteine (F80C) and then modifying the charge with either positively or negatively charged sulfhydryl reagents (MTSET+ or MTSES−) (Fig. 3F). Finally, certain mutations alter the affinity of the channel for Cl− channel blockers. Wild-type mBest2 is blocked by DIDS with an IC50 of ∼3 μM, whereas the S79C mutant is about five times less sensitive (168). These manipulations all show that bestrophin is an integral part of the Cl− channel that is induced when bestrophin is expressed heterologously.
Two different topology models have been proposed for hBest1 (130, 199; see also Ref. 121) (Fig. 4). Tsunenari et al. (199) performed a very detailed analysis of hBest1 topology and predicted that of the six hydrophobic domains predicted by hydropathy, TMD1, TMD2, TMD4, and TMD6 traversed the membrane and that TMD5 might be a reentrant loop (Fig. 4A). This prediction was based on data from three different methods. They substituted cysteine residues for selected amino acids in hBest1 and examined the effect of extracellular MTSET+ (a membrane impermeant sulfhydryl reagent) on heterologously expressed Cl− currents. Because MTSET+ is membrane impermeant, alteration of the current indicates that the substituted cysteine is accessible to the hydrophilic extracellular space. Thus one would expect MTSET+ to modify the current only if the cysteine is located in the extracellular space or in a transmembrane domain that lines the pore. In addition to substituted cysteine accessibility experiments, Tsunenari et al. (199) also examined the ability of inserted N-glycosylation sites to be glycosylated, which occurs in a topologically extracellular compartment (Fig. 4A; glycosylated, solid red residues; not glycosylated, red letter). Finally, TEV protease sites were inserted and their ability to be cleaved by protease added to the topologically cytoplasmic compartment were examined (Fig. 4A; TEV protease sites cleaved, solid cyan residues; not cleaved, cyan letter).
Weber and colleagues (130) tested the topology of hBest1 by examining the ability of various hBest1 fragments consisting of predicted transmembrane domains or groups of predicted transmembrane domains to be incorporated into the membrane in the presence of a Lep H1 membrane targeting domain. They found that TMD1, -2, -4, -5, and -6 were capable of being incorporated into the membrane when they were alone. TMD4 was able to incorporate into the membrane when it was presented alone or with TMD3; however, it was not able to incorporate when it was presented with TMD5, TMD2 + TMD3, TMD3 + TMD5, or TMD5 + TMD6. These results suggested that only TMD1, -2, -5, and -6 cross the membrane (Fig. 4B).
These two models are directly contradictory. The Tsunenari model places amino acids ∼200–231 in an extracellular loop, whereas in the Milenkovic model, this domain is included in a large cytoplasmic loop (amino acids 106–230). Our prejudice is that insertion of glycosylation sites will perturb the protein less than cutting it into fragments and measuring the ability of these fragments to insert into the membrane in the presence of another membrane-targeting signal. Tsunenari's evidence that N-glycosylation sites inserted after amino acids 212, 218, 223, and 227 are glycosylated and that H223C and A226C are modified by MTSET seem strong pieces of evidence that amino acids 212–227 are extracellular. Furthermore, the data of Milenkovic et al. about the ability of TMD4 to enter the membrane is a little ambiguous. Although TMD4 does not enter the membrane when it is combined with adjacent TMDs, it does enter the membrane alone and with TMD3. On the other hand, if these residues are extracellular and thus presumably not in the permeation pathway, how does MTSET+ modification affect anion conductance? The effect may be allosteric. Our finding that the Cl− currents induced by cysteine-substitution mBest2 residues R140, T146, K149, R150, F151, E160, K180, R196, R197, R200, D228, W229, or I230 have properties like wild type and are not affected by MTSET+ is consistent with, but certainly does not prove, a cytoplasmic location of these residues (Qu and Hartzell, unpublished data).
All the methods used to infer membrane topology have the limitation that artificial constructs may not fold and enter the membrane in the same way as the native protein. In this regard, these data are reminiscent of the controversies regarding the topology of ClC channels, which was not resolved until the crystal structure became available. Although experimental data for bestrophin transmembrane topology exist mainly for hBest1, the high conservation of the NH2-terminal 350 amino acids, which includes all the predicted transmembrane domains, suggests that the topologies will be similar for all vertebrate bestrophins.
C. Quaternary Structure
Because many ion channels are oligomers, one might expect that bestrophins are also oligomeric. Sun et al. (196) showed that bestrophins exist as tetramers or pentamers when expressed heterologously. Bestrophins tagged with myc or Rim3F4 epitopes were transfected into HEK cells. When myc-tagged hBest1 was cotransfected with Rim3F4-tagged hBest1, myc antibody precipitated a significant amount of Rim3F4-tagged hBest1. By transfecting with different ratios of myc- and Rim3F4-tagged hBest1, they estimated that the channel was comprised of four or five subunits. The immunoprecipitation was specific: individual subunits did not associate when mixed together in detergent, and C. elegans and Drosophila bestrophins associated with hBest1 at much lower efficiency than they associated with themselves. These studies show that hBest1 can exist as a homomeric tetramer or pentamer. hBest1 can also associate with hBest2, so the possibility exists that bestrophins can also exist as heteromers.
A different conclusion has been reached from hydrodynamic studies by Stanton et al. (188), who conclude that native bestrophin is a dimer. In their studies, pBest1 was extracted from porcine RPE cells with a buffer containing 1% Triton X-100. pBest1 exhibited a sedimentation coefficient of 4.9S in sucrose gradients and a Stokes radius of ∼7.3 nm in gel filtration. Using a modified Svedberg equation, the authors calculated the mass of the pBest1-detergent complex to be 206 kDa. Adjusting for the fraction of the complex that is detergent (31%), pBest1 was estimated to have a mass of 138 kDa, which is almost exactly twice the mass of the monomer.
In an attempt to resolve the discrepancy with the results of Sun et al. (196), Stanton et al. (188) performed similar measurements on pBest1 extracted from transfected HEK cells and found that a significant fraction of the protein was aggregated. They suggest that the tetrameric/pentameric stoichiometry proposed by Sun et al. (196) was an artifact of overexpression. An alternative possibility is that the detergent and ionic conditions used in the hydrodynamic studies disrupt a higher order structure.
Each of the approaches used in these studies has its own limitations, and determination of membrane protein stoichiometry has rarely been straightforward. Nevertheless, both studies agree that bestrophins can oligomerize, although the number of monomers that assemble may be in question. The conclusion that bestrophins oligomerize is consistent with functional studies that show that certain bestrophin mutants can inhibit the Cl− channel activity of wild-type subunits, as might be expected if mutant and wild-type subunits assemble together (see sect. viiA).
D. Cl− Channel Pore
Nathans and co-workers (199) suggested that TMD2 of hBest1 might be involved in forming the pore because cysteine-substituted mutants along the entire length of TMD2 between amino acids 69 and 99 were accessible to extracellularly applied MTSET+. Because hydropathy analysis predicted that TMD2 traversed the membrane, the most plausible explanation was that this region lined an aqueous channel that was continuous with the extracellular environment. The fact that intracellular cysteine did not quench the MTSET+ effect further supported the idea that MTSET+ was moving into the channel pore from the extracellular space.
TMD2 is one of the most highly conserved regions of the bestrophins. Among the animal bestrophins, amino acids 65–97 are ∼40% identical. P77, F80, L82, G83, F84, Y85, V86, R92, W93, and W94 are identical in all animal bestrophins and are likely to be critical for bestrophin function. Vertebrate bestrophins are identical from amino acid I76 to W94 with the exception of position 78 which can also be V, I, L, or M and position 79 which can be S or T. Furthermore, within this region, mutations at 10 different positions are associated with human disease.
The bestrophin pore has been most thoroughly analyzed for mBest2 (168, 169, 171). Using mutagenesis and cysteine-accessibility analysis of all amino acids from 69 to 105, Qu and co-workers (168, 169, 171) have shown that TMD2 very likely plays a role in ion selectivity of the pore. A73C, V78C, S79C, and F80C react most rapidly with anionic sulfhydryl reagents, suggesting that these residues may form the outer mouth of the channel. Amino acids in TMD2 closer to the COOH terminus react with MTS reagents more slowly than those closer to the NH2 terminus. This is consistent with the COOH-terminal end of the putative transmembrane domain being closer to the cytoplasm and deeper in the pore. Several other observations indicate that anionic selectivity is determined by this region of the protein. Replacement of F80 with amino acids of opposite charge had opposite effects on rectification of the current: F80R outwardly rectifies, whereas F80E inwardly rectifies (Fig. 3E). V78C exhibits the highest anionic selectivity as measured by rates of modification by anionic and cationic MTS reagents.
There is considerable structure-function similarity between hBest1 and mBest2 (Fig. 5); substitution of amino acids near the ends of the predicted transmembrane segment with cysteine makes both channels nonfunctional (169, 171, 199). Furthermore, MTSET+ stimulates the current when residues in the middle of the transmembrane domain are substituted with cysteine. However, there are some differences between hBest1 and mBest2 that are worth noting. Foremost is the finding that wild-type hBest1 current is modified by MTSET+ due to the endogenous cysteine at position 69 (199). However, wild-type mBest2, which also has C69, is not affected by MTSET+ (171). Furthermore, Tsunenari et al. (199) found that cysteine substitution of hBest1 residues V81, L82, Y85, L88, V89, and R92 produced nonfunctional channels, but the corresponding substitutions in mBest2 were functional (171). We have confirmed the results by Tsunenari et al. with the exception that we find V81C in hBest1 is functional. Finally, in mBest2, the conductance of SCN− is much lower than Cl−, but in hBest1, the conductance of SCN− and Cl− are very similar. These differences between hBest1 and mBest2 suggest that the fine structure of the pore may be slightly different between different bestrophins. Elucidating the mechanisms of anion permeation through bestrophin channels would benefit from more detailed studies on selectivity of different bestrophin subtypes.
Like many anion channels and unlike many cation channels, bestrophins are not very discriminating among permeant ions. mBest2, which is the only bestrophin whose anionic selectivity has been carefully examined, selects among anions with a lyotropic permeability sequence of SCN− > NO3− > I− > Br− > Cl− > F−. The most permeant ion, SCN−, is ∼10 times more permeant than Cl− (168). The lyotropic permeability sequence suggests that electrostatic interaction of the permeant ion with the pore is less important than hydrophobic interactions (46). This explains the observation that larger ions (SCN−), which are more easily dehydrated, are more permeant than smaller ions (F−) (169).
Because the selectivity filter of mBest2 is rather nondiscriminating, one might expect that the selectivity filter would have fewer structural requirements than other channels that are highly selective, like voltage-gated K+ channels. Mutations in 20/35 amino acids between 69 and 104 altered anion permeability or conductance (171). Most of these were clustered between residues 72 and 94. Mutations in 13 amino acids (amino acids 78–80, 82–87, 90, 92, 94, 98) increased relative SCN− conductance and decreased relative SCN− permeability, and mutations in 7 amino acids produced nonfunctional channels (amino acids 74–77, 91, 93, 101). By measuring the rates of modification of cysteine-substituted residues by anionic and cationic sulfhydryl reagents, it was concluded that residues 72–80 form an outer vestibule that is readily accessible to the extracellular aqueous environment and that V78, F80, and G83 exhibit the highest ability to select between anionic and cationic MTS reagents (171). From their results on MTSET+ modification of cysteine-substituted hBest1, Tsunenari et al. (199) proposed that TMD2 was an α-helix. A helical wheel representation of TMD2 suggested that amino acids on one side of the helix were affected by MTSET+. Presumably, this side of the helix would face the aqueous pore of the channel. Our cysteine-accessibility analysis of mBest2 extends this idea (171). A helical wheel plot of mBest2 TMD2 reveals that the residues that select most strongly between anions and cations are clustered on one side of the helix (Fig. 6). One apparent problem with this model is that V78, one of the residues that discriminates the best between anionic and cationic MTS reagents, is on the “wrong” side of the helix. However, because it is adjacent to proline 77, the helix may be disrupted at this point.
Significantly, none of the cysteine mutations (or other mutations that were tried) produced a dramatic change in ion selectivity, and no single amino acid residue exhibited priority in determining anion permeability. Mutations that altered permeability all had relatively similar effects: changes in permeability were significant only with SCN− as the permeant anion. Changes in relative conductance (G) were more robust (169, 171). GSCN/GCl and GNO3/GCl were changed approximately fivefold by some mutations. Like many Cl− channels, large anions can block bestrophin Cl− currents because they become lodged in the pore. Mutations S79C and F80C greatly reduce the ability of SCN− to block Cl− current, by reducing the apparent affinity of SCN− for the pore (168).
At first, we were disappointed that the effects of mutations in TMD2 were so modest. We had hoped that we could reverse the anionic-cationic selectivity of the channel by a point mutation, as has been shown for the glycine and nicotinic ACh receptors (63, 94). However, we soon came to appreciate that many anion channels have a less “specialized” pore than some other kinds of channels. The ClC family of Cl− channels offers a useful framework for thinking about bestrophin permeability, because the ClC crystal structure has been solved (40–42). In ClC channels, four separate, noncontiguous stretches of amino acids contribute to the selectivity filter (41, 88). Mutations of most (17/19) amino acid residues in one transmembrane segment of hClC-1 alter anionic permeability (50, 51), but the effects of these mutations are relatively modest. The G233A mutation in human ClC-1 channel increases PSCN/PCl only approximately eightfold and increases PNO3/PCl and PI/PCl only approximately threefold. In this regard, the present-day studies with mBest2 recall the early mutagenesis studies on hClC-1. Without knowledge of the tertiary and quaternary structure of bestrophins, interpreting mutagenesis data are like “working in the dark.” Although we have learned that certain residues are important in anion permeation, interpreting how these residues affect ion transport is a challenge without solid structural data. But, even with the crystal structure of the bacterial ClC channel, our understanding of anion permeation in ClC channels still remains rudimentary (e.g., Ref. 132). For example, although electrostatic interactions play an important role in the ClC selectivity filter (26), hydrophobic interactions apparently play a crucial, though less-well delineated, role. Unlike K+ channels where the permeant ion maintains its hydrophilic coordination in the pore (89, 217), the anions in the central and interior anion binding sites of the bacterial ClC channel are not coordinated by hydrophilic ligands but are instead partially in hydrophobic contact with aliphatic amino acid side chains (41, 43). Clearly, crystal structures of bestrophins would be a valuable addition for solving this problem.
The pores of anion channels are less discriminating than those of cation channels, possibly because Cl− is the predominant monovalent anion in biological systems, and thus there has been less evolutionary pressure to evolve a highly selective pore compared with cation channels. Dawson and colleagues (32, 185) have proposed that the detailed structure of the CFTR pore may not be a major factor determining anion selectivity because permeation is determined largely by how easily the anion exchanges its water of hydration with residues in the channel (32, 185). It seems that CFTR, ClCs, and bestrophins may share this characteristic. The suggestion that the specific structure of the channel pore is relatively unimportant in bestrophin selectivity is supported by the observation that qualitatively similar effects on channel permeability are produced by substitutions at position 79 with amino acids having a diversity of side chains. For example, PSCN/PCl and GSCN/GCl are similar in the S79A, S79E, and S79R mutants.
Electrostatic interactions are important in ClCs (26), CFTR (185), and ligand-gated anion channels (94). Thus one might expect bestrophins to also have charged residues in the pore. However, it is not obvious which residues might contribute electrostatic interactions in the bestrophins. The only amino acids in TMD2 of mBest2 that might be positively charged are H91 and R92. R92 is likely to be a critical residue because mBest2 macroscopic conductance is related to the charge at this position (171). Despite the evidence that TMD2 is involved in forming the pore, it is likely that other transmembrane domains also contribute. TMD5 and TMD6 are also very highly conserved, and thus might be involved in forming the pore.
E. Splice Variants
A number of splice variants of bestrophins have been identified by RT-PCR, but it is not known whether any of these produce functional protein. Petrukhin et al. (159) found splice variants of hBest1 in RPE by RT-PCR (Table 2). The mRNA of isoform 1 is transcribed from 11 exons. The protein is 585 amino acids long and corresponds to what we call “wild-type” hBest1. Isoform 2 has exon 7 alternatively spliced resulting in a frame shift at amino acid 290 and premature termination at amino acid 435. It is believed that the mRNA of this form is degraded by nonsense-mediated decay. In addition, there are two other putative alternatively spliced forms that can be found in the sequence databases represented by EST clones BC015220 (isoform 3) and BC041664 (isoform 4) deposited by the Mammalian Genome Collection (NIH). These are 498 and 604 amino acids in length, respectively. In both forms, exon 2 is skipped, resulting in an alternative start at methionine 61. The 498-amino acid variant also lacks exon 8, which codes for the highly conserved domain immediately after the last transmembrane domain (amino acids 290–316). It is unlikely that the 498-amino acid variant is functional as a Cl− channel because amino acids 290–316 are essential for Cl− channel activity. The 604-amino acid variant has an alternatively spliced exon 10 which results in a longer COOH terminus.
Putative splice variants have also been reported in mBest3 (98) and in hBest3 (192). The longest hBest3 form is 669 amino acids. There are two shorter forms that represent alternative splicing of the first and last exons. In addition, as with hBest1, there are several short forms, some of which lack predicted transmembrane domains and whose significance remains unclear.
F. Protein Expression
To date, the expression profiles of the various bestrophins have not been clearly established. Most of the available data come from measurement of mRNA levels by RT-PCR or Northern analysis; relatively limited information is available about protein levels determined by western blot or immunocytochemistry (Fig. 7). Each of the methods used to determine expression has its own limitations. Northern analysis has relatively low sensitivity and requires very high quality RNA for quantitative measurements. RT-PCR, on the other hand, is so robust and sensitive that it may amplify a pool of RNA that is expressed in only a very small subset of cells in a tissue. Contamination of tissue with blood or adjacent tissues can give false-positive results. Quantitative RT-PCR circumvents some of these problems but requires care in designing optimal primers and in assessing the quality of the template RNA. RNA expression can also be quantified by counting the fraction of bestrophin expressed sequence tags (ESTs) relative to the total number of ESTs from that tissue. Such data are available from UNIGENE using the expression profile viewer (http://www.ncbi.nlm.nih.gov/UniGene). This method suffers from limited sampling but does provide an estimate of the fraction of mRNAs encoding the gene of interest. Ideally one would like to know the level of protein expression, because RNA levels may not reflect protein levels. But, quantifying protein levels has its own problems. To date, there are relatively few antibodies available for bestrophins, and many of these have not been adequately characterized for specificity. The antibodies that provide the most confidence are ones that have been shown to have no reactivity in bestrophin knockout animals. In our hands, many of the commercially available antibodies fail to meet this test. But, even with good antibodies, detection of bestrophins has proven to be challenging because it seems that the levels of protein expression in most tissues are too low to detect reliably by western blot. Many of the antibodies that have been shown to work well in western blot do not work well in immunocytochemistry. In Figure 7, RNA levels were determined by Northern blot or RT-PCR without quantification with the exception of quantitative RT-PCR data from Reference 98. There are a number of generalizations that can be made from Figure 7.
There is rather poor correspondence between different techniques and laboratories. For example, by Northern blot and RT-PCR (159), the expression of hBest1 is highly restricted. The levels in retina and RPE are much higher than in any other tissue, although some expression is noted in brain, spinal cord, kidney, and testis. From EST counts, however, hBest1 is widely expressed in a variety of tissues. Protein expression for hBest1 has been reported only in RPE. Furthermore, there appear to be important differences between species. The pattern of tissue RNA expression measured in the same lab differs between human (124) and mouse (98). Although hBest1 protein and pBest1 protein has been found only in RPE, mBest1 protein expression has been reported in colon, kidney, and trachea in addition to RPE (13). Whether this reflects species differences or differences in the antibody sensitivity or specificity is not known.
Best2 seems to have a very limited tissue distribution. The distribution is similar in human and mouse. Best2 is clearly expressed in colon and testes in both species. There currently is a controversy whether mBest2 is expressed in olfactory epithelium. Using Western blot and immunostaining, one lab found that mBest2 is expressed in olfactory epithelium (162). However, another lab replaced the first two exons of mBest2 with Lac-Z and found expression in colon and ciliary epithelium but not olfactory epithelium (11). These results were confirmed with a mBest2 antibody in wild-type animals. Furthermore, these mBest2 knockout mice showed no obvious olfactory deficit. This controversy about mBest2 expression in olfactory epithelia highlights the murkiness that currently surrounds the area of bestrophin expression. The expression of mBest2 in ciliary epithelium in the eye is particularly interesting because mice with mBest2 knocked out have diminished intraocular pressure (11). This suggests that mBest2 may be involved in aqueous humor generation.
3. Best3 and −4
Best3 RNA is rather widely expressed, but there have been no studies published on protein expression. No studies have been published using Best3 or Best4 antibodies. Again, species differences and differences between labs and techniques are notable.
As bestrophins have been proposed to be candidate CaC channels, it is worthwhile to ask whether bestrophins are expressed in tissues that are known to have CaC currents (74). Such tissues include pancreas, salivary gland, and colon. Bestrophins are expressed in each of these tissues.
A. Regulation by Ca2+
hBest1 is Ca2+ activated with a Kd for Ca2+ of ∼150 nM, but the mechanism of regulation of the current by Ca2+ remains unknown. The channel could be activated directly by Ca2+ binding, indirectly via a Ca2+-binding protein like calmodulin (CaM), or even more indirectly by mechanisms that might involve phosphorylation or other enzymatic processes. Furthermore, it is not clear that all bestrophins are dependent on Ca2+ for activation. Only hBest1, xBest2, mBest2, and hBest4 have been examined carefully (168, 170, 199, 200). Each of these channels has an apparent Kd for activation by Ca2+ in the range of ∼200 nM (168, 170, 199, 200). If native bestrophin channels have the same Ca2+ sensitivity as these heterologously expressed channels, bestrophin current must be partially activated at all times, because basal free cytosolic Ca2+ is typically around 100 nM.
Tsunenari et al. (200) have suggested that hBest4 might be regulated directly by Ca2+ because hBest4 can be activated in excised membrane patches in the absence of substrates for phosphorylation. They also suggest that the Ca2+ binding site might be located in the COOH terminus immediately after the last transmembrane domain because this region contains a high density of acidic amino acids that could coordinate positively charged Ca2+. This region exhibits some similarity to the Ca2+ bowl of BK potassium channels (Fig. 8 A). Likewise, endogenous Drosophila bestrophins in excised patches can be activated by Ca, suggesting that Ca2+ acts on the channel directly (28). However, addition of ATP accelerates the activation (28).
B. Potential Regulation by Phosphorylation
hBest1 incorporates 32P when it is expressed in RPE-J cells and hBest1 coimmunoprecipitates with protein phosphatase 2A (122). The association with phosphatase 2A suggests that phosphorylation may play a role in regulation of hBest1. Human bestrophins have predicted high-stringency phosphorylation sites (scansite.mit.edu) for protein kinase A (PKA) (hBest1 T536, hBest2 S413, hBest3 S412), protein kinase C (PKC) (hBest1 S358, hBest2 S127, hBest4 S176 and S222), and various other kinases (Fig. 8B). There are no sites found for phosphorylation by CaM kinases at high stringency. Interestingly, with hBest1, Ca2+/CaM kinase II (CaMKII) sites are predicted only at the lowest stringency, and these are located in residues that are unique to hBest1. The only published evidence that bestrophins are regulated by phosphorylation is a study showing that nitric oxide activates a basolateral Cl− channel in Calu-3 cells via cGMP-dependent phosphorylation (38). Because hBest1 is expressed in these cells and because a DIDS-inhibited Cl− current is decreased by siRNA to hBest1, it has been suggested that this Cl− current is mediated by hBest1 (39). Nitric oxide stimulates 32P incorporation into immunoprecipitated hBest1. A major limitation of these studies, as well as those of Marmorstein (122), is that the stoichiometry of 32P incorporation into hBest1 was not quantified. Because it is not known what fraction of bestrophin molecules incorporate phosphate, it is not possible to say whether phosphorylation serves regulatory function.
C. Potential Regulation by Protein-Protein Interaction
The COOH termini of bestrophins are generally proline rich, suggesting that the COOH terminus is involved in protein-protein interaction (Figs. 2 and 8B). Both Scansite (scansite.mit.edu) and iSPOT (http://cbm.bio.uniroma2.it/ispot/) predict that bestrophins contain SH3-binding domains (Fig. 8B). SH3-binding domains generally contain the consensus sequence PxxP. The highest scoring hits (>0.8 for human amphiphysin, a score consistent with known SH3 binding partners) were found in hBest3 at residues 404–413 and 430–448. However, all the bestrophins tested had predicted SH3-binding domains in the COOH termini with scores >0.6. All four human bestrophins have a conserved SH3-binding domain immediately following TMD2. In mBest2, mutation of P101 to cysteine produces nonfunctional channels (171), and in hBest1, seven disease-causing mutations have been reported within this SH3-binding domain.
D. Autoinhibitory Domain
A region in hBest3, mBest3, and mBest2 has been shown to inhibit channel activity when the channels are expressed heterologously (167, 172). Wild-type mBest3 and hBest3 generate very small currents at physiological membrane potentials when expressed in HEK293 cells. At very negative potentials, they do develop current with a very slow time course. Deletion of the COOH terminus distal to amino acid 353 renders the channel functional at physiological potentials. By successive deletions and mutations, an autoinhibitory region was localized to include amino acids 356IPSFLGS362. Single amino acid substitutions in this region result in very large Best3 currents. Expression of a fragment of the COOH terminus of mBest3 including the inhibitory domain reduces the amplitude of currents induced by the mBest3 that lacks the COOH terminus (Δ353–668), suggesting that the COOH terminus binds to another portion of the protein to inhibit the current. It is presumed that the effect of this domain is regulated physiologically by some mechanism such as phosphorylation, but the mechanisms are not known. Whether this domain interacts with or influences the effect of the nearby SH3 domains is not known. All vertebrate bestrophins share the consensus SFxGS sequence (Fig. 8B). Mutation of this region in mBest2 results in much larger currents than wild type, suggesting that this domain may play a common regulatory function. In mBest2, this domain does not inhibit the wild-type current as completely as it does in mBest3, because there is apparently another domain between amino acids 405 and 454 that suppresses the inhibitory effect of the SFxGS domain (172).
V. PHYSIOLOGICAL ROLES OF BESTROPHINS
A. Are Bestrophins Classical Ca2+-Activated Cl− Channels?
There are important similarities between classical CaC channels (74) and bestrophins. Both channels exhibit the generic lyotropic anion selectivity sequence, and both are activated by cytosolic Ca2+ in the submicromolar range. Classical CaC channels can be gated directly by Ca2+ (24, 102, 112, 200). Other CaC channels are activated by Ca2+-dependent phosphorylation (7). Although dBest1 is activated by Ca2+ in excised patches, its activation is augmented by the presence of ATP (28).
But, there are also some important differences between CaC channels and bestrophins. Although some bestrophins are activated by Ca2+, it appears that not all are Ca2+ dependent. hBest2 and dBest1 expressed in HEK cells are not Ca2+ dependent (Yu, Qu, Chien, and Hartzell, unpublished data). Classical CaC channels exhibit voltage-dependent kinetics and outward rectification that is not seen with wild-type hBest1 or mBest2. But, interestingly, a few point mutants of hBest1 and mBest2 exhibit rectification, voltage sensitivity, and kinetics that resemble classical CaC channels (Qu and Hartzell, unpublished data). Possibly association of these subunits with accessory subunits could alter their biophysical properties. Our initial cloning of bestrophins was by RT-PCR from Xenopus oocyte mRNA (170). Xenopus oocytes are a rich source of classical CaC channels; however, it remains to be seen whether bestrophins are the molecular counterpart of CaC channels.
B. Volume Sensitivity of Bestrophins
Stretch-activated or volume-regulated ion channels (VRACs) are thought to play a key role in regulating cell volume. Cells placed in a hypotonic solution swell as water flows down its concentration gradient into the cell. The cell then usually undergoes a process of regulatory volume decrease (RVD). This process often involves the opening of volume-activated Cl− and/or K+ channels. Efflux of these ions is followed by water efflux as the osmolarity of the intracellular solution is decreased (79, 104, 146, 147, 179). When cells are placed in hypertonic solution, regulatory volume increase occurs by comparable mechanisms. The identity of the VRACs has been a steady controversy for many years (29, 88, 147, 179). P-Glycoprotein, ClC-3, ClC-2, pICln, phospholemman, and the Cl−/HCO3 exchanger have all been considered as channels involved in cell volume regulation, but none of these has received general acceptance as a VRAC.
Cl− currents induced by expression of hBest1 or mBest2 are strongly inhibited by hyperosmotic solutions and are stimulated by hyposmotic solutions (55). Increasing extracellular osmolarity 20% causes cell shrinkage and reduces hBest1 current ∼80%. Furthermore, when HEK cells are transfected with hBest1, the typical VRAC current was increased ∼10-fold (55). Recently, we have found that the Cl− currents encoded by native dBest1 are also sensitive to osmolarity (27). Our findings have led us to propose that dBest1 is a type of VRAC that is involved in control of cell volume. This suggestion is supported by several pieces of evidence. 1) Hyposmotic cell swelling precedes activation of the osmotically sensitive Cl− current in S2 cells. 2) The osmotically sensitive current is abolished or greatly reduced by four different RNAi constructs that reduce dBest1 protein levels. 3) Inhibition of the current by RNAi is rescued by overexpression of dBest1. 4) Cells with reduced dBest1 expression have an impaired ability to regulate cell volume in response to hyposmotic solutions. 5) An excellent correlation exists between the amplitude of the dBest1 current and the ability of cells to undergo RVD.
The classical VRAC current is outwardly rectifying, inactivates at positive potentials in a time-dependent manner, and exhibits an anion selectivity of SCN− > I− > NO3− > Br− > Cl− > gluconate (147). Although bestrophins have a very similar anionic selectivity (171), bestrophin currents exhibit little outward rectification and do not inactivate at positive potentials. Although this difference might indicate that bestrophins are not VRACs, VRAC rectification and inactivation seem to depend on the recording conditions and cell type (34, 147). For example, in endothelial cells, parotid acinar cells, T lymphocytes, neutrophils, and skate hepatocytes, inactivation is small or even completely absent (147).
The single-channel properties of bestrophins and VRACs also seem to be different. dBest1 has been reported to have a single-channel conductance of ∼2 pS (28). VRACs have been reported to have single-channel conductances of 2 to >100 pS (147). However, many of these estimates of single-channel conductance were obtained from stationary noise analysis. Jackson and Strange (85) have shown that in C6 glioma cells, stationary noise analysis yields a single-channel conductance of ∼1 pS at 0 mV, but nonstationary noise analysis and single-channel recording reveal a conductance of ∼15 pS. It appears that stationary noise analysis provides the wrong value because of the very high open probability of the channels when they are maximally activated. For this reason, it is generally believed that VRAC channels have a single-channel conductance of ∼15 pS at 0 mV (85). One should remain open-minded, because of the difficulty in relating single channels to macroscopic currents because of the sampling problem of single-channel recording. Furthermore, the gating of the dBest1 channels is complex (28). The 2-pS channels activated by Ca2+ in excised patches (28) may be sub-states: physiological activation by changes in cell volume in intact cells may gate the channel differently.
The Drosophila S2 VRAC current is apparently mediated by the same channels that mediate the CaC current, because the same RNAi treatments reduce both currents. But dBest1 VRAC currents in S2 cells are activated independently of Ca2+. This is consistent with previous reports that the activation of VRAC does not require an increase of intracellular Ca2+ (37, 77). However, there is considerable evidence that although elevation of Ca2+ is not necessary, a low level of Ca2+ (∼50–100 nM) is required for VRAC activation, at least in certain cell types (25, 147, 155, 197). hBest1 has an EC50 for Ca2+ of ∼150 nM, which is very close to the permissive level for VRAC activation. Also, increases in cell volume are often accompanied by increases in intracellular Ca2+ (126). Conversely, it seems that Ca2+ can activate the current in the absence of cell volume changes. These data indicate that cell volume and Ca2+ may be independent regulatory activators to bestrophin. It remains to be seen whether these two activators are truly independent or whether they converge on some common regulatory pathway.
Many observations suggest that VRACs participate in other cellular functions besides cell volume regulation (45). These other functions include mechanotransduction, cell cycle control, and apoptosis. Thus the possibility exists that under isotonic conditions, VRACs may be activated by other means. However, these functions have not been amenable to exploration largely because the molecular identity of VRACs remains in question.
C. Function of Bestrophins in the Eye
Although mutations in hBest1 have been clearly linked to eye disease, the exact function of hBest1 in the eye remains to be determined. hBest1 is expressed in the RPE (118). The RPE plays a number of essential roles in retinal homeostasis (195) that include regulation of the composition of the fluid surrounding the photoreceptor outer segments, regeneration of the visual pigment (the retinoid or visual cycle), and phagocytosis of shed photoreceptor discs (15, 105, 143). Bestrophin could possibly affect any of these processes as a result of changes in hydration and ionic composition of the subretinal space. This is discussed in more detail below in sections vi and vii. However, attempts at understanding the mechanisms by which hBest1 mutations cause BVMD have been complicated by the finding that mBest1 knockout mice do not phenocopy the human disease and have no gross visual deficit or retinal pathology (117, 123).
D. Olfactory Transduction
Olfactory sensory neurons (OSNs) express a Ca2+-activated Cl− channel that is important in olfactory transduction. Ca2+ influx that occurs via cyclic nucleotide-gated channels stimulates Ca2+-activated Cl− channels that amplify the response and are responsible for the largest fraction of the olfactory receptor potential (57, 74, 174). Recently, it has been proposed that mBest2 is responsible for these olfactory CaC currents (162). mBest2 expression in OSNs is detected by RT-PCR and is found on the cilia by immunocytochemistry. Furthermore, mBest2 expressed in HEK cells has similar biophysical and pharmacological properties as the OSN CaC channel. However, there are some important differences between the expressed mBest2 current and the endogenous CaC current. The endogenous CaC current has an EC50 for Ca2+ of 4.7 μM, whereas the EC50 for the expressed mBest2 current is 0.4 μM. Furthermore, although the anionic permeability sequences are the same for the expressed and the endogenous channels (I− > NO3− > Br− > Cl− > methanesulfonate−), there are quantitative differences. For example, PI/PCl is 1.8 for the expressed mBest2 current, whereas it is 4.7 for the endogenous CaC current. The expressed mBest2 current rectifies somewhat less than the endogenous CaC, and stationary noise analysis suggests that the single-channel conductances of the endogenous and expressed channels are different. If mBest2 is responsible for the CaC in OSNs, it seems likely that its biophysical properties are functionally modulated by accessory subunits or other bestrophin subunits. Another study using a mBest2 knockout/Lac-Z knockin concluded that there is no or very little expression of mBest2 in nasal tissues (11). Hopefully, the mBest2 knockout mice will provide more information about the possible role of mBest2 in olfactory transduction.
E. Epithelial Secretion
Many epithelial cells, for example, from salivary glands, pancreas, airway, and colonic epithelium, have Ca2+-dependent Cl− secretory pathways (74). Recently, Kunzelmann et al. (101) have shown by RT-PCR that mBest1 and mBest3 are expressed in salivary gland. They also show that expression of mBest1 and hBest1 parallels the expression of Ca2+-activated Cl− channels in different cell types (13). siRNA was shown to abolish ATP-activated Cl− currents in HT29 cells. These currents were shown to be insensitive to TEA and Ba2+, which block K+ channels, but a detailed characterization of these currents was not published. These data suggest the possibility that bestrophins play a role in epithelial secretion and CaC currents.
F. Potential Intracellular Functions
Cl− channels are known to function not only at the plasma membrane but also in intracellular membranes. For example, Cl− channels play an important role in acidification of endosomes and organelle trafficking (52, 87). A number of authors have commented that bestrophins are present not only on the plasma membrane but also in intracellular compartments (101, 168, 170, 199). Because this high level of intracellular localization has been reported mainly in transfected cells, it is possible that intracellular bestrophins are an artifact of overexpression. However, the fact that a characteristic feature of hBest1-linked diseases is accumulation of “lipofuscin” pigment (see sect. vi) and that accumulation of lipofuscin pigment is also a characteristic of lysosomal storage diseases like neuronal ceroid lipofuscinosis, it is tempting to speculate that bestrophins might be involved in lyosomal function (76). Because RPE cells are professional phagocytes that are loaded with lysosomes, this idea is even more attractive. However, there is no evidence to support such a function and, moreover, it has been reported that hBest1 is not found to concentrate in lysosomes when expressed heterologously (101).
VI. DISEASES ASSOCIATED WITH hBest1 MUTATIONS
Mutations in the human hBest1 gene are closely linked with Best vitelliform macular dystrophy (BVMD; OMIM no. 153700) (124, 159). In addition, a small fraction of patients diagnosed with adult-onset vitelliform macular dystrophy (AVMD; OMIM no. 608161) (3, 99, 215), Bull's eye maculopathy (OMIM no. 153870) (3), and autosomal dominant vitreoretinochoroidopathy (ADVIRC; OMIM no. 193220) (213) have mutations in hBest1. Recently, it has been found that bestrophin mutations in dogs cause canine multifocal retinopathy (CMR) (70).
A. Best Vitelliform Macular Dystrophy
BVMD is classically described as an autosomal-dominant, progressive, juvenile-onset macular degeneration associated with large deposits of a yellow pigmented material in the subretinal and sub-RPE spaces (17, 64, 137, 160). The deposits initially resemble an egg yolk (vitelliform). The lesion then often passes through a pseudohypopyon stage that is often accompanied by detachment of the neural retina from the RPE (86, 129, 160, 161, 201). With age, the deposits become disorganized (the vitelleruptive stage) and the RPE/choroid layer becomes thickened partly as a consequence of pigment accumulation in the RPE layer (160, 161). The retinal layer above the thickened area becomes thinner, the photoreceptors degenerate, and deterioration of central vision follows. As discussed in detail below, there is considerable variability in the age of onset and the expressivity of vitelliform lesions. However, almost all individuals with hBest1 mutations have abnormal electro-oculograms (3, 9, 23, 31, 36, 97, 99, 111, 116, 124, 159, 184, 202).
The nature of the yellow pigment in BVMD is not known, but from its histological appearance, it is often assumed to be lipofuscin. Lipofuscin is a yellow pigment that normally accumulates with age in granules in the lysosomal compartment of RPE cells (187). Lipofuscin originates at least partly from photoreceptor outer segments that are phagocytosed by the RPE and is composed partly of oxidized proteins and lipids as well as fluorescent compounds. The major fluorophores include a diretinal conjugate, N-retinylidene-N-retinylethanolamine (A2E), and its photooxidation products (22, 47). A2E may disrupt membrane integrity by a detergent-like effect (47), and its photooxidation products may activate complement (216) and promote apoptosis (186). It has been shown that A2E mediates blue-light-induced damage in the retina (186) and is believed to be an important causal factor in AMD.
The idea that the pigment that accumulates in BVMD is lipofuscin probably originated with O'Gorman et al. (152) who described the histology of a 69-yr-old man with BVMD. The pigment that accumulated in the RPE cells in the fundus was defined as lipofuscin by ultrastructural appearance, autofluorescence, and staining properties. Recent studies from Marmorstein's lab have shown that A2E (in pmol/eye) is increased ∼60% in a W93C homozygote and is increased threefold in a T6R heterozygote (8).
There are a variety of mechanisms that could be proposed to explain the dominant inheritance pattern of BVMD. These include reduced levels of functional protein caused by a nonfunctional gene product from one allele (haploinsufficiency), the acquisition of new or constitutive protein functions by the mutant protein (gain of function), and dominant negative effects caused by inhibition of an oligomeric complex composed of wild-type and mutant subunits (208). Haploinsufficiency has been suggested as a possible mechanism underlying the effects of at least two mutations, A243V (214) and P260fsX288 (a frameshift mutation after P260 that introduces a stop codon at amino acid 288) (97); however, this does not seem to be the predominant mechanism. Furthermore, two homozygotes that have been studied did not have more severe disease than typical heterozygotes (12, 151). There is good evidence that some mutations are dominant negative because they inhibit the Cl− channel function of wild-type hBest1 when wild type and mutant are coexpressed in HEK cells (116, 196, 214, 215). In contrast, Marmorstein et al. (123) have proposed that the dominant inheritance is due to a gain of function of bestrophin. The exact function that is affected is not yet known, although there is some evidence that it involves Ca2+ signaling. This is discussed in more detail in section viiB.
B. Adult Vitelliform Macular Dystrophy
AVMD is a heterogeneous group of diseases with widely variable clinical and histopathological features. A large fraction of AVMD patients have mutations in the RDS gene, but some have mutations in hBest1 (3, 9, 97, 99, 111, 140, 184, 218). Patients with AVMD have vitelliform lesions, but the lesions usually develop later in life, after the third or fourth decade, and may have variable morphology. The distinction between BVMD and AVMD caused by hBest1 mutations is not clear-cut. The EOG is generally normal in AVMD patients, but on average, the EOG is at the lower end of the normal range. Because the age of onset of BVMD in the same pedigree is variable (e.g., Refs. 151, 175), AVMD caused by hBest1 mutations should probably be considered a subset of BVMD.
C. Autosomal Dominant Vitreoretinochoroidopathy
ADVIRC is rather different from BVMD and AVMD. The retinal dystrophy includes circumferential peripheral retinal hyperpigmentation, punctate retinal and vitreous opacities, and choroidal atrophy. Three mutations in hBest1 have been associated with ADVIRC: V86M, Y236C, and V239M (213). The mutations are located near mutations that have previously been shown to cause BVMD. To explain why these mutations produce ADVIRC and others produce BVMD, it has been suggested that these mutations result in exon skipping. In support of this idea, it was shown that these mutations induce exon skipping in a reporter construct in vitro, but it remains unknown whether such exon skipping occurs in patients with these mutations. Many patients with ADVIRC have abnormal EOGs (73, 103), but in some patients, the EOG is normal (92). Corresponding genotype and EOG data have been published for only one family (103, 213), and in this family the EOG was abnormal for all affected individuals in the pedigree. ADVIRC is sometimes associated with developmental abnormalities of the eye, including nanophthalmos, microcornea, closed-angle glaucoma, and congenital cataract. This suggests that hBest1 might also play an important role in eye development. Mutations in hBest1 are also apparently responsible for a related rare autosomal dominant phenotype comprising microcornea, rod-cone dystrophy, cataract, and posterior staphyloma (MRCS) (213).
D. Age-Related Macular Degeneration
Several hBest1 mutations have been found in patients diagnosed with age-related macular degeneration (3, 111). Interestingly, these mutations occur in the large cytoplasmic loop predicted by Milenkovic et al. (130) (R105C, E119Q, K149X, T216I, V275F) or in the cytoplasmic COOH terminus (L567F). In all cases, these mutations were found in isolated cases with no clear family history of eye disease. Thus it remains unclear whether these mutations are causative or even increase susceptibility to age-related macular degeneration. These mutations were found in a very small percentage of patients with age-related macular degeneration, so even if they do confer susceptibility, this apparently would apply to only a minor fraction of cases.
E. Canine Multifocal Retinopathy
Best1 mutations in dogs are linked to canine multifocal retinopathy (CMR) (70). CMR has been recognized in several breeds of dogs including Great Pyrenees, Coton de Tulear, English Mastiff, and Bullmastiff. The disease usually develops before 4 mo of age and is characterized by multifocal areas of retinal elevation with subretinal accumulation of serous to pink-tan fluid. CMR lesions are larger and more numerous than those in most BVMD patients and are most prominent in the superior quadrant. Some lesions are delimited by a partial ring of retinal edema. cBest1 is 86% identical to hBest1, and two cBest1 mutations have been linked to CMR. One is a point mutation in a conserved glycine residue (G161D) that is located in the intracellular loop after TMD2. The other is a nonsense mutation (R25X) that introduces a stop at codon 25. This probably results in a null allele. Most interestingly, both canine mutations are recessive, while all the human mutations so far described are dominant. This is particularly interesting with regard to the R25X mutation. If this mutation is actually a null mutation, this would suggest that CMR is not caused by Best1 haploinsufficiency, because disease is seen only when both alleles are null. The presence of recessive inheritance of cBest1-linked diseases in dogs raises the possibility that recessive hBest1 mutations may be associated with additional human retinopathies besides BVMD and AVMD.
F. The Electro-oculogram and Light Peak
A defining feature of BVMD has long been considered to be an abnormal electro-oculogram (EOG) (31, 36, 198, 203). The ionic and cellular basis of the EOG has been well studied (Fig. 9). In the dark, there is a voltage difference across the retina-RPE-choroid of ∼6 mV with the anterior positive relative to the back of the eye (Fig. 9A) (61, 148, 149). The voltage difference increases in the light (Fig. 9B). This voltage difference can be measured in isolated preparations of retina-RPE-choroid using microelectrodes (59, 83, 90, 134, 173, 189, 190). Such microelectrode studies have shown that most of the voltage drop across the retina occurs across the RPE cell layer. Like many epithelia, RPE cells produce a transepithelial potential difference as a consequence of the polarized distribution of channels and transporters on the apical and basolateral membranes. The RPE cell apical (facing the photoreceptor) membrane is more hyperpolarized than the basolateral (facing the choroid) membrane because the apical membrane potential is dominated by K+ conductances, whereas the basolateral membrane potential has in addition other conductances with less negative equilibrium potentials (such as Cl−).
The transepithelial potential can be measured indirectly by the EOG (4–6, 131). Electrodes are placed on either side of the eye at the canthus. The subject is then instructed to look alternately left and right. At the extremes of the saccades, the voltage recorded by the external electrodes is different because of the different orientation of the dipole set up by the RPE transepithelial potential. Typically, the recorded voltage is on the order of 20–150 μV in the dark-adapted eye. The voltage increases about twofold within 5–15 min after turning on the lights. The dark-adapted value is called the dark trough, and the maximum voltage in the light is termed the light peak (LP). Because the voltages can depend on many factors including placement of the electrodes, the ratio of the dark trough to the LP [the Arden ratio (5)] is usually used as the relevant comparative parameter. The Arden ratio in normal individuals is usually >2. The Arden ratio is usually considered pathologic if it is <1.5. Still, there is considerable variability in the Arden ratio, at least partly as a consequence of variability in eye movements (176).
There is considerable evidence that the LP is generated by a Cl− channel in the basolateral membrane of the RPE (62, 83). When the basolateral Cl− conductance increases, the basolateral membrane depolarizes, which thus increases the transepithelial potential. Evidence supporting this mechanism includes the observations that the LP is blocked by the Cl− channel blocker DIDS, the reversal potential of the LP changes in concert with Cl− equilibrium potential (ECl), and the relative Cl− conductance increases during the LP. Because hBest1 immunoreactivity is located basolaterally in RPE cells (10, 118, 139), it has reasonably been assumed that BVMD is caused by a defect in the basolateral RPE Cl− channel that is hBest1 (76, 196).
The mechanisms that turn on the basolateral Cl− conductance in the light remain obscure. There is evidence that the LP, although it is generated by the RPE, requires intact photoreceptors. This has led to the suggestion that photoreceptors in light produce a diffusible signal (“the light peak substance”) that binds to receptors on the RPE to activate the Cl− conductance (Fig. 9B) (60, 110). A second messenger cascade is implied by the very slow time course of the LP (∼10 min). The identity of the diffusible signal and the second messenger pathway remains unknown. Although several candidates have been proposed for the LP substance, ATP is favored because ATP mimics the LP (157, 195). It has been suggested that ATP released by photoreceptors acts on G protein-coupled purinergic receptors (P2Y) in the RPE to elevate cytosolic Ca2+, which in turn activates the Cl− conductance (Fig. 9).
The involvement of metabotropic receptors coupled to Ca2+ release from intracellular stores in the generation of the LP is purely circumstantial, and the involvement of other second messengers has not been thoroughly examined. There are cAMP-activated Cl− channels in RPE (75, 84), and CFTR is expressed in RPE (209). For example, the LP might be partially generated by cAMP-activated channels. In mouse, the cAMP-activated Cl− channel CFTR seems to play an important role in the LP (210). The LP in mice homozygous for the ΔF508 mutation have severely depressed LPs. Because other components of the dc-electroretinogram (ERG) associated with the RPE are also reduced, the authors suggest that this mutation may affect the general health of the RPE. However, the effect on the LP was more pronounced than on the other components. The LP was also reduced in CFTR knockout mice, although it appeared that the reduction was not as great as for the ΔF508 mutant. These data suggest the possibility that CFTR contributes to the LP in mouse. To our knowledge, detailed EOG recordings in human CF patients have not been published, but a frequently cited abstract states that CF patients have normal LPs (133).
Although activation of the basolateral Cl− conductance is usually imagined as being activated by Ca2+ mobilization from internal stores by a Gq-coupled pathway, recent evidence suggests that voltage-gated Ca2+ entry may be important (Fig. 9). Marmorstein et al. (123) have shown that the LP in mice is suppressed by dihydropyridines that block voltage-gated Ca2+ channels. Furthermore, the LP is greatly reduced in lethargic mice which have a loss-of-function mutation in the voltage-gated Ca2+ channel β4 subunit (123) and in mice lacking the Cav α1.3 channel subunit (211). Although these data provide strong evidence that voltage-gated Ca2+ channels are important in the LP, it still remains unclear how they are activated and whether they are responsible for activation of basolateral Ca2+-activated Cl− channels.
G. Variable Expressivity and Penetrance
BVMD is characterized by variable expressivity of clinical symptoms. Before the hBest1 gene was identified, Nordstrom and Thorburn (151) described a pedigree of one man who was very likely homozygous for a bestrophin mutation. He had 11 children with 2 different women. All of the children had abnormal EOGs, but the age of onset of symptoms was very variable. For example, pigment accumulation was found before 1 yr of age in one child, whereas another child had completely normal vision and retinal structure at the age of 24. Although the variable expressivity of macular lesions is common, affected individuals are usually reliably identified by abnormal EOGs (150, 151). Thus it seems clear that the disease occurs in two steps, with the initial lesion being associated with a depressed EOG and a second step (or series of steps) leading to the vitelliform lesion. After genotypic diagnosis became available, a number of additional examples of variable expressivity have been reported. For example, a 59-yr-old woman who had the R218C mutation was asymptomatic even though her children had severe macular lesions (23).
These findings show that the disease is multi-factorial, but whether the variable penetrance is related to genetic or environmental factors remains unknown. Identification of other genetic susceptibility loci would be useful in understanding the incomplete penetrance of this disease; however, the low frequency of the disease may make such a study impractical. Although the mechanisms of variable penetrance and expressivity remain unknown for BVMD, they are common features of a variety of genetic diseases (219). The possibilities could run the gamut from purely genetic to environmental. For example, digenic inheritance has been demonstrated for a subset of families with autosomal dominant retinitis pigmentosa in which mutations in both ROM1 and RP11 are required to produce disease (127). On the other hand, environmental factors can explain the variable expressivity for diseases like maternally inherited aminoglycoside-induced ototoxicity in which mutations in mitochondrial rRNA confer susceptibility to aminoglycosides (164).
H. hBest1 Disease-Causing Mutations
The VMD2 mutations database (http://www-huge.uni-regensburg.de/VMD2_database/index.php?select_db=VMD2) has a complete up-to-date listing of the mutations and polymorphisms that have been reported in hBest1. These are summarized in Figure 4B. As of this writing, over 100 mutations have been found. Most (>90%) of the mutations are point mutations at 84 different positions (multiple substitutions have been identified at some positions). There are two deletions (ΔI295 and ΔD301) and three truncations (L149X, P260fsX288, and H490fsX514). The two latter truncations were generated by point deletions that introduced frame shifts at upstream residues.
The mutations are clustered in several different regions of the protein (Fig. 4B). The highest density of mutations surrounds amino acid 300 where there is, on average, >1 mutation per amino acid. The other hot spots are located at the NH2 terminus, within and following TMD2, and in TMD5. The region with the lowest density of mutations is the cytoplasmic COOH terminus. Only four mutations have been found distal to amino acid 312. One of these is a frame-shift truncation, and the others have not been definitely shown to cause disease as they were found in isolated cases with no family history of eye disease. The mechanisms by which the mutations lead to macular degeneration are obviously one of the major questions that investigators in the bestrophin field would like to answer.
I. Diseases Associated With Other Bestrophin Paralogs
To date, there have been no human diseases that have been found to be associated with mutations in paralogs of hBest1.
VII. MECHANISMS OF BEST VITELLIFORM MACULAR DYSTROPHY
A. Is BVMD Caused by Cl− Channel Dysfunction?
Although bestrophins function as Cl− channels, the question remains whether BVMD is caused by a defect in plasma membrane bestrophin Cl− channel function. It is possible that hBest1 has additional functions. For example, CFTR, the Cl− channel responsible for cystic fibrosis, is a multifunctional protein that regulates a variety of other channels and transporters and functions both at the plasma membrane and in intracellular membranes (128, 154).
A strong argument supporting the idea that hBest1 mutations produce macular dystrophy by altering hBest1 Cl− channel function is the finding that of the 31 disease-causing mutations that have been studied, all of them except 3 alter the Cl− channel function of hBest1 expressed in HEK cells (116, 196, 214, 215) (Table 3). Of the mutants that affect Cl− channel function, the current amplitudes of the mutants are zero or much smaller than wild type. Reduction of current amplitude could be caused by changes in protein synthesis or trafficking, alteration in single-channel conductance, or changes in channel gating. Protein expression and trafficking have been evaluated by membrane biotinylation for a few, but not most, mutants (214, 215). Qualitatively, protein expression and trafficking of the mutants that were studied were not affected, suggesting that the mutation altered channel gating or conductance. Often the mutants suppress the current produced by coexpression of wild-type channels. This is consistent with the dominant negative mechanism of inheritance, but data regarding the mechanisms of the dominant negative effects are lacking. One possible explanation is that the channels are oligomers (see sect. iiiC) and that oligomers containing one nonfunctional subunit are nonfunctional. However, another possibility is that the mutant subunit alters the trafficking of wild-type subunits. Mutations in transmembrane domains would be expected to affect the channel conductance or gating machinery, but the mechanisms by which mutations elsewhere in the protein (for example, in the NH2 terminus) produce their effect are less easy to guess. There are 17 disease-causing mutations in hBest1 that are located in the large intracellular loop between amino acids ∼106–230 in the model of Milenkovic et al. (Fig. 4B). Of these 17 mutations, few have been studied for their effects on Cl− channel function. R218S, G222E, and Y227N are nonfunctional (196). None of the mutations near the middle of this cytoplasmic loop has been tested.
The three mutations that do not alter Cl− channel function, E119Q, T216I, and L567F, were found in isolated cases having no family history of BVMD. The patients were diagnosed with age-related macular degeneration or Bull's eye maculopathy, so the possibility exists that these mutations are not actually disease causing. Furthermore, EOG tracings were not shown in these papers; only Arden ratios are given, so independent evaluation of the data is not possible.
Because >90% of the mutations tested altered Cl− channel function, the simplest conclusion is that vitelliform lesions are a consequence of hBest1 Cl− channel dysfunction. But, the story may be more complicated than this. There are several hBest1 mutations that have been reported to be associated with disease without LP changes (Table 3). This finding challenges the hypothesis that the LP is generated by the hBest1 Cl− channel. These “potential LP sparing mutations” include T216I, A243V, ΔI295, D312N, and L567F. Most of the “potential LP sparing mutations” are associated with adult-onset forms of macular dystrophy, and the pathological effects of these “potential LP sparing mutations” were less severe than those that produced frank BVMD. We have studied the Cl− channel function of these mutations (214, 215). Two of these mutations, ΔI295 and D312N, were nonfunctional and dominant negative. One produced small or negligible currents but was not dominant negative (A243V). Two were like wild type (T216I and L567F). As noted above, the T216I and L567F mutations were found in isolated cases with no family history of BVMD or AVMD. Thus these mutations could be polymorphisms that do not actually cause disease. The LP of patients with the D312N mutation is clearly subnormal, although it is not absent. Of the potential LP-sparing mutations, there is unambiguous evidence for normal or near-normal LPs for only A243V and ΔI295. For both A243V and ΔI295, some patients have normal EOGs, but others have subnormal EOGs. In our analysis of A243V (214), we found that this mutation depressed Cl− channel function, but it was not dominant negative like many of the mutations found in the BVMD group. We proposed that the normal EOG in some A243V patients was due to a gene dosage effect with decreased levels of wild-type hBest1 resulting in defective RPE ion transport while barely maintaining the LP. We presume that the defect in ion transport is not sufficiently large to be reliably detected by EOG. The adult onset of the disease may be a consequence of the ability of the wild-type allele to sustain near-normal RPE function until aging stresses the system.
The normal LP in the ΔI295 mutation is more difficult to explain (215). ΔI295 is a common BVMD mutation (99, 124, 202) and is typically associated with a depressed LP. However, in one study, it was reported that two patients had vitelleruptive lesions in eyes with perfectly normal EOGs. The EOGs then deteriorated after a period of 4–5 yr. The observation that the abnormal EOG follows, rather than precedes, the vitelliform lesion challenges the hypothesis that BVMD is directly caused by a defect in the LP. The normal LP in some patients could be explained if the LP is not generated by hBest1, as suggested by Marmorstein et al. (123). In this case, the decreased LP would be secondary to the hBest1 Cl− channel defect. Another possibility is that patients with vitelliform lesions and normal LPs have a compensatory conductance that maintains the LP within normal levels for a period of time even though transport across the RPE is defective. hBest2 is also expressed in RPE (192). Its upregulation could possibly compensate for the ΔI295 mutation. Although the ΔI295 mutant is dominant negative on wild-type hBest1, it is not dominant negative on wild-type hBest2 (215). It should also be emphasized that the LP is an indirect measure of Cl− channel activity. The LP reflects the difference between the voltage drops across the apical and basolateral RPE membranes produced by the increased Cl− conductance in the basolateral membrane (110). Thus decreases in LP amplitude caused by decreases in the basolateral LP conductance could be offset by changes in other conductances that could come about secondarily.
Another fact that one should include in thinking about the factors that affect LP amplitude is the observation that the LP requires diffuse illumination over the entire retina: even relatively large spots of light are not sufficient to evoke a LP (110). This suggests that whatever the mechanism that generates the LP may require the coordination of several components in both central and peripheral retina. Dysfunction of one of these components (hBest1) in a small region of the retina may not, by itself, abolish the LP.
In summary, the vast majority of hBest1 disease-causing mutations that have been examined affect Cl− channel function, and the vast majority of patients with BVMD have altered LPs. The simplest conclusion, therefore, is that BVMD is a Cl− channelopathy resulting from the dysfunction of hBest1. The few examples of patients with hBest1 mutations but normal LPs could be explained if the LP can be generated by other ion channels in individuals with different genetic backgrounds. In any case, the number of well-documented BVMD cases with normal LPs is small. The few examples of mutations that function normally as Cl− channels can be explained if these mutations are not actually disease-causing. These mutations were all found in isolated cases that were not diagnosed as BVMD or AVMD.
B. Animal Models of BVMD
The hypothesis that BVMD is caused by Cl− channel dysfunction has recently been challenged by Marmostein's lab (117, 119–121, 123, 177). The most compelling evidence is that mice with the mBest1 gene disrupted (mBest1−/− null) have no retinal pathology; the a-, b-, and c-waves of the electroretinogram are normal; and the LP amplitude at maximum light intensity is the same as in wild-type animals (123). A surprising and important conclusion that the authors draw from these data is that the LP is not generated solely by Best1, at least in mice. Furthermore, RPE CaC currents are not obviously affected by knockout of mBest1 (123). We have confirmed that CaC currents are present in RPE cells in a different mBest1−/− strain (Yu and Hartzell, unpublished data). This observation seems to contradict the hypothesis that hBest1 is a CaC channel and that the LP is dependent on Best1. Although the amplitude of the LP in response to maximum light intensity was unchanged in mBest1−/− mice, the luminance-response curve (the sensitivity of the LP to light intensity) was altered. At low illumination intensities, the LP was enhanced in the knockout. This suggests that mBest1 expression reduces the sensitivity of the LP to light and that knockout relieves this inhibition.
Additional information about the role of Best1 in regulating the LP is provided by experiments in a rat model in which hBest1 was overexpressed by subretinal injection of adenovirus encoding hBest1 (120). Overexpression of W93C and R218C mutants caused a decrease in LP amplitude. This is consistent with the observation that W93C and R218C are dominant negative in their effects on Cl− channel function (196). However, the authors disfavor the idea that Best1 is involved in generation of the LP because overexpression of wild-type hBest1 does not consistently increase the maximum amplitude of the LP as would be expected if hBest1 were the LP generator. Rather, wild-type hBest1 expression reduced the sensitivity of the LP to light. This reduction in light sensitivity with overexpression mirrors the increase in light sensitivity with knockout of mBest1 (123). These results suggest that hBest1 mutants are gain-of-function mutations that inhibit whatever mechanism is responsible for generating the LP. This idea would help explain why the mBest1−/− knockout mouse does not phenocopy human BVMD: loss of Best1 protein would clearly have a different effect than a gain-of-function mutation.
Although the ion channels that generate the LP in mice have not been identified, Marmorstein and co-workers (123, 211) have shown that voltage-gated Ca2+ (CaV) channels play an important role. The LP is reduced by the Ca2+ channel blocker nimodipine (123) and is also reduced in mice lacking the α1.3 CaV channel subunit (211) or the CaV channel β4 subunit (123). In addition, Rosenthal et al. (177) have shown that hBest1 expression in RPE-J cells affects the function of endogenous CaV channels in these cells. Overexpression of wild-type hBest1 shifts the voltage dependence of Ca2+ channel activation to the left and accelerates activation (177). Overexpression of the hBest1 W93C mutant reduces current amplitude, significantly slows inactivation, and shifts voltage dependence of activation to negative potentials. In contrast, the R218C mutation has little effect on current amplitude but significantly accelerates inactivation. Additional evidence that mBest1 is involved in regulating Ca2+ channels is provided by the observation that increases in intracellular Ca2+ induced by ATP (presumably acting on purinergic receptors) in RPE cells are augmented in mBest1−/− mice (123).
These results raise several important questions.
1. What is the mechanism of the LP?
Extensive work in gekko and cat reviewed in section viF has shown that the LP is generated by a Cl− conductance. But, the fact that the LP is not affected by knockout of mBest1 provides very compelling evidence that mBest1 is not the generator of the LP. This conclusion is also apparently supported by the fact that some individuals with hBest1 mutations have normal LPs as discussed in section viA. However, there are alternative explanations that should be considered. Perhaps the LP is generated by a different mechanism in mouse than in human or cat. Mice are nocturnal, do not have a macula, and the LP appears to be considerably smaller in mice than in cats or humans. For example, in human and cat, the LP is three to five times larger than the C-wave (caused by a hyperpolarization of the apical membrane of the RPE) (110), whereas in mice the LP is usually smaller than the C-wave (123, 212). This suggests that the ionic mechanisms of the dc-ERG in mouse and human are different. If mBest1 is not responsible for the LP, which ion channels are involved? The fact that the LP is greatly depressed in CFTR mutant mice (210) raises the possibility that CFTR might contribute to the LP in mice. Also, ClC-2, ClC-3, and ClC-5 are involved in generation of transepithelial potential in RPE (207, 209). For example, knockout of ClC-2 abolishes the RPE transepithelial potential (20), so the possibility exists that these channels contribute to the LP. Also, it should be considered that upregulation of any of these channels or other bestrophin subtypes might compensate for the loss of mBest1 in the knockout.
2. Is Best1 a Cl− channel?
The experiments of Marmorstein et al. (123) do not directly address the question whether mBest1 is a Cl− channel. Their results (123) show (and we have confirmed) that CaC currents in RPE cells are not reduced in mBest1−/− mice. This shows that mBest1 is not responsible for the RPE CaC current, but it does not exclude the possibility that mBest1 is responsible for a different RPE Cl− current (VRAC?). In fact, in low intracellular Ca2+, the I-V relationships of the wild-type and mBest1−/− knockout differ in amplitude, reversal potential, and rectification in a way that is consistent with a loss of a fraction of background Cl− current (123). The authors do not comment on this result. It is important to point out that the Ca2+ sensitivity of mBest1 is not known, because the properties of heterologously expressed mBest1 Cl− currents have not yet been published. Is it possible that mBest1 is not a Cl− channel, even though its human ortholog clearly is? This seems unlikely considering the very high similarity in sequence between mouse and human Best1.
3. Are hBest1 mutations loss of function/dominant negative or gain of function?
As discussed in section viA, the human disease-causing mutations that have been studied are dominant negative and/or loss of function with regard to Cl− channel activity. The suggestion that disease-causing mutations are gain of function (120, 123) is based on the observation that the light sensitivity of the LP in mouse is decreased with increasing mBest1 expression and that hBest1 mutations decrease the LP. However, exactly what function might be gained has not been clearly identified. Because hBest1 regulates CaV channels (177), the implication is that the gain of function may involve CaV channels. However, it is difficult to understand precisely how this would work. It is not obvious why the disease-causing mutations R218C and W93C have such different effects on Ca2+ currents (177) even though they produce similar disease symptoms (9, 23, 111, 116, 159). Although the W93C mutant would evidently inhibit Ca2+ influx through CaV channels, it is not clear that the R218C mutant would have any significantly different effect than wild-type hBest1. Nevertheless, if we consider only W93C and assume that the gain of function is to reduce Ca2+ current, this would be consistent with the observation that CaV channels are necessary for the LP (123, 211). But, this seems difficult to reconcile with the conclusion from the mBest1−/− knockout studies that wild-type mBest1 has an inhibitory effect on the LP and that Ca2+ signals are augmented in RPE cells of Best1−/− mice (123).
C. Hypothetical Mechanisms of BVMD
Regardless of whether hBest1 functions as a Cl− channel or a Ca2+-channel regulator, or both, it seems very likely that hBest1 participates in epithelial transport across the RPE. But, how might disrupted epithelial transport be linked to development of vitelliform lesions? RPE has a variety of essential functions in retina including paracrine hormone secretion, participation in the blood-eye barrier, epithelial transport, regeneration of visual pigment, and phagocytosis of photoreceptor outer segments (POS) (195). All of these functions might easily be affected by disruption of ionic homeostasis. Transepithelial Cl− transport is the driving force for control of hydration and composition of the subretinal space between photoreceptors and the RPE (61). Although the exact mechanisms by which fluid transport is coupled to electrolyte transport remain poorly understood, it seems clear that the channels and transporters that mediate RVD also participate in fluid transport across the epithelium (1, 2, 18, 61, 81, 93, 107). Fluid in the subretinal space is absorbed by the RPE and transported towards the choroid at a rate of several microliters per square centimeter per hour (62). If RPE ion and fluid transport is disrupted, the composition and fluid volume of the subretinal space is changed. Our hypothesis is that abnormal ion transport across the RPE and possibly abnormal cell volume regulation results in a functionally weakened interface between the RPE and photoreceptors (Fig. 10). Evidence supporting the idea that fluid transport is disrupted in BVMD is provided by the observation that detachment of the neural retina from the RPE is a common feature of the pseudohypopyon stage of BVMD (86, 129, 160, 161, 201). Although the pseudohypopyon stage is an advanced stage of the disease, weakening of the RPE-photoreceptor interface might begin much earlier and might not be detectable as a physical separation of the retina and RPE until much later.
Weakening of the interface between RPE and photoreceptors could affect both retinoid transport between photoreceptors and RPE and the ability of RPE cells to phagocytose photoreceptor outer segments (Fig. 10). Abnormal accumulation of retinal-derived pigments could very likely be a consequence of disruption of these processes. The development of BVMD, therefore, is likely to require two steps. Disruption of epithelial transport could result in a decrease in RPE-photoreceptor interaction which potentially results in accumulation of vitelliform pigment, depending on other factors. Because the development of vitelliform lesions has variable penetrance, vitelliform lesions are clearly not an obligatory consequence of disrupted epithelial transport but are governed by other environmental and/or genetic factors that confer susceptibility or protection.
The finding that bestrophins may be involved in cell volume regulation suggests some more specific, though quite speculative, hypotheses for hBest1 function. For example, the activation of hBest1 by changes in cell volume could play a variety of roles during the phagocytosis of POS. hBest1 channels could be involved in the phagocytic process itself, by participating in changes in cell shape and volume that occur during engulfment of outer segments. Also, hBest1 at the cell surface may help the cell deal with water and ion fluxes associated with dehydration and acidification of the phagolysosome. Cl− channels including VRACs have been shown to play a role in regulation of cell pH (147). A common theme of Cl− channels is that they seem to have functions both at the plasma membrane and in intracellular organelles, and it is possible that Best1 is also involved in regulation of phagolysosomal pH or trafficking (52).
1. A very speculative scenario
Just before outer segments are shed, the tips of the outer segments become permeable to Lucifer yellow, suggesting that the membrane becomes leaky (125). This leak is likely to permit intracellular amino acids to escape into the extracellular space. Intracellular taurine concentration in outer segments has been estimated to be 60–80 mM (181), and light stimulates taurine release from outer segments into the subretinal space (178, 180). This release of taurine from outer segments into the extracellular space may be a trigger for POS phagocytosis because 10 mM taurine added to eyecups will cause massive phagocytosis of photoreceptor outer segments by the RPE (69). How might taurine trigger phagocytosis? The apical membrane of RPE cells has a Na+-dependent taurine cotransporter (78, 100, 135, 158) that moves taurine into the cell. Uptake of taurine will induce RPE cell swelling as water follows taurine osmotically into the cell. Although taurine is a likely candidate osmolyte, accumulation of other amino acids or HCO3− could produce the same result. Glutamate, aspartate, and glycine all stimulate POS phagocytosis (69), and high HCO3 is required (16). RPE cells have transporters in the apical membrane for HCO3, glutamate, and other amino acids (19, 82, 109, 113, 136, 156). RPE swelling would then activate VRACs which would cause the cells to undergo RVD and shrink (147). The swelling and shrinkage of the RPE might play a role in the phagocytic process. The mechanical forces involved in dynamic cell volume changes could certainly play a role in the phagocytic event. There is accumulating evidence that Cl− channels are involved in cell shape changes and cell migration (33, 95, 138). In macrophages, phagocytosis is associated with extension of pseudopods (67), but it is not clear whether this also occurs in RPE cells (125, 143). Nevertheless, POS engulfment may be driven by restructuring of microvilli and pseudopods as a consequence of cell swelling. This restructuring might plausibly include integrin signaling (53). Activation of VRAC in cardiac myocytes is coupled to β1-integrins and focal adhesion kinases (21, 204), just like POS phagocytosis by RPE involves integrins (αvβ5) and focal adhesion kinases (53, 54, 141, 142). It is important to point out that because Best1−/− knockout mice do not exhibit a decrease in the ERG, the defect in POS phagocytosis is certain to be different from the defect in phagocytosis caused by knockout of β5 integrin, which does cause a decrease in the ERG (142). This scheme is consistent with the observations that POS phagocytosis is associated with increased adhesion of POS to the RPE (35) and that the subretinal space gets larger in light as a result of shrinkage of RPE cells (61). Interestingly, the LP, like bestrophin currents, is inhibited by hyperosmotic conditions (145). Cell swelling has an additional consequence: it causes an increase in intracellular Ca2+ (126). Ca2+ influx is required for POS phagocytosis, and blockers of voltage-gated Ca2+ channels inhibit POS phagocytosis (68). This scenario would link the Cl− channel and Ca2+ regulatory roles of bestrophin in renewal of photoreceptor outer segment.
VIII. CONCLUSIONS AND SIGNIFICANCE
In summary, the evidence is very strong that mutations in hBest1 produce macular dystrophy by affecting Cl− channel function or expression. However, it seems likely that hBest1 is not the (sole) generator of the LP. Also, a defective LP may not be a prerequisite for disease, as shown by the ΔI295 and A243V mutations. The fact that hBest1 mutations have been identified in several different forms of macular dystrophy, including BVMD, AVMD, ADVIRC, and AMD, suggests that different hBest1 alleles can cause different disease phenotypes. How this variability in phenotype relates to the detailed effect of the mutation on hBest1 function remains to be determined. Although hBest1 mutations do not play a major role in AMD (71), hBest1 mutations may contribute to AMD susceptibility in a very small (<1%) percentage of cases (3, 79a). Considering that there are an estimated 107 Americans with some form of macular dystrophy and that this number will increase as our population ages (71), hBest1 could possibly contribute to AMD in tens of thousands of individuals. Furthermore, the recent finding that recessive mutations in hBest1 are linked to retinopathy in dogs (70) raises the possibility that hBest1 will be found to be associated with other retinal diseases besides BVMD, AVMD, and ADVIRC. For example, retinal detachment might be favored by certain hBest1 polymorphisms.
But, beyond macular dystrophy, it is important to appreciate the roles of other Cl− channels in retinal homeostasis. Knockout of the Cl− channel genes ClC-2, ClC-3, and ClC-7 in mice results in retinal degeneration (20, 96, 191). The mechanisms by which knockouts of ClC genes produce retinal degeneration remain speculative, but retinal degeneration in the ClC-2 knockout is probably due to RPE dysfunction (20). Interestingly, ClC-7 and ClC-3 knockout animals show accumulation of yellow lipofuscin pigment in the brain and have a disease resembling human neuronal ceroid lipofuscinosis (NCL). The relationship between Cl− channel dysfunction and accumulation of lipofuscin-like material suggests that neural degeneration in NCL and photoreceptor degeneration in macular dystrophies may share common mechanisms. ClC channels have functions both on the plasma membrane and in intracellular compartments. Their plasma membrane roles include transepithelial transport, and dysfunction of these channels would likely result in abnormal composition of fluid surrounding photoreceptors. Intracellular roles of ClC channels include regulation of vesicular pH and membrane trafficking (52, 76, 87). We think that bestrophin, like other Cl− channels such as CFTR, may have multiple functions, both at the plasma membrane and in intracellular organelles. Understanding the functions of bestrophins will provide a greater understanding of the mechanisms of retinal homeostasis.
NOTE ADDED IN PROOF
Recently, a new human retinopathy linked to recessive mutations in hBest1 has been described (22a). Also, some patients with multifocal vitelliform dystrophy have been shown to have hBest1 mutations (19a).
This work was supported by National Institutes of Health Grants EY-14852 and GM-60448 and a grant from the American Health Assistance Foundation (to H. C. Hartzell) and grants from the American Heart Association and American Federation of Aging Research (to Z. Qu).
Address for reprint requests and other correspondence: H. C. Hartzell, Dept. of Cell Biology, Emory University School of Medicine, 615 Michael St., 535 Whitehead Bldg., Atlanta, GA 30322 (e-mail:).
- Copyright © 2008 the American Physiological Society