Blindness Caused by Deficiency in AE3 Chloride/Bicarbonate Exchanger

Background Vision is initiated by phototransduction in the outer retina by photoreceptors, whose high metabolic rate generates large CO2 loads. Inner retina cells then process the visual signal and CO2. The anion exchanger 3 gene (AE3/Slc4a3) encodes full-length AE3 (AE3fl) and cardiac AE3 (AE3c) isoforms, catalyzing plasma membrane Cl−/HCO3 − exchange in Müller (AE3fl) and horizontal (AE3c) cells. AE3 thus maintains acid-balance by removing photoreceptor-generated CO2 waste. Methodology/Principal Findings We report that Slc4a3−/− null mice have inner retina defects (electroretinogram b-wave reduction, optic nerve and retinal vessel anomalies). These pathologic features are common to most human vitreoretinal degenerations. Immunobloting analysis revealed that Na+/HCO3 − co-transporter (NBC1), and carbonic anhydrase II and CAXIV, protein expression were elevated in Slc4a3 −/− mouse retinas, suggesting compensation for loss of AE3. TUNEL staining showed increased numbers of apoptotic nuclei from 4–6 months of age, in Slc4a3 −/− mice, indicating late onset photoreceptor death. Conclusions/Significance Identification of Slc4a3 as underlying a previously unrecognized cause of blindness suggests this gene as a new candidate for a subset of hereditary vitreoretinal retinal degeneration.


INTRODUCTION
The human retina has extremely high rates of aerobic metabolism and oxygen consumption, which generates large CO 2 loads [1]. For optical reasons, the outer retina, where photoreceptors cells are localized, is avascular [2]. Efficient elimination of the considerable CO 2 and H + load produced by photoreceptor activity is necessary to maintain intracellular (pH i ) and extracellular (pH o ) pH homeostasis. Aspects of the phototransduction cascade, including photoreceptor light-sensitive currents [3], cGMP content of photoreceptors [4], and ion channels activity [5], are highly sensitive to pH.
Transient changes of intracellular pH (pH i ) and extracellular pH (pH o ) upon synaptic activity in the retina influence the response properties of neuronal circuits. Activity of plasma membrane acid and base transporters in neurons and glial cells of the retina regulate those changes. Na + -independent Cl 2 / HCO 3 2 exchangers (AE) likely contribute to maintain ionic homeostasis in the retina [6].
The AE family comprises ten genes [7] [8]. One of them, the AE3 gene (SLC4A3), encodes the full length AE3 (AE3fl) and the cardiac AE3 (AE3c) isoforms generated by alternative promoter usage [9]. The predicted cardiac AE3 (AE3c) polypeptide is 1030 amino acids in length and approximately 120 kDa, while the AE3fl variant consists of 1227 amino acids and ,160 kDa. The Cterminal 957 amino acids of both polypeptides are identical, but AE3c contains a unique N-terminal sequence of 73 amino acids, which replaces the first 270 amino acids of AE3fl. AE3 catalyzes electroneutral Cl 2 /HCO 3 2 exchange across the plasma cell membrane, regulating [Cl 2 ] i , [HCO 3 2 ] i , pH i , and volume [8]. In the central nervous system, neuronal and glial brain cells express only AE3fl isoform. Conversely, the inner retina, which is also part of the central nervous system, expresses both AE3fl and AE3c isoforms, in neuronal and glial cells [10].
To explore the role of proton trafficking in the central nervous system, it is important to complement physiological studies with its molecular counterpart. Therefore, we studied the functional role of AE3 in the inner retina, using a mouse model with a target disruption of the Slc4a3 gene.
While inheritance patterns have been clearly demonstrated for hereditary vitreoretinal degenerations (HVD), most genes have yet to be identified. In the few cases for which genes have been discovered, there is a very clear correlation between the function of the gene and the associated phenotype. The present study of the AE3 knockout mouse provides further evidence that AE3 plays an essential role in catalyzing Cl 2 /HCO 3 2 exchange across the plasma membrane of Müller and horizontal cells of the inner retina [10]. Therefore, AE3 contributes to the removal of photoreceptor-generated CO 2 waste, which contributes to maintaining acid-balance in the inner retina.
We report that mice with homozygous disruption of Slc4a3 present with inner retinal defects and late onset photoreceptor death, which are the pathologic features of most human HVD [11].

METHODS Animals
The animals were housed and handled with the authorization and supervision of the Institutional Animal Care and Use Committee at the University of Alberta. Experiments were carried out in accordance with the guidelines laid down by the NIH regarding the care and use of animals for experimental procedures (NIH Publications No. 80-23, revised 1996). The procedures also conformed to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. All efforts were made to minimize the number of animals used and their discomfort.
Generation and genotyping of Slc4a3 2/2 mice The Slc4a3 gene was disrupted by replacing a region of the gene that contained most of exon 6, intron 6, the cardiac-specific first exon and promoter, and exon 7 with the neomycin resistance gene (GE Shull, manuscript in revision). The region deleted contained codons 209-319 of the long form of AE3 and codons 1-72 of the cardiac-specific form. Expression of the cardiacspecific mRNA was eliminated. Due to the introduction of premature stop codons and frameshifts, the open reading frames of any mRNAs transcribed from the promoter for the longer variant of AE3 do not include the codons for amino acids 209-1227, which contain the anion transport domains. Genomic DNA was extracted from 3.0 mm ear notch biopsies with a Qiagen Kit (Qiagen Inc., ON, CA) and this DNA was used for genotyping by polymerase chain reaction. For the detection of the wild type allele (+) the oligonucleotides AE3wt.for 59-GAT GAA GAT GAC AGC CCA GGC CTT CC, and AE3wt.rev 59-CCG GCT CTT CTG TGT GGA GAT TCG GG, were used as forward and reverse primers, and amplified a 593 bp fragment. The forward primer corresponds to a portion of the deleted AE3 gene; thus the mutant allele was not amplified. The AE3wt.for primer, and AE3dNEO.rev primer, 59-GAC AAT AGC AGG CAT GCT GG amplified a 654 bp fragment of the mutant allele (2).

Preparation of mouse retina membranes
Freshly isolated mouse retinas were homogenized by 12 strokes of a Dounce homogenizer in 0.5 ml/retina of ice-cold 0.32 M sucrose, 1 mM EGTA, 0.1 mM EDTA, 10 mM HEPES, pH 7.5, containing protease inhibitors (MiniComplete, Roche). Homogenates were centrifuged at 1,440 g for 5 min in a Beckman G5-6K centrifuge. Supernatants were removed and centrifuged at 66,700 g for 30 min at 4uC in a Beckman TLA 100.4 rotor. Resulting membrane fraction was resuspended in 25 ml/retina, of PBS (140 mM NaCl, 3 mM KCl, 6.5 mM Na 2 HPO 4 , 1.5 mM KH 2 PO 4 , pH 7.5). Non-membranous fractions were kept for whole retinal lysate preparation. Protein was quantified by Bradford assay, and 50 mg of protein used for immunoblots.

Immunostaining of mouse retinas and analysis by confocal microscopy
For immunofluorescence experiments, mice were euthanized with pentobarbital, intravenously (200 mg/kg). Eyes were excised and immediately frozen at 280uC in Shandon Cryomatrix TM (Thermo Electron Corporation, PA). Cryostat sections (20 mm thick) of retinas were then cut onto glass slides. Following washing (265 min with PBS), and blocking (10% Chicken Serum in PBS, 30 min), retinas were incubated with primary antibodies in PBS buffer, containing 0.5% Triton X-100 (overnight, in a humidified chamber, 25uC), washed (365 min in PBS) and incubated with secondary antibody as above (1 h, in a humidified chamber, 25uC). Primary rabbit polyclonal anti-AE3 (AP3) antibody [10], rabbit anti-AE3c [15], and rabbit anti-PKC-aC-20, Santa Cruz, USA), were used at 1:100 dilution. Primary goat polyclonal anti-CAII (C14, Santa Cruz, USA), and goat polyclonal anti-CAXIV (N-19, Santa Cruz, USA), were used at 1:100 dilution. Mouse monoclonal anti-GFAP (SMI-22, Sternberger Monoclonals Inc.), and mouse monoclonal anti-Bassoon (VAM-PS003, Stressgen), were used at 1:1000 and 1:500 dilutions, respectively. Secondary chicken anti-rabbit conjugated to Alexa Fluor 488 (green), or secondary chicken anti-goat conjugated to Alexa Fluor 594 (red), were used at 1:100 dilutions, or secondary goat anti-rabbit conjugated to Alexa Fluor 488 (green), or secondary goat antimouse conjugated to Alexa Fluor 594 (red), was used at 1:1000 dilutions. In control experiments, slides were incubated only with secondary antibodies (data not shown). Slides were washed three times in PBS and mounted and viewed using confocal microscopy. Immunostained retinas on slides were mounted in Prolong Antifade solution containing DAPI for nuclear staining (Molecular Probes, OR, USA). Slides were imaged with a Zeiss LSM 510 laser scanning confocal microscope imaging system mounted on an Axiovert 100 M controller. Images were collected using an oil immersion 436 objective, at a resolution of 0.5-0.7 mm field depth. Filtering was used to integrate the signal collected over 4-8 frames to decrease noise (scan time of 7 s/frame). Multiple sections from different mice were examined. Some retinas were stained for GFAP and imaged as above, directly on retinal flat mounts.

Electron microscopy
Mice were euthanized as above. Eyes were fixed by transcardiac perfusion with 1% formaldehyde and 2% glutaraldehyde 0.1 M sodium phosphate, pH 7.2, then removed and fixed additionally for 1 h with 1% osmium tetroxide in 0.1 M sodium phosphate, pH 7.2. Tissues were then dehydrated and embedded in a 1:1 mixture of Epon 812 (Tousimis) and Araldite 502 (Electron Microscopy Sciences) epoxy resins. Sections of quadrants of the retina were cut, including the optic nerve head or at 0.5 mm peripherally, and stained with 1% toluidine blue. Sections were viewed and imaged with a Hitachi Transmission Electron Microscope H-7000.

Apoptotic nuclear staining
After eye removal as described for immunostaining, eyes were post-fixed in 4% paraformaldehyde for 1 h and then immersed in 10% (1 h), 20% (1 h), and 30% (12-16 h) sucrose. In situ cell death detection kit TUNEL Label Mix (Roche) was used to detect apopotic nuclei (green). Briefly, frozen 20 mm tissue sections were immersed in 20 mg/ml of proteinase K nuclease free buffer (Roche) with 10 mM Tris/HCl, pH 7.4, for 2 min at 4uC. Sections were rinsed repeatedly with phosphate-buffered saline (PBS) and then permeabilized with 0.1% Triton X-100 in 0.1% sodium citrate for 2 min on ice. After washing twice with PBS, 50 ml of the TUNEL reaction mixture was applied to the samples. Slides were incubated in a humidified chamber atmosphere for 45 min at 37uC in the dark. At the end of the incubation, slides were washed repeatedly with PBS, and mounted under coverslips, in the presence of DAPI. For control slides, TUNEL Label only was used, omitting the TUNEL enzyme in the reaction mixture, and no staining was detected.

Funduscopy
Under ketamine (150 mg/kg i.p.) and xylazine (10 mg/kg i.p.) anesthesia, pupils were dilated with a drop of Tropicamide (Alcon Laboratory). A digital camera was used in conjunction with a 78D lens (Volk) mounted between the camera and the mouse eye. Photographs were taken using conscious mice to avoid corneal clouding. Vibrissae were gently and partially removed with fine scissors to prevent them from obscuring the photograph.

Fluorescein angiography
Immediately after euthanasia (as described above), mice were transcardiacally perfused with 2 ml of a fresh solution of Fluorescein Isothiocyanate Dextran 500,000-conjugate (Sigma; 15 mg/ml of distilled water, filtered through 0.8 mm filter). The eyes were removed, post-fixed for 30 minutes and then retina flat mounts dissected to be post-fixed for one hour on filter paper (to force a flat shape during fixation). Tissues were mounted on slides and visualized with a Zeiss LSM 510 confocal microscope.

Electroretinography
After overnight dark adaptation (12-15 h), animals were prepared for bilateral ERG recordings under dim red light. Under anesthesia with a mixture of ketamine (150 mg/kg i.p.) and xylazine (10 mg/kg i.p.), the head was secured with a stereotaxic holder and the body temperature was maintained at 38uC, using a homeothermic blanket. Pupils were dilated using 1% Tropicamide. A drop of 0.9% saline was applied on each cornea to prevent dehydration and to allow electrical contact with the recording electrode (gold wire loop). A 25-gauge platinum needle inserted subdermally behind each eye, served as reference electrode. Amplification (at 1-1000 Hz bandpass, without notch filtering), stimulus presentation, and data acquisition were provided by the Espion E 2 system from Diagnosys LLC (Lowell, MA). Stimuli consisted of single white (6500 K) flashes (10 ms duration), repeated 3-5 times to verify the responsiveness reliability. For intensity responses, stimuli were presented at 19 increasing intensities varying from 25.7 to 2.9 log cds/m 2 in luminance. To allow for maximal rod recovery between consecutive flashes, inter-stimuli-intervals was increased (as the stimulus intensities were progressively increased) from 10 s at lowest stimulus intensity up to 2 min at highest stimulus intensity. Amplitude of the b-wave was measured from the a-wave negative peak up to the b-wave positive apex, and not up to the peak of oscillations, which can exceed the b-wave apex. Following 10 min of photopic adaptation (30 cd/m 2 background), cone-driven intensity responses were studied, using single flashes with intensities ranging from 21.6 to 2.9 log cds/m 2 along 11 steps of incremental intensities. Flicker ERG was then recorded, starting at 3 Hz, then 5 Hz and up to 45 Hz (along steps of 5 Hz) to establish the critical flicker fusion, according to 10 mV criterion amplitude. Peak-to-peak amplitudes were plotted as a function of the flicker frequency.

Statistics
Data points represent mean6s.e.m. Statistical significance between groups was assessed using ANOVA, with P,0.05 considered significant.

URL
Detailed information regarding the potential genes causing vitreoretinal degenerations can be obtained using the web-based server, OMIM (Online Mendelian Inheritance in Man). http:// www.ncbi.nlm.nih.gov/entrez/query.fcgi?db = OMIM
Consistent with anatomical observations in an independently generated Slc4a3-null mouse line [22], Slc4a3 2/2 retinas appeared normal with all layers present and of comparable thickness to agematched Slc4a3 +/+ littermates at 4 months of age (Fig. 2a, 2c). While light microscopic studies revealed a normal retina, they are not optimal for assessing the integrity of outer retina component such as outer segment discs and the intricate relationship between outer segments and retinal pigment epithelial cells. Ultrastructural analysis showed the presence of full-length outer segments with invaginating retinal pigmented epithelium villi as well as normal appearance of Bruch's membrane. The discs from the outer segments were perfectly aligned (Fig. 2b, 2d). Together, the light and electron microscopy data confirm the absence of any gross retina morphology disturbances in Slc4a3-null mice.
The absence of any cataract formations in the eyes of Slc4a3null mice allowed additional anatomical analysis using funduscopy (Fig. 2e, 2f). The retina, optic disc, choroid, and blood vessels were comparable between mutant and wild type mice. There were also no signs of any pigment migration or vitreous condensation.
To obtain a more detailed scrutiny of the retina, we probed specific structures on retinal flat mounts. Immunolabeling of astrocytes with specific glial fibrillary acidic protein (GFAP) antibody, showed that inner retina vessels were wrapped by dense astrocytic processes at 8 months of age in Slc4a3 2/2 mice (Fig. 2g,  h). This pathological feature, which was absent in WT mice, is analogous to the vascular sheathing previously described on histological sections of human retinas [23]. This observation prompted us to characterize inner retina blood vessels in greater detail than possible with funduscopy. Transcardiac perfusion with fluorescein confirmed normal appearance of inner retina vessels in 8-month-old WT mice (Fig. 2i, j). Conversely, age-matched Slc4a3 2/2 mice had inner retina blood vessels that formed sporadic loops (Fig. 2k). These loops were typical of pathological cases in which venules and arterioles are shunted, i.e. not separated by capillaries. Overall, inner retina blood vessels followed tortuous trajectories (Fig. 2l) and the major vessels in the central area of the retina tended to have thinner diameters.
Detailed characterization of the optic disc, beyond the level possible with funduscopy was undertaken using GFAP immunolabeling of astrocytes on cross sections of the retina. Astrocytic processes had marked disorganization at the level of optic nerve head (Fig. 2m). In Slc4a3 +/+ mice, the optic nerve head consisted of well-aligned bundles (not shown).
To test for potential retina dysfunction in Slc4a3 2/2 mice, photoreceptor activity was distinguished from post-synaptic inner retina activity by measuring the electroretinogram (ERG) a-wave and b-wave, which respectively represent photoreceptor and inner retina activity. Reduction of any of these components implies visual losses. We found a pronounced and selective ERG b-wave depression up to ,60%, that was age-independent (Fig. 4a, 4b,  4d, 4e, 4g, 4h). Another indicator of inner retina function, the flicker ERG, also paralleled b-wave reductions in Slc4a3 2/2 mice (Fig. 4f, 4i). Moreover, oscillatory potentials (wavelets on ascending b-wave), which also reflect inner retina functional integrity, were absent at all ages tested in Slc4a3 2/2 mice. In addition to specific inner retina defects at all ages in Slc4a3 2/2 mice, there was a progressive scotopic a-wave amplitude reductions (reaching statistical significance only by 4 months; see Fig. 4b, 4c, 4g) implying phototransduction failure in the already compromised vision of these mice, secondary to the ongoing inner retina defect.
AE3fl and AE3c proteins are found in a specific type of glial cells (Müller), and of neuronal cells (horizontal), respectively [10]. Both NBC1 [24] and carbonic anhydrase enzymes II (CAII) and XIV (CAXIV) [25], which are involved in bicarbonate metabolism and transport, are found in these same two cell types selectively expressing AE3. Therefore, we examined the expression of CAII (whole lysate), and CAXIV and NBC1 (membrane) cellular fractions of freshly isolated Slc4a3 2/2 null and Slc4a3 +/+ wild type mouse retinas (4-8 months), by immunoblotting. Samples were probed with a-tubulin as a loading marker. Cytoplasmic CAII, and CAXIV, with catalytic domain anchored to the extracellular surface, showed strong expression in Slc4a3 2/2 and Slc4a3 +/+ retinas (Fig. 5A). NBC1 was also detectable in Slc4a3 2/2 and Slc4a3 +/+ retinas (Fig. 5A). Notably, expression of CAII, CAXIV, and NBC1 increased by 30%, 35%, and 50%, respectively, in Slc4a3 2/2 compared to Slc4a3 +/+ retinas (Fig. 5B). This abnormal elevation in CAII, CAXIV, and NBC1 (not shown) protein levels, was confirmed by immunohistochemistry combined with confocal microscopy, in retinas of Slc4a3 2/2 and Slc4a3 +/+ mouse (Fig. 6). Frozen vertical sections of Slc4a3 2/2 and Slc4a3 +/+ mouse retina were mounted on the same slide and used to study localization and immunoreactivity of CAII and CAXIV, in adult mice. CAII was found in horizontal cells, predominantly associated with cell membranes (somas in inner nuclear layer and processes in outer plexiform layer) (Fig. 6A), whereas CAXIV showed intense labeling of Müller cells (somas and processes in the INNER NUCLEAR LAYER, and processes in the inner plexiform layer), and somas in the ganglion cell layer (Fig. 6B), in both Slc4a3 2/2 and Slc4a3 +/+ mice. Interestingly, under same experimental and imaging capture conditions, immunoreactivity of CAII and CAXIV labeling in Slc4a3 2/2 null mouse was stronger than in wild type littermates, confirming elevated expression of CA proteins in the pathologic retinas.
Increased expression of HCO 3 2 transporter (NBC1), and HCO 3 2 regulatory proteins (CAII, CAXIV) suggests the occurrence of compensatory changes in response to the loss of AE3 in Slc4a3 2/2 mouse retinas. Functional defects in the Slc4a3 2/2 null mice (ERGs) indicates that the ability to maintain acid-base balance is dramatically compromised in the inner retina and cannot be fully compensated by other pH regulatory proteins. We examined whether this leads to cell death in the retina. Staining for apoptotic cells, using TUNEL, revealed increased number of apoptotic nuclei from 4-6 months, in both the outer and inner nuclear retina layers, in Slc4a3 2/2 but not Slc4a3 +/+ retinas (Fig. 7A).
We have confirmed a key role played by the AE3 Cl 2 /HCO 3 2 exchangers in the mammalian retina, proving that bicarbonate transport and metabolism are key elements of normal ocular function. The specific visual phenotype resulting from AE3 deficiency, however, is unexpected, considering that the AE3  anion exchanger is expressed in other excitable tissues such as the heart and brain.

DISCUSSION
In this work, we have demonstrated that the apoptotic cascade is initiated as a result of absence of AE3-mediated HCO 3 2 fluxes. AE3 activity normally contributes to the regulation of pH i , [Cl 2 ] and cell volume. In addition to contributing to cytoplasmic pH regulation in Müller glial cells, AE3 may act as an additional HCO 3 2 efflux mechanism working in concert with CAII and CAXIV to remove CO 2 produced by photoreceptors. Along with AE3c, CAII and CAXIV are also involved in tight regulation of horizontal cells pH i , increasing the effectiveness of the bicarbonate buffer, forming metabolon-like systems [26] (Fig. 7B). The consequence of AE3 deficiency is a unique visual defect phenotype associated with a deregulation in pH mechanisms: a selective inner retina defect followed at later ages by photoreceptor degeneration. As for the human counterpart, all aspects of retina dysfunction reported here share analogies with

Specific inner retina defect: pH implication
In the visual system, AE3 activity contributes to regulation of pH i , [Cl 2 ], and cell volume. In addition to regulating pH i in Müller glial and horizontal neuronal cells, AE3 may contribute to the removal of photoreceptor-generated CO 2 waste.
Herein, we have demonstrated the occurrence of functional defects in the AE3 2/2 null mice by electroretinogram (ERG) (Fig. 4). Potential retina dysfunction in AE3 2/2 mice was evaluated by measuring ERG a-wave and b-wave, which respectively represent photoreceptor and inner retina functional activity. Reduction of any of these components implies visual losses. We found a pronounced and selective ERG b-wave depression that was present at all ages tested (1-8 months) in the AE3 2/2 mice. Selective b-wave depression in AE3 2/2 mice (evidenced by a drop in b/a ratio; Fig. 4a) is likely related to lower extracellular pH in the inner retina and reduced buffering power, following failed HCO 3 2 secretion by AE3. Consistent with this observation, ERG recordings from isolated toad retina indicate that acidification selectively suppresses the b-wave [27]. Small reductions in extracellular pH (on the order of 0.25) impede the function of the Cav1.4 presynaptic calcium channels [28], involved in synaptic transmission between photoreceptors and bipolar cells. Dysfunction in Cav1.4, such as occurs in congenital stationary night blindness (type 2), leads to selective bwave diminution, with sprouting of bipolar cell dendrites [29] as reported here in Slc4a3 2/2 retina. Therefore, both the absence of synaptic inputs on aberrantly extended bipolar cell dendrites, and Cav1.4 dysfunction in the photoreceptor to bipolar cell synapses would contribute to selective b-wave reduction in Slc4a3 2/2 mice. Furthermore, loss of AE3 in Müller cells likely affects their function, impacting on b-wave generation such as seen in Müller cell sheen dystrophy [30]. Lack of AE3 in horizontal cells might also contribute to b-wave reductions, since exclusive loss of horizontal cells causes selective b-wave reductions [31]. In addition, horizontal cell inhibitory feedback to cones (responsible for receptive field antagonistic surround) is regulated by pH changes at the synaptic level. Changes in pH resulting from H + accumulation modulate the presynaptic Cav1.4 voltage-gated calcium channels, in fish retina [32], and in transfected cells [28]. Also, reduction in flicker ERG and oscillatory potential amplitudes, at a time when scotopic a-waves remain unaffected, both support the occurrence of a selective inner retina defect. A progressive scotopic a-wave amplitude reduction, however, was observed in AE3 2/2 null mice only at 4 month of age and later, suggesting initiation of photoreceptor dysfunction and potentially photoreceptor death (Fig. 4c). The delayed onset of photoreceptor pathologies could be attributable, in part, to prolonged Müller cell dysfunction [33]. The elevated expression of GFAP by Müller cells reflects the induction of gliotic responses characteristic of retinal degeneration. Müller cells are essential to retina ionic homeostasis and impairment of their supportive functions have been associated with retinal dystrophies [33].
Ocular anomalies in AE3 2/2 and CAXIV 2/2 null mice Functional abnormalities in AE3 2/2 mice retinas were accompanied by increased expression of CAII and CAXIV enzymes, and NBC1 Na + / HCO 3 2 co-transporter, suggesting the occurrence of compensatory changes for the loss of AE3 (Fig. 5, 6). While we found compensatory changes in the AE3 2/2 knockout mice, the ability to maintain acid-base balance in the inner retina is likely to remain dramatically compromised given the phenotype of these mice.
Finally, the AE3 2/2 null mice presented with late onset photoreceptor cell death (Fig. 7A). Presumably, following functional ERGs and anatomical abnormalities, the apoptotic cascade was initiated in the AE3 2/2 null mice, as a result of absent AE3-mediated HCO 3 2 fluxes. In the eye, CAXIV, a membrane bound isozyme, is expressed within the retina, in Müller glial cells and retinal epithelial cells [25,34]. Because CAXIV is expressed in astrocytes, Müller cells, and the retinal epithelial cells, its expression spans the entire thickness of the retina. CAXIV distribution pattern is consistent with involvement in pH regulation and trans-retinal transport functions. Studies using CA inhibitors suggest that CAs, most likely CAII, CAIV and CAXIV, buffer the excess acidification in retina under certain metabolic conditions by preventing increase in H + concentration [35][36][37]. Therefore, a role for CAXIV in buffering the subretinal space volume could be predicted.
A mouse deficient for CAXIV was recently described [37]. Of interest, flash ERGs performed at 2, 7, and 10 months of age showed that the rod/cone a-wave, b-wave, and cone b-wave were significantly reduced (up to 45%) in the CAXIV 2/2 compared to wild-type mice [38], matching with findings of AE3 2/2 mice. Moreover, reductions in the dark-adapted response were not progressive between 2 and 10 months, as reported here for AE3 2/2 mice, suggesting a functional linked between AE3 and CAXIV in the normal eye function. Although CAXIV 2/2 and AE3 2/2 null mice shared very similar abnormal electrophysiological responses in the eye, the more profound pathology observed in the AE3 2/2 mice suggests additional roles played by the AE3 Cl 2 /HCO 3 2 exchanger in the retina.
Finally, we have found that the AE3/CAXIV forms a physical and functional complex in the mouse retina (Alvarez et al. Manuscript in preparation). The AE3-mediated HCO 3 2 flux is maximized by CAXIV interaction. Tethering of CAXIV [39] and CAII [12] to AE3 maximizes the transmembrane HCO 3 2 gradient  local to AE3, thereby activating transport rate. Association of the AE3 Cl 2 /HCO 3 2 exchanger and carbonic anhydrases II and XIV may represent a mechanism for the disposal of high CO 2 and H + production, and for pH regulation in the inner retina and brain. Recurrent phenotypes found for AE3 2/2 and CAXIV 2/2 null mice, and on direct interaction of AE3 and CAXIV, suggest that the association of AE3 Cl 2 /HCO 3 2 exchanger and CAXIV enzyme play a key role in normal ocular function.

Clinical implications
To date, only one human variant (Ala867Asp substitution), has been identified in the SLC4A3 gene [21,40]. The Ala867Asp variant was associated with common subtype idiopathic generalized epilepsy (IGE) in humans [40]. The effect of the variant on AE3 function has not been assessed. Eye examination and heart function of patients with the SLC4A3 Ala867Asp variant resulting in IGE have yet to be reported.
This paper provides the first documentation for a role of AE3 in normal ocular function. In addition, our findings demonstrate that lack of AE3 causes retinal pathology stereotypical of most vitreoretinal degenerations. Therefore SLC4A3 (AE3) is a novel candidate gene that must now be considered for the diagnosis of HVD.
HVD are a complex group of genetic disorders manifesting abnormalities of the vitreous and retina, affecting approximately 1 in 3000 people [41]. HVD includes lattice degeneration, snowflake vitreoretinal degeneration, X-linked juvenile retinoschisis, vitreoretinopathies (chromosome 5q, e.g. Wagner syndrome), chondrodysplasias (e.g. Stickler syndrome), and enhanced S-cone syndrome/Goldmann-Favre vitreotapetoretinal degeneration [42][43][44][45][46][47][48]. HVD can be difficult to distinguish because of subtle and Figure 7. Apoptosis in retinas of Slc4a3 2/2 mice. (A) Absence of apoptotic nuclei in retinas of 4 and 6-month-old Slc4a3 +/+ mice. Nuclei (blue) are stained with DAPI dye. Apoptotic nuclei are shown (green) in retinas of age-matched Slc4a3 2/2 mice littermates, with predominant apoptotic nuclei found at 6-months of age. (B) Model demonstrating functional significance of AE3fl expression in Mü ller glial cells, and functional significance of AE3c in horizontal cells. In Mü ller cells, cytoplasmic carbonic anhydrase (CAII) can trap CO 2 release by the photoreceptors intracellularly by converting to HCO 3 2 and H + . AE3fl located in Mü ller cells end feet facilitate the removal intro the vitreous or the blood of the HCO 3 2 and H + , in a process that is maximized by the CAXIV. In horizontal cells, AE3c may contribute to pH i homeostasis by exchanging intracellularly produced HCO 3 2 with Cl 2 . CAII and CAXIV are also involved in tight regulation of horizontal cells pH i , increasing the effectiveness of the bicarbonate buffer system. doi:10.1371/journal.pone.0000839.g007 overlapping clinical findings. Nevertheless, nearly all HVD present with altered ERG, in particular a reduction or absence of the scotopic ERG b-wave, different degrees of retinal lesions, poor visual acuity, and retinal blood vessel abnormalities [49]. Abnormal development of the vitreous progresses to increased vitreoretinal traction and retinal detachment, with a concomitant loss of vision. Additional ocular and systemic features of HVD may occur, depending on the underlying cause.
Our results provide compelling evidence linking AE3 Cl 2 / HCO 3 2 exchangers, pH imbalance in the inner retina, and vitreoretinal degeneration.