Characterization of Cav1.4 Complexes (α11.4, β2, and α2δ4) in HEK293T Cells and in the Retina*

Background: The properties of voltage-gated Ca2+ channels are regulated by auxiliary β and α2δ subunits. Results: Retinal β2 and α2δ4 subunits interact with Cav1.4 and differentially modulate channel properties compared with other Cav subunits. Conclusion: β2 and α2δ subunits are major determinants of Cav1.4 function. Significance: Cav1.4 channels in retinal photoreceptors are composed of auxiliary subunits with distinct properties compared with other Cav1 channels. In photoreceptor synaptic terminals, voltage-gated Cav1.4 channels mediate Ca2+ signals required for transmission of visual stimuli. Like other high voltage-activated Cav channels, Cav1.4 channels are composed of a main pore-forming Cav1.4 α1 subunit and auxiliary β and α2δ subunits. Of the four distinct classes of β and α2δ, β2 and α2δ4 are thought to co-assemble with Cav1.4 α1 subunits in photoreceptors. However, an understanding of the functional properties of this combination of Cav subunits is lacking. Here, we provide evidence that Cav1.4 α1, β2, and α2δ4 contribute to Cav1.4 channel complexes in the retina and describe their properties in electrophysiological recordings. In addition, we identified a variant of β2, named here β2X13, which, along with β2a, is present in photoreceptor terminals. Cav1.4 α1, β2, and α2δ4 were coimmunoprecipitated from lysates of transfected HEK293 cells and mouse retina and were found to interact in the outer plexiform layer of the retina containing the photoreceptor synaptic terminals, by proximity ligation assays. In whole-cell patch clamp recordings of transfected HEK293T cells, channels (Cav1.4 α1 + β2X13) containing α2δ4 exhibited weaker voltage-dependent activation than those with α2δ1. Moreover, compared with channels (Cav1.4 α1 + α2δ4) with β2a, β2X13-containing channels exhibited greater voltage-dependent inactivation. The latter effect was specific to Cav1.4 because it was not seen for Cav1.2 channels. Our results provide the first detailed functional analysis of the Cav1.4 subunits that form native photoreceptor Cav1.4 channels and indicate potential heterogeneity in these channels conferred by β2a and β2X13 variants.

Ca 2ϩ ions are important for many cellular functions, including neurotransmitter release, muscle contraction, and gene transcription. In neurons, depolarizing stimuli initiate Ca 2ϩ signaling mainly through activation of voltage-gated (Ca v ) Ca 2ϩ channels. Neuronal Ca v channels are multiprotein complexes composed of a major pore-forming ␣ 1 subunit, and two auxiliary subunits, ␤ and ␣ 2 ␦, which not only regulate the functional properties of the ␣ 1 subunit but also are important for the proper trafficking of the ␣ 1 subunit to the plasma membrane and protection of the ␣ 1 subunit from proteosomal degradation (reviewed in Refs. [1][2][3][4]. Ten different genes encode the Ca v ␣ 1 subunits, whereas the ␤ and ␣ 2 ␦ subunits are encoded by four genes each (5,6). Alternative splicing further adds to the molecular and functional diversity of ␣ 1 and ␤ subunits (1,7).
In photoreceptor terminals, Ca v 1 L-type channels are concentrated near the synaptic "ribbon," a structure specialized for high throughput and tonic exocytosis (8). At the depolarized photoreceptor membrane potential in darkness, Ca v 1-mediated Ca 2ϩ influx triggers the release of glutamate, which is required for transmission of visual stimuli to second-order neurons (9,10). Of the different classes of Ca v 1 channels (Ca v 1.1-Ca v 1.4), multiple lines of evidence indicate that Ca v 1.4 is the major Ca v 1 channel in rod and cone photoreceptors. Antibodies against Ca v 1.4 label both rod and cone terminals (11)(12)(13)(14). In mice, inactivation of CACNA1F, the gene encoding Ca v 1.4 (Ca v 1.4 KO), disrupts photoreceptor synaptic transmission and presynaptic calcium signaling (15) and prevents the maturation of photoreceptor synapses (13,16). In addition, human mutations in CACNA1F cause vision disorders, including incomplete congenital stationary night blindness 2, which is characterized by impaired rod photoreceptor transmission and low visual acuity in darkness (17)(18)(19)(20). Antibody labeling for the Ca v 1.3 ␣ 1 subunit has also been detected in the cones from tree shrew (21,22) and chick (23). However, in mice lacking Ca v 1.3, morphological changes in photoreceptor synapses are observed, but visual function is largely normal (24). The auxiliary Ca v 1.4 subunits in photoreceptors are most likely ␤ 2 and ␣ 2 ␦ 4 , because mice lacking functional ␤ 2 or ␣ 2 ␦ 4 subunits was then cloned between the XbaI and NotI sites of the pcDNA3.1 mammalian expression vector by ligating fragments XbaI-ClaI (F1-F3) with ClaI-HindIII (F4) and HindIII-NotI (F5).
␤ 2 Subunits-The ␤ 2X13 coding sequence was cloned from human retina cDNA library using PCR with primers FH1168 (5Ј-CACCATGCAGTGCTGCGGGCTGGT-3Ј) and FH 1169 (5Ј-TCATTGGCGGATGTAAACATCCCTG-3Ј) that hybridizes on the initiation and stop codons, respectively, of ␤ 2a and ␤ 2X13 . After subcloning into the pCR-blunt II vector and confirmation of correct sequence, ␤ 2X13 was transferred into the XbaI and BamHI sites of pcDNA3.1 vector using the XbaI and BamHI restriction sites provided by the pCR-blunt II vector. The ␤ 2a subunit was cloned from a human retina cDNA library using PCR in two fragments: F1 with primers FH1168 and FH1291 (5Ј-GGTTTAGGGGACCGGTGGTTTGC-3Ј) that inserts a silent mutation (underlined) at nucleotide 567, creating an AgeI site, and F2 with primers FH1290 (5Ј-GCAA-ACCACCGGTCCCCTAAACC-3Ј) and FH1169. The human ␤ 2a full coding sequence was then cloned between the XbaI and BamHI sites of the pcDNA3.1 vector by ligating fragments F1 Xba-AgeI (ATG-563) and F2 AgeI-BamHI (564-Stop codon). For bacterial expression of partial ␤ 2 protein fused to a His 6 tag, a fragment covering from nucleotide 1225 to the stop codon of human ␤ 2 was excised with BclI and HindIII from the pcDNA3.1-␤ 2X13 plasmid and cloned into the BamHI and HindIII sites of the pET30B vector (EMD Millipore, Billerica, MA).
␤ 1b and ␣ 2 ␦ 1 Fragments-For bacterial expression, a fragment of ␤ 1b corresponding to the sequence used to generate anti-␤ 2 antibodies was fused to a His 6 tag. A PCR fragment was amplified with primers 5Ј-GACCGGGCCACTGGGGAG-CAT-3Ј and 5Ј-TCAGCGGATGTAGACGCCTTGTC-3Ј and then cloned into the pdest17 vector (Invitrogen). Similarly, a partial fragment of ␣ 2 ␦ 1 corresponding to the sequence used to generate the anti-␣ 2 ␦ 4 antibody was amplified by PCR using primers (5Ј-ATGGCTGCTGGCTGCCTGCTG-3Ј) and (5Ј-TCATGTAAAACGTGGTGTCAATCT-3Ј) and cloned into the pET-SUMO vector.

Quantitative RT-PCR
Human retinas were obtained from donors without known eye diseases from the Oregon Lions eye bank as allowed by the Institutional Human Subjects Division of the University of Washington. Total RNAs were isolated from human retina using the UltraSpec RNA isolation system (Biotecx, Houston, TX). A two-step quantitative PCR was then carried out to determine the relative expression of splice variants. Total RNA (1 g) was subjected to first strand cDNA synthesis using Superscript III reverse transcriptase and oligo(dT) in a volume of 20 l according to the manufacturer's protocol (Invitrogen). For the human ␤ 2X13 quantitative PCR (qPCR), 2 primers were designed on exon 7B (FH1280, 5Ј-GCTAAGCAGAAGCA-GAAATCGAC-3Ј) and ϳ290 bp downstream of FH1280 on the exon 9-exon 10 junction (FH1284, 5Ј-TTCACTCTGAACTT-CCGCTAAG-3Ј). For human ␤ 2a , primers were designed on exon 7A (FH1279, 5Ј-GCTATAGACATAGATGCTACT-GGC-3Ј) and ϳ400 bp downstream of FH1279 on the exon 9-exon 10 junction (FH1284, 5Ј-TTCACTCTGAACTTCCG-CTAAG-3Ј). For normalization of the qPCR products, primers FH812 (5Ј-TCAACGGATTTGGTCGTATTGGGC-3Ј) and FH813 (5Ј-AGTGATGGCATGGACTGTGGTCAT-3Ј) were used to amplify human glyceraldehyde 3-phosphate dehydrogenase (GAPDH). Triplicate quantitative PCRs were carried out using 0.5 l of cDNA, a 400 nM concentration of each primer, and 10 l of QuantiTect SYBR Green PCR mix (Qiagen) in a 20-l total reaction volume. After an initial incubation at 95°C for 15 min, the qPCR was carried out for 40 cycles of denaturation at 95°C for 15 s, annealing at 68°C for 30 s, and extension at 72°C for 1 min on a real-time PCR system (ABI PRISM 7000, Applied Biosystems). Single bands of the predicted size were verified post-PCR by agarose gel electrophoresis. Threshold cycle (Ct) was determined using the ABI Prism 7000 software. Data were analyzed by comparing Ct values. All cDNAs were normalized relative to the Ct values of the internal control, GAPDH.

RNA-seq
A retinal RNA-seq experiment using samples from donor eyes (35) was used to determine the expression of human retinal ␤ 2 splice variants. These sequences were aligned to the human genome (release GRCh 37) using the Tuxedo pipeline (36). The resulting alignments were visually evaluated using IGV (37) to determine the specificity of transcriptional inclusion for exons 7A, 7B, and 7C of the CACNB2 gene.
For electrophysiological experiments, HEK293T cells were grown to 70 -80% confluence and cotransfected with human Ca v 1.4 or Ca v 1.2 ␣ 1 , ␤ 2X13 or ␤ 2a , and ␣ 2 ␦ 4 or ␣ 2 ␦ 1 cDNAs cloned in pcDNA3.1 and pEGFP cDNA as a transfection marker. Fugene transfection reagent (Promega) was used according to the manufacturer's protocol. After transfection, cells were maintained at 30°C. Cells were dissociated 18 -24 h after transfection and plated at low density for electrophysiological recording of single cells.

Immunohistochemistry
C57Bl/6J mouse eyecups were fixed in 4% paraformaldehyde in 0.1 M phosphate buffer, pH 7.4 (PB) for 30 min to 1 h. After fixation, tissues were incubated with increasing concentration of sucrose to 20% sucrose in PB and then embedded in 33% OCT compound (Miles, Elkhart, NY) diluted with 20% sucrose in PB. Eye tissues were cut in 12-m sections. To block nonspecific labeling, retinal sections were incubated with 3% normal goat serum in PBST buffer (10 mM sodium phosphate, 150 mM NaCl, 0.1% Triton X-100, pH 7.4) for 20 min at room temperature. Sections were incubated overnight at 4°C in a mix of diluted primary antibodies: mouse anti-CaBP4 (1:100) plus rabbit anti-Ca v 1.4 (1:1000), rat anti-␤ 2 (1:25) plus rabbit anti-Ca v 1.4 (1:1000), or rat anti-␣ 2 ␦ 4 (1:25) plus rabbit anti-Cav1.4 (1:1000). A mix of Alexa Fluor 555-conjugated goat anti-mouse IgG or Alexa Fluor 555-conjugated goat anti-rat IgG and Alexa 488-conjugated goat anti-rabbit IgG was reacted with the sections for 1 h at room temperature. Then the sections were rinsed in PBST and mounted with Prolong antifade reagent (Molecular Probes, Inc., Eugene, OR). Sections were analyzed under a confocal microscope (Zeiss LSM710). Immunofluorescent images were obtained with a Plan-Neofluar ϫ40/1.3 numerical aperture (Zeiss) objective lens. For immunohistochemistry experiments with rat anti-␤ 2 and rat anti-␣ 2 ␦ 4 , antigen retrieval was performed by incubating the sections at 80°C for 20 min in 10 mM sodium citrate, pH 6.0, 0.05% Tween 20 and washing with PBS before blocking and incubation with the antibodies. Quantification of colocalization was performed using the JACoP plugin for ImageJ (National Institutes of Health), which uses correlation analysis based on Pearson's coefficient (40). A Pearson's coefficient near 1 suggests perfect correlation, whereas a number near 0 indicates no correlation.

Coimmunoprecipitation
Three days after transfection, whole cell lysates were prepared by incubation of transfected cells at 4°C for 1 h in 20 mM Tris, pH 7.5, 150 mM NaCl, 1% Nonidet P-40, 0.5% sodium deoxycholate, and inhibitors of proteases (Sigma-Aldrich) with or without 0.1 mM MgCl 2 and 0.1 mM CaCl 2 . Lysates were centrifuged at 22,000 ϫ g for 30 min and incubated with mouse IgG or anti-FLAG (5 g). After a 1-h incubation at 4°C, 20 l of protein G-magnetic beads (Invitrogen) were added, and the incubation proceeded for 3 h at 4°C. After four washes with lysis buffer, proteins were eluted with SDS-sample buffer and analyzed by Western blotting with specific antibodies.
For coimmunoprecipitations from retinas, 12 retinas from WT or Ca v 1.4 KO mice were lysed in 20 mM Hepes, pH 7.4, 150 mM NaCl, 1 mM EDTA, 1% n-dodecyl-␤-D-maltoside and a mix of protease inhibitors (Sigma-Aldrich) at 4°C for 30 min. Lysates were centrifuged at 10,000 ϫ g for 10 min, and 500 g of supernatant was incubated with 5 g of anti-␣ 2 ␦ 4 or anti-␤ 2 at 4°C. After 1 h of incubation, 30 l of protein A-Sepharose bead slurry was added to the mix and incubated for 1 h at 4°C. Beads were washed three times with lysis buffer without detergent, and the proteins were eluted with SDS-sample buffer plus 10 mM DTT before analysis by Western blotting with Ca v 1.4 antibodies.

Proximity Ligation Assay
The proximity ligation assay was performed using the Duolink kit (Sigma-Aldrich). Mouse retina sections were prepared as described above for immunohistochemistry and incubated overnight with rabbit anti-Ca v 1.4 and mouse anti-CaBP4, rat anti-␤ 2 , or rat anti-␣ 2 ␦ 4 . For the assay using rat anti-␤ 2 and rat anti-␣ 2 ␦ 4 , the sections were then incubated for 1 h at room temperature with mouse anti-rat antibody. Anti-rabbit PLUS probe and anti-mouse MINUS probe were then added to the sections for 1 h at 37°C. Ligation, amplification, and detection of the probes were carried out according to the manufacturer's protocol. The sections were mounted with antifade reagent and analyzed under a confocal microscope as described above. To quantify proximity ligation assay (PLA) signals, an arbitrary area of identical size (x ϭ 50 m ϫ y ϭ 10 m) was selected in the outer plexiform layer (OPL) of single z plane images as well as in the outer nuclear layer (ONL) to be able to normalize the measurements. Indeed, because there is variability in the background signals from one section to the other, we subtracted the number of PLA signals in the ONL from the number of PLA signals in the OPL. Quantification of PLA spots was performed using the "analyze particles" tools in ImageJ version 1.48.

Electrophysiological Recordings
Whole-cell patch clamp recordings of cells transfected with Ca v 1 channel subunits were performed 36 -48 h after transfection. Data were acquired with an EPC-9 patch clamp amplifier driven by Pulse software (HEKA Elektronik, Lambrecht/Pfalz, Germany) and analyzed with Igor Pro software (Wavemetrics, Lake Oswego, OR). Extracellular recording solutions contained 140 mM Tris, 1 mM MgCl 2 , and 20 mM BaCl 2 . Intracellular solutions consisted of 140 mM N-methyl-D-glucamine, 10 mM HEPES, 2 mM MgCl 2 , 2 mM Mg-ATP, and 5 mM EGTA. The pH of intracellular and extracellular recording solutions was adjusted to 7.3 with methanesulfonic acid. Electrode resistances were typically 1-2 megaohms, and series resistance compensated up to 80%. All averaged data are presented as the mean Ϯ S.E. Statistical significance of differences between groups was determined by Student's t test as indicated (Sigma-Plot, SPSS Science, Chicago, IL). Normalized tail currents were fit with the Boltzmann equation: where I is the maximal current, V is the test voltage, V h is the voltage of half-activation, k is the slope factor, and b is the baseline).

Molecular Identification of Ca v 1.4 Subunits in Human
Retina-To characterize the molecular constituents of Ca v 1.4 complexes in the retina, we performed RT-PCR from human retinas using primers specific for Ca v 1.4 ␣ 1 , ␤ 2 , and ␣ 2 ␦ 4 . We isolated full-length wild-type Ca v 1.4 ␣ 1 and ␣ 2 ␦ 4 cDNAs that were identical to those described previously (GenBank TM accession numbers AF201304 and NM_172364, respectively). The large majority (Ն90%; Fig. 1) of PCR products for ␤ 2 corresponded to an alternative splice variant that includes a differ-ent exon 7. This retinal ␤ 2 variant includes palmitoylation sites known so far to be specific to ␤ 2a and includes exon 7B instead of the exon 7A that is incorporated in ␤ 2a (Fig. 1A, ␤ 2a GenBank TM number NM_000724; exon numbering according to Buraei et al. (1)). Exon 7B is analogous to the exon 7B included in ␤ 1b and ␤ 1c mRNAs (1). This new retinal variant has here been named ␤ 2X13 (Fig. 2). A ␤ 2 variant with sequence identical to the retinal ␤ 2X13 was isolated from human jejunum (GenBank TM number AF465485) (41). Although ␤ 2X13 is expressed at higher levels than ␤ 2a in the retina, we were also able to detect the expression of ␤ 2a transcripts that include exon 7A (Fig. 1, B    expression of ␤ 2a and ␤ 2X13 was variable between human retinas, with ␤ 2X13 being 9 -77 times more abundant than ␤ 2a ( Fig.  1B and Table 1). ␤ 2a was undetectable in one human retina (data not shown). By RT-PCR, we also detected ␤ 2X13 expression in mouse retina (Fig. 1C), where its expression was also greater than that of ␤ 2a (data not shown). ␤ 2X13 was not detected in mouse brain and is thus probably tissue-specific (Fig. 1C). Further evidence for the predominance of ␤ 2X13 in the human retina was obtained in a retinal RNA-seq data set (35), which was specifically evaluated for the inclusion rates of exons 7A, 7B, and 7C. A ␤ 2 variant including the 7B exon and concomitantly excluding the 7A and 7C exons was observed to be the dominant retinal isoform in both the macular and peripheral retina (Fig. 1D). In general, exon 7B was found in greater than 95% of sampled transcripts. The mouse ␤ 2X13 sequence was submitted to GenBank TM and assigned accession number KJ789960 (Fig. 2).  anti-␤ 2 , anti-␤ 2 preadsorbed with specific ␤ 2X13 or nonspecific ␤ 1b (left), or anti-␣ 2 ␦ 4 , or anti-␣ 2 ␦ 4 preadsorbed with specific ␣ 2 ␦ 4 or nonspecific ␣ 2 ␦ 1 (right). Immunoreactivity was blocked by preadsorption with specific antigen only.
Using these antibodies for double-labeling of mouse retina, we observed selective labeling of the outer plexiform layer containing the photoreceptor terminals; no labeling was observed in other retinal layers (Fig. 4). High magnification images showed anti-Ca v 1.4 ␣ 1 labeling of elongated and horseshoeshaped structures corresponding to the photoreceptor synaptic ribbon that arches around the postsynaptic terminals, as described previously (13,33). The anti-␤ 2 or anti-␣ 2 ␦ 4 signals colocalized at these structures with anti-Ca v 1.4 ␣ 1 staining, as indicated by a Pearson's correlation coefficient of Ն0.710 or Ն0.608, respectively. These findings are consistent with the presence of Ca v 1.4 ␣ 1 -␤ 2 -␣ 2 ␦ 4 complexes that are concentrated near photoreceptor synaptic ribbons.
Ca v 1.4 ␣ 1 , ␤ 2X13 , and ␣ 2 ␦ 4 Subunits Interact in HEK293 Cells and in Mouse Retina-We next tested whether Ca v 1.4 ␣ 1 associates with ␤ 2X13 and ␣ 2 ␦ 4 in transfected HEK293 cells. The Ca v 1.4 ␣ 1 subunit was tagged with a FLAG epitope, and anti-FLAG antibodies were used to immunoprecipitate FLAG-Ca v 1.4 and associated proteins. In these experiments, we also tested for the co-immunoprecipitation of CaBP4, a known Ca v 1.4-interacting protein (28,42). Because the association of CaBP4 with Ca v 1.4 can be affected by Ca 2ϩ , coimmunoprecipitation experiments were done in the presence and absence of Ca 2ϩ . Due to the limited sensitivity of FLAG antibodies, FLAG-Ca v 1.4 ␣ 1 was not detected in cell lysates by Western blot (Fig.  5, left panels). However, FLAG-Ca v 1.4 ␣ 1 was visualized after enrichment using immunoprecipitation (Fig. 5, right panels). Western blotting with the corresponding antibodies revealed the co-immunoprecipitation of ␤ 2 , ␣ 2 ␦ 4 , and CaBP4 with FLAG-Ca v 1.4 both with and without Ca 2ϩ (Fig. 5, A and B). This result was not reproduced in cells transfected without FLAG-Ca v 1.4, which argued against nonspecific immunoprecipitation of ␤ 2 , ␣ 2 ␦ 4 , and CaBP4 by FLAG antibodies (Fig. 5A). No immunoprecipitated proteins were detected either when control mouse IgG was used instead of FLAG antibodies (Fig. 5,  A and B). We also tested whether Ca v 1.4 ␣ 1 associates with ␤ 2 and ␣ 2 ␦ 4 in mouse retina using coimmunoprecipitation with anti-␤ 2 and anti-␣ 2 ␦ 4 . Ca v 1.4 ␣ 1 coimmunoprecipitated with ␤ 2 and ␣ 2 ␦ 4 (Fig. 5C).
To test whether Ca v 1.4 subunits interact in photoreceptor synapses, we performed in situ PLAs, a technique that allows visualization of protein interactions in fixed tissue. With this method, antibodies against two potentially interacting proteins are coupled to oligonucleotides. If the two proteins interact, the antibodies that recognize these proteins in fixed tissue should be in close proximity (Ͻ40 nm) such that their coupled oligonucleotides can be ligated and further visualized with fluorescent probes. As proof of principle, we first tested the interaction of Ca v 1.4 ␣ 1 and CaBP4, which is known to colocalize with Ca v 1.4 at the photoreceptor terminals in the OPL (Fig. 6A) (28). PLA performed with anti-CaBP4 and anti-Ca v 1.4 ␣ 1 produced strong signals in the OPL, where both proteins are colocalized (Fig. 6, A and B). These signals were specific for the presence of CaBP4 because they were strongly reduced in the retina of mice lacking CaBP4 (CaBP4 KO; Fig. 6C). We have shown previously that immunolabeling for Ca v 1.4 channels is still present in the OPL of CaBP4 KO retina (13), so the absence of PLA signal is probably due to the lack of CaBP4/Ca v 1.4 interactions. A few fluorescent puncta were observed in the outer and inner nuclear layers of WT retina, but these were probably nonspecific because they were also seen in CaBP4 KO retina (Fig. 6C). With anti-Ca v 1.4 ␣ 1 and either anti-␤ 2 or anti-␣ 2 ␦ 4 antibodies, PLA generated strong signals in the OPL of WT but not Ca v 1.4 KO mice (Fig. 6, E-L). Again, only a few puncta were observed in the outer and inner nuclear layers of both WT and Ca v 1.4 KO retina, so they were probably nonspecific (Fig. 6, G and K). Quantification showed significantly greater density of PLA signals in WT compared with Ca v 1.4 KO retina using antibodies for Ca v 1.4-CaBP4, Ca v 1.4-␤ 2 , or Ca v 1.4-␣ 2 ␦ 4 complexes (Fig. 6, D, H, and L; p Ͻ 0.0001). Additional negative controls for PLA were done by omission of one or both antibodies (Fig. 6, M-P) or by preadsorption of anti-␤ 2 antibody with ␤ 2a or anti-␣ 2 ␦ 4 with ␣ 2 ␦ 4 (Fig. 6, Q and R). No PLA signals were observed under these conditions. Taken together, our results confirm that Ca v 1.4 channel complexes at the photoreceptor synapse are composed of Ca v 1.4 ␣ 1 , ␤ 2 , and ␣ 2 ␦ 4 and CaBP4.
Electrophysiological Analysis of Ca v 1.4 Channels in HEK293T Cells-Despite the molecular characterization of ␣ 2 ␦ 4 as a Ca v subunit (43), how ␣ 2 ␦ 4 affects the biophysical properties of Ca v channels, particularly Ca v 1.4, is unknown; previous electrophysiological studies of Ca v 1.4 utilized the ␣ 2 ␦ 1 subunit (27, 30 -32, 44). To better understand the functional properties of the native photoreceptor Ca v 1.4 channel complex, we performed whole-cell patch clamp recordings of Ca v 1.4 channels containing ␣ 2 ␦ 4 in transfected HEK293T cells. For comparison, we also carried out recordings of Ca v 1.4 with the ␣ 2 ␦ 1 subunit. Based on our findings that ␤ 2X13 is more highly expressed in the retina than ␤ 2a (Fig. 1), we used ␤ 2X13 in these experiments. Ba 2ϩ was used as the charge carrier because the greater permeation of Ba 2ϩ compared with Ca 2ϩ increases the resolution of Ca v 1.4 currents, which are significantly smaller compared with those mediated by other Ca v channels in HEK293T cells (31,32). The properties of Ba 2ϩ currents are largely similar to Ca 2ϩ currents because Ca v 1.4 channels containing the distal C-terminal domain inhibitory module exhibit little Ca 2ϩ -dependent inactivation (31,32,44,45). In cells cotransfected with Ca v 1.4 ␣ 1 -␤ 2X13 -␣ 2 ␦ 4 , Ba 2ϩ currents were generally smaller than in cells expressing Ca v 1.4 ␣ 1 -␤ 2X13 -␣ 2 ␦ 1 , although this difference did not reach statistical significance in plots of current density against test voltage (Fig.  7, A and B). However, analyses of normalized tail current-voltage curves revealed that Ca v 1.4 ␣ 1 -␤ 2X13 -␣ 2 ␦ 4 channels exhibited weaker voltage dependence of activation compared with channels containing Ca v 1.4 ␣ 1 -␤ 2X13 -␣ 2 ␦ 1 (Fig. 7C). Boltzmann fits indicated more positive half-maximal activation voltage (V h ) and greater slope (k) for channels containing ␣ 2 ␦ 4 compared with those with ␣ 2 ␦ 1 ( Table 2). In these experiments, tail currents were normalized to those measured upon repolarization from a ϩ80-mV step, at which channel open probability should be maximal. Interestingly, the normalized tail current amplitude exceeded 1 between 0 and ϩ30 mV with ␣ 2 ␦ 1 but not with ␣ 2 ␦ 4 (Fig. 7C), similar to that due to Ca 2ϩ -dependent facilitation of Ca v 2.1 channels (46). Because we used Ba 2ϩ as the charge carrier, this "overshoot" in the normalized tail current-  . Differential modulation of Ca v 1.4 properties by ␣ 2 ␦ 1 and ␣ 2 ␦ 4 . A and B, Ba 2ϩ currents (I Ba ) were evoked by 50-ms depolarizations from a holding voltage of Ϫ80 mV in HEK293T cells transfected with Ca v 1.4 containing ␣ 2 ␦ 1 (n ϭ 8) or ␣ 2 ␦ 4 (n ϭ 9) and ␤ 2X13 . A, representative I Ba traces during a 50-ms depolarization to Ϫ70, 0, and ϩ30 mV. B, current-voltage (I-V) plots for data obtained as in A. C, tail currents were evoked after 20-ms depolarization to various voltages from a holding voltage of Ϫ80 mV and repolarization to Ϫ60 mV. Tail currents were normalized to that obtained after a ϩ80-mV pulse and plotted against test voltage. D, inactivation of I Ba evoked by 5-s pulses from Ϫ80 to 0 mV. Left, representative traces. Right, inactivation was measured by dividing residual current amplitude (I res ) by the peak amplitude (I peak ) in cells transfected with Ca v 1.4 containing ␣ 2 ␦ 1 (n ϭ 5) or ␣ 2 ␦ 4 (n ϭ 6), p ϭ 0.59, by Student's t test. The time constant, , was obtained by fitting I Ba decay with a single exponential function and was not different for ␣ 2 ␦ 1 and ␣ 2 ␦ 4 ; p ϭ 0.88, Student's t test. Error bars, S.E.
We next compared the impact of ␤ 2a and ␤ 2X13 on Ca v 1.4 function. Compared with other Ca v ␤ subunits, ␤ 2a significantly slows VDI of Ca v channels (47). Structure/function analyses indicate that this property of ␤ 2a is mediated by the HOOK domain (48,49). Because ␤ 2X13 and ␤ 2a differ in the HOOK domain (Fig. 1A), we predicted that these splice variants may have distinct effects on VDI of Ca v 1.4. Consistent with previous findings that the HOOK domain did not affect voltage-dependent activation (48), there were no differences in V h or k obtained from normalized tail current-voltage curves in cells transfected with Ca v 1.4 ␣ 1 -␤ 2X13 -␣ 2 ␦ 4 or Ca v 1.4 ␣ 1 -␤ 2a -␣ 2 ␦ 4 (Fig. 8, A and B, and Table 2). However, VDI during a 5-and 10-s depolarizing pulse was significantly greater for Ca v 1.4 ␣ 1 -␤ 2X13 -␣ 2 ␦ 4 than for Ca v 1.4 ␣ 1 -␤ 2a -␣ 2 ␦ 4 , although the time constants for VDI were not different (see legend to Fig. 8D for values). These results show that the ␤ 2X13 subunit influences the amount but not the rate of VDI of Ca v 1.4 channels.
To determine whether ␤ 2X13 had similar effects on other Ca v channels, we compared properties of Ba 2ϩ currents in cells transfected with Ca v 1.2 ␣ 1 -␤ 2X13 -␣ 2 ␦ 1 and those transfected with Ca v 1.2 ␣ 1 -␤ 2a -␣ 2 ␦ 1 . In these experiments, ␣ 2 ␦ 1 was used because native Ca v 1.2 channels in the brain are not likely to associate with ␣ 2 ␦ 4 based on the near absence of ␣ 2 ␦ 4 , but prev-alence of ␣ 2 ␦ 1 , in mouse brain (50). Ba 2ϩ current density was generally lower for Ca v 1.2 ␣ 1 -␤ 2X13 -␣ 2 ␦ 1 than for Ca v 1.2 ␣ 1 -␤ 2a -␣ 2 ␦ 1 but not significantly different. Unlike its lack of effect on Ca v 1.4 activation, ␤ 2X13 caused a small but significant (ϳϩ5 mV) shift in the V h obtained from normalized tail current curves compared with ␤ 2a for Ca v 1.2 channels (Fig. 9, A-C, and Table 2). However, unlike for Ca v 1.4, ␤ 2X13 and ␤ 2a had similar effects on VDI (Fig. 9D). Therefore, ␤ 2X13 and ␤ 2a differentially modulate the functional properties of Ca v 1.2 and Ca v 1.4.

DISCUSSION
In this paper, we extend previous work indicating the importance of the auxiliary Ca v subunits, ␤ 2 and ␣ 2 ␦ 4 , in the retina. First, we provide the first evidence that Ca v 1.4 ␣ 1 , ␤ 2 , and ␣ 2 ␦ 4 physically and functionally interact both in transfected HEK293T cells and in photoreceptor synaptic terminals in the retina. Second, we identify the ␤ 2X13 splice variant, which is distinct from the ␤ 2a subunit expressed in the brain, as the major ␤ 2 subunit in human retina. Finally, we show that both ␤ 2X13 and ␣ 2 ␦ 4 have distinct effects on the biophysical properties of Ca v 1.4 compared with ␤ 2a and ␣ 2 ␦ 1 . Our results reveal unexpected differences in the modulatory capabilities of Ca v ␤ and ␣ 2 ␦ subunits, which may influence the properties of native Ca v 1.4 channels in photoreceptors.
Role of Ca v ␤ and ␣ 2 ␦ 4 as Auxiliary Subunits of Ca v 1.4 in the Retina-The notion that ␤ 2 and ␣ 2 ␦ 4 associate with Ca v 1.4 channels in the retina is supported by immunochemical and genetic evidence. First, antibodies against ␤ 2 and ␣ 2 ␦ 4 strongly label the OPL of the retina, where photoreceptor synapses are localized (51)(52)(53). In contrast, antibodies against ␤ 1, ␤ 3 , and ␤ 4 label other regions in the retina but not the OPL (51). Second, mice lacking expression of ␤ 2 , but not mice lacking ␤ 1 , ␤ 3 , or ␤ 4 , exhibited strongly reduced b-waves in electroretinograms (25). The b-wave indicates efficiency of transmission from photoreceptors to bipolar neurons and is highly dependent on proper function of Ca v 1.4. The b-wave is absent in Ca v 1.4 KO mice (15) and in mice expressing a mutation in the ␣ 2 ␦ 4 gene, Cacna2d4, in which a truncated ␣ 2 ␦ 4 protein is expressed at very low levels compared with the full-length protein in wild-type mice (26). In the ␣ 2 ␦ 4 mutant mice, there is also delayed degeneration of cones, and human mutations in CACNA2D4 are associated with slowly progressive cone dystrophy (54). Our findings that ␤ 2 and ␣ 2 ␦ 4 colocalize with Ca v 1.4 ␣ 1 at structures resembling photoreceptor synaptic ribbons (Fig. 4) and interact with Ca v 1.4 ␣ 1 in transfected cells (Fig. 5) and in the OPL (Fig. 6) verify that these auxiliary subunits are primary components of Ca v 1.4 channel complexes at the photoreceptor synapse.
Molecular and Functional Characterization of a New Retinal Ca v ␤ 2X13 Splice Variant-Ca v ␤ subunits regulate multiple aspects of Ca v channel function, including their gating properties and levels at the cell surface (reviewed in Refs. 1 and 55). Of the four classes of Ca v ␤ that have been characterized, ␤ 2 is subject to the most significant alternative splicing, with 13 variants thus far identified (1,41). Despite the evidence in favor of ␤ 2 as a Ca v 1.4 subunit, there has been no characterization of Ca v ␤ 2 variants expressed in human retina. In our RT-PCR analyses of human retina, we detected the ␤ 2a variant that is present in the brain (56) as well as a ␤ 2 splice variant, ␤ 2X13 , which has not  (57) and containing a shorter HOOK domain partially encoded by exon 7B (Fig. 1A). Whereas ␤ 2X13 was consistently detected at higher levels by qPCR than ␤ 2a in RNA isolated from three different individuals, there was large variability in the ratio between ␤ 2X13 and ␤ 2a ( Fig. 1B and Table 1). Previous analyses indicate that ␤ 2 splice variant expression can vary significantly during development (58). Because we did not have information regarding the age of the individuals from which the samples were isolated, it is possible that our results reflect age-related differences in ␤ 2X13 and ␤ 2a levels. We detected ␤ 2X13 in both mouse and human retina but not in mouse brain (Fig. 1C), which further supports the physiological importance of ␤ 2X13 as a partner for photoreceptor Ca v 1.4 channels. In addition, RNA-seq analysis of the retinal ␤ 2 transcripts showed that ␤ 2X13 is the major transcript in the cone-rich macula and in the peripheral retina, indicating that ␤ 2X13 is the main isoform in both rods and cones. The inclusion of exon 7B in ␤ 2X13 creates a shorter HOOK domain (7 amino acids) compared with the exon 7A-containing ␤ 2a (45 amino acids; Fig. 1A). The HOOK domains of ␤ 1b , ␤ 1c , ␤ 3 , and ␤ 4 are the same length as (and contain the same sequence (AKQKQK) that is present in) the ␤ 2X13 exon 7B (1). Transfer of the HOOK domain from ␤ 1b to ␤ 2a causes strong VDI of Ca v 2.1 channels, typical of ␤ 1b (48). Moreover, deletion of the HOOK domain from ␤ 2a enhances VDI of Ca v 2.2 channels (49). Consistent with these findings, we found that, compared with ␤ 2a , ␤ 2X13 causes stronger VDI of Ca v 1.4 channels (Fig. 8D). Curiously, this was not the case for Ca v 1.2 (Fig. 9D). The effect of ␤ 2a in slowing VDI of Ca v 1.2 is less profound than observed for other Ca v channels (59), so it may be that Ca v 1.2 VDI is somehow more resistant to modulation by ␤ 2a subunits than Ca v 1.4, such that the distinct effects of ␤ 2a and ␤ 2X13 on VDI are more readily resolved in Ca v 1.4 than Ca v 1.2.
The distinct exon 7B present in ␤ 2X13 is orthologous to exon 9 of the zebrafish ␤ 2.1 gene, which is found in a number of ␤ 2.1 variants that are expressed at distinct developmental ages and in different tissues, including the eye (60). Whereas ␤ 2X13 includes the two palmitoylated N-terminal cysteines present in ␤ 2a (57), these ␤ 2.1 variants do not. Although it is tempting to speculate on the physiological significance of ␤ 2X13 -containing Ca v 1.4 channels, it is important to note that VDI is still relatively limited for these channels, with ϳ35% of the current remaining after 10 s of depolarization (Fig. 9D). Moreover, maximal depolarization of photoreceptors is likely not to exceed ϳϪ40 mV, which should not induce significant VDI of Ca v 1.4 channels. Therefore, whether the difference in VDI due to ␤ 2X13 and ␤ 2a is physiologically relevant for photoreceptor synaptic transmission is debatable. An alternate possibility is that ␤ 2X13 may affect modulation of Ca v 1.4 by other factors. For example, Ca v 1.3 channels containing ␤ 2a but not other ␤ subunits are relatively resistant to inhibition by arachidonic acid (61). Because synaptic Ca 2ϩ currents in salamander photoreceptors are strongly inhibited by arachidonic acid (62), it will be of interest to determine whether ␤ 2X13 alters the sensitivity of Ca v 1.4 channels to this arachidonic acid or other neuromodulators.
Functional Characterization of Ca v 1.4 Channels Containing the ␣ 2 ␦ 4 Subunit-Ca v ␣ 2 ␦ subunits are composed of ␣ 2 and ␦ proteins that are encoded by a single gene; proteolytic cleavage of the ␣ 2 -␦ preprotein separates ␣ 2 and ␦ proteins, which remain linked under native conditions by disulfide bonds (reviewed in Ref. 3). Ca v ␣ 2 ␦ subunits enhance the cell surface trafficking and decrease the turnover of Ca v 1 and Ca v 2 channels (63,64). These effects are due in part to a von Willebrand factor A (VFA) domain and in particular a metal ion adhesion motif (MIDAS), the mutation of which prevents the cell surface trafficking function of ␣ 2 ␦ 2 on Ca v 2.1 currents (64). In addition, glycosylation of the extracellular domain is important for the current-enhancing effects of ␣ 2 ␦ (65). ␣ 2 ␦ 4 contains fewer glycosylation sites compared with the three other ␣ 2 ␦ subunits (43), which may explain why Ca v 1.4 current densities were generally smaller with ␣ 2 ␦ 4 than with ␣ 2 ␦ 1 (Fig. 7, A and B). Notably, ␣ 2 ␦ 4 does not have the asparagine corresponding to residue 184, which, when mutated to glutamine, inhibits the enhancement of Ca v 2.2 channels caused by ␣ 2 ␦ 1 (66).
Ca v ␣ 2 ␦ subunits have been shown to have variable effects on voltage-dependent activation of Ca v channels, which may depend on the particular Ca v ␣ 1 subunit with which they associate (reviewed in Ref. 3). For Ca v 1.2 channels, ␣ 2 ␦ 1 causes a negative shift in the voltage dependence of activation, which depends largely on the ␦-encoding sequence (67). Our findings that Ca v 1.4 channels containing ␣ 2 ␦ 4 exhibit weaker voltage dependence of activation than with ␣ 2 ␦ 1 may result from relatively weak conservation (ϳ28% sequence identity) in the ␦-encoding sequence for ␣ 2 ␦ 4 and ␣ 2 ␦ 1 . A further distinction between ␣ 2 ␦ 4 and ␣ 2 ␦ 1 was in the facilitation of currents evoked by moderate voltages that was unique to ␣ 2 ␦ 1 -containing Ca v 1.4 channels (Fig. 7C). Although further analyses are required, this facilitation may result from the binding of permeant Ba 2ϩ ions in the channel protein, similar to what has been proposed for ion-dependent inactivation of Ba 2ϩ currents through Ca v 1.2 channels (68). The absence of this facilitation, in addition to relatively weak voltage dependence of activation, in ␣ 2 ␦ 4 -containing Ca v 1.4 channels might be considered maladaptive in terms of supporting synaptic Ca 2ϩ signals that drive photoreceptor transmission. However, native Ca v 1.4 channels would also be associated with CaBP4, which significantly enhances voltage-dependent activation of Ca v 1.4 (28,42). Thus, native ␣ 2 ␦ 4 -containing Ca v 1.4 channels in photoreceptors may be more strongly modulated by CaBP4 than ␣ 2 ␦ 1containing Ca v 1.4 channels. ␣ 2 ␦ subunits also play roles in regulating neurotransmitter release probability (69) as well as synaptogenesis (70,71), independent of effects on Ca v function. Understanding the precise roles of ␣ 2 ␦ 4 in regulating photoreceptor signaling and how its dysregulation leads to vision impairment in humans and mice (26,54) remains an important challenge for future studies.