Alternative Splicing Governs Cone Cyclic Nucleotide-gated (CNG) Channel Sensitivity to Regulation by Phosphoinositides*

Background: Variants of photoreceptor CNG channels are produced by alternative splicing of precursor mRNA. Results: Inclusion of an optional exon in human CNGA3 transcripts produces channel isoforms with enhanced sensitivity to phosphoinositides via an allosteric mechanism. Conclusion: Alternative splicing of CNGA3 transcripts controls phosphoinositide regulation of cone CNG channels. Significance: These findings reveal the functional importance of CNGA3 alternative splicing. Precursor mRNA encoding CNGA3 subunits of cone photoreceptor cyclic nucleotide-gated (CNG) channels undergoes alternative splicing, generating isoforms differing in the N-terminal cytoplasmic region of the protein. In humans, four variants arise from alternative splicing, but the functional significance of these changes has been a persistent mystery. Heterologous expression of the four possible CNGA3 isoforms alone or with CNGB3 subunits did not reveal significant differences in basic channel properties. However, inclusion of optional exon 3, with or without optional exon 5, produced heteromeric CNGA3 + CNGB3 channels exhibiting an ∼2-fold greater shift in K1/2,cGMP after phosphatidylinositol 4,5-biphosphate or phosphatidylinositol 3,4,5-trisphosphate application compared with channels lacking the sequence encoded by exon 3. We have previously identified two structural features within CNGA3 that support phosphoinositides (PIPn) regulation of cone CNG channels: N- and C-terminal regulatory modules. Specific mutations within these regions eliminated PIPn sensitivity of CNGA3 + CNGB3 channels. The exon 3 variant enhanced the component of PIPn regulation that depends on the C-terminal region rather than the nearby N-terminal region, consistent with an allosteric effect on PIPn sensitivity because of altered N-C coupling. Alternative splicing of CNGA3 occurs in multiple species, although the exact variants are not conserved across CNGA3 orthologs. Optional exon 3 appears to be unique to humans, even compared with other primates. In parallel, we found that a specific splice variant of canine CNGA3 removes a region of the protein that is necessary for high sensitivity to PIPn. CNGA3 alternative splicing may have evolved, in part, to tune the interactions between cone CNG channels and membrane-bound phosphoinositides.

Vertebrate visual transduction requires the activity of cyclic nucleotide-gated (CNG) 2 channels. Photoreceptor light responses generate a decrease in cGMP concentration that is converted into an electrical signal (membrane hyperpolarization) by the closure of CNG channels, leading to decreased neurotransmitter release onto second-order cells (1). Photoreceptor CNG channels are members of the large superfamily of cation-selective, "pore-loop" channels (2,3). They are tetrameric proteins, with each subunit possessing six transmembrane segments, amino-terminal (N) and carboxyl-terminal (C) cytoplasmic domains, a central pore element contributing to the ion conduction pathway, and a cyclic nucleotide-binding domain located in the C-terminal region (4). Channel sensitivity to cGMP is tuned by several pathways, including the action of calcium sensor proteins such as calmodulin (5)(6)(7)(8)(9), serine/threonine and tyrosine phosphorylation (10,11), extracellular matrix metalloproteinases (12,13), and changes in the concentration of plasma membrane phosphoinositides such as PIP 2 and PIP 3 (14 -16). Adjustments of CNG channel ligand affinity can contribute to photoreceptor adaptation and to paracrine and/or circadian changes in photoreceptor sensitivity (17)(18)(19).
The functional diversity of CNG channels is generated via several fundamental mechanisms. First, multiple genes encode CNG channel subunits, and cell type-specific expression of combinations of CNG channel subunits produce heteromeric channel assemblies having properties optimized for the role they play in those cells (4). For example, in cone and rod photoreceptors, the channels are composed CNGA3 and CNGB3 subunits or CNGA1 and CNGB1 subunits, respectively. Second, protein variants of CNGB1 (20,21) and CNGA3 (22)(23)(24) are known to arise from alternative splicing of precursor mRNAs. Alternative splicing of primary transcripts is a prevalent mechanism for generating proteome complexity from a limited number of genes and is thought to have particular importance in the central nervous system (25). For CNGB1, a short variant (CNGB1b) is expressed in olfactory receptor neu-* This work was supported by NEI, National Institutes of Health grant EY12836 (to M. D. V.). 1 To whom correspondence should be addressed: Dept rons where it forms part of the olfactory CNG channel, whereas a long variant (CNGB1a) is expressed in rod photoreceptor cells. In addition, the CNGB1 transcriptional unit generates splice variants encoding soluble GARP1 and GARP2 proteins, short isoforms lacking the channel-forming domain but serving structural and regulatory functions. Soluble GARP proteins and the GARP region of CNGB1a are thought to interact with proteins in the rim of outer segment disc membranes, including phosphodiesterase (26) and peripherin (27). Furthermore, soluble GARP2 and the GARP region intrinsic to CNGB1a can act as gating inhibitors of the rod CNG channel (28). For human CNGA3, four protein isoforms can arise from alternative splicing because of different exon combinations within the 5Ј region of the mRNA encoding the N-terminal cytoplasmic domain of the subunit (24,29). Alternative splicing of CNGA3 occurs in multiple species with many species-specific variants, including, in some cases, unique alternative cassette exons (22)(23)(24). Unlike CNGB1, the functional importance of CNGA3 alternative splicing has remained an unresolved issue. For many CNG channel subunits, the cytoplasmic N-terminal region can critically influence channel regulation by second messengers such as Ca 2ϩ /CaM and phosphoinositides (9, 15, 16, 30 -32). Previous studies in our laboratory have revealed two PIP n regulation modules within CNGA3 subunits. One is located within the N-terminal region, overlapping with a functionally silent Ca 2ϩ -CaM binding site, and the other is located within the C-terminal leucine zipper domain distal to the cyclic nucleotide-binding domain (16). Both of these PIP n regulation modules are characterized by clusters of positively charged amino acids. Because alternative splicing produces changes in the N-terminal region of CNGA3, we hypothesized that these CNGA3 variants would exhibit altered channel sensitivity to regulation by PIP n . Here we show that heteromeric channels formed with CNGA3 subunits having an optional e3 exon-encoded sequence exhibit enhanced sensitivity to regulation by phosphoinositides. Surprisingly, this enhancement of regulation depended on the PIP n regulation module located in the C-terminal region of CNGA3, suggesting that changes in interdomain N-C coupling may influence channel gating and regulation.

EXPERIMENTAL PROCEDURES
Channel Subunit cDNA and Mutagenesis-Human CNGA3 was a gift from Dr. K.-W. Yau (Johns Hopkins University, Baltimore, MD). Subunit cDNAs were subcloned into pGEMHE, where they are flanked by the Xenopus ␤-globin gene 5Ј-and 3Ј UTRs for heterologous expression in Xenopus oocytes. Point mutations were generated by overlapping PCR (cassette) mutagenesis. All mutations were confirmed by fluorescencebased DNA sequencing.
Quantitative PCR-Human eyes were obtained from donors through SightLife with the approval of the Institutional Review Board at Washington State University. The eyes were immersed in RNAlater immediately after extraction. The retinas were subsequently isolated for RNA extraction. Retinal tissues, including the macula, were homogenized, and RNA extraction was performed (RNeasy mini kit, Qiagen, Valencia, CA), followed by on-column DNase treatment. The quantity of RNA was determined spectrophotometrically by the A 260 /A 280 ratio (Nanodrop 1000, Thermo-Scientific, Wilmington, DE).
Standard PCR was performed on all four variants of CNGA3 plasmid cDNA and on human retinal cDNA using JumpStart REDTaq ReadyMix (Sigma-Aldrich, St. Louis, MO). The primers used to detect the presence or absence of specific exons were as follows. For inclusion of exon 3, ATC GCC AGG TTT GGA AGA AT (forward) and CAT CCT CAG CAC ACC ACT GT (reverse); for skipping of exon 3, GAG ACC AGA GGA CTG GCT (forward) and GGC GCG ACA GCC TGG CGA (reverse); for inclusion of exon 5, AGG GAC CGG ACT CTT TTC CT (forward) and GGG CCA GGC GCT TCT CCC TC (reverse); for skipping of exon 5, GCA GAC AGA GGG AGA AGG AA (forward) and AAG ACA GGC AGG GCG ATG (reverse); and for constitutive exon 8, GTC CTG TAT GTC TTG GAT GTG C (forward) and CGT CTT GTA ATG CTG CCA CA (reverse). All primers were designed using Primer Express 3.0 (Applied Biosystems). The following conditions were used: initial denaturation, 94°C (2 min); denaturation, 94°C (30 s); annealing, 66°C (30 s); extension, 72°(2 min); and final extension, 72°C (5 min). 21 cycles for plasmid cDNAs and 39 cycles for human retinal cDNA were used.
Expression levels for inclusion of alternative exons in human retinal tissue were quantified by RT quantitative PCR. For each human retinal sample, first-strand cDNA was synthesized with 500 ng of total RNA, 100 ng of random primers (Invitrogen), and 50 units of reverse transcriptase (Superscript II, Invitrogen) according to the protocol of the manufacturer. RT quantitative PCR was performed using an iCycler version 2.039 thermocycler (Bio-Rad) according to the protocol of the manufacturer. The specificity of the reactions was verified via melt curve analysis. Three biological replicates and three technical replicates were performed for each sample. cDNA corresponding to constitutive exon 8 was used as an endogenous control. Before statistical analysis, Ct values were averaged across technical replicates.
Immunohistochemistry-Human retinal sections were purchased from Abcam (Cambridge, MA). Polyclonal goat anti-CNGA3 (1:500) was from Santa Cruz Biotechnology (catalog no. sc-21927, Dallas, TX). Rabbit polyclonal antibody specific to the exon 3-encoded sequence (1:300) was generated using the following peptide antigen: RFGRIQKKSQPEKV-[C] (Gen-Script, Piscataway, NJ). Fluorescein isothiocyanate-coupled peanut agglutinin (1:1000), which labels cone outer segments (33), was purchased from Sigma-Aldrich. Four slides, each containing two human retinal sections, were stained. The images were captured using a Leica TCS SP8 X point-scanning confocal microscope (Leica Microsystems, Buffalo Grove, IL). PIP 2 -and PIP 3 -binding Assays-Purified GST fusion proteins were used for in vitro PIP 2 -and PIP 3 -binding assays in buffer containing 10 mM HEPES (pH 7.4), 50 mM NaCl, 2 mM EDTA, 0.25% v/v Triton X-100, and 0.25% v/v Nonidet P-40. PIP 2 -and PIP 3 -agarose beads were from Echelon Biosciences (Salt Lake City, UT). Briefly, 40 l of a 50% slurry of PIP n or control agarose beads and purified protein (2 g/ml) was incubated in 0.5 ml of binding buffer for 2 h at 4°C with rocking. Beads were pelleted gently and washed five times with excess binding buffer. PIP n -interacting proteins were eluted with 1ϫ NuPAGE sample buffer (Invitrogen). Protein samples were then separated under reducing conditions in 4 -12% BisTris gels and blotted onto nitrocellulose by using the NuPAGE transfer buffer system (Invitrogen). Blotted proteins were detected overnight using B-14 anti-GST monoclonal antibody (Santa Cruz Biotechnology) at a dilution of 1:2000 in 1% milk, 500 mM NaCl, 20 mM Tris-HCl (pH 7.5), and 0.05% Tween 20. Horseradish peroxide-conjugated anti-mouse IgG secondary antibodies were used subsequently for 1 h. Chemiluminescent detection was performed using the Super-Signal West Dura extended duration detection kit (Pierce).
Electrophysiology-For expression in Xenopus laevis oocytes, channel subunit cDNAs were linearized by digestion with NheI or SphI, and cRNA was synthesized in vitro using an upstream T-7 promoter (mMessage mMachine, Ambion). Stage IV oocytes were isolated as described previously (34). The animal use protocols were consistent with the recommendations of the American Veterinary Medical Association and were approved by the Institutional Animal Care and Use Committee of Washington State University. Approximately 20 ng of channel RNA was microinjected per oocyte. To efficiently generate heteromeric channels, an excess of CNGB3 relative to the CNGA3 subunit RNA was injected (35). 2-7 days after microinjection of in vitro-transcribed cRNA into oocytes, patch clamp experiments were performed using an Axopatch 200B amplifier (Axon Instruments, Foster City, CA) and pulse acquisition software (HEKA, Bellmore, NY) in the inside-out configuration. Initial pipette resistances were 0.30 -0.80 M⍀. Intracellular (bath) and extracellular (pipette) solutions contained 130 mM NaCl, 0.2 mM EDTA, and 3 mM HEPES (pH 7.2). Cyclic nucleotides were added to the intracellular solution as needed. Intracellular solutions were changed using an RSC-160 rapid solution changer (Molecular Kinetics, Indianapolis, IN). Currents in the absence of cyclic nucleotide were subtracted. For channel activation by cGMP, dose-response data were fitted using the Hill equation, where I is the current amplitude at ϩ80 mV, I max is the maximum current elicited by a saturating concentration of ligand, [cGMP] is the cGMP concentration, K 1/2 is the concentration of cGMP producing half-maximal current, and h is the Hill coefficient. To confirm the formation of heteromeric CNG channels, we tested the expressed channels for block by 25 M L-cis-diltiazem (Sigma-Aldrich) in the intracellular solution in the presence of 1 mM cGMP at ϩ80 mV. For Ca 2ϩ -CaM experiments, 250 nM CaM was used in intracellular solution containing 130 mM NaCl, 3 mM HEPES (pH 7.2), 2 mM nitrilotriacetic acid, and 704 mM CaCl 2 for a final buffered free [Ca 2ϩ ] of 50 M. Niflumic acid (500 M) was added to the pipette solution for Ca 2ϩ -CaM experiments. Recordings were made at 20 -22°C. PIP n solutions were prepared with FVPP (a phosphatase inhibitor mixture) (36) containing 5 mM sodium fluoride, 0.1 mM sodium orthovanadate, and 10 mM sodium pyrophosphate, as described previously (37). PIP 3 and PIP 2 analogues were phosphatidylinositol 3,4,5-triphosphate, dipalmitoyl, sodium salt and phosphatidylinositol 4,5-biphosphate, dipalmitoyl, and sodium salt (Matreya LLC, Pleasant Gap, PA). PIP n solutions were applied to the cytoplasmic face of membrane patches approximately 5-10 min or more after patch excision. Data parameters were expressed as mean Ϯ S.E. of n experiments unless indicated otherwise. Statistical significance was determined by using Student's t test or Mann-Whitney U test. p Ͻ 0.05 was considered statistically significant.

Alternative Splicing of CNGA3 Generates Channel Isoforms
with Similar Fundamental Properties-Four variants of human CNGA3 transcripts are generated by alternative splicing involving binary exclusion or inclusion of optional cassette exons 3 and 5 (24,29). To determine the expression levels for these variants in the human retina, we designed variant-specific primers to detect the presence or absence of the optional exons (Fig. 1A). To confirm their specificity, we first tested these primers on plasmid cDNA representing the CNGA3 splice variants. RT quantitative PCR was used to assess the expression level of the variant CNGA3 transcripts in human retina. We found that ϳ25% of retinal CNGA3 transcripts contained exon 3, whereas more than 80% contained exon 5 ( Fig. 1, B and C). Furthermore, we performed immunohistochemical studies using human retinal sections and a specific antibody targeting the exon 3-encoded sequence ( Fig. 1, D--G). Fluorescein-coupled peanut agglutinin was used to label cone photoreceptor outer segments, the predominant location for mature cone CNG channels. We found that exon 3-containing isoforms were present in the retina and located within the outer segment of cone photoreceptors, similar to the distribution of total CNGA3 subunits detected using an antibody directed against an epitope within the C-terminal region (encoded by a constitutive exon). The observed overlap between the general CNGA3 signal and the e3-specific signal was consistent with the presence of e3 variant CNGA3 subunits in most cone photoreceptor subtypes. It is noteworthy that exon 3 appears to be unique to humans, even compared with other primates, where the related genomic sequence is not an open reading frame and the 5Ј (donor) splice site does not match the consensus sequence ( Fig. 2A).
Next, we expressed the four CNGA3 isoforms individually with CNGB3 subunits in Xenopus oocytes for functional analysis of the splice variants. We designated the CNGA3 isoform that lacked exon 3 but contained exon 5 as the reference CNGA3 isoform because this isoform was the first cloned and characterized (38). It is reported to be the most abundant isoform in the human retina and largely absent from other tissues (29). We found that cGMP-or cAMP-dependent activation of heteromeric CNGA3 ϩ CNGB3 channels was not altered significantly by alternative splicing (Fig. 2, B and C). We evaluated channel activation by cGMP by measuring the concentration of cGMP that produced the half-maximal current (K 1/2,cGMP ), whereas we used the ratio of maximal currents elicited by saturating concentrations of cAMP and cGMP (I max cAMP/ cGMP) to assess the efficacy of the partial agonist cAMP. For cone CNG channels, cGMP is nearly a full agonist, whereas cAMP is a partial agonist. For heteromeric channels containing the reference CNGA3, the K 1/2,cGMP was 14.9 Ϯ 0.8 M (n ϭ 12), and the I max cAMP/cGMP was 0.26 Ϯ 0.01 (n ϭ 14). We found that neither the apparent cGMP affinity (K 1/2,cGMP ) nor the relative cAMP efficacy (I max cAMP/cGMP) of heteromeric CNG channels were altered significantly by inclusion or exclusion of the optional exons (Fig. 2, B and C, p Ͼ 0.05 compared with channels containing the reference CNGA3). Further-more, because there is a functionally silent Ca 2ϩ -CaM binding site within the N-terminal region of CNGA3 (located within the sequence encoded by constitutive exon 4) (30), we examined the effect of alternative splicing on the Ca 2ϩ -CaM sensitivity of heteromeric and homomeric channels. No differences in the Ca 2ϩ -CaM sensitivity of heteromeric channels were observed with the CNGA3 splice variants (Fig. 2D). Application of 250 nM CaM in the presence of 50 M free Ca 2ϩ moderately attenuated the current induced by 10 M cGMP to a similar extent for channels containing all four CNGA3 isoforms (ϳ25% suppression). In addition, similar to reference homomeric CNGA3 channels (30), CNGA3e3only channels remained insensitive to CaM. The current induced by 10 M cGMP was not reduced significantly by CaM (the I CaM /I ratio was 0.98 Ϯ 0.03, n ϭ 6). Together, these results indicate that alternative splicing of human CNGA3 does not modify the fundamental gating properties of the channel nor sensitivity to regulation by calmodulin. Furthermore, these results show that CNGA3 alternative splicing does not alter heteromeric channel assembly.
Alternative Splicing Tunes the Phosphoinositide Sensitivity of Heteromeric Cone CNG Channels-We tested the hypothesis that optional exon-encoded sequences might alter the phosphoinositide sensitivity of heteromeric CNGA3 ϩ CNGB3 channels by applying diC8-PIP 2 or diC8-PIP 3 to the cytoplasmic face of the inside-out patches. Phosphoinositides were applied to patches approximately 5-10 min (or more) after patch excision. We have shown previously that application of these diC8-PIP n analogs produced similar effects on cone CNG channel gating as natural phosphoinositides (16). Also, manipulations of endogenous phosphoinositides have been shown to alter the ligand sensitivity of expressed cone CNG channels (37). We measured currents elicited by a subsaturating concentration of cGMP (5 M), which approximates the intracellular cGMP concentration within vertebrate photoreceptors in the dark (17,39). 10 M PIP 2 suppressed the currents to a greater extent for channels formed with the optional exon 3 variant of CNGA3 (A3e3ϩB3) compared with channels comprised of the reference CNGA3 with CNGB3 (A3ϩB3) (Fig. 3A). For reference A3ϩB3 channels, the cGMP dose-response curve was shifted about 1.5-fold to higher concentrations of cGMP after PIP n application (Fig. 3, B and C). K 1/2,cGMP increased from 14.3 Ϯ 0.8 M (n ϭ 14) to 20.6 Ϯ 1.6 M (n ϭ 8) after 1 M diC8-PIP 3 (PIP 3 ) or to 22.1 Ϯ 1.7 M (n ϭ 6) after 10 M diC8-PIP 2 (PIP 2 ). Variant A3e3ϩB3 channels exhibited enhanced sensitivity to PIP n (Fig. 3, B and C). There was an ϳ2-fold greater shift in the cGMP dose-response relationship (p Ͻ 0.001 compared with reference channels). For A3e3ϩB3 channels, K 1/2,cGMP was increased from 14.5 Ϯ 0.9 M (n ϭ 16) to 27.1 Ϯ 1.3 M (n ϭ 10) after 1 M PIP 3 or to 32.7 Ϯ 2.3 M (n ϭ 6) after 10 M PIP 2 . A similar enhancement of sensitivity to regulation by natural PIP 2 was observed for A3e3ϩB3 channels. K 1/2,cGMP was increased from 12.5 Ϯ 0.9 M to 29.2 Ϯ 2.1 M (n ϭ 4) after application of 10 M natural PIP 2 . Several minutes after patch excision, the contribution of endogenous PIP n to channel regulation is thought to be negligible under control conditions (37). Application of 25 g/ml poly-lysine to control patches expressing A3ϩB3 channels (5-10 min after patch Representative PCR products generated using primers sets are shown on the left for plasmid cDNA representing the four CNGA3 splice variants (NM_001298, NM_001079878, AK_131300, and XM_006712243) (24,29) and for cDNA generated from human retinal mRNA. Control, no template negative control; CNBD, cyclic nucleotide binding domain. B and C, percentage expression levels in human retinal tissue for alternative exons 3 and 5, respectively, were determined by quantitative real-time RT-PCR (n ϭ 3 biological replicates with 2-3 technical replicates each). The primers used are shown in A. D-G, representative images of stained human retinal sections demonstrating the presence of CNGA3-e3 isoforms in cone outer segments. D, expression of total CNGA3 protein (green), detected using a general anti-CNGA3 antibody, with DAPI counterstaining for nuclei (blue). E, expression of the CNGA3 e3 isoform, detected using an antibody specific to the e3-encoded protein sequence (red). F, staining using the cone-specific marker peanut agglutinin (PNA), which labels cone outer segments (purple). G, merged image (green and red). Scale bar ϭ 25 m (D-G). excision) did not increase apparent cGMP affinity. K 1/2,cGMP was 12.7 Ϯ 0.7 M before and 15.7 Ϯ 1.2 M after poly-lysine (p ϭ 0.074. n ϭ 4). In contrast to the exon 3 variant, the presence or absence of exon 5-encoded sequence did not influence PIP n regulation of cone CNG channels (Fig. 3D). Fig. 3E shows the concentration-response relationship for the percent increase in K 1/2,cGMP induced by different concentrations of PIP 2 and PIP 3 for heteromeric A3e3ϩB3 channels compared with A3ϩB3 channels. These results indicate that incorporation of the exon 3-encoded sequence significantly increased both the affinity and efficacy of PIP 2 and PIP 3 for inhibition of cGMP-dependent gating of A3ϩB3 channels.
The Exon 3-encoded Sequence Does Not Enhance PIP n Regulation of Homomeric CNGA3-only Channels-Homomeric CNGA3 channels are not sensitive to PIP n -dependent changes in apparent cGMP affinity but, instead, show an ϳ2.5-fold increase in relative cAMP efficacy (I max cAMP/cGMP) (16).
Truncation of the C-terminal region (613X) distal to the cyclic nucleotide-binding domain, mutations disrupting the structure of the C-terminal leucine zipper domain, or charge neutralizations within the C-terminal PIP n regulation module all eliminated the increase in I max cAMP/cGMP produced by PIP n for CNGA3-only channels (16,40). We expressed the CNGA3e3 subunit without CNGB3 and found that the increase in I max cAMP/cGMP after PIP n application for the homomeric channels was similar to the reference CNGA3-only channels. The fold increase in I max cAMP/cGMP was 2.6 Ϯ 0.3 (n ϭ 4) after 1 M PIP 3 or 2.2 Ϯ 0.2 (n ϭ 4) after 10 M PIP 2 . In addition, homomeric CNGA3e3 channels remained insensitive to PIP n regarding apparent cGMP affinity (Fig. 4A). We have shown previously that truncation of CNGA3 at amino acid 613 (CNGA3 613X) unmasks the N-terminal component of PIP n regulation for homomeric channels, producing a more than 3-fold increase in K 1/2,cGMP after PIP n application (16). How-FIGURE 2. Basic gating properties and Ca 2؉ -calmodulin sensitivity of heteromeric CNGA3؉CNGB3 channels were not altered by alternative splicing of CNGA3. A, alignment of genomic sequences for parts of human optional exon 3 and adjacent intronic sequence compared with corresponding genomic sequences of three other primates. Only human exon 3 presents an open reading frame and an intact consensus 5Ј (donor) splice site. Only scant homology was apparent between human exon 3 and genomic sequences from mouse, rat, dog, cow, pig, opossum, platypus, chicken, zebrafish, Xenopus, and fly. Some homology was present with horse, dog, and giant panda sequences, but splice junctions were not intact (data not shown). B, initial apparent affinity for cGMP (K 1/2,cGMP ) was not significantly different for the four human CNGA3 variants coexpressed with CNGB3 in Xenopus oocytes. n.s., not significantly different. C, relative cAMP efficacy compared with cGMP was not affected by alternative splicing of CNGA3 transcripts. I max was measured in 10 mM cAMP and 1 mM cGMP. D, Ca 2ϩ -CaM inhibition of CNGA3ϩCNGB3 channels was unaffected by alternative splicing of CNGA3 transcripts. The value below each symbol indicates the number of experiments. Data are mean Ϯ S.E. ever, incorporation of the region encoded by exon 3 did not significantly change the PIP n -induced increase in K 1/2,cGMP for 613X channels (Fig. 4, A and B). K 1/2,cGMP was increased from 16.3 Ϯ 1.0 M (n ϭ 11) to 64.7 Ϯ 3.6 M (n ϭ 6) after 1 M PIP 3 or to 53.5 Ϯ 5.0 M (n ϭ 5) after 10 M PIP 2 for A3e3-613X channels. These results suggest that the potentiation of PIP n sensitivity mediated by the exon 3-encoded sequence in CNGA3 requires assembly with CNGB3 subunits and/or structural elements within the distal C-terminal domain of CNGA3.
Next, we sought to determine whether the exon 3-encoded sequence was sufficient to support PIP n regulation of homomeric CNGA3 channels. To test this, we neutralized the positively charged arginines within the sequence encoded by constitutive exon 4 (R72A, R75A, R81A, R82A, and R86A in reference CNGA3; N5R-A). These changes have been shown previously to abolish the PIP n sensitivity of CNGA3-613X channels (16). We found that the PIP n -induced increase in K 1/2,cGMP of CNGA3e3-613X channels was similarly eliminated by N5R-A mutations (Fig. 4 C). This result indicated that the region encoded by alternative exon 3 alone was not sufficient to confer phosphoinositide sensitivity to homomeric CNGA3 channels. Furthermore, we carried out in vitro PIP n -binding assays using PIP 3 -and PIP 2 -linked agarose beads together with GST-tagged N-terminal fragments of CNGA3 with or without the region encoded by exon 3. We found that the e3 isoform did not alter in vitro binding of the N-terminal fragment to PIP 2 or PIP 3 (Fig. 4D).
Potentiation of PIP n Sensitivity by the e3 Variant Depends on the C-terminal PIP n Regulation Module of CNGA3-PIP n regulation of heteromeric CNGA3 ϩCNGB3 channels is supported by two regulatory modules (N and C) (Fig. 5A) having additive effects on K 1/2,cGMP (16). We silenced both N-(N5R-A) and C-terminal (R643Q and R646Q in reference CNGA3, RR-QQ) PIP n regulation modules in CNGA3e3. These mutations together have been shown previously to abolish the PIP n sensitivity of heteromeric CNGA3ϩCNGB3 chan-FIGURE 3. CNGA3 optional exon 3 inclusion produces heteromeric channels with enhanced PIP 2 and PIP 3 sensitivity. A, representative currents elicited by 5 M cGMP (an approximately physiological ligand concentration) using the voltage protocol shown on the left, before (black) and after (red) 10 M PIP 2 , for heteromeric channels composed of B3 plus reference A3 or variant A3e3 expressed in Xenopus oocytes. The formation of heteromeric channels was confirmed by sensitivity to block by 25 M L-cis-diltiazem. B, representative cGMP dose-response relationships for currents elicited at ϩ80 mV before (black) and after (red) 10 M diC8-PIP 2 application for activation of heteromeric A3ϩB3 channels, fitted using the Hill equation. Inclusion of exon e3 enhanced the rightward shift of K 1/2,cGMP after PIP 2 . The Hill parameters were as follows. For A3ϩB3 channels, before PIP 2 , K 1/2,cGMP ϭ 12.9 M, h ϭ 2; after PIP 2 , K 1/2,cGMP ϭ 18.3 M, h ϭ 2.1. For A3e3ϩB3 channels, before PIP 2 , K 1/2 ϭ 11.9 M, h ϭ 1.8; after PIP 2 , K 1/2 ϭ 28.0 M, h ϭ 1.8. C, representative cGMP dose-response relationships for currents elicited at ϩ80 mV before (black) and after (red) 1 M diC8-PIP 3 application for activation of heteromeric A3ϩB3 channels, fitted using the Hill equation. The Hill parameters were as follows. For A3ϩB3 channels, before PIP 3 , K 1/2,cGMP ϭ 15.8 M, h ϭ 2; after PIP 3 , K 1/2,cGMP ϭ 22.9 M, h ϭ 2. For A3e3ϩB3 channels, before PIP 3 , K 1/2 ϭ 12.5 M, h ϭ 1.9; after PIP 3 , K 1/2 ϭ 29.6 M, h ϭ 1.8. D, summary of the enhancement of percent increase in K 1/2 cGMP after 10 M diC8-PIP 2 or 1 M diC8-PIP 3 with inclusion of exon e3 but not e5. Error bars denote the mean Ϯ S.E. (n ϭ 6 -16). *, p Ͻ 0.05 compared with channels containing the reference CNGA3 subunit, which includes exon e5. E, DiC8-PIP 2 and diC8-PIP 3 concentration-response relationships for the percentage of increase in K 1/2,cGMP for activation of heteromeric A3ϩB3 or A3e3ϩB3 channels. Continuous curves represent fits using the Hill equation. The Hill parameters are as follows. For PIP 2 ,   (16). We found that channels made up of CNGA3e3-N5R-A, RR-QQ ϩ CNGB3 subunits were no longer sensitive to PIP n (Fig. 5B). The percentage change in K 1/2,cGMP was 6.3 Ϯ 11% (n ϭ 4) after 1 M PIP 3 and Ϫ2.5 Ϯ 5.5% (n ϭ 6) after 10 M PIP 2 . This observation is consistent with the idea that the e3-encoded sequence is not sufficient to confer PIP n sensitivity to heteromeric channels. The next question we addressed was whether the enhancement of PIP n sensitivity by the e3 isoform depends on the N-, C-, or both N-and C-terminal regulatory modules. Thus, we ablated either the N-terminal or C-terminal PIP n regulation sites in CNGA3 combined with inclusion or exclusion of the region encoded by exon 3. We found that the e3 isoform still enhanced the PIP n -induced increase in K 1/2,cGMP for CNGA3ϩCNGB3 channels when the C-terminal PIP n -regulation module was intact but the N-terminal site was silenced (N5R-A) (Fig. 5C, p Ͻ 0.05 for both PIP 2 and PIP 3 ). In contrast, the e3 variant was not able to potentiate the PIP n -induced increase in K 1/2,cGMP for CNGA3ϩCNGB3 channels when the N-terminal PIP n regulation module was intact but the C-terminal PIP n regulation module was silenced (RR-QQ) (Fig. 5D, p ϭ 0.941 for PIP 2 and p ϭ 0.610 for PIP 3 ). For all mutant CNGA3 subunits described here, the formation of functional heteromeric channels was confirmed by sensitivity to block by L-cisdiltiazem and by enhanced relative cAMP efficacy compared with homomeric channels. Therefore, these data indicate that, even though the exon 3-encoded sequence is in close proximity to the N-terminal PIP n regulation module, the CNGA3e3 variant enhances phosphoinositide sensitivity by influencing the C-terminal PIP n regulation module.
Control of PIP n Sensitivity by Alternative Splicing of Canine CNGA3-We discovered that transcripts encoding canine CNGA3 subunits also are subject to alternative splicing. There are three alternative exons encoding a region within the N-terminal cytoplasmic domain of canine CNGA3: exons 3, 4, and 5. Canine exons 3 and 4 are orthologous to human exons 4 and 5, respectively (Fig. 6A). Canine exon 5 has no equivalent exon in humans but is similar to exon 5 of chicken CNGA3 (22). RT-PCR using canine retinal mRNA revealed that three splice variants of canine CNGA3 were generated: the reference canine CNGA3 with all optional exons included (cCNGA3), a variant lacking exons 3 and 4 (cCNGA3 ⌬e3,4), and a variant lacking exons 4 and 5 (cCNGA3⌬e4,5). We tested the PIP n sensitivity of all three canine CNGA3 isoforms after expression with human CNGB3 (Fig. 6B). We found that channels containing cCNGA3 having a sequence encoded by all the optional exons exhibited a considerable PIP n -induced increase in K 1/2,cGMP . For cCNGA3ϩCNGB3 channels, K 1/2,cGMP increased from 13.9 Ϯ 0.7 M (n ϭ 8) to 27.6 Ϯ 4.0 M (n ϭ 4) after 1 M PIP 3 or to 22.9 Ϯ 1.8 M (n ϭ 4) after 10 M PIP 2 . Interestingly, the PIP n -induced increase in K 1/2,cGMP was eliminated for cCNGA3⌬e3,4 ϩ CNGB3 channels (p Ͻ 0.001 for both PIP 2 and PIP 3 compared with the reference channel). For A, representative cGMP dose-response relationships before and after diC8-PIP 2 (red) for activation of A3e3 wild-type or A3e3-613X channels. Red and green boxes indicate silent and active PIP n interaction sites, respectively. Blue box, region encoded by alternative exon 3. cG, cGMP. B, summary illustrating the impact of inclusion of exon 3 on the percentage of increase in K 1/2,cGMP after diC8-PIP 2 or diC8-PIP 3 for homomeric wild-type A3 and A3-613X channels. Exon 3 did not produce a statistically significant effect on PIP n sensitivity for homomeric A3 channels (n ϭ 4 -6). n.s., not significant. C, representative dose-response relationships for activation of A3e3, N5R-A, and 613X channels by cGMP. A red box indicates silencing of the N-terminal component for PIP n regulation of cone CNG channels by mutation (N5R-A: R72A, R75A, R81A, R82A, and R86A). D, in vitro pulldown assay using PIP 2 , PIP 3 , or control agarose beads combined with GST fusion proteins representing the N-terminal regions of CNGA3 either with or without the region encoded by exon 3.
cCNGA3⌬e4,5 ϩ CNGB3 channels, the PIP n -induced increase in K 1/2,cGMP was attenuated (p Ͻ 0.05 for PIP 3 ). Presumably, the region encoded by exon 3 in canine CNGA3, similar to exon 4 in human CNGA3, contains a PIP n interaction site. Skipping this exon abolished the PIP n sensitivity of the channel. Together, these results indicate that alternative splicing of both canine and human CNGA3, with divergent splice patterns, similarly tunes the PIP n sensitivity of heteromeric cone CNG channels.

DISCUSSION
For genes encoding CNG channel subunits, three are known to exhibit alternatively spliced transcripts: CNGB1, CNGA3, and CNGA1. The functional and physiological consequences of alternative splicing of CNGB1 have been elucidated previously (26 -28, 41-43). In this work, we found that alternative splicing of CNGA3 tunes the phosphoinositide sensitivity of cone CNG channels. Inclusion of the human-specific alternative exon 3 generated CNGA3 isoforms that significantly potentiated the PIP 2 or PIP 3 sensitivity of CNGA3ϩCNGB3 channels. The region encoded by exon 3 alone was not sufficient to support PIP n sensitivity in the absence of the other known PIP n regulation modules. The effect of the exon 3-encoded sequence on PIP n sensitivity appears to be mediated by an allosteric mechanism whereby the modified N-terminal cytoplasmic domain enhances the change in K 1/2,cGMP mediated by the C-terminal PIP n regulation module of CNGA3.
Previous studies in our lab have demonstrated the importance of intersubunit N-C interactions for control of the PIP n sensitivity of cone CNG channels (16,40). This idea is consist-FIGURE 5. Enhancement of PIP n sensitivity by the CNGA3 e3 isoform depends on the C-terminal structural component of PIP n regulation. A, schematic illustrating the PIP n regulation modules within the N-and C-terminal regions of CNGA3, characterized by positively charged arginines (16). B, representative dose-response relationship for activation of heteromeric CNGA3 N5R-A, RR-QQ ϩ B3 channels by cGMP. In this scenario, both the N-terminal (N5R-A) and C-terminal (RR-QQ: R643Q and R646Q) structural components underlying PIP n regulation were silenced by mutation. cG, cGMP. C, summary showing that the enhancement of PIP 2 (10 M) or PIP 3 (1 M) sensitivity by the e3 isoform was not prevented by mutation of the N-terminal PIP n regulation module (N5R-A). The green and red boxes indicate that the designated PIP n regulation module was active or silenced, respectively. Error bars denote the mean Ϯ S.E. (n ϭ 4 -7). *, p Ͻ 0.05 compared with reference heteromeric channels. D, summary showing that mutation of residues critical for the C-terminal PIP n regulation module (RR-QQ) eliminated the enhancement of PIP 2 or PIP 3 sensitivity in K 1/2,cGMP conferred by the e3 isoform. No significant differences were observed. FIGURE 6. Alternative splicing of canine CNGA3 transcripts also tunes PIP n sensitivity. A, CNGA3 splicing pattern for constitutive (green) and alternative cassette exons across representative species. Exon (box) lengths, but not intron (line) lengths, are shown to scale. For simplicity, removal of introns between constitutive exons is not shown. B, summary of PIP 2 and PIP 3 effects on K 1/2,cGMP for heteromeric CNGA3ϩCNGB3 channels containing canine CNGA3 isoforms generated by alternative splicing. All subunit combinations exhibited sensitivity to blocking by L-cis-diltiazem, confirming the formation of functional heteromeric channels. Data are mean Ϯ S.E. of four independent experiments. **, p Ͻ 0.01; *, p Ͻ 0.05 compared with channels having all regions encoded by e3, e4, and e5. ent with the results described here for enhancement of PIP n sensitivity by exon 3 inclusion, which inserts 55 additional amino acids into the N-terminal domain but depends on the C-terminal PIP n regulation module rather than the adjacent N-terminal PIP n regulation site. Although the exact mechanism underlying potentiation of PIP n sensitivity in the e3 isoform is not fully understood, we hypothesize that, with the assistance of CNGB3, the conformational difference in the CNGA3 N-terminal region caused by inclusion or exclusion of the region encoded by exon 3 is coupled to the C-terminal region of the adjacent CNGA3 subunit, altering the ability of the C-terminal PIP n regulation module to respond to membrane-bound phosphoinositides (Fig. 7). In other words, incorporation of the exon 3-encoded sequence modulates the N-C interactions of CNGA3 subunits, thus potentiating the PIP n sensitivity of cone CNG channels. The exon 3-encoded sequence may also influence CNGA3-CNGB3 interactions because these are also thought to modulate PIP n sensitivity and the competency of the C-terminal PIP n regulation module in CNGA3 to confer a change in K 1/2,cGMP (16). Interestingly, in addition to well characterized N-C-terminal interactions for CNG channels (32, 44 -47, 28), a recent crystal structure of the EAG domain-cyclic nucleotide-binding homology domain (CNBHD) complex of homologous mEAG1 (KCNH) channels revealed that the post-CNBHD region (equivalent to the region adjacent to the C-terminal PIP n regulation module in CNGA3) is interacting directly with the N-terminal EAG domain (48).
Moreover, we have shown previously that truncations and point mutations in this post-cyclic nucleotide-binding domain region can influence the N-terminal PIP n regulation module of CNGA3 (16,40). We propose that the coupling between N-and C-terminal regions of CNGA3 (and presumably CNGB3) is an essential element of cone CNG channel regulation. However, more sophisticated structural approaches may be required to elucidate the precise interdomain contacts and detailed conformational changes associated with the e3 variant.
We found that alternative splicing of canine CNGA3 also regulates the phosphoinositide sensitivity of the resulting channels. Together, these findings suggest that the functional significance of CNGA3 alternative splicing might be conserved across species, despite the divergence in exon combinations for the orthologous genes. Considering that exon 3-containing transcripts were expressed in 23 human tissues, including heart, kidney, and brain (29), enhanced phosphoinositide sensitivity by CNGA3e3 may have widespread physiological significance. However, how tuning of PIP n regulation of CNG channels by alternative splicing is related to function in cone photoreceptors and other cell types is not understood. The functional importance of optional exon 5 remains elusive. On the basis of previous work showing that the human CNGA3 e5-containing transcripts were expressed primarily in the retina (29), exon 5 might contain features favoring its inclusion here, such as exonic splicing enhancers recognized by neuronor retina-specific splicing factors expressed in a tissue-specific manner. The splicing factors and cis-regulatory features driving e3 inclusion also are unknown. The sequence upstream of exon 3 presents a weak poly-pyrimidine tract that likely necessitates splicing enhancer sequences that facilitate e3 inclusion.
The functional CNGA3e3 variant seems to represent a recent innovation along the human evolutionary pathway. Comparative analysis of alternative splicing patterns in humans and chimpanzees reveal that, for at least 4% of genes, one or more cassette alternative exons display pronounced splicing differences between the species (49). Human-specific patterns of alternative splicing have been found to be particularly widespread within the brain (50). Also, it has been estimated that humans and chimpanzees differ by at least 6% in their complement of genes (51). The trichromatic visual system is one of the dramatic specializations present in the human lineage compared with other mammals. It is not entirely surprising that unique patterns of splicing and/or gene regulation may be related to these specializations (52, 53). The model illustrates how incorporation of the e3-encoded sequence within the N-terminal region of CNGA3 may enhance the C-terminal PIP n regulation module by altering intersubunit interactions. The green and yellow boxes indicate basal and enhanced PIP n regulation modules, respectively. We propose that the interface mediating N-C coupling for CNGA3 is altered in the e3 isoform. For clarity, the CNGB3 subunits are not shown, but CNGA3-CNGB3 interactions may also be altered for channels containing the e3 isoform. CNBD, cyclic nucleotide binding domain.