The β Subunit of Voltage-gated Ca2+ Channels Interacts with and Regulates the Activity of a Novel Isoform of Pax6*

Ca2+ channel β subunits (Cavβs) are essential for regulating the surface expression and gating of high voltage-activated Ca2+ channels through their interaction with Ca2+ channel α1 subunits. In efforts to uncover new interacting partners and new functions for Cavβ, we identified a new splicing isoform of Pax6, a transcription factor crucial for the development of the eye, nose, brain, and pancreas. Pax6 contains two DNA binding domains (paired domain and homeodomain), a glycine-rich linker connecting these two domains and a C-terminal proline-, serine-, and threonine-rich transactivation domain. The protein sequence and function of Pax6 are highly conserved from invertebrate to human. The newly isolated isoform, named Pax6(S), retains the paired domain, linker, and homeodomain of Pax6, but its C terminus is composed of a truncated classic proline, serine, and threonine domain and a unique S tail. Pax6(S) shows a similar level of transcriptional activity in vitro as does Pax6, but only in primates is the protein sequence highly conserved. Its spatial-temporal expression profiles are also different from those of Pax6. These divergences suggest a noncanonical role of Pax6(S) during development. The interaction between Pax6(S) and Cavβ is mainly endowed by the S tail. Co-expression of Pax6(S) with a Ca2+ channel complex containing the β3 subunit in Xenopus oocytes does not affect channel properties. Conversely, however, β3 is able to suppress the transcriptional activity of Pax6(S). Furthermore, in the presence of Pax6(S), β3 is translocated from the cytoplasm to the nucleus. These results suggest that full-length Cavβ may act directly as a transcription regulator independent of its role in regulating Ca2+ channel activity.

complexes, which include L-, N-, P/Q-, and R-type Ca 2ϩ channels. It plays an essential role in chaperoning the channel complex to the plasma membrane and normalizing its gating properties (1)(2)(3)(4). Crystal structures of Ca v ␤ in complex with its high affinity binding site in the principal poreforming ␣ 1 subunit (Ca v ␣ 1 ) show that much of the exposed surface of Ca v ␤ is unoccupied and is available to engage in interactions with other regions of Ca v ␣ 1 or with other proteins (5)(6)(7). Indeed, an increasing number of proteins has been shown to directly interact with Ca v ␤, including the Rem/Rad/Gem/Kir (RGK) family of small monomeric GTPases (8,9), RIM1 (10), ryanodine receptors (11), Ahnak (12,13), bestrophin-1 (14), and dynamin (15). Many of these proteins have been reported to regulate the activity of HVA Ca 2ϩ channels. To search for other potential Ca v ␤-interacting proteins, we carried out yeast two-hybrid screens using the ␤ 3 subunit as bait. Among the candidate target proteins we isolated, one was related to Pax6.
Pax6 is a transcription factor that belongs to the paired box (Pax) family (16 -25). It is widely expressed in the eye, nose, pancreas, and the central nervous system in both embryonic and adult mammals, and it plays important roles in regulating the development of these tissues and organs (18, 20, 26 -31). The protein sequence of Pax6 is highly conserved throughout vertebrates, lower chordates, and invertebrates (24). The function of Pax6 is also highly conserved, as suggested by the induction of ectopic eye structures after the overexpression of Drosophila or mouse Pax6 genes in Drosophila or Xenopus laevis embryos (32)(33)(34)(35).
The human Pax6 gene is located on chromosome 11p13 and occupies 14 exons (exons 1-13 plus exon 5␣ between exons 5 and 6) in a 22-kb genomic region (36). There are at least three Pax6 isoforms produced by alternative splicing (24). The canonical Pax6 is generated from a transcript composed of exons 1-13 (Fig. 1A). It contains a paired domain (PD), a homeodomain (HD), a glycine-rich linker connecting the above two domains, and a C-terminal proline, serine, and threonine (PST)-rich domain (Fig. 1A). The second Pax6 isoform contains 14 extra amino acids encoded by exon 5␣ in the PD and is named Pax6(5␣). Protein translation of the above two isoforms begins in exon 4 and terminates in exon 13 (Fig. 1A). The third isoform is generated by translational initiation in exon 8. It does not contain the PD and, therefore, is named paired-less Pax6 (Pax6(⌬PD)).
The PD and HD are two domains where Pax6 interacts with its DNA targets. The C-terminal PST domain plays a key role in regulating Pax6 transcriptional activity but does not bind DNA directly. Missense mutations and partial or complete truncation of the PST domain decrease the transcriptional activity of Pax6 (37,38). Fusion of the Pax6 PST with the transcription factor GAL4 increases the activity of GAL4 (37,39,40), suggesting that the transactivity of the PST domain can be independent of the PD and HD. The PST domain encompasses 152 amino acids encoded by exons 10 -13 (Fig. 1A). Studies have revealed that these four exons synergistically stimulate transcriptional activation and that the transactivation potential is not localized but spread throughout the PST domain (37). It has been suggested that the transactivity of the PST domain stems from its interaction with other regulatory proteins, which enhances the assembly of the transcriptional preinitiation complex (41). Recent studies demonstrate that the high proportions of serine and threonine residues in the PST domain allow phosphoryl and dephosphoryl modulation (42)(43)(44), which may fine-tune the protein-protein interactions.
In this study we identified a novel splicing isoform of Pax6 named Pax6(S). It contains the canonical PD and HD, but it has a different C terminus composed of the N-terminal half of the canonical PST and a unique S tail encoded by the intron between exons 11 and 12. In contrast to Pax6, Pax6(S) is completely conserved only in human and chimpanzee, and it seems to be expressed only at the early stages of development, suggesting a yet-to-be-defined and perhaps noncanonical function during development. Pax6(S) retains transcriptional activity, but its C terminus shows less transactivity compared with the canonical PST domain. In addition, we found that Pax6(S) interacted with a full-length Ca v ␤ through its S tail. This interaction did not alter Ca 2ϩ channel properties, but it decreased Pax6(S) activity in vitro and resulted in the translocation of Ca v ␤ from the cytoplasm to the nucleus. Our results suggest a novel function of full-length Ca v ␤ as a suppressor of Pax6(S).
A human adult brain cDNA library (Clontech) was screened with the pGBKT7-␤ 3 _ core construct. The library was constructed in the pGADT7-rec vector. All procedures were carried out according to the manufacturer's instructions (Clontech). Briefly, pGBKT7-␤ 3 _ core was transformed into Y187 and grown in medium lacking tryptophan. The AH109 yeasts pretransformed with the human adult brain cDNA library were then mated to these Y187 cells and grown in a medium lacking adenine, histidine, tryptophan, and leucine. After growing for 10 days, the cells were plated on selective plates lacking adenine, histidine, tryptophan, and leucine with 5-bromo-4chloro-3-indolyl-␣-D-galactopyranoside (X-␣-gal) and incubated at 30°C for 2 weeks. The colonies were then picked, and the plasmids were extracted and transformed into the bacterial strain Escherichia coli DH5␣ for amplification. The plasmids were extracted from DH5␣ and checked with sequencing. The specificity of the interactions was tested by transforming competent AH109 yeast cells with one bait construct (in pGBKT7) and one target construct (in pGADT7) and examining the resulting colonies for activation of the ADE2, HIS3, and MEL1 reporters on selective plates as described above.
Protein Synthesis and GST Pulldown Assay-The S tail of Pax6(S) (Val-345-Asp-401) subcloned into a modified pET26b vector was expressed in BL21(DE3) to obtain the S tail protein. cDNA encoding ␤ 3 was subcloned into a modified pGEX4T-1 vector and expressed in BL21(DE3) bacteria to obtain the GST_␤ 3 protein. The No. 8 protein was synthesized in vitro with the TNT Coupled Transcription/Translation Systems (Promega). GST_␤ 3 was immobilized on glutathione-Sepharose 4B beads (Novagen). The No. 8 protein bound to the immobilized GST_␤ 3 was eluted from the beads with glutathione and detected with Coomassie Blue staining on SDS-PAGE.
5Ј Rapid Amplification of cDNA Ends-The full-length Pax6(S) was obtained by 5Ј-rapid amplification of cDNA end (5Ј RACE) reaction with the 5Ј/3Ј RACE kit (Roche Applied Science). The procedures were carried out according to the manufacturer's instruction. The following primers were used: SP1 (5Ј-GGGCATGAATTAATGAGT-3Ј), SP2 (5Ј-TCTCCG-ACTTGACTGGTC-3Ј), SP3 primer (5Ј-GGGAAAGUCCAC-CACCAGCCGCACTTAC-3Ј), oligo(dT) anchor primer (5Ј-GACCACGCGTATCGATGTCGACTTTTTTTTTTTTTT-TTV-3Ј, V ϭ A, C, or G), and PCR anchor primer (5Ј-GGAG-ACAUGACCACGCGTATCGATGT-3Ј). The product of the second round of PCR was inserted into the lacZ␣ gene in the NEB206A vector according to the instruction of USER TM Friendly Cloning kit (New England Biolabs). The ligation mixture was transformed into DH5␣ cells and plated on a LB plate with isopropyl 1-thio-␤-D-galactopyranoside and X-␣-gal to perform a white/blue selection. The white colonies, which indicate an insertion in the vector, were selected. The plasmids were extracted and sequenced.
BLAST Search of Nucleotide Databases-The nucleotide sequence of No. 8 and Pax6(S) was scanned against the human genomic plus transcript database, human Pax6 mRNA (GenBank TM accession number M93650), human DNA sequence from clone XX-A1280 on chromosome 11(Z83307), and databases of high throughout genomic sequences and whole-genome shotgun reads.
Cell Culture and Transfection-Human embryonic kidney (HEK) 293T cells were maintained in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% fetal bovine serum (Invitrogen). CHO cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and 2 mM glutamine. Both cell cultures were incubated in an incubator at 37°C under 5% CO 2 and were subcultivated every 2-3 days with a ratio of 1:5.
HEK 293T or CHO cells were transfected at a confluence of over 90%. Transfections were performed with plasmid DNA coated with Lipofectamine TM 2000 (Invitrogen) according to the manufacturer's instructions. For HEK 293T cells, which were plated on poly-D-lysine-coated coverslips immersed in 35-mm dishes (BD Biosciences), 10 l of Lipofectamine TM 2000 and 4.0 g of total DNAs were transfected per dish. For CHO cells, which were plated onto 24-well plates (Corning), 2 l of Lipofectamine TM 2000 and 0.8 g of total DNAs were transfected per well.
Cell and Tissue Slide Imaging-24 -40 h after transfection, cells transfected with FLAG-Pax6 or FLAG-Pax6(S) (with or without pEGFP-C3-␤ 3 co-transfected) were fixed with 5% polyformaldehyde and permeabilized with 0.5% Triton X-100. They were then subjected to immunostaining. 1 g/ml anti-FLAG monoclonal antibody (Sigma) or anti-Pax6(S) polyclonal antibody with a dilution of 1:100 was used. After staining, cells were mounted on glass slides with ProLong Gold Antifade reagent (with DAPI) (Invitrogen). Cells transfected with pEGFP-C3-␤ 3 were only fixed with 5% polyformaldehyde and mounted with the same reagent. Paraffin-embedded human retina and brain tissue slides (BioChain) were deparaffinized with xylene and then rehydrated through a series of graded ethanol (100, 95, 70, and 50%). The slides were subsequently immersed in an antigen retrieval solution containing 10 mM Tris base, 1 mM EDTA, and 0.05% Tween 20 (pH 9.0) at 60°C overnight. In the next 2 days the slides were permeabilized and stained with the same procedures described above. Anti-Pax6 polyclonal antibody (Covance) with a dilution of 1:500 was used to stain the endogenous Pax6. All the above cells or tissue slides were imaged with an Olympus FluoView 500 confocal microscope (Olympus) or Nikon 80i upright epifluorescence microscope (Nikon).
Western Blot Assay-A preblotted polyvinylidene difluoride membrane containing ϳ50 g of protein lysates per lane from eight different human tissues (brain, kidney, lung, small intestine, heart, liver, skeletal muscle, and placenta) was obtained from BioChain and then analyzed by hybridization with antibodies according to the instruction of Odyssey Infrared Imaging System (LI-COR Biosciences). Anti-Pax6(S) polyclonal antibody with a dilution of 1:100 was used as the primary antibody. 0.2 g/ml IR Dye 800 (LI-COR Biosciences) was used as the secondary antibody. The membrane was scanned with Odyssey Infrared Imager under 800-nm channels at 169-m resolution (LI-COR Biosciences).
Luciferase Assay-To test the transcriptional activity of Pax6 and Pax6(S), pGL3-OFLuc-CD19-2, FLAG-Pax6 or FLAG-Pax6(S), and pRL-SV40 were transfected into CHO cells with a ratio of 2:14:0.1. To examine the effects of different fragments of the PST N S tail on GAL4 activity, p5XGAL4-E1b-Luc, pCG-GAL4 containing PST N S fragment, or pCG-GAL4 itself and pRL-SV40 were transfected into CHO cells with a ratio of 20:1: 0.4. To examine the effects of ␤ 3 on Pax6(S) activity, pGL3-OFLuc-CD19-2, FLAG-Pax6(S) (or FLAG-Pax6(⌬S) or empty FLAG vector), FLAG-␤ 3 (or empty FLAG vector), and pRL-SV40 were co-transfected into CHO cells with a ratio of 10:40: 15:0.8. pRL-SV40, which constitutively expresses Renilla firefly driven by a SV40 promoter, was used as an internal control. 24 -40 h after transfection the medium was removed, and CHO cells were briefly washed with phosphate-buffered saline solution twice. Luciferase assay was performed at room temperature using the Dual-Luciferase Reporter Assay System (Promega).
The solutions and protocols for two-electrode voltage clamp and patch clamp are as described before (45). Currents were sampled at 10 kHz and filtered at 2.5 kHz. The holding potential for all the following protocols was Ϫ80 mV. Macroscopic currents were evoked by 20-ms depolarizations ranging from Ϫ40 mV to ϩ100 mV in 10-mV increments at a 6-s interval. Tail currents were always recorded by repolarization to Ϫ30 mV regardless of the preceding test pulse. To obtain the currentvoltage relationship, peak currents evoked by depolarizations were plotted against the test potentials. To obtain the activation curves, tail currents were normalized by that after the depolarization to ϩ100 mV and plotted against the test potentials. To examine deactivation, a 10-ms depolarization to ϩ100 mV was applied to fully open the channels followed by repolarizations ranging from Ϫ80 mV to ϩ80 mV in 10-mV increments at a 5s interval to obtain tail currents. Steady-state inactivation was determined by a three-pulse protocol in which a 20-ms normalizing pulse to ϩ30 mV (pulse A) was followed sequentially by a 25-s conditioning pulse (ranging from Ϫ80 mV to ϩ50 mV) and a 20-ms test pulse to ϩ30 mV (pulse B). The interval between each protocol was 2 min. Peak current evoked by pulse B, nor-malized by that evoked by pulse A, was plotted against the conditioning potentials to obtain the voltage dependence of inactivation.
The voltage dependence of activation and inactivation was fitted with the Boltzmann function 1/(1 ϩ exp(Ϫ(V Ϫ V1 ⁄ 2 )/k), where V1 ⁄ 2 and k are the midpoint of activation or inactivation and the slope factor, respectively. To obtain the time constant of deactivation ( deact ), tail currents were fitted with a single exponential function. deact was then plotted against the test potentials to produce the voltage-dependent deactivation curve. Data were represented as the mean Ϯ S.D. (number of observations). Significance was determined using two-tailed Student's t tests.

RESULTS
Cloning of Full-length Pax6(S)-In an yeast two-hybrid screen of a human adult brain cDNA library using the core domain Gly-16 -Gly-366 of Ca 2ϩ channel ␤ 3 subunit as bait, 12 targets were identified (data not shown). A BLAST search of the human genome revealed that the cDNA sequence of No. 8 matched part of Pax6. More careful comparison between No. 8 and Pax6 cDNA sequences showed that the No. 8 cDNA spans the major part of exon 7, entire exons 8 -11, and the 5Ј half of the intron between exons 11 and 12 (renamed exon 11␣, Fig.  1B) of the Pax6 gene. Exon 11␣ contains a stop codon that is in-frame with the preceding exons. Hence, the protein encoded by No. 8 contains the last 12 amino acids of the PD, the linker region, the HD, and the N-terminal half of the PST domain (PST N ) present in the canonical Pax6 plus a novel C-terminal tail encoded by exon 11␣ (Fig. 1B). Compared with the canonical PST, which contains 15.1% proline, 17.8% serine, and 12.5% threonine in mammalian species, this new tail also has a high proportion of serine (19.3%) but a lower proportion of proline (7%) and threonine (1.8%). It was, thus, named the S tail, and S tail plus the PST N was named PST N S (Fig. 1B).
The amino acid sequence of No. 8 suggests the existence of a new splicing isoform of Pax6 that was designated as Pax6(S). The DNA sequence preceding exon 7 was missing in No. 8 (Fig.  1B). We next used 5Ј rapid amplification of cDNA ends to identify the nucleotide sequence of full-length Pax6(S) (for details, see "Experimental Procedures"). As shown in Figs. 1B and 2, the Pax6(S) transcript includes exons 4 -11␣, producing a 401amino acid protein with a calculated molecular mass of ϳ45 kDa. Thus, the only difference between Pax6(S) and Pax6 proteins lies in the C terminus (compare the lower panels of Fig. 1,  A and B). The 3Ј-untranslated region (UTR) of Pax6(S) is different from that of Pax6; however, it does contain a classic polyadenylation signal (Fig. 2). The 5Ј-UTR of the Pax6(S) transcript is composed of only 48 nucleotides (Fig. 2), whereas Pax6 has a 417-nucleotide 5Ј-UTR. Because UTRs carry information for regulation of translation and mRNA stability (46 -53), the divergence in the UTR sequence implies that translation and/or mRNA stability of Pax6(S) and Pax6 are differentially modulated.
The protein sequence of the canonical PST domain is 100% conserved from rodents to human. Do different species, if they produce the S tail homologues, also share a high identity in this region? We performed a BLAST search with the DNA sequence of exon 11␣ and indeed pulled out homologues from various species. However, the protein sequences of these homologues were 100% conserved only between human and chimpanzee (supplemental Fig. S1). The conservation remained high among the primates we examined, decreasing to 92% for sumatran orangutan and 91% for olive baboon and rhesus monkey, but it decreased dramatically in rabbit (25%) and mouse (31%) (supplemental Fig. S1). In fact, the predicted rabbit and mouse S tails are 14 and 22 amino acids shorter than the human S tail, respectively (supplemental Fig. S1). Using embryonic and adult mouse brain cDNA libraries, we performed reverse transcription-PCR analysis with primers that specifically annealed to the DNA encoding the predicted mouse S tail. We did not detect any reverse transcription-PCR products (data not shown), suggesting low, if any, expression of the S tail in mouse.
Expression of Pax6(S) Protein in Situ-To study Pax6(S) protein expression, a rabbit polyclonal antibody against an epitope (HNPGPREVRSGSGP) located within the unique S tail (Fig. 2) was synthesized. To test the specificity of this Pax6(S) antibody, Pax6(S) or Pax6 (both tagged with FLAG on the N terminus) were individually expressed in HEK 293T cells. The Pax6(S) antibody stained the cells expressing FLAG-Pax6(S) (Fig. 3A, upper leftmost) but not those expressing FLAG-Pax6 (Fig. 3A,  lower leftmost). The lack of staining in the latter was not due to a lack of protein expression, as strong FLAG signals were detected with an anti-FLAG antibody (Fig. 3A, the second left panel in the lower row).
Two nuclear localization signals are present in Pax6, one in the PD and the other in the linker region (54). The PST domain has been suggested to contain glycosylation sites (55) that may also be involved in the nuclear transport of Pax6 (56,57). Overexpressed Pax6(S) was localized in the cell nucleus as was Pax6, as indicated by colocalization with a nuclear dye, DAPI (Fig. With a specific anti-Pax6(S) antibody in hand, we went on to examine Pax6(S) expression in situ. The critical role of Pax6 in eye development has been well established (32)(33)(34)(35). We indeed detected Pax6 in the inner nuclear layer of human retina (Fig. 3B, left), in agreement with previous reports (58). However, we did not detect Pax6(S) in retina (Fig. 3B, right). Because the Pax6(S) transcript was originally isolated from a human adult brain cDNA library, we then screened Pax6(S) expression in various regions of human adult brain, including the frontal, temporal, parietal, and occipital lobes, pons, thalamus, and corpus callosum. None of these regions showed any detectable Pax6(S) signal (data not shown), suggesting that Pax6(S) expression is low in adult brain. We next examined Pax6(S) expression in human embryonic brain. Interestingly, nucleus-like signals were detected in 4-month-old (Fig. 3C, left) but not in 5-month-old (Fig.  3C, right) embryonic brain. Western blots detected a protein band of ϳ45 kDa in the lysate of human embryonic brain (Fig. 3D). In addition, protein bands of ϳ50 kDa were detected in lysates of various human embryonic tissues including the lung, small intestine, heart, liver, skeletal muscle, and placenta (Fig.  3D). This higher-than-calculated molecular weight could be due to phosphoryl or glycosyl modification of Pax6(S) in these tissues. In fact, it has been reported by several groups that phosphorylation or glycosylation of Pax6 upper-shifted its protein band in Western blot (42)(43)(44)55). Extra bands of ϳ80 or ϳ37 kDa were present in the small intestine, heart, skeletal muscle, and placenta, suggesting that apart from Pax6(S), Pax6 might have other isoforms sharing part of the S tail that can be recognized by the Pax6(S) antibody. Alternatively, these bands may simply be generated by a nonspecific binding of the antibody. However, because they were not universally stained in every tissue (Fig. 3D) and were completely absent in human cell line lysates we tested (data not shown), they are less likely to be artifacts. To verify this, further experiments such as peptide competition assays need to be performed.
Pax6(S) Retains Transcriptional Activity in Vitro-As a transcription factor, Pax6 governs the expression of diverse genes through binding to their cis-elements via its PD or HD. The C-terminal PST domain is not directly involved in DNA binding but plays a role in regulating the transcriptional activity (37)(38)(39)(40). We investigated whether Pax6(S) retains transcriptional activity and, if it does, whether its unique S tail confers a different level of transcriptional activity. The activity of Pax6(S) and Pax6 was examined in vitro with a luciferase assay. CD19-2, one of the consensus DNA binding sequences for the PD (24), was used as the promoter of the luciferase reporter gene. As shown in Fig. 4A, Pax6(S) enhanced the luciferase-induced fluo-rescence by ϳ4-fold, reflecting its ability in driving luciferase transcription. This ability is as potent as that of Pax6 (Fig. 4A), indicating that the S tail does not confer a different transcriptional activity to Pax6(S) in this context.
The PST N S Domain Shows Transactivity-The PST domain of Pax6 has been shown to have a transactivity independent of the PD and HD, as it increases the activity of GAL4 when fused to GAL4 (37,39,40). Using a luciferase assay, we investigated the transactivity of the PST N S domain of Pax6(S) by fusing its coding sequence, exons 10 -11␣ (Fig. 1B), to the GAL4 coding sequence. As a control, exons 10 -13, which encode the canonical PST domain (Fig. 1A), were also fused to the GAL4 coding sequence in a separate expression construct. In the reporter construct p5XGAL4-E1b-Luc, a GAL4 binding sequence was introduced as the promoter of the luciferase gene. GAL4 itself was able to slightly drive the expression of luciferase (Fig. 4B,  compare bars a and b). Its activity was greatly increased when GAL4 was fused with PST (encoded by exons 10 -13) (Fig. 4B,  compare bars c and b). PST N S (encoded by exons 10 -11␣) also enhanced the activity of GAL4 (Fig. 4B, compare bars d and b), albeit to a much less degree than PST did (Fig. 4B, compare bars  d and c). Consistent with its inability to bind DNA by itself,  neither PST nor PST N S was able to induce luciferase expression in the absence of GAL4 (Fig. 4B, bars h and i).
A study has shown that exons 10 -13 synergistically stimulate transcriptional activation and that the transactivation potential is not localized but spread throughout PST (37). We confirmed this finding by individually fusing exons 10 -11 or exons 12-13 to the GAL4 coding sequence. Both were able to stimulate the GAL4 activity, with exons 12-13 being more potent than exons 10 -11 (Fig. 4B, compare bars e and f). However, neither shows a transactivity comparable with that produced by exons 10 -13 (Fig. 4B, compare bars c, e, and f). Exon 11␣, which is uniquely present in Pax6(S), by itself enhanced the GAL4 activity but to a lesser extent than exons 10 -11␣ did (Fig. 4B, compare bars g and d), suggesting that exon 11␣ also works synergistically with exons 10 -11 to stimulate transcriptional activation. In the absence of GAL4, exons 10 -11, 12-13, and 11␣ had no effect (Fig. 4B, bars j, k, and l).
Why did the PST N S domain show a lower transactivity than the PST when fused with GAL4 (Fig. 4B, bars c and d), whereas Pax6(S) exhibited transcriptional activity as potent as that of Pax6 (Fig. 4A)? One explanation is that PST and PST N S produce different allosteric effects on GAL4 but not in Pax6 and Pax6(S) and, thus, lead to different regulations of GAL4 activity.
Pax6(S) Interacts with Ca v ␤ Mainly through the S Tail-The interaction between Pax6(S) and ␤ 3 was first identified from a yeast two-hybrid assay (Fig. 5A) and was subsequently confirmed by a pulldown assay. The No. 8 protein synthesized in vitro was pulled down by GST-tagged ␤ 3 (Fig. 5B). This effect was not an artifact caused by the GST tag, as the No. 8 protein could not be pulled down by GST itself (Fig. 5B).
By performing a pairwise yeast two-hybrid assay, we next examined whether the S tail of Pax6(S) was involved in this interaction. No. 8 was dissected into two fragments. Fragment 1 (Fig. 5A, No. 8(⌬S)) contains the regions that are common in both No. 8 and Pax6. Fragment 2 (Fig. 5A, S) is entirely the S tail. Fragment 1 showed a weak interaction with ␤ 3 , whereas fragment 2 (i.e. the S tail) conferred a much stronger interaction. On the contrary, the C-terminal half of the PST domain (PST C ), which is missing in Pax6(S) and is the S tail counterpart in Pax6, did not seem to interact with ␤ 3 (Fig. 5A).
Pax6(S) Does Not Affect the Biophysical Properties of Ca 2ϩ Channels-It is surprising that Pax6(S), a transcription factor, can interact with a channel protein. Ca v ␤s profoundly regulate gating and the surface expression of HVA Ca 2ϩ channels (1)(2)(3)(4). To test whether the interaction between Pax6(S) and Ca v ␤ affects HVA Ca 2ϩ channel properties, we expressed the channel complex containing Ca v 2.1, ␣ 2 -␦, and ␤ 3 in Xenopus oocytes with or without Pax6(S) being present and examined a host of channel biophysical properties with cell-attached patch clamp. Pax6(S) did not affect the current-voltage relationship, voltagedependent activation, deactivation speed (indicated by deactivation constant, deact ), voltage-dependent inactivation, and inactivation speed (Fig. 6, A-E). Similarly, No. 8 did not affect FIGURE 5. Interaction between No. 8 and Ca 2؉ channel ␤ subunit. A, a summary of the results from yeast two-hybrid assay is shown. Positive and negative results are shown by plus and minus signs, respectively. Compared with one plus sign, three plus signs indicate more yeast cell growing on the selective plates, suggesting a stronger interaction. EV, empty vector. B, re-test the interaction between No. 8 and ␤ 3 with the GST pulldown assay. these biophysical properties (supplemental Fig. S2). However, Pax6(S) moderately reduced the current amplitude measured by two-electrode voltage clamp (Fig. 6F), suggesting that Pax6(S) may interfere with channel trafficking to the plasma membrane. In contrast, co-expression of No. 8 did not significantly change the current amplitude (Fig. 6F), suggesting that No. 8 does not interfere with channel trafficking. The current reduction caused by Pax6(S) could also be due to toxicity of Pax6(S) RNA or protein, as it occurred only when large amounts of Pax6(S) cRNA were injected.
One could argue that the lack of effects of Pax6(S) on the channel properties in the above experiments could be because Pax6(S) cRNA failed to produce protein products in Xenopus oocytes or Pax6(S) proteins were completely sequestered in the nucleus. To exclude these possibilities, purified S tail protein, which by itself is capable of interacting with Ca v ␤ (Fig. 5A), was directly applied to the intracellular side of the channels in the inside-out macropatch configuration. This treatment did not affect the current amplitude and voltage-dependent activation (Fig. 6, G and H). Altogether, these results suggest that the interaction between Pax6(S) and Ca v ␤ does not affect the activity of HVA Ca 2ϩ channels.
Ca v ␤ Decreases the Transcriptional Activity of Pax6(S) in Vitro-We next investigated whether Ca v ␤ affects the transcriptional activity of Pax6(S), as determined by the luciferase assay described in Fig. 4A. Pax6(S) robustly stimulated luciferase expression (Fig. 7, compare bars a and b), and co-expression of ␤ 3 decreased this activity by ϳ50% (Fig. 7, compare bars b  and c). Suppression of Pax6(S) activity by ␤ 3 was a specific effect, because the activity of Pax6(S) was unchanged in the presence of bacterial maltose-binding protein (MBP) (Fig. 7,  compare bars b and d).
We further investigated whether the suppression of Pax6(S) activity by ␤ 3 is through the S tail. A Pax6(⌬S) construct, in which the S tail was truncated from Pax6(S), was capable of stimulating luciferase activity to a larger level than did Pax6(S) (Fig. 7, compare bars b and e). However, ␤ 3 did not dampen the transcriptional activity of Pax6(⌬S) (Fig. 7, compare bars f and e). These results indicate that ␤ 3 can suppress Pax6(S) activity, and it does so through interacting with the unique S tail of Pax6(S).

Overexpressed Pax6(S) and Ca v ␤ Co-localize in the Nuclei of HEK 293T Cells-Does Ca v ␤ regulate Pax6(S) activity in cells?
As an initial step to address this question, we investigated the subcellular localization of Ca v ␤ in the absence and presence of Pax6(S) or No. 8, the Ca v ␤-interacting component of Pax6(S). Pax6(S) is localized in the nuclei, as shown in Fig. 3A. As a subunit of HVA Ca 2ϩ channels, Ca v ␤ is predominantly distributed in the cytoplasm (1)(2)(3)(4). Does the interaction between Pax6(S) and Ca v ␤ change their respective subcellular localization? Pax6(S) and ␤ 3 were tagged with FLAG and EGFP, respectively. When they were expressed in HEK 293T cells separately, Pax6(S) was always restricted in the nucleus (Fig. 8A, left) and ␤ 3 in the cytoplasm (Fig. 8A, right) as expected. However, in the presence of Pax6(S), ␤ 3 was translocated into the nucleus and co-localized with Pax6(S) (Fig. 8B). Similar observations were obtained when ␤ 3 was with coexpressed with No. 8 (supplemental Fig. S3). In the presence of ␤ 3 , a larger amount of No. 8 became aggregated along the nuclear membrane and exhibited a punctuate expression pattern (supplemental Fig. S3B). This, however, was not the case for Pax6(S).

Pax6(S) Is a Novel Splicing Isoform of Pax6-Pax6
is critical for the development of various tissues and organs, particularly the eye and the nervous system (18,20,(27)(28)(29)(30)(31). Here we reported a new splicing isoform of Pax6, Pax6(S), whose C terminus (PST N S) is composed of a truncated canonical PST domain and a unique S tail. Compared with Pax6, Pax6(S) differed in tissue distribution, temporal expression profile, and the transactivity of the PST N S domain. Our results suggest a yet-to-be-defined noncanonical role of Pax6(S) during development.
The protein sequence of classic Pax6 is highly conserved from invertebrates to human, which also leads to functional conservation. For example, overexpression of the mouse or Drosophila Pax6 gene induces an ectopic eye structure in both vertebrates and invertebrates (32)(33)(34)(35). In contrast, the protein sequence of Pax6(S) is highly conserved only among primates (supplemental Fig. S1).  The critical role of Pax6 in eye development was consistent with its expression in the inner nuclear layer of the human retina (Fig. 3B,  left (58)). However, we did not detect any Pax6(S) expression in the human retina (Fig. 3B, right). The divergence in sequence conservation between Pax6 and Pax6(S) and the lack of Pax6(S) expression in the retina imply that whereas Pax6 is involved in developmental events common in both lower and higher animals, Pax6(S) probably plays roles in regulating more specific functions existing solely in primates.
The temporal expression profile of Pax6(S) during development also seems to be different. Classic Pax6 is expressed in both embryos and adults (24). However, Pax6(S) seems to be preferably expressed in embryonic tissues. Even during embryonic development, Pax6(S) expression appears to be strictly regulated. It was detected in 4-month-but not in 5-month-old embryonic human brain (Fig. 3C). This different expression profile of Pax6(S) could be at least partially related to the unique 5Ј-and 3Ј-UTRs of its mRNA (Fig. 2).
Ca v ␤ Acts as a Transcription Regulator-As a channel accessory subunit, Ca v ␤ is predominantly localized in the cytoplasm (1-4) (also see Fig. 8A). However, co-expression of Pax6(S) redistributed Ca v ␤ into the nucleus (Fig. 8B). This could be a direct effect of Pax6(S) (i.e. Pax6(S) enters the cytoplasm and transport Ca v ␤ to the nucleus) or involves an unknown chaperone protein(s). The first scenario requires colocalization of Pax6(S) and Ca v ␤ in the cytoplasm, which we never observed under our experimental conditions; however, it could be that the cytoplasmic expression of Pax6(S) was too transient to be detected.
We also showed that the transcriptional activity of Pax6(S) was reduced through its interaction with Ca v ␤. One could propose at least two scenarios to explain how Ca v ␤ reduces Pax6(S) activity. First, Ca v ␤, Pax6(S), and Pax6(S)-regulated DNA form a complex. The binding of Ca v ␤ allosterically suppresses Pax6(S) activity. Second, instead of associating with the Pax6(S)-DNA complex, Ca v ␤ forms a complex with Pax6(S) and removes Pax6(S) from its DNA targets. These scenarios need to be examined in further studies.
The potential role of Ca v ␤ in directly regulating gene expression was first suggested by the study of Hibino et al. (59). In that study a short splice variant of ␤ 4 , named ␤ 4c , in the chicken cochlea and brain was found to interact directly with the chromo shadow domain of chromobox protein 2/heterochromatin protein 1␥ (CHCB2/HP1␥), a nuclear protein involved in gene silencing and transcriptional regulation. Co-expression of this protein specifically recruited ␤ 4c to the nuclei of mammalian cells. Furthermore, ␤ 4c dramatically attenuated the genesilencing activity of CHCB2/HP1␥. This effect was ␤ 4c -specific, as a longer isoform, ␤ 4a , did not affect CHCB2/HP1␥ activity (59). These findings establish ␤ 4c as a likely transcription regulator. However, ␤ 4c is severely truncated and lacks all the amino acids that are involved in the high affinity interaction between a full-length Ca v ␤ and the I-II loop of the Ca 2ϩ channel ␣ 1 subunit (5)(6)(7), an interaction that is critical for Ca v ␤ regulation of Ca 2ϩ channel surface expression and biophysical properties. Indeed, ␤ 4c has little effect on Ca 2ϩ channel activity (59).
Our results provide evidence that a full-length Ca v ␤ can directly interact with a transcription factor and regulate its activity. Transcriptional regulation may be a general function of Ca v ␤ as, apart from ␤ 3, all other Ca v ␤ species (␤ 1b , ␤ 2a , and ␤ 4 ) we tested were also able to interact with Pax6(S) (data not shown). Consistent with this notion, ␤ 4 , and to a lesser extent, ␤ 1b and ␤ 3 , are translocated into the nucleus when exogenously expressed in cardiac cells (60), and ␤ 3 localizes in the nucleus when it is co-expressed in PC12 cells with Rad and Rem, two members of the RGK family of Ras-related monomeric small GTP-binding proteins (61).
CASK, a membrane-associated guanylate kinase involved in cell junction, has been shown to have a transcription regulation function that lies in its guanylate kinase domain (62,63). This further suggests a general role of Ca v ␤ in transcriptional modulation, as all Ca v ␤s contain a homologous guanylate kinase domain (5)(6)(7). CASK can either act as a co-activator of Tbr-1, a T-box transcription factor, or as an independent transcription factor through binding to a specific DNA sequence (the T-element) (62). Can Ca v ␤ also act as an independent transcription factor? In our luciferase assay system, where pGL3-OFLuc-CD19-2 was used as the reporter construct, Ca v ␤ was unable to induce luciferase expression in the absence of Pax6(S) (Fig. 7, bar h). However, it cannot be excluded that Ca v ␤ could govern gene expression through binding to an unknown, specific DNA region, the sequence of which could be identified using chromatin immunoprecipitation.
Function and Regulation of Pax6(S)-Pax6(S) is fully conserved only in human and chimpanzee. Therefore, to examine the function of Pax6(S) under physiological conditions, cell lines must be utilized because animal models are impractical. We have screened more than 10 cell lines developed from diverse human tissues including the eye, lung, brain, and colon. Unfortunately, none of them showed detectable expression of Pax6(S) (data not shown). If a suitable cell line is found, the endogenous Pax6(S) can be knocked down with the RNA interference technique, and its physiological functions can be explored by examining changes in gene expression, cell differentiation, and cell morphology.
Phosphorylation and dephosphorylation of Pax6 are important modulations that fine-tune its functions (42)(43)(44). Several phosphorylation sites have been identified within the PST domain, including Thr-281, Thr-304, and Thr-373 (42). On the other hand, protein serine/threonine phosphatase-1 dephosphorylates Pax6 and attenuates its activity in human lens epithelial cells (44). Because the S tail of Pax6(S) has a similar proportion of serine but a much lower proportion of proline and threonine compared with the PST tail in Pax6, it is likely that Pax6(S) is regulated differently by protein kinases or phosphatases. In addition, protein phosphorylation has been shown to affect nuclear import or export of various proteins, including the SV40 large T antigen (64), cyclin B1 (65), nucleolin (66), mitogen-activated protein kinase-activated protein kinase 2 (67), tumor suppressor protein 53 (68), and human double minute 2 oncogene (69). It will be interesting to examine whether phosphorylation of Pax6(S) affects its subcellular localization and, if it does, whether this plays a role in the cytoplasm-to-nucleus redistribution of Ca v ␤ (Fig. 8). These studies will provide further insights into the mechanism and physiological regulation of the interaction between Pax6(S) and Ca v ␤.