BARP suppresses voltage-gated calcium channel activity and Ca²⁺-evoked exocytosis

Exocytosis in response to action potential–evoked membrane depolarization has been extensively characterized in the nervous system, in which neurotransmitters or hormones are released after extracellular Ca²⁺ influx at synapses in neurons or in neuroendocrine cells, respectively. In pancreatic islet β cells, for example, glucose elevation results in the closure of K_ATP channels, membrane depolarization, opening of voltage-gated calcium channels (VGCCs), and, in response to Ca²⁺ influx, secretion of insulin (Yang and Berggren, 2006). At neuronal synapses, neurotransmitter-containing vesicles are docked in close vicinity to VGCCs at the presynaptic active zone (Neher, 1998; Zhai and Bellen, 2004; Atwood, 2006). Although the spatial proximity of VGCCs and exocytic vesicles undergoing fusion with the plasma membrane is well documented, the detailed molecular mechanisms involved in the spatial and temporal coupling of exocytosis and VGCC activation and inactivation remain to be elucidated.

VGCCs are composed of an ion pore–forming Ca_vα1 subunit associated with several auxiliary subunits (Ca_vβ, Ca_vα2δ, and Ca_vγ; Arikkath and Campbell, 2003). Among the Ca_vα1 subunits, the P/Q-type Ca_v2.1 and the N-type Ca_v2.2 define the main channel subtypes important for presynaptic neurotransmitter release (Spafford and Zamponi, 2003; Evans and Zamponi, 2006), and the L-type Ca_v1.2 subtype triggers Ca²⁺-dependent secretion in neuroendocrine cells (Catterall, 2000). Four Ca_vβ subunit isoforms (Ca_vβ1, Ca_vβ2, Ca_vβ3, and Ca_vβ4) show distinct tissue and subcellular distributions (Dolphin, 2003; Buraei and Yang, 2010). Ca_vβ subunits interact with the 18-aa α1 interaction domain (AID) of the cytoplasmic linker between internal repeats I and II of the pore-forming α₁ subunit (Pragnell et al., 1994; Chen et al., 2004; Opatowsky et al., 2004; Van Petegem et al., 2004). Ca_vβ subunits enhance VGCC channel activity (Mori et al., 1991; Chien et al., 1995; Josephson and Varadi, 1996; Kamp et al., 1996; Brice et al., 1997; Jones et al., 1998; Colecraft et al., 2002), not only by facilitating cell surface transport of VGCCs and by preventing ER-associated protein degradation (Altier et al., 2011) but also by modulating their gating properties (Buraei and Yang, 2010).

VGCCs interact via the Ca_vα1 subunit with several pre- and postsynaptic proteins, including SNAP-25, synaptotagmin, syntaxin, Mint, and calcium/calmodulin-dependent serine protein kinase (Sheng et al., 1994; Bezprozvanny et al., 1995; Zhong et al., 1999; Maximov and Bezprozvanny, 2002; Spafford and Zamponi, 2003; Nishimune et al., 2004; Kang et al., 2006). The interaction and clustering of VGCCs with components of the secretory vesicle docking and fusion machinery by multiprotein adaptors highlights the importance of the spatial and temporal coordination of Ca²⁺ entry and neurosecretion (Yang and Berggren, 2006). The Ca_vβ subunits also interact with regulatory proteins that inhibit (e.g., RGK proteins, calcium, heterotrimeric G proteins, opioid receptor–like receptor 1, and several synaptic proteins) or facilitate VGCC activity (e.g., Rim1) or both (e.g., calmodulin; Herlitze et al., 1996; Ikeda, 1996; Lee et al., 1999; Béguin et al., 2001, 2005a,b, 2006, 2007; Beedle et al., 2004; Chen et al., 2005; Finlin et al., 2005; Evans and Zamponi, 2006; Jarvis and Zamponi, 2007; Kiyonaka et al., 2007; Buraei and Yang, 2010; Flynn and Zamponi, 2010; Yang et al., 2010).

Here, we describe a previously uncharacterized protein, which we term the VGCC–β-anchoring and -regulatory protein (BARP), and characterize its role in the regulation of VGCC activity and Ca²⁺-regulated exocytosis. BARP is highly expressed in several specific neuronal populations and neuropeptide secretory cells, plays a role in the recruitment of Ca_vβ subunits to the plasma membrane, and negatively regulates VGCCs by interfering with the association of the Ca_vβ subunit with the Ca_vα1 subunit. We hypothesize that BARP serves as an adaptor protein that modulates Ca_vβ subunit localization and their association with Ca_vα1 subunits to regulate VGCC activity.

BARP was identified in a yeast two-hybrid screen of a mouse insulin-secreting MIN6 cell cDNA library using Ca_vβ3 as bait. BARP is encoded by an open reading frame of unknown function, C19orf26, which, based on its chromosomal location, has also been referred to as Dos (downstream of Stk11 kinase; Gerhard et al., 2004). Sequence analysis of EST clones and cDNA cloned from libraries revealed a ∼3-kb transcript, coding for a 698-aa protein. BARP contains no known functional domains except for a single putative transmembrane domain and a putative N-glycosylation site (Figs. 1 A and S1 A). High BARP mRNA levels were found in brain, pancreatic islets, and neuroendocrine cell lines (MIN6 and PC12), with undetectable or weak expression in other tissues (Fig. 1 B).

The presence of BARP protein was confirmed in brain and PC12 cells using a polyclonal (72) and two affinity-purified mAbs (12B1 and 8B2) raised against different regions of the protein (Fig. 1 C). Specificity of the different antibody (Ab) was validated by the lack of staining in untransfected cells (Fig. S1 B) or after preabsorbing the Ab on a GST-BARP fusion protein (Figs. S1 C and S2 B) and by the absence of reactivity in pancreas-specific BARP knockout mice (Fig. S2 C). In COS-1 cells transfected with a Myc-tagged BARP cDNA, the labeling of the three Abs predominantly colocalized with that of the anti-Myc Ab (Fig. S1 B), but the BARP Ab did not colocalize with a Myc-tagged β-galactoside used as a negative control.

The predicted initiation methionine for BARP (M1; Fig. S1 A) is in accordance with the high quality annotation of the protein coding regions of the mouse and human genome by the Consensus Coding Sequence Project (Pruitt et al., 2009). This is the first conserved Met and the site where the high degree of amino acid identity among BARPs from different species starts (Fig. S1 D). In agreement with M1 being the initiation methionine, expression in COS-1 cells of mutants lacking either M7 or M80, the only two other conserved putative translation initiation sites in BARP, showed the same electrophoretic mobility as wild-type (WT) BARP, consistent with initiation of translation from M1 (Fig. S1 D). In contrast, the mobility of the M1A mutant was detectably faster than WT BARP, presumably because in the absence of M1, initiation can occur from M7. A truncated translation product was also obtained from the M1A/M7A mutant, presumably as a result of translation from M80 because only mutation of all three conserved methionine residues abolished translation. Thus, although M1, M7, and M80 can act as independent initiation sites, in the full-length cDNA, the codon for M1 serves as the main site for the initiation of translation.

BARP overexpressed in COS-1 or human embryonic kidney–derived tsA201 cells migrated as a doublet of higher molecular mass as compared with the in vitro translated BARP (Fig. 1 D), suggestive of posttranslational modifications. In PC12 cells, the upper band was more prominent. Tunicamycin led to a reduction of the apparent molecular mass for WT BARP but not for a mutant lacking the putative N-glycosylation site (²⁵N-X-S/T), consistent with the presence of N-linked glycosylation (Fig. 1 E). For better resolution of the different bands by SDS-PAGE, a C-terminal truncation of BARP (C-tr145, consisting of aa 1–145) was analyzed. After tunicamycin, PNGase F, or Endoglycosidase H treatment, the truncated form of BARP co-migrated with the truncated form lacking the N-glycosylation site (Fig. S1 E), confirming the presence of this glycosylation. After these treatments, however, BARP still migrated as a doublet, indicating the existence of additional unknown posttranslational modifications or internal initiation of translation.

The presence of N-terminal carbohydrate chains suggested that BARP is a type I membrane protein, with the N terminus located on the extracellular side. This topology was validated by cell surface labeling experiments, in which N- or C-terminally Myc-tagged BARP was expressed in COS-1 cells and the cells were labeled with antitag Ab either before (cell surface labeling) or after (cell expression) cell permeabilization. Only N-terminally tagged BARP was detected by incubating nonpermeabilized intact cells with Ab to Myc, thus confirming the extracellular exposure of the N terminus (Fig. 1 F).

Immunolabeling of mouse tissue sections revealed BARP protein in the cortex, cerebellum, and hippocampus of the brain (Fig. 2 A, a) and in pancreatic islets (Figs. 2 A, b; and S2 C), consistent with the Northern blot analysis (Fig. 1 B). During mouse development, BARP expression in the brain peaked between embryonic day 18 (E18) and postnatal day 7 (P7; Fig. S2 A). In the cortex and hippocampus, BARP was detected in pyramidal cell bodies and dendrites. In the cerebellum, BARP was exclusively expressed in Purkinje cells, uniformly in soma and the main dendritic shaft and as a patchy staining along the distal dendrites (Figs. 2 A, a; and S2 B). Costaining of cultured primary cells isolated from hippocampus and cerebellum with BARP and neuronal (MAP2 and calbindin) or glial (glial fibrillary acidic protein [GFAP]) markers confirmed the neuron-specific expression of BARP (Fig. 2 B). In cerebellar primary cells, BARP was present in calbindin-positive Purkinje cells, where it localized to the cell soma, the dendritic shaft, and along the axon, including the presynaptic button, and colocalized with the dense core vesicle marker brain-derived neurotrophic factor (BDNF).

In PC12 cells, BARP partially colocalized with the Ca²⁺-dependent secretory vesicle markers synaptotagmin I and, upon NGF-induced differentiation, localized to the growth cone (Fig. 2 C). In contrast to the more prominent vesicular staining of endogenous BARP in PC12 cells (Fig. 3 C, b), BARP overexpressed in PC12 or other cell lines was enriched at the plasma membrane (Fig. S2, D and F). Such a difference in distribution was also reported for endogenous versus overexpressed synaptotagmin I (Vega and Hsu, 2001) and could be reproduced in our PC12 cells (compare Figs. 2 C and S2 D). This was interpreted to reflect a shift in the steady-state distribution from a more prominent vesicular to a more pronounced plasma membrane localization after overexpression (Han et al., 2004; Atiya-Nasagi et al., 2005).

Coimmunoprecipitation experiments using BARP and the different Ca_vβ subunit isoforms overexpressed in COS-1 cells established that BARP binds to all the different Ca_vβ isoforms (Fig. 3 A). These associations were corroborated in intact cells using immunofluorescence experiments. When overexpressed in PC12 (Fig. 3 B) or COS-1 (Fig. S2 E) cells, the Ca_vβ subunits showed, with the exception of Ca_vβ2a, a cytosolic distribution with some degree of nuclear labeling for Ca_vβ3 and Ca_vβ4a (Dolphin, 2003). Remarkably, overexpression of BARP localized the cytosolic Ca_vβ isoforms to the plasma membrane. This confirms that the two proteins interact in a cellular context and shows that BARP can influence the subcellular distribution of Ca_vβ subunits in the absence of Ca_vα1. The colocalization index, which in this context reflects the efficiency of recruitment of the different Ca_vβ subunits by BARP, did not drastically differ, indicating a similar effect of BARP for all isoforms. As a control, BARP M1A/M7A, which utilizes M80 for initiation of translation (Fig. S1 D) and thus lacks the first 79 aa and hence the transmembrane domain, failed to localize Ca_vβ3 to the plasma membrane (Figs. 3 B, a; and S2 E).

When BARP was constitutively and stably overexpressed from a cytomegalovirus promoter in PC12 Tet-On cells (BARP10; Fig. 3 C, a), the localization of endogenous BARP changed from a vesicular pattern in parental cells to a prominent peripheral staining in BARP-overexpressing cells (Fig. 3 C, b). Concomitantly, endogenous Ca_vβ1 and Ca_vβ3, which poorly colocalized with endogenous BARP in untransfected cells, redistributed to the cell periphery upon BARP overexpression (Fig. 3 C, b).

To obtain additional insight into the mechanism by which BARP associates with VGCCs, the domains in BARP required for its interaction with the Ca_vβ subunit were identified. Deletion and alanine-scan mutagenesis of BARP combined with yeast two-hybrid screening (Fig. S3, A and B) and GST pull-down experiments (Fig. S3, C and E) narrowed down the interaction with Ca_vβ subunits to two domains, termed domain I and domain II. Based on in silico molecular dynamics simulations, domain I (aa 422–442; Fig. 4 A, a) is predicted to fold into an α helix (Fig. S3 D). Mutation of L426, W427, or R430 in BARP abolished the interaction of domain I with the Ca_vβ subunit (Fig. 4 A, b). The AID of the Ca_vα1 subunit and mutants carrying substitutions of aa Y467, W470, and I471, known to be important for binding to Ca_vβ (Richards et al., 2004), served as a control (Fig. 4 A, a and b).

Previous studies, including crystallographic analysis, established that the AID is associated as an α helix with a hydrophobic groove in Ca_vβ, also termed the AID-binding pocket (ABP; Pragnell et al., 1994; Chen et al., 2004; Richards et al., 2004; Van Petegem et al., 2004). To explore whether domain I could also bind to the ABP, mutations in this region of Ca_vβ were generated, and their effect on the interaction with BARP was tested. Substitutions of several amino acids in the hydrophobic pocket of the Ca_vβ subunit impaired its association with BARP domain I (Fig. 4 A, c). This was not caused by an effect of the mutations on the folding of Ca_vβ because, with the exception of M196A, the different Ca_vβ mutants still bound the AID.

Because a truncated BARP that lacked domain I still interacted with Ca_vβ, analysis of additional BARP mutants (Fig. S3 E) led to the identification of a second binding region, domain II (aa 525–563; Fig. 4 B, a). Amino acid substitutions in domain II revealed two leucine-phenylalanine pairs (L545-F546 and F549-L550) as important for efficient Ca_vβ binding (Fig. 4 B). Interestingly, the isolated BARP domain II interacted with Ca_vβ3, Ca_vβ2a, Ca_vβ2b, and Ca_vβ4a but not with Ca_vβ1a, whereas domain I bound all Ca_vβ subunits tested (Fig. S3 F).

The effect of the amino acids substitutions in domain I (L426A and W427A) and/or domain II (L545A, F546A, F549A, and L550A) on the association with Ca_vβ was next analyzed in full-length BARP. Simultaneous mutation of both domains abolished the association of the mutated BARP with the Ca_vβ (Fig. 4 C). Although the interaction between BARP and Ca_vβ was more sensitive to the disruption of domain I, the presence of one intact domain was sufficient to confer not only detectable binding but also plasma membrane localization of the Ca_vβ subunits in cells (Figs. 4 D and S3 H). However, the colocalization index suggests that domain II alone mediates a less efficient localization of Ca_vβ3 to the plasma membrane (0.68 ± 0.05) compared with WT BARP (0.80 ± 0.02) or BARP with domain II mutated (0.80 ± 0.02, P < 0.05).

Because BARP domain I binds to the ABP in Ca_vβ, BARP may interfere with the interaction between Ca_vβ and Ca_vα1. To test this hypothesis, we monitored the stability of a preassembled complex between the Ca_vβ₃ and a GST-AID fusion protein after addition of competitive peptides coding for domain I or, as a control, the AID. Indeed, Ca_vβ3 was displaced by soluble domain I or AID peptides from the immobilized GST-AID and recovered in the supernatant (Fig. 5 A), consistent with BARP domain I and the AID binding in a mutually exclusive manner to the same or an overlapping site in Ca_vβ. Slightly higher concentrations of domain I peptide than AID peptide were required for disruption of the complex.

The biochemical results were corroborated in a cellular context by coexpressing BARP, Ca_vβ3, and Ca_vα1 and monitoring their associations in coprecipitation experiments. In the presence of WT BARP, the Ca_vβ3 and Ca_vα1 subunits no longer associated (Fig. 5 B). In contrast, coprecipitation was not affected in the presence of BARP with both domain I and II mutated. Mutation of either domain individually partially interfered with the association between the Ca_vβ3 and Ca_vα1, showing the importance of domain I and II. Interestingly, mutation of domain I alone allowed the detection of a ternary complex containing BARP, Ca_vα1, and Ca_vβ3 (Fig. 5 B, lane 3). As a control, neither Ca_vβ3 nor BARP associated with a mutated Ca_vα1 subunit unable to bind Ca_vβ3 (Fig. S4 A), suggesting that BARP does not bind to Ca_vα1 directly. Similar results were also obtained for Ca_vβ3 in combination with other α1 subunit subtypes (e.g., Ca_v2.1 and Ca_v2.2; Fig. S4, B and C) and for other Ca_vβ subunit isoforms (e.g., Ca_vβ1a, Ca_vβ2a, Ca_vβ2b, and Ca_vβ4a; Fig. S4 D).

Coimmunoprecipitation experiments from PC12 cells and the brain confirmed that also endogenous BARP and Ca_vβ3 associate with each other (Fig. 5 C). Interestingly, comparison of the amount of endogenous Ca_vβ3 that was bound to either endogenous or overexpressed BARP in PC12 cells suggested the presence of a significant pool of Ca_vβ3 that is not associated with endogenous BARP. In addition to Ca_vβ3, BARP associated with Ca_vβ4 in cerebellum and cerebrum and to a lesser extent with Ca_vβ1 in the latter (Figs. 5 D and S4 E).

Ca_vβ subunits modulate Ca²⁺ channel activity by increasing channel current density at the plasma membrane and/or facilitating the trafficking of newly synthesized Ca_vα1 subunits from the ER to the plasma membrane (Chien et al., 1995; Dolphin, 2003). To explore the functional role of BARP in regulating VGCC activity, BHK cells stably expressing Ca_vα1 (Ca_v2.1 or Ca_v2.2), Ca_vβ1a, and Ca_vα2δ subunits to reconstitute P/Q- or N-type VGCCs (Niidome et al., 1994) were transfected with or without WT or mutated BARP cDNAs (Fig. 6, A and B) and subjected to electrophysiological analysis. Compared with controls, in cells overexpressing WT BARP, but not BARP with domains I and II mutated, a drastic reduction in P/Q- or N-type channel Ca²⁺ currents was recorded (Figs. 6, A and B, a and c; and S5 A). Mutation of domain I or II individually resulted in a partial reduction of Ca²⁺ channel activity (Figs. 6, A and B, b and c; and S5 A). Inactivation kinetics in the presence or absence of WT BARP did not significantly differ (P/Q-type channel: control = 78 ± 12 ms, n = 20 [Kameyama et al., 1999]; WT BARP = 80 ± 23 ms, n = 17).

To analyze whether a defect in Ca²⁺ channel surface expression in the presence of BARP accounts for its inhibitory effect on Ca²⁺ currents, we monitored Ca_v1.2 trafficking in tsA201 cells. Similar to BHK cells expressing N or P/Q Ca²⁺ channel subtypes (Fig. 6, A and B), BARP inhibited Ca²⁺ channel activity in tsA201 cells coexpressing Ca_v1.2 and Ca_vβ3 (Fig. 6 C). This inhibition was abrogated if both domains I and II were mutated and partially reduced if domain I or II was mutated independently.

Detection of Ca_v1.2 on the surface of nonpermeabilized intact tsA201 cells (Altier et al., 2002; Béguin et al., 2006) required coexpression with one of the Ca_vβ subunits (Figs. 6 D, a; and S5 B), and accordingly, mutation of the AID abolished Ca²⁺ channel cell surface transport (Fig. 6 D, a). Coexpression of Ca_v1.2 subunits with BARP, either in the absence or presence of Ca_vβ3, did not significantly affect Ca²⁺ channel surface expression (Fig. 6 D, a). Relative cell surface expression of Ca_v1.2 in the presence or absence of Ca_vβ3 and/or BARP was corroborated and quantified by coexpressing in tsA201 cells a Ca_vα1 subunit carrying both a luminal (HA) and a cytosolic (EGFP) tag and measuring relative pixel intensities in intact and permeabilized cells (Fig. 6 D, b and c). Although Ca_vβ3, either in the absence of presence of BARP, significantly increased the fraction of Ca_vα1 at the cell surface (0.12 ± 0.02 vs. 0.39 ± 0.03, P < 0.01), BARP had no significant influence on Ca_v1.2 distribution (Fig. 6 C, c). Thus, BARP does not significantly interfere with the role of Ca_vβ in facilitating cell surface expression of Ca²⁺ channels and thus most likely inhibits channel activity at the plasma membrane.

To elucidate the role of BARP in physiological processes linked to VGCC function, we explored the effects of BARP on VGCC activity and Ca²⁺-dependent hormone secretion in PC12 cells (Béguin et al., 2001). First, we took advantage of the observation that endogenous BARP expression varies in different PC12 clones and is higher in ATCC PC12 cells (CRL-1721) than in PC12 Tet-On cells (Takara Bio Inc.; Fig. 7 A, a). Correlating with the different BARP expression levels in these cells, endogenous Ca²⁺ channel current densities were lower in ATCC PC12 cells and significantly inhibited in PC12 Tet-On cells stably expressing BARP (Fig. 7 A, b). These results were corroborated in PC12 cells transiently expressing BARP. An almost complete reduction of endogenous Ca²⁺ currents was recorded in cells overexpressing WT BARP, whereas BARP with both domains I and II mutated had no significant effect on VGCC activity (Fig. 7 B).

Importantly, inhibition of VGCC activity by BARP had an effect on Ca²⁺-triggered exocytosis (Fig. 7 C). Correlating with the Ca²⁺ channel recordings, overexpression of BARP strongly inhibited Ca²⁺-dependent growth hormone secretion. Mutation of either domain I or II individually led to intermediate effects, whereas BARP carrying mutations in both Ca_vβ subunit binding sites only marginally affected growth hormone secretion.

Relative cell surface expression of Ca_v1.2 in the presence or absence of Ca_vβ3 and/or BARP was also analyzed in PC12 cells as described for tsA201 cells (Fig. 6 D, b and c). Although Ca_vβ3, either in the absence or presence of BARP, significantly increased the fraction of Ca_vα1 at the cell surface (0.22 ± 0.01 vs. 0.45 ± 0.01, P < 0.01), BARP had no significant influence on Ca_v1.2 channel distribution in PC12 cells (Fig. 7 D). Collectively, these data establish BARP as a negative regulator of Ca²⁺-dependent exocytosis, most likely by modulating VGCC activity at the plasma membrane.

We next analyzed the effect of silencing BARP in ATCC PC12 cells because these express significant levels of the protein endogenously (Fig. 7 A, a). Three shRNAs (A, B, and C) that target different regions of the BARP mRNA (Fig. 8 A, a) were stably transfected into PC12 cells, and two independent clones for each shRNA were selected. Expression levels of BARP in the six PC12 clones analyzed ranged between 10 and 25% of controls (Fig. 8 A, b and c), and VGCC Ca²⁺ current densities in these clones were between 50 and 140% higher than in controls (Fig. 8 A, d). The stimulatory effect of the shRNAs on VGCC activity could be suppressed by transfecting a rescue mouse BARP cDNA not targeted by the shRNAs (Fig. 8 A, d, red bars).

Endogenous BARP expression was also silenced by transient transfection of PC12 cells with a combination of two siRNAs (Fig. 8 B, a and b). Neurotransmitter release in these cells was analyzed to determine whether the Ca²⁺-evoked secretion was also affected by silencing BARP. No significant differences in the basal secretion of acetylcholine were apparent between control and BARP siRNA–transfected cells (Fig. 8, c). However, upon stimulation with high K⁺ to depolarize the plasma membrane and in turn activate VGCCs, an almost twofold higher secretion of the neurotransmitter was observed in BARP siRNA–transfected PC12 cells as compared with controls. These results are thus consistent with a negative regulatory role for BARP on VGCC activity and Ca²⁺-regulated secretion.

The biochemical and functional data presented establish BARP as a novel VGCC regulatory protein that exerts its effect by binding to Ca_vβ subunits and thereby interferes with the association of Ca_vβ with Ca_vα1, leading to the inhibition of Ca²⁺ channel activity. Two domains in BARP associate with Ca_vβ. Mutation of domain I to abolish its interaction with Ca_vβ allows BARP to associate, via domain II, with Ca_vβ and Ca_vα1 to form a ternary complex. This ternary complex is not detected in the presence of a functional domain I, likely because domain I prevents a stable association between Ca_vα1 and Ca_vβ, and could thus in vivo be short lived or transient. Domain I is predicted to acquire an α-helical structure that interacts with the ABP in Ca_vβ. In vitro, the binding between domain I and the ABP occurs in the nanomolar range and is of lower apparent affinity than the binding of the AID to Ca_vβ, which ranges from 2 to 54 nM, depending on the particular Ca_vα1 and Ca_vβ species (De Waard et al., 1995; Bell et al., 2001; Cantí et al., 2001; Geib et al., 2002; Opatowsky et al., 2003). In cells, however, evidence suggests that the Ca_vα1 and Ca_vβ interaction has a lower affinity and is reversible (Hidalgo et al., 2006; Buraei and Yang, 2010). A local concentration of BARP (e.g., domain I) could thus be sufficiently high to modulate the Ca_vα1 AID–Ca_vβ interaction. Alternatively, a cooperative binding of domains I and II could increase the affinity of BARP for the Ca_vβ, and/or domain II binding could alter the conformation of Ca_vβ and thereby lower the affinity of Ca_vβ for the AID (Fig. S5 D). Consistent with this hypothesis, cellular overexpression of BARP abolished the association between Ca_vβ and Ca_vα1 only if domain II–mediated binding was preserved. Posttranslational modifications of BARP or a binding protein may also regulate the cooperation between domains I and II for Ca_vβ association. The region in Ca_vβ that interacts with domain II has not been identified but could involve the SH₃ and/or HOOK domain because they show the highest divergence between Ca_vβ1a, which does not bind domain II, and the other Ca_vβ isoforms (Fig. S3 G).

In the hippocampus and Purkinje cells, in which BARP is expressed, Ca_vβ1 is mostly present in soma and dendrites, whereas Ca_vβ3 and Ca_vβ4 are found in axons and other parts of the neurons (Obermair et al., 2010). Interestingly, in the cerebrum and cerebellum, BARP coprecipitated with Ca_vβ3 and Ca_vβ4 but not, or only to a lesser extent, with Ca_vβ1, suggesting that BARP may associate with specific Ca_vβ subunits in particular subcellular domains of neurons. For instance, BARP colocalized with BDNF, indicating its presence on dense core vesicles (Dieni et al., 2012). Intriguingly, during mouse development, BARP expression in the brain peaks between E18 and P7, a period crucial for neural circuit formation and synaptogenesis (Ullian et al., 2004; Christopherson et al., 2005).

Ca_vβ subunits facilitate surface expression of L-type VGCC by preventing ER-associated protein degradation of Cav1.2 (Altier et al., 2011). In tsA201 and PC12 cells, BARP inhibits VGCC activity without significantly altering the distribution of the Ca_vα1 subunit between the cell surface and intracellular compartments and is thus unlikely affecting ER exit and stability of Ca_vα1. However, it is well documented that Ca_vβ plays an important role in modulating, mainly via the IS6-AID linker, not only VGCC trafficking but also gating, including voltage-dependent activation, inactivation, and facilitation kinetics (Buraei and Yang, 2010). For instance, RGK GTPases, through association with Ca_vβ and/or Ca_vα1, not only regulate cell surface expression of VGCCs but also directly modulate currents of the channel at the plasma membrane by lowering channel opening probabilities and/or immobilizing VGCC voltage sensors (Fan et al., 2010; Yang et al., 2010). The time course of the macroscopic current in the presence of BARP was not modified, suggesting that single channel conductance, opening probability, or sensor movement rather than the kinetic or activation threshold is affected. More detailed electrophysiological studies of single channel dynamics will be required to elucidate how BARP affects VGCC activity. Although BARP does not alter surface expression of the Ca_vα1 subunit in tsA201 and PC12 cells, we cannot rule out an effect on VGCC trafficking in other cell types and/or under certain conditions.

One possible function of BARP could be that of an acceptor/donor for the reversible transfer of Ca_vβ from/to the Ca_vα1 subunit. Such a reversible transfer could provide an attractive mechanism to retain the Ca_vβ in the active zone to allow for the rapid modulation of VGCC activity. Thus, BARP may provide the first example of a negative modulation though the displacement of Ca_vβ from the Ca_vα1 subunit. The presynaptic protein Rim1 also interacts with Ca_vβ to anchor it to secretory vesicles (Kiyonaka et al., 2007; Gandini et al., 2011; Weiss et al., 2011), but in contrast to BARP, Rim1 stimulates Ca²⁺-dependent secretion by preventing voltage-dependent VGCC inactivation (Kiyonaka et al., 2007).

BARP is abundant in the brain and pancreas, in which Ca²⁺-regulated exocytosis through activation of L-type VGCC plays important roles in the release of neurotransmitters and hormones. Calcium channelopathies are congenital or noninherited muscular, neurological, and cardiac diseases associated with the gain or loss of VGCC function (Bidaud et al., 2006). Ca²⁺ channel inhibitors represent one of the most active areas of pharmacological drug development. Overexpression of BARP, or peptides encoding domains I and/or II, may be used as Ca²⁺ channel modulators. The identification of BARP may thus open new avenues for the design of novel therapeutic VGCC blockers.

The yeast two-hybrid screens were performed as follows. A yeast strain L40 (MATa trp1 leu2 his3 LYS2::lexA-HIS3 URA3::lexA-lacZ) was transformed with a derivative of pBTM116 encompassing Ca_vβ3 subunit amino acid residues 50–484 fused to the LexA DNA-binding domain. A mouse MIN6 cell cDNA library was then screened, and after histidine selection, positive clones were further confirmed to be true positives by β-galactosidase activity measurement. Eight positive clones that presented fragments of BARP cDNA were found. A BARP alanine mutagenesis scan for aa 280–485 was assessed by substituting three by three the amino acids of BARP to alanine and processed for β-galactosidase activity measurements using paper filters stained with 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (Béguin et al., 2001). Full-length mouse BARP was isolated by conventional screening of MIN6 cDNA libraries with a partial mouse cDNA. Sequence analysis of EST clones from mouse, rat, and human (see legend of Fig. S1 A) confirmed that BARP is translated from a 3-kb transcript. Rat Ca_vβ1b, Ca_vβ21 (e.g., β2A), Ca_vβ3, and Ca_v1.2 were originally cloned in S. Seino’s laboratory, and the Ca_v2.1 and Ca_v2.2 cDNAs were a gift from T.W. Soong (National University of Singapore, Singapore) and T.P. Snutch (University of British Columbia, Vancouver, British Columbia, Canada). Mouse Ca_vβ4a was obtained from the I.M.A.G.E. Consortium (4501980). Epitope-tagged constructs (Flag, HA, Myc, GST, and EGFP) as well as deletion and point mutants were generated by PCR-based methods and subcloned into the pME18S vector containing an SRα promoter. The internally HA-tagged Ca_v1.2 has been described elsewhere (Altier et al., 2002; Béguin et al., 2006). In brief, the HA sequence was introduced in the S5-H5 loop in position 697 of rabbit Ca_v1.2. The amino acid sequence is defined as MQTRH-HA-MQTR (MQTR are amino acids of Ca_v1.2, which were duplicated, underlined is an additional amino acid, and the HA sequence is without the first methionine). Northern blot analysis was performed under standard stringency hybridization and washing conditions using mouse and human BARP cDNA probes.

Polyclonal anti-BARP Ab 72 was custom made (BioGenes) by injecting rabbits with a BARP peptide (¹³¹NEAALFEQSRK¹⁴¹) conjugated to hemocyanin and affinity purified. A GST-BARP fusion protein (aa G125-A698) was injected into mice to generate mAb 12B1 and 8B2. Epitope mapping using truncated forms of BARP located the epitopes recognized by 8B2 and 12B1 to a region between aa 380 and 698.

COS-1, standard (CRL-1721), and Tet-On PC12 cells as well as BHK and tsA201 cells were grown and transiently transfected with WT or mutated cDNAs using Lipofectamine (LTX; Invitrogen) and jetPRIME (Polyplus Transfection) for biochemical and immunofluorescence experiments, respectively (Béguin et al., 2001, 2005b, 2006, 2007; Mahalakshmi et al., 2007a,b). A plasmid carrying a hygromycin resistance gene and the BARP cDNA downstream of a cytomegalovirus promoter was transfected into PC12 cells using Lipofectamine LTX. PC12 clones stably overexpressing BARP were selected and maintained in 0.2 µg/µl hygromycin. BHK cells expressing functional Ca²⁺ channels reconstituted by expression of rabbit cDNAs for the different subunits have been characterized (Mori et al., 1991; Fujita et al., 1993). Silencing of BARP in PC12 cells was achieved using the pSUPER RNAi System (Oligoengine) system and shRNAs A (5′-TTCTCAAGTCCATATACGG-3′), B (5′-TAGTGTTGATTGTCCTCCT-3′), and C (5′-CTTTGTAGCAACTGTACCT-3′) with selection and maintenance of stable clones in 400 µg/ml G418. Alternatively, PC12 cells were transiently transfected using DharmaFECT (Thermo Fisher Scientific) with siRNAs A (5′-GGAUUUCCAUCACCUCAAG-3′) and B (5′-CAUGCUGACUUCAUUCAAU-3′) designed by Thermo Fisher Scientific and used for analysis 48 h after transfection. For cell surface expression analysis, PC12 cells were electroporated (1,410 V at 30 ms) using the Neon Transfection System (Invitrogen). Primary hippocampal neurons were purchased from Cambrex and cultured according to the manufacturer’s instructions. Cerebellar primary cultures were prepared as previously described (Launey et al., 2004) with substitution of B27 by the N21 supplement (Chen et al., 2008). In brief, cerebella were removed from 19-d Wistar rat fetuses, minced in Ca²⁺/Mg²⁺-free Hank’s saline (Gibco), and digested with 0.01% trypsin (15 min at 37°C). After trituration, the suspension was plated at 5,000 cells/mm² on 18-mm glass coverslip (Thermo Fisher Scientific) coated with poly-l-lysine and poly-l-ornithine. Culture medium (37°C at 5% CO₂) consisted of glutamate/aspartate-free DMEM/F12 supplemented with 10 mg/ml bovine insulin, 100 mg/ml BSA, 1 mg/ml gentamycin, 200 mg/ml glutamine, 100 mg/ml human apotransferrin, 40 nM progesterone, 100 nM putrescine, 30 nM sodium selenite, 500 pg/ml triiodothyronine, 3% heat-inactivated horse serum (Gibco), and 25% of astrocyte-conditioned medium (Sumitomo Corp.), renewed by half twice a week. Cells were used at 3–4 wk. All animal experimentation was approved by the RIKEN or Institute of Molecular and Cell Biology Institutional Animal Care and Use Committees. Cells were maintained in culture for 3–4 wk before immunofluorescence microscopy experiments. Tunicamycin treatment was performed by incubating COS-1 cells immediately after transfection for 2 d with 10 µg/ml tunicamycin (Sigma-Aldrich).

Cell homogenates were prepared in lysis buffer (50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 1 mM MgCl₂, and 0.5% Triton X-100) supplemented with protease inhibitors and used for coimmunoprecipitation and WB analysis as previously described (Béguin et al., 2006). For better resolution of higher molecular weight proteins, some samples were run on Tris-acetate gels (7%) according to the manufacturer’s instructions (Invitrogen). Rat mAb to HA (Roche), mouse mAb to Flag (M2; Sigma-Aldrich), Ca_vα1 (NeuroMab), GAPDH, MAP2, calbindin (EMD Millipore), synaptotagmin (Stressgen), GST (Santa Cruz Biotechnology, Inc.), Ca_vβ3 (Alomone Labs), GFAP (Sigma-Aldrich), and Ab to BARP (72 and 8B2) were used. Rat cerebrum and cerebellum were manually dissociated and immediately homogenized in the same lysis buffer using 15 strokes of a glass Teflon homogenizer followed by one freeze/thaw cycle. Insoluble material was removed by centrifugation. Brain, skeletal muscle, and heart lysates were purchased from Zyagen Laboratories. At least three independent experiments were performed, and representative examples are shown.

BARP was synthesized in vitro using the quick-coupled transcription/translation system (TnT; Promega) according to the manufacturer’s protocol.

For dissociation experiments, AID (⁴⁴⁵AKARGDFQKLREKQQLEEDLKGALDAATQAED⁴⁷⁶) and BARP domain I (⁴²²SYRDLWSLRASLELHAATASD⁴⁴²) peptides were synthesized (Mimotopes). Ca_vβ3 was pulled down from cell lysates with a GST-AID (aa A445–D476) fusion protein. After extensive washing in lysis buffer (50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 1 mM MgCl₂, and 0.5% Triton X-100) supplemented with protease inhibitors to remove unbound Ca_vβ3, increasing concentrations of peptides (30, 60, 300, and 600 nM) were added in lysis buffer. After a 4-h incubation on ice, 10% of the supernatant containing any dissociated Ca_vβ3 was analyzed by SDS-PAGE and WB. In the absence of competitive peptides, the GST-AID–Ca_vβ3 subunit complex remained stable over the 4-h incubation period. At least three independent experiments were performed, and representative examples are shown.

N-Flag–Ca_vβ subunits and BARP or N-Myc–BARP overexpressed in PC12 and COS-1 cells were stained with mouse anti-Flag (M2; Sigma-Aldrich) and rabbit anti-BARP Ab 72 or a rabbit Ab to Myc (Sigma-Aldrich) followed by Cy3-labeled donkey anti–rabbit IgG (Jackson ImmunoResearch Laboratories, Inc.) and Alexa Fluor 488 goat anti–mouse IgG (Molecular Probes) secondary Ab as previously described (Béguin et al., 2005b). PC12, primary hippocampal, or cerebellar cells were stained with rabbit Ab to MAP2 (EMD Millipore), calbindin (EMD Millipore), and GFAP (Sigma-Aldrich), BDNF (EMD Millipore), or with mouse mAb to synaptotagmin (Stressgen). Rat brains were mechanically cut into 0.3-µm-thick sections and incubated in 10% TCA for 30 min before staining. Mouse pancreas was fixed in 4% PFA and subjected to standard ethanol/xylene processing, embedded into paraffin, cut into 0.5-µm sections, and rehydrated in graded ethanol. Slices were then washed three times with 30 mM PBS-glycine and incubated 1 h in a blocking buffer (10% goat serum, 2% BSA, and 0.4% Triton X-100), and Ab incubations were performed as described in this paragraph. For the 72 BARP Abs, antigen retrieval was achieved by incubating fixed slices for 30 min in 10 mM Tris-HCl, pH 9.0, 1 mM EDTA, and 0.05% Tween 20 at 80°C followed by three washes in PBS before blocking. Antigen retrieval in pancreas sections was performed in citrate buffer (10 mM sodium citrate, pH 6, and 0.05% Tween 20) in an autoclave (Retriever 2100; Prestige Medical). Mouse mAb to glucagon (Abcam) or guinea pig Ab to insulin (Abcam) was used to stain α and β cells, respectively, in pancreatic islets. Labeled specimens were mounted in FluorSave (Vector Laboratories) or ProLong Gold (Invitrogen) and visualized using a confocal microscope (LSM 510 Meta [Carl Zeiss] or FluoView FV1000 [Olympus]). Antigen preabsorption experiments were performed by incubating GST-BARP (G125-A698) fusion protein (1 µg) linked to Sepharose beads with anti-Myc and anti-BARP Ab (72, 12B1, or 8B2) in lysis buffer for 2 h at 4°C. After a short centrifugation, the supernatant was collected and mixed with blocking buffer (1:1) before being applied to PFA-fixed cells expressing N-Myc–BARP or cerebellum slices fixed in 10% TCA.

To detect cell surface expression of N- or C-terminally tagged BARP or a Ca_v1.2 carrying a tag in an extracellular loop (Béguin et al., 2001; Altier et al., 2002), we took advantage of the fact that Abs to the tags when added to nonpermeabilized intact cells only bind and label the cells if the tag is exposed on the cell surface. Thus, intact transfected COS-1, tsA201, or BHK cells were incubated with 2 µg/ml rat anti-HA (Roche) and/or 1 µg/ml rabbit anti-Myc Ab (Sigma-Aldrich) for 1 h and then washed twice in ice-cold PBS before fixation. In some experiments, the cells were subsequently permeabilized and incubated with a different Ab to detect an intracellular protein such as the Ca_vβ subunits and EGFP-Ca_v1.2 in the same cells. Visualization by immunofluorescence microscopy was as described in the previous section.

Quantification of surface channel expression was performed as follows: a random 540 × 540–µm area was scanned for each coverslip at 488 nm (EGFP-Ca_v1.2 cell expression), 546 nm (Ca_v1.2 surface expression), and 633 nm (N-Myc–BARP surface expression) using a microscope (Eclipse Ti; Nikon; 20×, 1.0 NA oil objective) with a motorized stage with NIS element AR software version 4.0 (Nikon). Cells expressing Ca_v1.2 alone or with Ca_vβ3 were first selected blindly based on EGFP fluorescence only (reflecting Ca_v1.2 total expression) and segmented, without looking at the surface expression. The mask thus defined was then applied to all fluorophores, and mean pixel intensity for each cell was calculated, yielding Ca_v1.2 total expression (488 nm) and relative surface localization (546/488 nm). Cell surface expression of Ca_v1.2 with Ca_vβ3 together with BARP WT or domain I and II mutated was performed by first blindly selecting cells expressing BARP (633 nm), without looking at the total or surface expression of EGFP-Ca_v1.2. Again, the thus defined regions of interest were used as a mask to measure mean pixel intensity for each cell and each channel. To ascertain that quantification is not affected by Cav1.2 cellular expression, a cutoff corresponding to half of the mean pixel intensity of EGFP-Ca_v1.2 in the absence of BARP was applied.

Fluorescence colocalization was quantified using the ImageJ/Fiji (National Institutes of Health; Schneider et al., 2012) with the plugin Coloc 2 (version December 2011), after confocal image acquisition with 62-nm/pixel spatial oversampling (FluoView FV1000 confocal with 100×, 1.4 NA oil objective, software version 4.0). For each cell, five independent optical sections were acquired at low noise (Kalman filter 5), sequentially for each channel, with negligible bleed through between channels. For accurate sampling of the different cell compartments, the five sections were projected into single images (one per channel) before background subtraction and colocalization analysis. Region of interest was defined automatically by applying a logical OR operation the two segmented channels, to select pixels that are above background in each channel. The colocalization index is reported as the thresholded Pearson’s correlation coefficient, as it is more robust than other metrics (Adler and Parmryd, 2010). For each condition, at least 13–30 cells from three independent replicates were analyzed.

Whole-cell patch clamp recordings were made on BHK, PC12, and tsA201 cells using bath solution containing 40 mM Ba²⁺ at 37°C as previously described (Béguin et al., 2001, 2006). In brief, to obtain expression of EGFP and BARP, the WT or mutated BARP cDNAs were transfected into the P/Q or N type–expressing BHK cell lines or PC12 cells using the pCMS-EGFP vector (Takara Bio Inc.). Cells expressing the EGFP were selected for measurements. For expression in the tsA201 cells, WT and mutated BARP cDNA were cloned into a pCMS-EGFP vector in which EGFP had been replaced by mCherry and cotransfected with a pCMS-EGFP vector carrying the cDNAs for Ca_v1.2 and Ca_vβ3 separated by an internal ribosomal entry site. Cells expressing EGFP and mCherry were selected for measurements. For each cell, current density (I_Ba) was calculated by dividing the total current by the membrane capacitance. The holding potential was –60 mV, and test pulses of 400 ms at potentials between −40 and 60 mV in steps of 10 mV were applied every 4 s. The membrane potential was measured by the perforated patch clamp method in the current clamp mode as previously described (Gonoi et al., 1994).

Secretion of transfected human growth hormone (hGH) or acetylcholine by PC12 cells were assayed as described previously (Béguin et al., 2001; Kiyonaka et al., 2007). In brief, PC12 cells were cotransfected with pXGH5 vector (Nichols Institute) containing an SRα promoter driving WT or mutated BARP, using Lipofectamine (Invitrogen) according to the manufacturer’s instructions. After 3 d, PC12 cells were washed with a physiological salt solution (PSS; 140 mM NaCl, 4.7 mM KCl, 2.5 mM CaCl₂, 1.2 mM MgCl₂, 1.2 mM KH₂PO₄, 20 mM Hepes, pH 7.4, and 11 mM glucose) and incubated for 10 min with a high K⁺ solution (PSS containing 60 mM KCl and 85 mM NaCl) or a low K⁺ solution (PSS containing 4.7 mM KCl and 140 mM NaCl). Growth hormone was measured with a colorimetric immunoassay kit (Roche). Acetylcholine secretion in PC12 cells (Kiyonaka et al., 2007) was assayed with the similar basic procedure with the exception that pXGH5 was replaced by pEFmChAT cDNA, and acetylcholine release was measured using HPLC with electrochemical detection (HTEC-500; EiCOM).

In silico modeling, molecular dynamics, and energy refinements were performed using the Sybyl 7.2 software package (Tripos, Inc.). The Ca_vβ subunit crystal structure (Protein Data Bank accession no. 1VYT) was used as a template to dock the BARP domain I. This domain was modeled as an α helix in a reverse orientation. Amino acid W427 of BARP was positioned similarly to residue W440 of AID followed by molecular dynamics simulations (1,000 fs with 1-fs steps at 300 K) between residues of the Ca_vβ subunit and domain I within 6 Å. The lowest energy conformation was then obtained by energy minimization using Powell’s method (Fletcher and Powell, 1963). The α-helical model of domain I shown in Fig. S3 D was generated using Phyre2 (Kelley and Sternberg, 2009).

Statistical significances were tested using unpaired and paired Student’s t tests, and results were expressed as means ± SEM for the indicated n values. For coprecipitation, pull-down, WB, and immunofluorescence microscopy experiments, at least three independent experiments were performed, and a representative example is shown.

Fig. S1 shows alignment of BARP protein sequences of different species, evidence for the specificity of antibodies to BARP, and the assignment of the initiation methionine and glycosylation of BARP. Fig. S2 shows expression of BARP during mouse development, evidence for the specificity of the antibodies for immunohistochemistry localization of BARP in brain and pancreas, localization of overexpressed BARP and overexpressed synaptotagmin I, and membrane localization of Ca_vβ subunit isoforms by BARP. Fig. S3 shows the identification of domains and amino acids in BARP that mediate the interaction with Ca_vβ and molecular dynamics modeling of the interaction of BARP domain I with the ABP of Ca_vβ. Fig. S4 characterizes the effect of BARP on the association of the Ca_vβ with the Ca_vα1 subunit and the association of BARP with different Ca_vβ subunit isoforms in brain. Fig. S5 analyzes the effect of BARP on VGCC activity and surface expression and presents a working model for BARP as a Ca_vβ-anchoring protein.

We gratefully acknowledge the technical support from the Support Unit for Bio-Material Analysis, Brain Science Institute Research Resources Center for DNA sequencing and cell sorting and the Institute of Molecular and Cell Biology DNA Sequencing Facility for DNA sequencing.

This work was supported by the Agency for Science, Technology and Research, Singapore and the RIKEN Brain Science Institute, Japan.

The authors declare no competing financial interests.