Auxiliary Ca2+ channel β subunits (CaVβ) regulate cellular Ca2+ signaling by trafficking pore-forming α1 subunits to the membrane and normalizing channel gating. These effects are mediated through a characteristic src homology 3/guanylate kinase (SH3–GK) structural module, a design feature shared in common with the membrane-associated guanylate kinase (MAGUK) family of scaffold proteins. However, the mechanisms by which the CaVβ SH3–GK module regulates multiple Ca2+ channel functions are not well understood. Here, using a split-domain approach, we investigated the role of the interrelationship between CaVβ SH3 and GK domains in defining channel properties. The studies build upon a previously identified split-domain pair that displays a trans SH3–GK interaction, and fully reconstitutes CaVβ effects on channel trafficking, activation gating, and increased open probability (Po). Here, by varying the precise locations used to separate SH3 and GK domains and monitoring subsequent SH3–GK interactions by fluorescence resonance energy transfer (FRET), we identified a particular split-domain pair that displayed a subtly altered configuration of the trans SH3–GK interaction. Remarkably, this pair discriminated between CaVβ trafficking and gating properties: α1C targeting to the membrane was fully reconstituted, whereas shifts in activation gating and increased Po functions were selectively lost. A more extreme case, in which the trans SH3–GK interaction was selectively ablated, yielded a split-domain pair that could reconstitute neither the trafficking nor gating-modulation functions, even though both moieties could independently engage their respective binding sites on the α1C (CaV1.2) subunit. The results reveal that CaVβ SH3 and GK domains function codependently to tune Ca2+ channel trafficking and gating properties, and suggest new paradigms for physiological and therapeutic regulation of Ca2+ channel activity.
Introduction
High voltage–activated calcium channels, comprised of pore-forming (α1) and auxiliary (α2δ, CaVβ, and sometimes γ) subunits, transduce electrical signals into Ca2+ fluxes that regulate essential biological processes such as neurotransmission, muscle contraction, and gene expression (Tsien and Tsien, 1990; Catterall, 2000). Auxiliary CaVβ subunits (CaVβ1–CaVβ4) exert a powerful influence on cellular Ca2+ signaling by regulating multiple functional properties of α1 subunits (α1A–α1F, α1S; CaV1.1–1.4, CaV2.1–2.3) (Ertel et al., 2000), including trafficking the channel complex to the plasma membrane (Chien et al., 1995; Brice et al., 1997; Gao et al., 1999), producing hyperpolarizing shifts in the voltage dependence of channel activation (Singer et al., 1991; Perez-Reyes et al., 1992; De Waard et al., 1994), increasing channel open probability (Po) (Gerster et al., 1999; Hullin et al., 2003), and producing characteristic effects on the rate of channel inactivation (Wei et al., 2000; Stotz and Zamponi, 2001; Colecraft et al., 2002; Dolphin, 2003; Takahashi et al., 2003). Unraveling how CaVβs regulate these myriad effects is fundamental to understanding the operation of Ca2+ channels in both physiological and disease states, as well as to efforts to target the α1–β interaction to generate a new class of therapeutically useful Ca2+ channel inhibitors (Young et al., 1998). Nevertheless, mechanistic understanding of how CaVβs regulate multiple Ca2+ channel properties is lacking, despite years of intense research focus. One ambiguous area is how structural determinants on CaVβ combine to regulate multiple functional properties of α1 subunits. Fortunately, recent high resolution crystal structures of CaVβs (Chen et al., 2004; Opatowsky et al., 2004; Van Petegem et al., 2004) provide new opportunities for increased mechanistic understanding of how CaVβs tune Ca2+ channel behavior.
The CaVβ crystal structures directly reveal src homology 3 (SH3) and guanylate kinase-like (GK) motifs within two domains conserved among distinct CaVβ isoforms (C1 and C2; Fig. 1, A and B) (De Waard et al., 1994; Birnbaumer et al., 1998). These structural features establish a link between CaVβs and the membrane-associated guanylate kinase (MAGUK) family proteins that play a vital role in organizing intracellular signaling pathways; SH3–GK is an essential functional module conserved among MAGUKs (Anderson, 1996; Craven and Bredt, 1998). The CaVβ SH3–GK core is sufficient to reconstitute the bulk of CaVβ functional effects, including promoting membrane transport of pore-forming α1 subunits (Gao et al., 1999) and robustly increasing recombinant whole-cell Ca2+ channel currents when coexpressed with α1 (Opatowsky et al., 2003). Therefore, this structural unit holds the key for mechanistic insights into how the α1–β interaction regulates multiple channel properties.
How does the two-pronged nature of the SH3–GK module impact functional operation of CaVβs? For nearly a decade, it was believed that the second conserved domain of CaVβ (C2), now known to constitute a GK domain, was primarily responsible for channel regulation (De Waard et al., 1994; Walker and De Waard, 1998). Consistent with this idea, high-resolution crystal structures of CaVβ subunits in complex with a conserved α1 interaction domain (AID) peptide, indicate that the AID interacts exclusively with the GK domain (Chen et al., 2004; Opatowsky et al., 2004; Van Petegem et al., 2004). However, recent functional experiments implicate a more prominent role for the SH3 domain than previously suspected (Opatowsky et al., 2003; Takahashi et al., 2004; Yue, 2004; Maltez et al., 2005). Specifically, using a split-domain approach, we (Takahashi et al., 2004) showed that CaVβ2a fragments containing either the SH3 (NSH3) or GK (GKC) domains (Fig. 1 C) reconstituted wild-type CaVβ function only when both fragments were coexpressed with recombinant α1C in HEK 293 cells. Qualitatively similar results were obtained for split-domain CaVβ regulation of CaV2.1 and CaV1.2 in Xenopus oocytes (Opatowsky et al., 2003; Maltez et al., 2005). The unexpected importance of the CaVβ–SH3 domain in Ca2+ channel modulation raises new questions regarding how structural determinants on CaVβ cooperate to regulate channel properties. Do SH3 and GK domains act independently to reconstitute CaVβ functions, or does the SH3 domain act allosterically via interaction with GK? Is the CaVβ SH3–GK interaction necessary for channel modulation? Does CaVβ SH3 domain bind independently to the α1 subunit?
Here, we combine a split-domain approach with site-directed mutagenesis, electrophysiology, and FRET determination of protein interactions to investigate these issues. We find that the configuration of the SH3–GK interaction is a dominant determinant of CaVβ function. Subtle changes in the relative affinity and orientation of the CaVβ SH3–GK domain interaction produces significant effects on channel properties by preferentially preventing the increased-Po and shift in activation gating functions, whereas the channel trafficking role is retained. Moreover, selectively ablating the SH3–GK interaction disrupted both trafficking and gating modulation, indicating that the two motifs act codependently to reconstitute CaVβ function. The results shed new light on how the structural design of CaVβs relates to their regulation of multiple Ca2+ channel functions, and reveal an unexpected signaling versatility of the SH3–GK module.
Materials And Methods
DNA Cloning
α1C-GFP was described previously (Takahashi et al., 2004). Split-domain β2a constructs (NSH3, NSH3-V2, GKC, GKC[trunc]) were generated by PCR amplification using rat β2a/pGW1 (Perez-Reyes et al., 1992) as template, followed by subcloning into pcDNA4.1 (Invitrogen) using BamHI and EcoRI sites. The deletion mutant GKC[ΔPYDVV] was generated by overlap extension PCR, followed by subcloning into pcDNA4.1 using BamHI and EcoRI sites. To generate XFP-tagged channel subunit fragments, full-length cyan fluorescent protein (CFP) and yellow fluorescent protein (YFP) (minus the stop codon) were amplified by PCR and cloned into pcDNA4.1 using KpnI and BamHI sites. The various channel fragments were then PCR amplified and cloned in-frame to the 3′ end of the appropriate XFP using BamHI and EcoRI sites. All PCR constructs were verified by sequencing.
Cell Culture and Transfection
For electrophysiology experiments, low-passage number HEK 293 cells plated on coverslips in 10-cm tissue culture dishes were transiently transfected with 8 μg α1C-GFP, 8 μg each of the appropriate split-domain CaVβ fragment, and 3 μg T-antigen, using calcium-phosphate precipitation. To permit unambiguous identification of functional effects specific to the split-domain CaVβ2a fragments, α2δ was omitted from transfections. For FRET experiments, HEK 293 cells were plated in culture dishes with No. 0 glass coverslip bottoms (MaTek) and transiently transfected with 0.5–3 μg XFP-tagged proteins using Fugene (Roche).
Electrophysiology
Whole-cell recordings were performed at room temperature 2–3 d after transfection using an EPC8 or EPC10 patch-clamp amplifier (HEKA Electronics) controlled by Pulse software (HEKA Electronics). Micropipettes were fashioned from 1.5-mm thin-wall glass with filament (WPI Instruments); series resistance was typically 2–4 MΩ and compensated 50–70%. A standard I–V protocol with 20-ms step depolarization (−40 to +120 mV) was used to evoke currents from a −100-mV holding potential. Tail currents were measured at −50-mV repolarization potential. Currents were sampled at 25 kHz and filtered at 10 kHz. Leak and capacitive transients were subtracted by P/8 protocol. External solution contained (in mM) 140 TEA-MeSO3, 10 HEPES, 5 BaCl2 (pH 7.3). Internal solution contained (in mM) 135 Cs-MeSO3, 5 CsCl, 5 EGTA, 1 MgCl2, 4 MgATP, 10 HEPES (pH 7.3). Recordings were analyzed off-line using PulseFit software (HEKA Electronics) and Ms Excel. I-V relations for each cell were fit to the following equation:
where I is the whole-cell current density (pA/pF), G is the specific conductance (pA pF−1 mV−1), Vrev is the reversal potential (mV), V1/2 is the voltage of half-maximal activation (mV), and k is a slope factor (mV).
Normalized tail-activation data were fit to a double-Boltzmann function of the form
where G is the tail current amplitude, G100 is the tail current amplitude elicited by a 100-mV test-pulse depolarization, Flow is the fraction of the low-threshold component, V is the test-pulse potential, V1/2,low and V1/2,high, and klow and khigh are the half-activation potentials and slope factors for the low- and high-threshold components, respectively.
FRET Measurements
Detection of protein interactions in live HEK 293 cells was accomplished using the three-cube FRET algorithm (Erickson et al., 2001, 2003). Transfected cells were washed with Tyrode solution and mounted on an inverted Nikon TE300 Eclipse microscope equipped for epifluorescence. A 150-W Xenon arc lamp served as excitation light source, and was gated by a computer-controlled shutter. Epifluorescence emission signals from individual cells were measured with a photomultiplier tube, integrated and filtered (10 kHz) by a fluorometer, and digitized. Autofluorescence and background levels were assessed from averages of individual untransfected cells and subtracted from experimental values for each cube. For each cell, measurements were taken with the following three-cube FRET filter cubes (excitation, dichroic, emission): CFP (440, 455, 480), YFP (500, 525, 530), and FRET (440, 455, 535). FRET ratio (FR) was calculated as
where SCUBE(SPECIMEN) represents an intensity measurement, CUBE represents the filter cube, and SPECIMEN denotes whether the cell is expressing donor alone (D; CFP), acceptor alone (A; YFP), or both (DA). RD1 and RA1 are constants determined from cells expressing only CFP- or YFP-tagged molecules, respectively. Effective FRET efficiency (EEFF) was determined from
where E is the intrinsic FRET efficiency when particular XFP-tagged molecules associate with each other, Ab is the fraction of YFP-tagged molecules bound by CFP-tagged molecules, and the bracketed term is the ratio of the YFP and CFP molar extinction coefficients scaled for the FRET cube excitation filter.
Confocal Imaging
XFP-tagged fusion proteins were transiently transfected into HEK 293 cells cultured in MaTek tissue culture dishes. Confocal images were acquired 48 h post-transfection using an Olympus Fluoview laser scanning confocal microscope.
Statistics
Experimental parameters for various constructs were compared with those obtained with NSH3+GKC, and statistically significant differences (defined as P < 0.05) determined by the Student's t test. Linear regression was conducted using Excel with intercept set to zero. Multiple-regression analysis was performed in MS Excel to determine significant difference in fitted slopes. For ON-gating charge measurements, currents were leak subtracted using smooth curves fitted to leak pulses in IgorPro and analyzed in PulseFit. Data are mean ± SEM.
Results
Location of Split Site in Split-domain CaVβ2a Proteins Is an Important Determinant of Functional Channel Reconstitution
CaVβs contain SH3 and GK motifs in two conserved domains (Fig. 1, A and B). Recently, using a split-domain approach, we found that CaVβ fragments containing either the SH3 (NSH3) or GK (GKC) domains (Fig. 1 C) could reconstitute wild-type CaVβ functions only when both fragments were coexpressed with recombinant α1C (Takahashi et al., 2004). Fig. 2 A shows robust whole-cell currents from channels reconstituted with α1C/NSH3/GKC that confirms the previous results. Moreover, we and others found that NSH3 and GKC (or similar split-domain proteins) can interact in trans (Opatowsky et al., 2003; McGee et al., 2004; Takahashi et al., 2004), raising the question of whether and how the CaVβ SH3–GK domain interaction is important for channel modulation (Takahashi et al., 2004). The most direct way to address this issue would be to selectively interfere with the CaVβ SH3–GK interaction while leaving the GK–AID interaction unchanged.
To explore the feasibility of this approach we initially truncated 84 residues from the amino terminus of GKC, generating GKC[trunc] (Fig. 1 C). The rationale for this strategy stemmed from information provided by the CaVβ crystal structures showing that the AID–GK interaction surface is spatially removed from the SH3–GK interface (Fig. 1 B) (Chen et al., 2004; Opatowsky et al., 2004; Van Petegem et al., 2004). Before probing the actual effect of the 84-residue deletion on the status of the CaVβ SH3–GK interaction, we first determined whether channels reconstituted with NSH3+ GKC[trunc] exhibited any gross functional distinctions compared with those formed with NSH3+GKC. Interestingly, whole-cell currents from NSH3+GKC[trunc]-reconstituted L-type channels were significantly smaller in amplitude (Fig. 2 B) compared with their NSH3+ GKC-reconstituted counterparts (Fig. 2 A). The deficit in current amplitude extended across the entire range of test pulse voltages, with no change in reversal potential between the two channel types as indicated by population current density (Jpeak) versus voltage (J–V) plots (Fig. 2 C). Beyond the deficits in whole-cell current amplitude, normalized tail activation plots demonstrated that NSH3+GKC[trunc] channels displayed a right-shifted voltage dependence of channel activation compared with channels reconstituted with NSH3+GKC (Fig. 2 D), revealing an inability to recapitulate this functional hallmark (i.e., a hyperpolarizing shift in channel activation) of CaVβ regulation. Overall, NSH3+GKC[trunc] channels displayed an over threefold decrease in peak current density, a 6-mV rightward shift in V1/2 of activation, and a significant decrease in the steepness of voltage dependence of channel activation, compared with channels reconstituted with NSH3+GKC (Table I).
What molecular mechanisms underpinned the functional differences between channels reconstituted with NSH3+GKC and NSH3+GKC[trunc], respectively? Mechanistic insights were gained from considering that whole-cell current (I) is related to microscopic channel properties by the relation I = N × i × Po, where N is the number of channels at the membrane, and i is the unitary current amplitude. CaVβs do not affect i (Hullin et al., 2003), as might be expected from cytosolic auxiliary proteins that do not alter the permeation pathway of the channel. Hence, differential alterations of N and Po likely underlie the distinctions in how NSH3+GKC and NSH3+GKC[trunc] reconstitute whole-cell currents. Fortunately, further analyses of our whole-cell experimental records permit useful comparative evaluation of the relative effects of NSH3+GKC and NSH3+ GKC[trunc] on N and Po. The time integral of the ON gating current isolated at the reversal potential (Fig. 2, A and B, bottom traces) yields the ON gating charge (QON), which provides a convenient index of N (QON = Nq; Wei et al., 1994; Takahashi et al., 2004). Exemplar traces (Fig. 2, A and B, bottom traces) hinted that despite the clearly divergent whole-cell current amplitude (as reported by tail current size), gating-current size was quite similar between NSH3+GKC and NSH3+ GKC[trunc] channels. This impression was confirmed in the population data (Fig. 2 E). As previously reported (Takahashi et al., 2004), neither NSH3 nor GKC alone could efficiently transport α1C to the membrane, whereas NSH3+GKC restored trafficking to wild-type CaVβ2a levels, as gauged by QON measurements. As expected, GKC[trunc] alone was also unable to rescue channel trafficking. The surprising new result was that NSH3+GKC[trunc] also restored N (Fig. 2 E), belying its modest effects on whole-cell current amplitude (Fig. 2 C). This discrepancy suggested that NSH3+GKC and NSH3+GKC[trunc] might have rather divergent effects on channel Po. We sought direct confirmation of this difference using a slope analysis method (Fig. 2 F). Here, the regression line slope of a plot of tail current amplitude obtained at the reversal potential (typically +50 mV; Itail) versus QON provides a convenient metric for relative Po (Wei et al., 1994; Takahashi et al., 2004). By this analysis, channels reconstituted with NSH3+ GKC[trunc] displayed a lower Po, as indicated by a significantly shallower slope compared with NSH3+GKC (Fig. 2 F; slope = 16.3 ± 3.2 pA/fC [95% CI] for NSH3+ GKC, ▵; slope = 5.7 ± 2.1 pA/fC for NSH3+ GKC[trunc], ▴; P < 0.001 by multiple regression). Therefore, the deficit in whole-cell current amplitude obtained with NSH3+GKC[trunc] is due specifically to an inability of this split-domain pair to reconstitute the increased-Po function of CaVβ.
Overall, the functional differences between NSH3+ GKC and NSH3+GKC[trunc]-reconstituted channels could be consistent with one of two competing hypotheses. The first stems from the fact that NSH3+GKC[trunc] lacks 81 amino acids (spanning the latter region of the first conserved domain [C1] and the entire alternatively spliced second variable domain [V2]) compared with either wild-type β2a or NSH3+GKC (Fig. 1). Hence, the lack of an effect of NSH3+GKC[trunc] on activation gating and Po could simply reflect an essential role of this absent region in transducing these gating properties. Alternatively, the functional distinctions between NSH3+GKC and NSH3+GKC[trunc] could be directly due to variations in the CaVβ SH3–GK domain interaction between the two split-domain configurations. To discriminate between these two hypotheses, we generated a new SH3-containing split-domain CaVβ2a variant (termed NSH3-V2) that included the first three domains of β2a (Fig. 1 C). Accordingly, NSH3-V2+GKC[trunc] now spanned all of the CaVβ2a sequence, thereby addressing ambiguities arising from the loss of 81 amino acids in NSH3+GKC[trunc].
Functionally, electrophysiological data obtained from channels reconstituted with NSH3-V2+GKC[trunc] (Fig. 3) were essentially identical to that obtained with NSH3+GKC[trunc] (Table I). NSH3-V2+GKC[trunc] was unable to recover the bulk of whole-cell current amplitude (Fig. 3, A–C), did not recapitulate the hyperpolarizing shift in channel activation gating (Fig. 3 D), and did not restore the increased-Po property (Fig. 3 F, ▴; slope = 5.96 ± 1.69 pA/fC; P < 0.001 compared with NSH3-V2+GKC, ▵). However, NSH3-V2+GKC[trunc] fully restored the channel trafficking function (Fig. 3 E). By contrast, NSH3-V2+GKC recovered wild-type whole-cell current amplitude (Fig. 3, A–C) by reconstituting both the trafficking (Fig. 3 E) and enhanced-Po functions (Fig. 3 F, ▵), and also recapitulated the hyperpolarizing shift in the voltage dependence of channel activation (Fig. 3 D; Table I). Together, these results affirmed the unexpected result that a simple difference in the cut site of split-domain CaVβ constructs can uncouple the trafficking and gating-modulation functions. Moreover, these results ruled out the explanation that NSH3+GKC[trunc] did not recover the gating-modulation functions simply because of the mere absence of an 81–amino acid central segment of CaVβ2a. This raised the idea that functional disparities between channels reconstituted with NSH3 (or NSH3-V2) and GKC or GKC[trunc], respectively, might be due to differences in the status of the CaVβ SH3–GK domain interaction. Accordingly, we next sought direct evidence for potential variations in the way in which GKC and GKC[trunc] interacted with NSH3.
Probing CaVβ Domain Interactions by Live-cell FRET Reveals Subtle Distinctions between GKC and GKC[trunc] in their Binding to NSH3
We adopted the optical FRET two-hybrid approach (Erickson et al., 2003) to probe for interactions between split-domain CaVβ constructs (Fig. 4). This approach offered the advantage of reporting on the status of split-domain CaVβ SH3–GK interactions in the relevant setting of live HEK 293 cells, permitting a direct correlation of binding properties and functional electrophysiological experiments. Coding sequences for enhanced CFP and enhanced YFP were fused upstream of those for NSH3 and the two GK domain constructs, respectively. Confocal images of live cells show the subcellular localization of the distinct XFP-tagged proteins upon their cotransfection in HEK 293 cells (Fig. 4 A). CFP-NSH3 localized uniformly throughout the cell, with similar fluorescent intensity in the cytoplasm and the nucleus (Fig. 4 A, left). This distribution presumably reflects the fact that the relatively small size of CFP-NSH3 permits passive diffusion into the nucleus. By contrast, both YFP-GKC and YFP-GKC[trunc] were uniformly present in the cytosol, and excluded from the nucleus (Fig. 4 A, right). Importantly, the mean fluorescence intensities for YFP-GKC and YFP-GKC-[trunc] were not significantly different (Fig. 4 B; P = 0.07), indicating no dramatic differences in expression levels between these two proteins. This finding rules out the trivial possibility that the observed functional divergence between GKC and GKC[trunc] (Figs. 2 and 3) could be due to large differences in expression levels or protein stability between the two.
To evaluate interactions between split-domain CaVβ SH3 and GK proteins, we used a three-cube FRET algorithm (Erickson et al., 2001). This method enables calculation of FRET efficiencies despite excitation cross talk between CFP and YFP, and inescapable variability in the ratio of CFP- and YFP-tagged molecules in a given cell (Erickson et al., 2001, 2003). The three-cube FRET method permits sensitive detection of protein interactions as measured by the FRET ratio (FR; the fold increase in YFP fluorescence due to FRET), or effective FRET efficiency (EEFF): FR = 1 indicates no FRET, whereas FR > 1 indicates detection of FRET, and hence, an interaction between CFP- and YFP-tagged proteins. Application of this analysis to cells cotransfected with CFP-NSH3 and YFP-GKC yielded FR = 2.8 ± 0.1 (EEFF = 10.2 ± 0.4%; n = 25), indicating an interaction between these two proteins in HEK 293 cells (Fig. 4 C). This was reassuring given previous results that identical or similar split-domain CaVβ fragments interact, as determined by in vitro protein binding and yeast two-hybrid assays (Opatowsky et al., 2003; McGee et al., 2004; Takahashi et al., 2004). Control experiments featuring cells expressing CFP-NSH3+YFP, CFP+YFP-GKC, and CFP+YFP-GKC[trunc] displayed uniformly low FR (≤1.2) and EEFF (≤1%) values (Fig. 4 C), demonstrating the specificity of the assay. The important new result was that cells cotransfected with CFP-NSH3 and YFP-GKC[trunc] also exhibited robust FRET (FR = 3.7 ± 0.1; EEFF = 15.5 ± 0.4%, n = 26), indicating an interaction between these two CaVβ2a fragments. This result showed that the deficiency of NSH3+GKC[trunc] in rescuing whole-cell current and increased-Po was not due to an outright ablation of the CaVβ SH3–GK domain interaction.
We next sought to determine whether there were quantitative distinctions between the binding of GKC and GKC[trunc] to NSH3 that might account for the observed functional differences. To accomplish this we took advantage of the capability of the three-cube FRET method to yield an estimate of the in situ dissociation constant governing interaction between XFP-tagged molecules (Erickson et al., 2003). The approach exploits the inevitable variability in the expression ratios of CFP- and YFP-tagged molecules in distinct transfected cells. FR reflects both the intrinsic FRET coupling between interacting pairs of molecules and the fraction of YFP-tagged molecules bound by a CFP-tagged molecule (Ab). Fitting FRET measurements from different cells to a 1:1 binding model yields two parameters, the relative dissociation constant Kd,EFF (which is proportional to the actual Kd) and the maximal FR (FRmax) attained when Ab is unity (Erickson et al., 2003). Alternatively, FR can be plotted versus Dfree, a measure of the relative concentration of tagged donor molecules. Application of this analysis to CFP-NSH3+YFP-GKC (Fig. 4 D) and CFP-NSH3+YFP-GKC[trunc] (Fig. 4 E) yielded Kd,EFF values of 1868 and 3090, respectively, indicating a maintained high-affinity interaction in both split-domain SH3–GK pairs. By contrast, FRmax was significantly different in the two cases (2.94 for CFP-NSH3+YFP-GKC and 4.17 for CFP-NSH3+YFP-GKC[trunc]), suggesting distinctions in the relative geometric orientations of the SH3 and GK domains in the two split-domain configurations. Overall, these results show that GKC and GKC[trunc] display only subtle distinctions in how they bind NSH3. Nevertheless, it appears that even such small differences in the arrangement of the CaVβ SH3–GK interaction can produce disproportionately large functional effects on Ca2+ channel behavior.
Deletion of a Five-residue β5 Strand Selectively Ablates the CaVβ SH3–GK Domain Interaction while Preserving the GK–AID Association
To gain insight into whether the SH3 and GK domains could act independently to reconstitute any aspect of CaVβ function, it was necessary to selectively ablate the CaVβ SH3–GK interaction, while preserving the GK–AID interaction. Our previous work hinted that a point mutation in the CaVβ2a GK domain (P234R) effectively disrupted the intramolecular SH3–GK interaction (Takahashi et al., 2004). A similar result was obtained with a truncation that eliminated five carboxy-terminal residues from the CaVβ GK domain (McGee et al., 2004). However, neither of these mutations is suitable to determine the independence of the CaVβ SH3 and GK domains because the disruption of the SH3–GK interaction in both cases is not selective; both mutants are expected to disrupt the global GK fold, and thus also ablate the GK–AID interaction (De Waard et al., 1994; Chen et al., 2004; Opatowsky et al., 2004; Van Petegem et al., 2004). Likewise, a point mutation in the CaVβ2a SH3 domain (L93P) also disrupted the CaVβ SH3–GK interaction (McGee et al., 2004; Takahashi et al., 2004). However, the effect of this mutation on the ternary structure of the SH3 domain is unknown. It is also unknown whether the L93P mutation abolishes potentially relevant interactions of CaVβ SH3 with the α1 subunit. Hence, to date, there is a lack of CaVβ mutations that unambiguously permit assessment of whether the SH3 and GK domains can act independently to reconstitute CaVβ functions.
To address this deficiency, we focused attention on a five-residue β5 strand (that occurs immediately after the variable V2 domain and is formally a part of the SH3 domain) that the crystal structures suggested is crucial for keeping the two domains together by interacting with residues in both the SH3 and GK domains (Opatowsky et al., 2003; Chen et al., 2004; Van Petegem et al., 2004). We hypothesized that deleting the β5 strand would selectively ablate the CaVβ SH3–GK interaction. To test this, we deleted the β5 strand from GKC, generating GKC[ΔPYDVV] (Fig. 5 A). YFP-GKC[ΔPYDVV] was diffusely localized throughout the cytosol and excluded from the nucleus (Fig. 5 B, right), similar to the subcellular localization of YFP-GKC (Fig. 4 A). Three-cube FRET experiments revealed that YFP-GKC[ΔPYDVV] did not interact with CFP-NSH3 (FR = 0.9 ± 0.07, n = 6) indicating a successful ablation of the SH3–GK interaction (Fig. 5 C), and affirming the critical role of the β5 strand in keeping the two domains together.
It was important to our goals to ensure that deleting the β5 strand did not cause a global disruption of the GK domain. To address this, we probed for the ability of YFP-GKC and YFP-GKC[ΔPYDVV] to interact with CFP-α1C[I-II loop] by FRET. CFP-α1C[I-II loop] had an unexpected, but fortuitous subcellular localization (Fig. 5 D, left). The protein was present in the nucleus, likely reflecting its small size permitting passive diffusion across the nuclear membrane. More surprising was its localization to the plasma membrane, with little representation in the cytosol (Fig. 5 D, left). YFP-GKC, when cotransfected with CFP-α1C[I-II loop] no longer localized diffusely in the cytosol, as observed when it was cotransfected with CFP-NSH3 (Fig. 4 A), but instead was highly enriched at the plasma membrane (Fig. 5 D, top right). This relocation provided an immediate visual affirmation that YFP-GKC was interacting with CFP-α1C[I-II loop], a fact that was confirmed by a robust FRET signal (FR = 5.4 ± 0.2; EEFF = 25.6 ± 1.4%, n = 5) between the two proteins. Importantly, YFP-GKC[ΔPYDVV] behaved similarly to YFP-GKC with respect to sub-cellular relocation (Fig. 5 D, bottom right) and FRET signal (FR = 5.5 ± 0.2; EEFF = 26.1 ± 1.4%, n = 5) when cotransfected with CFP-α1C[I-II loop]. Given the precise three-dimensional architecture of the α1-binding pocket on CaVβs (Chen et al., 2004; Opatowsky et al., 2004; Van Petegem et al., 2004), these findings indicate that deleting the β5 strand does not disrupt the global folding of the GK domain. Therefore, the ΔPYDVV deletion mutation fulfills the crucial criteria of selectively ablating the CaVβ SH3–GK interaction while preserving the GK–AID association, enabling the first direct test of whether SH3 and GK domains could act independently to reconstitute aspects of CaVβ function.
Selective Ablation of the SH3–GK Interaction Abolishes the Synergism between Split-domain CaVβ SH3- and GK-containing Fragments
To determine the functional effects of ablating the CaVβ SH3–GK interaction we turned to whole-cell electrophysiological experiments. GKC[ΔPYDVV] expressed alone with α1C was deficient in its ability to rescue whole-cell currents (Fig. 6, A and B; Table I) and to induce a hyperpolarizing shift in activation gating (Fig. 6 C; Table I), despite its demonstrated normal binding to the α1C[I-II loop] (Fig. 5 E). These results confirm our previous finding that the GK domain is insufficient to reconstitute the bulk of CaVβ modulatory properties in HEK 293 cells (Takahashi et al., 2004). The new insight came with studying the effects of NSH3+GKC [ΔPYDVV] on Ca2+ channel currents. Channels reconstituted with NSH3+GKC[ΔPYDVV] behaved similarly to those obtained with GKC[ΔPYDVV] alone, with respect to whole-cell current amplitude (Fig. 6, D and E; Table I) and the voltage dependence of channel activation (Fig. 6 F; Table I). These results contrast sharply with the synergism observed between NSH3 and GKC (Takahashi et al., 2004), and revealed that the SH3–GK interaction is requisite for CaVβ modulation of channel trafficking and gating. Overall, these results are consistent with a co-dependent model for CaVβ SH3 and GK domains to modulate channel function.
NSH3 Can Interact Independently with the α1C Subunit
The lack of effect of NSH3 on channel properties when cotransfected with GKC[ΔPYDVV] gave reason to wonder whether NSH3 could actually bind to α1C independently of an interaction with a CaVβ GK domain. Previous in vitro studies have indicated that the CaVβ SH3 domain may interact weakly with the α1C I-II loop (Maltez et al., 2005) or the carboxy tail (Gerhardstein et al., 2000), but an association with α1 subunit in the context of living cells has not been demonstrated. To address this, we probed for an interaction between CFP-NSH3 and YFP-α1C by FRET. We used an α1C subunit that was truncated at the carboxy terminus (α1C[1671]) to maximize the chances of observing FRET. Previous studies had shown that this truncated α1 was still dependent on CaVβ for trafficking and gating modulation (Erickson et al., 2001). Compared with control cells expressing CFP+YFP-α1C, test group cells coexpressing CFP-NSH3 and YFP-α1C displayed a significantly elevated FRET (FR = 1.68 ± 0.11, n = 9 for test-group cells; FR = 1.01 ± 0.04, n = 9 for control cells; P < 0.001). Therefore, NSH3 is able to independently associate with α1C in live HEK 293 cells. Furthermore, fits of the FRET data to a 1:1 binding model yielded a Kd,EFF = 11,708, indicating that the NSH3/α1C interaction was of a significantly lower affinity than the NSH3/GKC interaction. Moreover, this value is dramatically higher than one previously estimated for full-length CaVβ2a binding to α1C (Kd,EFF = 43) (Erickson et al., 2001). This is in agreement with the idea that the GK domain is primarily responsible for high-affinity binding of CaVβ to α1 subunits, compared with the SH3 domain. To determine whether NSH3 interacted with α1C primarily via the cytoplasmic domain I-II loop (Maltez et al., 2005), we investigated an association between YFP-NSH3 and CFP-α1C[I-II loop] by FRET. This experiment yielded a low FR value (Fig. 7 A, FR = 1.13 ± 0.06) that was not significantly different from the control case of YFP + CFP-NSH3 (Fig. 4 C, FR = 1.10 ± 0.04, P = 0.67). Hence the low-affinity interaction between NSH3 and the α1-subunit I-II loop may be below the threshold detectable by the FRET assay. Together with the identified FRET interaction between YFP-α1C and CFP-NSH3, the results suggest that NSH3 primarily binds to α1C at a site that is different from the I-II loop.
Overall, the results indicate that the inability of NSH3 to influence channel properties when coexpressed with GKC[ΔPYDVV] is not because it cannot interact with α1C. Instead, taken together, the results reveal the dominance of the SH3–GK interaction in dictating CaVβ modulation of multiple functional properties of high voltage–activated Ca2+ channels.
Discussion
This study makes two main findings. (1) Although both the SH3 and GK domains are necessary to fully recapitulate CaVβ modulation of Ca2+ channels, they cannot act independently to reconstitute CaVβ function. (2) Subtle alterations in the inter-relationship of the SH3 and GK domains can uncouple the CaVβ trafficking and gating-modulation functions. The data have important implications for insights into the structure–function of the Ca2+ channel α1–β interaction, and are also noteworthy for describing new mechanistic paradigms for an SH3–GK module that is widely conserved among MAGUK proteins. We discuss these ramifications of the work in relation to previous findings.
Ca2+ Channel Structure–Function Implications
It was previously believed that the second conserved domain of CaVβs, which constitutes a GK motif, predominantly encoded the channel regulation properties of CaVβs, via its high-affinity interaction with the AID (De Waard et al., 1994; Pragnell et al., 1994; Chen et al., 2004; Opatowsky et al., 2004; Van Petegem et al., 2004). However, recent functional studies pointed to an unexpectedly prominent role of the CaVβ SH3 domain in modulating both channel trafficking and gating (Opatowsky et al., 2003; Takahashi et al., 2004; Maltez et al., 2005). A new insight from the present work is that the CaVβ SH3 and GK domains do not act independent of each other to reconstitute function. This finding is at odds with a prevailing model that CaVβs promote channel trafficking by masking an ER retention signal in the α1 subunit domain I-II linker (Bichet et al., 2000). Specifically, our results are inconsistent with the sufficiency of this mechanism because binding to the I-II loop can occur without appreciable channel trafficking (as observed with GKC and GKC[trunc] alone; Fig. 2 E and Fig. 5 E). Moreover, our results are also inconsistent with the idea that the SH3 domain simply masks another ER retention signal elsewhere in the channel to synergistically promote channel trafficking. If that were the case, then the SH3 and GK domains would be expected to be able to reconstitute channel trafficking in an independent fashion. Instead, we found that NSH3 and GKC[ΔPYDVV] (which do not interact) could not reconstitute efficient α1C trafficking to the membrane, even though both split-domain motifs were capable of engaging their respective binding sites on the α1C subunit. The requirement for the CaVβ SH3–GK interaction to reconstitute the channel trafficking function could signify two things. First, it could serve to increase the effective local concentration of the SH3 domain so that it engages its relatively low-affinity binding site on α1 (Maltez et al., 2005). This scenario appears somewhat unlikely since we demonstrate that freely expressed NSH3 is quite capable of interacting with α1C by itself. Alternatively, the SH3–GK interaction could ensure that CaVβ binding to α1 subunit occurs according to an induced fit model, effectively promoting an α1 conformation favorable for channel trafficking to the membrane.
Another important finding was that the α1C subunit was exquisitely attuned to the status of the CaVβ SH3–GK interaction; relatively subtle changes in the configuration of the CaVβ SH3–GK domain interaction selectively ablated gating-modulation (increased-Po and hyperpolarizing shift in voltage dependence of activation gating) functions, whereas channel trafficking capabilities were preserved. Thus, trafficking and gating modulation are separable functions of CaVβ based on the arrangement of the SH3–GK interaction. Previous studies have also found that trafficking and gating modulation are independent functions of CaVβ. In one study, the two processes could be separated based on different CaVβ concentration dependencies; low CaVβ concentrations selectively recapitulated channel trafficking, whereas higher amounts were needed to reconstitute gating modulation (Canti et al., 2001). A second study found that a mutation in the AID of α1C (α1C[Y467S]) disrupted the ability of CaVβ to traffick the channel, whereas gating modulation of whole-cell and single-channel currents was maintained (Gerster et al., 1999). Our findings raise the exciting possibility that the CaVβ SH3–GK interaction may be targeted by physiological and, possibly, pharmacological messengers as a means to modulate Ca2+ channel activity. In this regard, our development of a FRET-based method to monitor CaVβ SH3–GK interactions in live cells should hasten discovery in such new dimensions of Ca2+ channel research.
Comparisons to MAGUK SH3–GK Structure–Function
CaVβs and MAGUK family proteins share the common design principle of containing an SH3–GK structural module (Anderson, 1996; Funke et al., 2004). In addition, MAGUKS typically have one or more PDZ domains that are lacking in CaVβs. The crystal structure of the SH3–GK core from PSD-95 revealed an intramolecular SH3–GK interaction (McGee et al., 2001; Tavares et al., 2001), although the structural details differ somewhat from that observed with CaVβs (Chen et al., 2004; Opatowsky et al., 2004; Van Petegem et al., 2004). The idea that an intramolecular SH3–GK interaction is an important determinant of function in MAGUKs is well established; ablation of this interaction in Dlg generates a tumorigenic phenotype of larval imaginal discs in Drosophila (Woods et al., 1996); its disruption in PSD-95 inhibits the clustering of KV1.4 potassium channels (Shin et al., 2000). Classically, the cis SH3–GK interaction in MAGUKs is envisioned as having an autoinhibitory role by preventing the two motifs from interacting in trans with other proteins (Masuko et al., 1999; McGee and Bredt, 1999; Nix et al., 2000; Shin et al., 2000; Seabold et al., 2003). Specific cellular signals can break the cis SH3–GK interaction, permitting the SH3 and GK domains to engage with other ligand-binding sites. For example, a trans interaction between the PSD-95 GK domain and the SH3 domain of another MAGUK, NE-dlg, is normally prevented by an intramolecular interaction, but can be turned on by Ca2+-calmodulin (Masuko et al., 1999). In this regime, the SH3–GK interaction essentially behaves as an ON–OFF molecular switch. This behavior is analogous to what we find for CaVβ in that when the SH3–GK interaction is ablated, functional modulation of α1 is lost. By contrast, the finding that finer adjustments of the CaVβ SH3–GK module can preferentially modulate distinct functions is a new twist on the classical conception. It will be interesting to determine whether analogous regulatory paradigms exist in other MAGUK proteins. Overall, the CaVβ SH3–GK module is distinguished from those of MAGUKs by having a precisely defined binding partner (namely the α1 subunit) and functional signature (channel trafficking and gating modulation). Hence, further mechanistic understanding into the Ca2+ channel α1–β interaction structure–function could yield new insights into the functional operation of the widely conserved SH3–GK structural module.
Acknowledgments
This study was supported by a grant from the National Institutes of Health (NIH) (RO1 HL-69911) to H.M. Colecraft, by a shared-instrumentation grant from the NIH (S10 RR16859) to D.T. Yue, and a Medical Scientist Training Program Fellowship to S.X. Takahashi.
Olaf S. Andersen served as editor.
References
Abbreviations used in this paper: AID, α1 interaction domain; CaVβ, Ca2+ channel β subunit; CFP, cyan fluorescent protein; FR, FRET ratio; FRET, fluorescence resonance energy transfer; GK, guanylate kinase; MAGUK, membrane-associated guanylate kinase; Po, open probability; SH3, src homology 3; YFP, yellow fluorescent protein.