The C-terminal residues in the alpha-interacting domain (AID) helix anchor CaV beta subunit interaction and modulation of CaV2.3 channels.

The alpha-interacting domain (AID) in the I-II linker of high voltage-activated (HVA) Ca(2+) channel alpha1 subunits binds with high affinity to Ca(V)beta auxiliary subunits. The recently solved crystal structures of the AID-Ca(V)beta complex in Ca(V)1.1/1.2 have revealed that this interaction occurs through a set of six mostly invariant residues Glu/Asp(6), Leu(7), Gly(9), Tyr(10), Trp(13), and Ile(14) (where the superscript refers to the position of the residue starting with the QQ signature doublet) distributed among three alpha-helical turns in the proximal section of the I-II linker. We show herein that alanine mutations of N-terminal AID residues Gln(1), Gln(2), Ile(3), Glu(4), Glu(6), Leu(7), and Gly(9) in Ca(V)2.3 did not abolish [(35)S]Ca(V)beta 1b or [(35)S]Ca(V)beta 3 subunit overlay binding to fusion proteins nor did they prevent the typical modulation of whole cell currents by Ca(V)beta 3. Mutations of the invariant Tyr(10) with either hydrophobic (Ala), aromatic (Phe), or positively charged (Arg, Lys) residues yielded Ca(V)beta 3-responsive whole cell currents, whereas mutations with negatively charged residues (Asp, Glu) disrupted Ca(V)beta 3 binding and modulation. In contrast, modulation and binding by Ca(V)beta 3 was significantly weakened in I14A (neutral and hydrophobic) and I14S (neutral and polar) mutants and eradicated in negatively charged I14D and I14E or positively charged I14R and I14K mutants. Ca(V)beta 3-induced modulation was only preserved with the conserved I14L mutation. Molecular replacement analyses carried out using a three-dimensional homology model of the AID helix from Ca(V)2.3 suggests that a high degree of hydrophobicity and a restrained binding pocket could account for the strict structural specificity of the interaction site found at position Ile(14). Altogether these results indicate that the C-terminal residues Trp(13) (1) and Ile(14) anchor Ca(V)beta subunit functional modulation of HVA Ca(2+) channels.

positions to achieve a strong interaction with Ca V ␤ subunits. In contrast, there is very little information regarding the role of the N-terminal pair of residues Glu 6 /Leu 7 in establishing the functional interaction with Ca V ␤ subunits although the crystal structures have demonstrated a strong interaction between the two subunits at this site. It thus remains to be seen whether all of these Ca V AID-Ca V ␤ interaction sites are required to confer the typical Ca V ␤ subunit-induced functional modulation of Ca V 1 and Ca V 2 currents.
Herein we show that the alanine mutation of each of the AID residues Gln 1 , Gln 2 , Glu 3 , Glu 4 , Glu 6 , Leu 7 , Gly 9 , and Glu 18 did not eliminate [ 35 S]Ca V ␤1b or [ 35 S]Ca V ␤3 subunit overlay binding to fusion proteins nor did it prevent the typical modulation of Ca V 2.3 whole cell currents by Ca V ␤3. Mutations of Tyr 10 with either aromatic (Phe) or positively charged (Arg, Lys) residues yielded whole cell currents that responded to Ca V ␤3 essentially like the wild-type channel whereas mutations with negatively charged residues (Asp, Glu) disrupted Ca V ␤3 binding and modulation. The structural requirements were more stringent for Ile 14 since the modulation and binding by Ca V ␤3 were only preserved with the conserved Iso to Leu mutation. Altogether these results suggest that the C-terminal residues Trp 13 (1) and Ile 14 play a more critical role than Gly 9 and Tyr 10 in Ca V ␤ subunit functional modulation of HVA Ca 2ϩ channels.

EXPERIMENTAL PROCEDURES
Recombinant DNA Materials-Standard methods of plasmid DNA preparation were used (29). cDNAs coding for the auxiliary rat Ca V ␤3 (GenBank TM M88751) and rat brain Ca V ␤1b (GenBank TM X61394) were kindly donated by Dr. E. Perez-Reyes (30 -32). The wild-type human ␣1E or Ca V 2.3 (GenBank TM L27745) was a gift from Dr Toni Schneider (33). The rat brain Ca V ␣2b␦ subunit was provided by Dr. T. P. Snutch.
Point Mutations and RNA Transcription-Point mutations were performed with 40-mer synthetic oligonucleotides using the QuikChange TM XL-mutagenesis kit (Stratagene, La Jolla, CA). Briefly, mutations were achieved by cassette cloning using the NotI/XhoI Ca V 2.3 fragment (positions 111 and 1275 nt) subcloned into pBluescript (Stratagene). Constructs were verified by restriction mapping after religation of the mutated fragment into the naturally occurring NotI/XhoI sites of the wild-type Ca V 2.3. The nucleotide sequence of the mutant channels was confirmed through automated fluorescent DNA sequencing (BioST, Lachine, Qué). cDNA constructs for wild-type and mutated Ca V ␣1 subunits were linearized at the 3Ј-end by HindIII digestion whereas the Ca V ␤3 subunit was digested by NotI. Run-off transcripts were prepared using the methylated cap analog m 7 G(5Ј)ppp(5Ј)G and T7 RNA polymerase with the mMessage mMachine® transcription kit (Ambion, Austin, TX). The final cRNA products were resuspended in DEPCtreated H 2 O and stored at Ϫ80°C. The integrity of the final product and the absence of degraded RNA were determined by denaturing agarose gel electrophoresis stained with ethidium bromide.
Ca V ␤3 and Ca V ␤1b Overlay Assays onto pGFP-uv Fusion Proteins-A fragment of 105 bp including the whole AID region of Ca V 2.3 (25-residue AID E peptide) was generated by polymerase chain reaction and cloned in-frame into the HindIII-KpnI sites of pGFPuv vector (Gen-Bank TM U62636) (Clontech, BD Biosciences, Mississauga, Ontario) to give a fusion protein with an estimated molecular mass of 32 kDa. The AID-fusion protein was expressed in the chemically competent Escherichia coli strain DH5␣. The synthesis of the fusion proteins was induced at 37°C using 0.5 mM isopropyl ␤-D-thiogalactoside in an overnight liquid culture, and bacteria were collected by centrifugation. As the GFP protein cloned in-frame with the AID E peptide has been modified to fluoresce under UV light, the fusion protein confers a green color to the DH5␣ bacterial extracts when monitored under UV light. For overlay assays, crude DH5␣ bacterial extracts (200 l) were boiled for 2 min in Laemmli's loading buffer and separated on a denaturing 10% SDS-polyacrylamide gel. Total proteins were either visualized with Red Ponceau or were loaded in duplicate to be visualized by Coomassie staining in addition to the autoradiogram. This half of the gel was transferred onto a polyvinylidene difluoride membrane (Millipore, Fisher, Whitby, Ontario) using Towbin buffer (in mM: 25 Tris,192 glycine, 20% methanol, and 0.05% SDS). The membrane was blocked with 1% bovine serum albumin in HBS-Tween (in mM: 137 NaCl, 3 KCl, 10 HEPES, and 0.05% Tween-20 pH 7.4), washed once with HBS-Tween and incubated for 1 h at room temperature in 5 ml of HBS-Tween with 20 l of [ 35 S]methionine labeled Ca V ␤3or Ca V ␤1b subunit prepared as described below. The blots were washed twice for 10 min with 1 mM CaCl 2 in HBS-Tween, air-dried, and radioactive signals were detected by autoradiography.
[ 35 S]Methionine-labeled Ca V ␤3 or Ca V ␤1b in pBluescript (0.5 g) was synthesized by coupled in vitro transcription and translation (TNT; Promega, Madison, WI) in a 50-l reaction volume for 1 h, and the reaction mixture was applied without further treatment to the overlay membrane.
Functional Expression of Wild-type and Mutants Channels-Oocytes were obtained from female Xenopus laevis clawed frog (Nasco, Fort Atkinson, WI) as described previously (1,34,35). Briefly, stage VI oocytes free of follicular cells were injected with 46 nl of a solution containing 28 ng of cRNA coding for the wild-type or mutated Ca V ␣1 subunit. The Ca V ␣1 subunit was always co-injected with cRNA coding for the rat brain Ca V ␣2b␦ (6) and with or without the rat brain Ca V ␤3 (36) in a 6:2:3 or 6:2 weight ratio. The final amount of exogenous Ca V ␤3 injected corresponds to 14 ng, which represents a 23-fold enrichment over the estimated concentration of 0.56 Ϯ 0.02 ng of endogenous Ca V ␤3xo per oocyte (22). Oocytes were incubated at 19°C for 1-5 days after injection in Barth's solution (in mM): 100 NaCl; 2 KCl; 1.8 CaCl 2 ; 1 MgCl 2 ; 5 HEPES; 2.5 pyruvic acid; 100 units/ml of penicillin; 50 g/ml gentamicin; pH 7.6.
Electrophysiological Recordings in Oocytes-Wild-type and mutant channels were screened at room temperature for macroscopic Ba 2ϩ currents using a two-electrode voltage clamp amplifier (OC-725C, Warner Instruments, Hamden, CT) as described earlier (1,4,34). In order to minimize series resistance problems associated with voltage clamping large currents in Xenopus oocytes, experiments were performed on cells with peak currents smaller than Ϫ5 A in most cases. Hence, the wild-type channel was recorded 3-4 days after cRNA injection whereas whole cell Ba 2ϩ currents were generally measured 1-2 days after injection for AID mutant channels in the absence of Ca V ␤3 and 2-3 days in the presence of Ca V ␤3. Oocytes were first impaled in a modified Ringer solution (in mM): 96 NaOH; 2 KOH; 1.8 CaCl 2 ; 1 MgCl 2 ; 10 HEPES titrated to pH 7.4 with methanesulfonic acid CH 3 SO 3 H (MeS), then the bath was perfused with the 10 Ba 2ϩ solution (in mM; 10 Ba(OH) 2 ; 110 NaOH; 1 KOH; 20 Hepes titrated to pH 7.3 with MES). To minimize kinetic contamination by the endogenous Ca 2ϩ -activated Cl Ϫ current, oocytes were injected with 18.4 nl of a 50 mM EGTA (Sigma) 0.5-1 h before the experiments. Oocytes were superfused by gravity flow at a rate of 2 ml/min that was fast enough to allow complete chamber fluid exchange within 30 s. Experiments were performed at room temperature (20 -22°C).
Electrophysiological Data Acquisition and Analysis-PClamp software, Clampex 6.02 and Clampfit 6.02 (Axon instruments, Foster City, CA) was used for on-line data acquisition and analysis as previously described (1,34,35). Unless stated otherwise, data were sampled at 10 kHz and low pass-filtered at 5 kHz using the amplifier built-in filter. For all recordings, a series of voltage pulses were applied from a holding potential of Ϫ80 mV at a frequency of 0.2 Hz from Ϫ40 to ϩ60 mV. Isochronal inactivation data (hϱ or hinf) were obtained from tail currents generated at the end of a 5-s prepulse (4,37). Tail current amplitudes were estimated using the function Analyze in Clampfit 6.0 from the peak current arising during the first 10 ms after the capacitive transient (20 data points). Each of these currents was then normalized to the maximum current obtained before the prepulse voltage (I/I max ) and was plotted against the prepulse voltage. For the isochronal inactivation figures, pooled data points (mean Ϯ S.E.) were fitted to Equation 1 using user-defined functions and the fitting algorithms provided by Origin 6.1 and 7.0 (Microcal Software, Northampton, MA) analysis software. Equation 1 accounts for the fraction of non-inactivating current with E 0.5,inact mid-point potential; z, slope parameter; Y 0 , fraction of noninactivating current; V m , the prepulse potential, and RT/F with their usual meanings. The fitting process generated values estimating errors on the given fit values. Activation potentials were estimated from the normalized I-V curves obtained for each channel combination (22). Although this calculation was not exempt from gating contamination, it provided a qualitative approximation of the Ca V ␤3 modulation on I-V parameters. The I-V relationships were normalized to the maximum amplitude and were fitted to Equation 2, a Boltzmann equation coupled to a linear function.
E 0.5,act is the potential for 50% activation; G rel is the normalized conductance; z, slope parameter; V m , the test potential, V rev , the apparent reversal potential and RT/F with their usual meanings. Inactivation kinetics were quantified using R300 values, that is the ratio of the whole cell current remaining at the end of a 300-ms pulse (34,35,38). As inactivation kinetics can vary with current density, comparisons between constructs and mutants were generally restricted to whole cell currents lower than Ϫ5 A as much as possible. Furthermore, this range of current densities made it easier to voltage clamp the oocyte uniformly thus decreasing the possibility of series resistance artifacts contaminating the current kinetics data. Capacitive transients were erased for clarity in the final figures. Statistical analyses and Student's t test were performed using the fitting routines provided by Origin 6.1 and 7.0 (Microcal Software, Northampton, MA).
Homology Modeling of Wild-type and Mutant Ca V 2.3-Computations were performed using the InsightII package (version 2000, Accelrys) as described elsewhere (39,40). Briefly, the BIOPOLYMER and BUILDER modules were used in particular to build or modify molecular structures, and all energy minimizations were performed with the DISCOVER module using the consistent valence force field (CVFF). For all minimizations, one thousand steps of Steepest Descent minimization were performed, followed by a Conjugate gradient minimization until a convergence of 0.001 kcal/mol/Å was reached.
The starting coordinates were taken from the PDB file 1T0J.pdb (Ca V ␤2a ϩ AID from Ca V 1.2) (19). Molecular replacement of the following eight residues (L430I, E432R, D433E, K435N, L438R, D439A, T442D, and Q443D) to their reciprocal residues in the human Ca V 2.3 was undertaken using BIOPOLYMER to obtain the complex Ca V ␤2a ϩ AID wt from Ca V 2.3. The crystallographic water molecules were not retained. Hydrogen atoms were added using the BIOPOLYMER module at the normal ionization state of the amino acids at pH 7.0. Potentials as well as partial and formal charges were fixed with CVFF. A first series of energy minimizations was then performed with the DIS-COVER module using the CVFF with a distance-dependent dielectric constant of 80 (implicit H 2 O). Minimizations were carried out while keeping the heavy atoms of the protein fixed. The structure obtained was then submitted to a second series of energy minimizations performed as explained above, after removing all constraints on the atoms and with a 20 Å radius sphere of explicit water molecules centered around AID residue Asn 435 . These two series of minimizations yielded the energy-minimized configuration of Ca V ␤2a ϩ AID wt from Ca V 2.3. This wild-type configuration was then used to produce six individual mutants: Y437F, Y437R, Y437D, I441L, I441R, and I441D that were also minimized as explained above with a 20 Å radius sphere of explicit water molecules.

RESULTS
A Three-dimensional Model of the AID Peptide from Ca V 2.3-By homology with the three crystal structures obtained by x-ray diffraction of the AID region in Ca V 1.1 and Ca V 1.2 channels, the AID region of the Ca V 2.3 channel (AID E ) should form an ␣-helix upon binding to Ca V ␤ subunits (19 -21). The primary sequence of AID E, Gln 1 -Gln 2 -Ile 3 -Glu 4 -Arg 5 -Glu 6 -Leu 7 -Asn 8 -Gly 9 -Tyr 10 -Arg 11 -Ala 12 -Trp 13 -Ile 14 -Asp 15 -Lys 16 -Ala 17 -Glu 18 conserves 9 out of 18 residues when compared with the AID helix from Ca V 1.1 and Ca V 1.2 ␣1 subunits ( Table I). The threedimensional representations of AID E obtained by homology modeling with either Ca V 1.1/Ca V ␤2a (21) (not shown), Ca V 1.2/ Ca V ␤3 (20) (not shown), or Ca V 1.2/Ca V ␤2a (19) (Fig. 1) all show that AID E assumes an ␣-helical structure in the presence of Ca V ␤. The orientation of the residues facing the hydrophilic side of the AID helix (Gln 2 , Glu 4 , Arg 5 , Asn 8 , Arg 11 , Ala 12 , Asp 15 , and Lys 16 ) was seen to vary slightly but the residues mostly located on the hydrophobic face of the AID E helix that includes the invariant LGWYI residues (Glu 6 , Leu 7 , Gly 9 , Tyr 10 , Trp 13 , and Ile 14 ) are well buried in the Ca V ␤ subunit fold. As seen in Fig. 1, the front (A), the back (B), and the bottom (C) perspectives of the AID E model obtained using the crystal coordinates of Ca V 1.2/Ca V ␤2a (19) clearly show that the side-chains of Tyr 10 , Trp 13 , and Ile 14 are entrenched within the Ca V ␤ subunit. Hence, the C-terminal residues appear to interact extensively with the Ca V ␤ subunit. The Gly 9 residue is significantly enclosed by the Ca V ␤ subunit and could be needed to provide some degree of flexibility to the AID helix. The crystal structures further show that the aromatic side-chain of Tyr 10 is significantly surrounded by residues of the Ca V ␤ subunit suggesting that the interaction might require an aromatic residue at this position. Located on the adjacent turn, the extreme concealment of Trp 13 is compatible with our previous observations that mutations considered to be even conservative (W13F, W13Y), eliminated Ca V ␤3 binding and modulation of Ca V 2.3 (1). In contrast, Ile 14 situated next to Tyr 13 and the N-terminal residues Glu 6 and Leu 7 appear to form fewer interactions with Ca V ␤ suggesting these three positions might be more compliant to structural substitutions than Tyr 10 and Trp 13 . We thus undertook a systematic analysis of the physicochemical requirements of key AID positions to achieve functional modulation of Ca V 2.3 by Ca V ␤ subunits.
Alanine Scan of the N-terminal AID residues (Gln 1 -Leu 7 ) in Ca V 2.3-Ca V ␤ subunit binding and modulation of N-terminal AID residues (Gln 1 , Gln 2 , Ile 3 , Glu 4 , Glu 6 , Leu 7 ) in Ca V 2.3 ( Fig.  2A) were investigated with alanine mutations of these residues. Mutations of the nonconserved Arg 5 were extensively studied and reported before (1, 34). Fig. 2, B and C shows the overlay binding assays for point mutations Q1A, Q2A, I3A, E4A, E6A, L7A performed in the 18-residue AID peptide from Ca V 2.3 (AID E ) inserted in the pGFPuv vector that corresponds respectively to mutants Q374A, Q375A, I376A, E377A, E379A, and L380A in Ca V 2.3. Ca V ␤ subunit binding was evaluated with both [ 35 S]methionine-labeled Ca V ␤3 (Fig. 2B) and Ca V ␤1b (Fig. 2C) subunits. As seen, strong signals were obtained in both cases at the expected molecular mass of 32 kDa for the wild-type AID peptide Ϸ Q374A Ϸ Q375A Ϸ I376A Ϸ E377A Ϸ E379A Ͼ L380A mutants indicating that both Ca V ␤3 and Ca V ␤1b interacted strongly with the alanine mutants under denaturing conditions. In contrast, the empty pGPFuv vector (29 kDa) and the well documented W386A (WA) mutant failed to interact with either [ 35 As seen in Fig. 2, D-G, the binding data correlated well with the functional data. The six alanine mutants were expressed in Xenopus oocytes in the presence and in the absence of Ca V ␤3. Under both conditions, the N-terminal mutants yielded robust inward Ba 2ϩ currents with current-voltage relationships typical of voltage-gated Ca 2ϩ channels. Typical current traces are shown for E6A (E379A) and L7A (L380A) in the absence (left panels) and in the presence (right panels) of Ca V ␤3. To ensure that the current density was not unduly influencing the VDI (voltage-dependent inactivation) gating, the RNA concentrations were adjusted to yield peak currents in the Ϫ3 to Ϫ5 A range when possible. Co-expression with Ca V ␤3 slightly hyperpolarized the activation potential for I377A, E379A, and L380A at p Ͻ 0.01 but had little or insignificant effect on Q374A and Q375A (see Table II for detailed values). The VDI gating was estimated from the whole cell currents remaining after 300 ms (R300). In the absence of Ca V ␤3, the R300 values were similar for most channels (wild-type and mutants) although they were significantly larger at p Ͻ 0.001 for the N-terminal mutants Gln 1 , Gln 2 , Ile 3 , and Glu 4 ( Fig. 2F) suggesting that the Nterminal end of the AID E helix could be an intrinsic determinant of VDI gating in this channel (44). Co-injection with Ca V ␤3 significantly decreased the R300 values for the wildtype channel, Q374A, Q375A, I376A, E377A, and E379A with less than 10% of the whole cell currents remaining at 300 ms ( Fig. 2G). Hence, the VDI gating of N-terminal mutants was not significantly different than the wild-type channel (p Ͼ 0.05). L380A turned out to be the only mutant with kinetics slightly significantly faster than the wild-type channel at p Ͻ 0.05. By comparison, the R300 values obtained for W386A under the same conditions were 3-times larger, with roughly 30% of the whole cell currents remaining at the end of a 300-ms pulse and significantly different from the wild-type channel at p Ͻ 10 Ϫ10 . Altogether, these data indicate that the VDI gating of the alanine mutants was intrinsically similar to the wildtype channel in the absence of Ca V ␤3 and further show that they were modulated by Ca V ␤3 in a wild-type fashion.
The N-terminal alanine mutants displayed the additional

FIG. 1. Three-dimensional model of the AID helix of the human Ca V 2.3 (AID E ).
Model was obtained with INSIGHT II using the atomic coordinates for the human Ca V 1.2 AID C crystal structure co-crystallized with Ca V ␤2a at a 2.0 Å resolution (1T0J. pdb) as a template (19). The AID C peptide (Gln 1 -Gln 2 -Leu 3 -Glu 4 -Glu 5 -Asp 6 -Leu 7 -Lys 8 -Gly 9 -Tyr 10 -Leu 11 -Asp 12 -Trp 13 -Ile 14 -Thr 15 -Lys 16 -Ala 17 -Glu 18 ) underwent 8 substitutions to obtain the primary sequence of AID E (Gln 1 -Gln 2 -Ile 3 -Glu 4 -Arg 5 -Glu 6 -Leu 7 -Asn 8 -Gly 9 -Tyr 10 -Arg 11 -Ala 12 -Trp 13 -Ile 14 -Asp 15 -Lys 16 -Ala 17 -Glu 18 ) as described under "Experimental Procedures" (see also Table I). The core of the AID E helix is colored in white. hallmarks of Ca V ␤-modulation. Co-expression with Ca V ␤3 significantly hyperpolarized by Ϸ Ϫ30 mV the mid-potential of inactivation of the N-terminal mutants replicating the behavior of the wild-type channel from a E 0.5,inact ϭ Ϫ36 Ϯ 3 mV (10) in the absence of Ca V ␤ to E 0.5,inact ϭ Ϫ64 Ϯ 1 mV (10) with Ca V ␤3 as previously published (1,4,34) (Table II). Considering that the fifth position (Arg 378 ) was previously studied and reported (1,34), these data indicate that alanine substitutions at any of the first seven positions of the AID E helix failed to alter significantly Ca V ␤ subunit binding and modulation of Ca V 2.3 channels. Hence, the N-terminal Glu 6 and Leu 7 residues located in the first ␣-helical turn of the AID E helix are tolerant to alanine substitutions.
Alanine Scan of the GYI Residues in the ␣-Helix C-terminal Turns in Ca V 2.3-The role of the two remaining pairs of critical residues Gly 9 /Tyr 10 and Trp 13 /Ile 14 located on the C-terminal ␣-helical turns was also investigated after mutation with alanine. As explained earlier, Ca V ␤ subunit binding was investigated in overlay assays whereas modulation by Ca V ␤ subunits was studied after functional expression in Xenopus oocytes. Fig. 3A shows the overlay binding assays for Y383A, G382A, and I387A performed in the 25-residue AID E peptide inserted in the pGFPuv vector. Ca V ␤ subunit binding is shown for (Fig. 3A). The empty pGPFuv vector and wild-type AID E peptide were tested along as additional controls. Strong bands were obtained at the expected [ 35 S]Ca V ␤1b binding could not be detected for the empty vector pGFPuv and the AID E mutant W386A. D, typical whole cell current traces obtained after the expression of the E379A (E6A) mutant in Xenopus oocytes with Ca V ␣2b␦ without exogenous Ca V ␤3 (Ϫ␤3) and after co-injection with Ca V ␤3 (ϩ␤3). Time and current scales are identical for D and E. E, typical whole cell current traces obtained after the expression of the L380A (L7A) mutant in Xenopus oocytes with Ca V ␣2b␦ without exogenous Ca V ␤3 (Ϫ␤3) and after co-injection with Ca V ␤3 (ϩ␤3). F, mean R300 ratios (the fraction of the whole cell current remaining at the end of a 300-ms pulse) are shown Ϯ S.E. from 0 to ϩ20 mV for the wild-type and the mutant Ca V 2.3 channels (Q374A, Q375A, I376A, E377A, E379A, L380A, and W386A) in the absence of Ca V ␤3. G, R300 graph for the wild-type and the mutant Ca V 2.3 channels (Q374A, Q375A, I376A, E377A, E379A, L380A, and W386A) after co-injection with Ca V ␤3. As seen, the VDI kinetics of the N-terminal mutants are significantly accelerated in a manner comparable to the wild-type channel. In contrast, the kinetics of W386A were not increased by Ca V ␤3 suggesting that the first turn of the AID helix is more tolerant to alanine mutations. Whole cell currents were recorded using the two-electrode voltage clamp technique in the presence of 10 mM Ba 2ϩ after injection of EGTA. Holding potential was Ϫ80 mV. Xenopus oocytes were pulsed from Ϫ40 mV to ϩ60 mV using 10 mV steps for 450 ms. All mutants expressed significant inward currents with typical current-voltage properties. Capacitive transients were erased for the first ms after the voltage step. The mutant properties (mid-potentials of activation and inactivation as well as peak currents) are provided in Table II. molecular weight of 32 kDa for the wild-type AID E peptide as well as for G382A indicating that the alanine mutation at position Gly 9 did not impede the interaction with [ 35 S]Ca V ␤1b under denaturing conditions. In contrast, [ 35 S]Ca V ␤1b binding onto Y383A and I387A peptides was similar to the background level as compared with the lane loaded with the empty vector pGPFuv suggesting that substitution with alanine disrupted the interaction between Tyr 383 or Ile 387 and Ca V ␤1b. Similar results were obtained with [ 35 S]Ca V ␤3 (not shown). The binding data were compared with the electrophysiological data obtained for the same mutants in the presence and in the absence of Ca V ␤3. As expected from the binding studies, Ca V ␤3 modulated the whole cell currents of G382A in a fashion reminiscent of the wild-type channel (Fig. 3B) with a Ϸ5-fold decrease in the R300 values in the presence of Ca V ␤3 (p Ͻ 10 Ϫ5 ) (Fig. 3D). Table II reports the estimated mid-potentials of activation and inactivation under the same conditions.
inactivation was also hyperpolarized by Ca V ␤3 although the Ϫ20 mV shift was significantly smaller at p Ͻ 10 Ϫ4 than the Ϫ30 mV routinely observed for the wild-type channel (Table II). The data for I387A appeared to be more easily reconciled with the binding studies. The modulation of VDI gating kinetics as measured at 300 ms was milder for I387A although the R300 values measured with Ca V ␤3 remained significantly smaller (p Ͻ 0.05 at 0 mV) (Fig. 3F). The milder modulation of the I387A mutant by Ca V ␤3 was also apparent in the smaller shift in the voltage dependence of inactivation with E 0.5 inact experiencing a Ϫ13 mV shift in average as compared with the Ϫ30 mV displacement reported for the wild-type channel (Table II). Hence, measuring Ca V ␤ binding to the AID peptide under denaturing conditions does not appear to entirely account for the interaction between the I-II linker and Ca V ␤ subunits. Aromatic and Positively Charged Residues at Tyr 10 Preserve Ca V ␤ Binding and Modulation-Because the substitution of the aromatic tyrosine residue Tyr 10 at position 383 by the small and hydrophobic alanine (Ala) residue was shown to disrupt Ca V ␤3 subunit binding as well as decreasing the Ca V ␤-induced modulation of Ca V 2.3, the structural requirements for Ca V ␤ subunit binding and modulation were further investigated after substitutions with aromatic (Phe, Trp) as well as positive (Arg, Lys) and negative (Asp, Glu) residues. Ca V ␤ subunitinduced modulation of inactivation was also preserved with the Y10S mutant in L-type Ca V 1.1 and Ca V 1.2 channels (26) whereas in vitro binding of Ca V ␤4 to AID A was demonstrated in the presence of the Y10S and Y10F mutants (25). We had previously reported that the substitution with the neutral but polar serine (Ser) residue at the same position yielded a channel with VDI gating slower than the wild-type channel in the presence of Ca V ␤3 (1). Nonetheless, the voltage dependence of inactivation of Y383S was clearly shifted in the hyperpolarized direction by Ϫ20 mV in the presence of Ca V ␤3 (1). These data were replicated over a 2-year period, and Table II provides ). B, typical whole cell current traces obtained after the expression of the G382A mutant in Xenopus oocytes with Ca V ␣2b␦ without exogenous Ca V ␤3 (Ϫ␤3) and after co-injection with Ca V ␤3 (ϩ␤3). Time and current scales are identical for panels B and C. C, typical whole cell current traces obtained after the expression of the I387A (I14A) mutant in Xenopus oocytes with Ca V ␣2b␦ without exogenous Ca V ␤3 (Ϫ␤3) and after co-injection with Ca V ␤3 (ϩ␤3). D, mean R300 Ϯ S.E. values of G382A were significantly decreased by the co-injection of Ca V ␤3 (p Ͻ 10 Ϫ8 at 0 Ͻ V m Ͻ 20 mV). R300 values for the wild-type channel have been added for comparison. E, mean R300 Ϯ S.E. values of Y383A were decreased by the co-injection of Ca V ␤3 (p Ͻ 10 Ϫ3 at 0 Ͻ V m Ͻ 20 mV). R300 values for the wild-type channel have been added for comparison. F, mean R300 Ϯ S.E. values of I387A were decreased by the co-injection of Ca V ␤3 (p Ͻ 0.05 at 0 Ͻ V m Ͻ 20 mV). R300 values for the wild-type channel have been added for comparison. The numbers to the right of the mutants refer to the numbers of experiments used for statistical analysis. The voltage dependence of activation and inactivation are detailed in Table II. tated to Y383F, Y383E, Y383K, and Y383R is shown in Fig. 4A whereas [ 35 S]Ca V ␤3 binding to pGFPuv-AID E mutants Y383F, Y383S, Y383E, Y383R, and Y383W in Fig. 4B. For the Ca V ␤1b data, the wild-type AID E fusion protein was used as a positive control whereas W386A was loaded to assess the background signal. E391A (E18A) was used as a positive control in the presence of Ca V ␤3 since it remained functionally modulated by Ca V ␤3 in a wild-type fashion (Table II). Altogether, the signal intensity for the AID E fusion peptides ranked from AID E wt Ϸ E391A Ͼ Ͼ Y383F Ϸ Y383W Ϸ Y383K Ϸ Y383R ϾY383S Ͼ Ͼ W386A. The binding signal for Y383E was somewhat stronger in the presence of Ca V ␤3. As seen, the signal was similar for the two aromatic mutations Y383F and Y383W confirming that the aromatic side-chain is a key structural requirement. Furthermore, positively charged residues were more tolerated than the negatively charged ones suggesting that electrostatic-based interactions are important for [ 35 S]Ca V ␤1b and [ 35 S]Ca V ␤3 binding at least under our experimental conditions where the AID helix has been denatured.
To establish a functional correlation with the binding studies, Tyr 383 mutants (Phe, Trp, Ala, Lys, Arg, Ser, Asp, Glu) were expressed Ϯ Ca V ␤3 and characterized by the double electrode voltage clamp approach in Xenopus oocytes. All Tyr 383 mutants, with the exception of Y383W, expressed robust inward currents in the presence of 10 mM Ba 2ϩ (Fig. 4, C and D). The activation potentials of the mutant channels were compa- . C, typical whole cell current traces obtained after the expression of the Y383R (Y10R) mutant in Xenopus oocytes with Ca V ␣2b␦ without exogenous Ca V ␤3 (Ϫ␤3) and after co-injection with Ca V ␤3 (ϩ␤3). Time and current scales are identical for C and D. D, typical whole-cell current traces obtained after the expression of the Y383D (Y10D) mutant in Xenopus oocytes with Ca V ␣2b␦ without exogenous Ca V ␤3 (Ϫ␤3) and after co-injection with Ca V ␤3 (ϩ␤3). E, bar graph showing the mean R300 Ϯ S.E. values for the wild-type, Y383F, Y383A, Y383K, Y383R channels (from left to right) with the R300 value obtained with Ca V ␤3 plotted before the value obtained in the absence of the subunit. VDI gating were significantly decreased by the co-injection of Ca V ␤3 for these mutants at p Ͻ 10 Ϫ4 for 0 Ͻ V m Ͻ 20 mV. The Y383W did not yield detectable inward Ba 2ϩ currents in response to membrane depolarization (see Table II). F, bar graph showing the mean R300 Ϯ S.E. values for the wild-type, Y383S, Y383D, Y383E channels (from left to right) with the R300 value obtained with Ca V ␤3 plotted before the value obtained in the absence of the subunit. The VDI gating of Y383D and Y383E were not significantly altered by the co-injection of Ca V ␤3 (p Ͼ 0.05) for 0 Ͻ V m Ͻ 20 mV. For Y383S, the R300 values were decreased in the presence of Ca V ␤3 (p Ͻ 0.01) for 0 Ͻ V m Ͻ 20 mV but remained significantly higher than for the wild-type channel (p Ͻ 0.001) for 0 Ͻ V m Ͻ 20 mV. The numbers to the right of the mutants refer to the numbers of experiments used for statistical analysis. The voltage dependence of inactivation for these mutants were computed using the Boltzmann Equation 1 after 5-s prepulses. The estimated values for the mid-potentials of activation and inactivation are provided in Table II. rable to the wild-type channel although their mean currentvoltage relationships were not significantly shifted by Ca V ␤3 in the hyperpolarized direction (Table II). The VDI kinetics (Fig.  4, E and F) and voltage dependence of inactivation (Table II) of Tyr 383 mutants were modulated by Ca V ␤3 in a manner that mimics the binding data. The R300 values plotted as a function of voltage indeed revealed that substitutions with aromatic (Phe), neutral (Ala), and positively charged (Arg, Lys) residues reproduced the phenotype of the wild-type channel shown as a control (Fig. 4E). Ca V ␤3 compellingly increased the VDI gating kinetics of these mutants with a 4-fold acceleration that is similar to the wild-type channel (p Ͻ 10 Ϫ4 ). Furthermore, the voltage dependence of inactivation of these mutants was indistinguishable from the wild-type channel whether measured in the presence or in the absence of Ca V ␤3 with the exception of Y383A that inactivated at E 0.5 inact ϭ Ϫ54 Ϯ 1 mV (10) in the presence of Ca V ␤3 (Table II). In contrast, mutations with the negatively charged aspartate and glutamate residues eliminated the functional modulation of VDI gating (Fig. 4F) and blunted the Ca V ␤3-induced hyperpolarizing shift in the voltage dependence of inactivation (Table II). Mutation with the neutral but polar serine residue significantly restrained the extent of Ca V ␤3-induced modulation. Altogether, it can be concluded that the Tyr 10 position in the second turn of the AID helix is relatively tolerant to substitutions with aromatic, small hydrophobic, or positively charged residues.
The Iso to Leu Substitution Is Tolerated at Position Ile 14 -The stereochemical requirements for Ca V ␤ binding and modulation at position Ile 14 were investigated after substitutions with the conserved leucine residue (Leu), the neutral but polar serine (Ser) as well as positive (Lys, Arg) and negative (Asp, Glu) residues. As seen earlier, the substitution of the hydrophobic isoleucine residue at this position by the likewise hydrophobic but smaller alanine residue was shown to disrupt Ca V ␤3 subunit binding as well as decreasing the Ca V ␤-induced modulation of Ca V 2.3 suggesting that the volume of the residue plays a critical role in establishing essential interactions with Ca V ␤ subunits.
[ 35 S]Ca V ␤1b binding to pGFPuv-AID E fusion proteins mutated to I387D, I387R, and I387K is shown in Fig. 5A. As shown before, the wild-type AID E fusion protein and W386A were used as positive and negative controls, respectively. The signal was unambiguous for AID E wt but was indistinguishable from the background noise for I387D Ϸ I387R Ϸ I387K Ϸ W386A. Similar data were obtained with [ 35 S]Ca V ␤3 with I387S Ϸ I387D Ϸ I387E Ϸ I387R Ϸ I387K (not shown). As seen, none of the polar mutations at this position, including the mutation by the neutral serine, preserved binding to [ 35 S]Ca V ␤ under denaturing conditions.
The functional characterization was carried out as discussed before in the presence of Ca V ␤3. All the Ile 14 mutants activated within the same voltage range as the wild-type channel (Table  II). In agreement with the binding data, the Ca V ␤-induced modulation of VDI kinetics (Fig. 5, B-E) and voltage dependence of inactivation (Table II) was preserved to a certain extent after substitution with neutral residues but was lost with the charged substitutions. Typical current traces are shown for I387L Ϯ Ca V ␤3 (Fig. 5B) and I387K Ϯ Ca V ␤3 (Fig. 5C). As shown in the R300 graph, the inactivation kinetics of I387L, I387A, I387S, and the wild-type channel were not significantly different in the absence of Ca V ␤ (p Ͼ 0.05) (Fig. 5D). Although co-injection with Ca V ␤3 sped up to a certain extent the inactivation kinetics of the three neutral mutants (Leu, Ser, Ala), I387A ϩ Ca V ␤3 and I387S ϩ Ca V ␤3 remained significantly slower at p Ͻ 10 Ϫ5 than the wild-type and the I387L channel measured under the same conditions (p Ͻ 10 Ϫ5 ) suggesting that the functional interaction with Ca V ␤3 was fully preserved with the conserved I387L mutation but not optimal with the I387A and the I387S mutants.
Substitutions with positively charged (Arg, Lys) or negatively charged (Asp, Glu) residues yielded functional channels with typical I-V properties that were ineffectually modulated by Ca V ␤3 (Fig. 5E). Although Ca V ␤3 appeared to significantly speed up the VDI kinetics of I387E at 0 mV (p Ͻ 10 Ϫ5 ), this effect vanished at ϩ20 mV. Furthermore the R300 values for I387E ϩ Ca V ␤3 were systematically larger than the wild-type channel under the same conditions. Finally, the voltage dependence of inactivation of the charged mutants was not significantly hyperpolarized by the co-injection with Ca V ␤3 (Table  II). Altogether these data suggest that the position Ile 14 specifically requires a bulky, neutral, and hydrophobic residue to achieve Ca V ␤-induced modulation of Ca V channels.

DISCUSSION
The C-terminal Turn of the AID Helix Anchors the Interaction with Ca V ␤-The four known auxiliary Ca V ␤ (Ca V ␤1-4) subunits can potentially associate with any of the six Ca V ␣1 pore-forming subunits of HVA VDCC (Ca V 1.1-1.3; Ca V 2.1-2.3) via AID located on the I-II linker of the Ca V ␣1 subunit. About half of the residues of the AID helix Gln 1 -Gln 2 -X 3 -Glu 4 -X 5 -X 6 -Leu 7 -X 8 -Gly 9 -Tyr 10 -X 11 -X 12 -Trp 13 -Ile 14 -X 15 -X 16 -X 17 -Glu 18 ) are strictly conserved among the six Ca V ␣1 subunits but homology was found to be higher within members of the Ca V 1 and Ca V 2 families (Table I). The crystal structures of the Ca V ␤/AID peptide complex from L-type Ca V ␣1.1 and Ca V ␣1.2 channels have recently revealed that the AID region forms an ␣-helix upon binding Ca V ␤ subunits (19 -21). The interaction between the AID helix and the Ca V ␤ protein appears to be primarily secured by the projection of three pairs of hydrophobic residues onto the Ca V ␤ fold with Tyr 10 and Trp 13 being almost completely buried within the Ca V ␤ protein (19 -21). The threedimensional model of AID E obtained by homology modeling with Ca V 1.2/Ca V ␤2a (19) suggests similarly that the ␣-helical structure of the AID peptide is preserved in Ca V 2.3 with the three pairs of residues Glu 6 /Leu 7 , Gly 9 /Tyr 10 , and Tyr 13 /Ile 14 projecting into Ca V ␤ residues with Leu 7 , Gly 9 , Tyr 10 , Trp 13 , and Ile 14 being invariant in all HVA Ca V ␣1 channels. The sixth position is always occupied by a negatively charged residue with the aspartate residue in Ca V 1 being replaced by a glutamate residue in Ca V 2 channels. The three-dimensional model of AID E shows that Tyr 10 and Trp 13 are deeply enmeshed within residues of the Ca V ␤ subunit suggesting that these residues could play a more critical role in Ca V ␤ modulation of Ca V 2.3 currents than Gln 1 , Gln 2 , Ile 3 , Glu 4 , Glu 6 , Leu 7 , Gly 9 , and Ile 14 .
Our mutational analysis herein demonstrated that alanine mutants of Gln 1 , Gln 2 , Ile 3 , Glu 4 , and Glu 18 did not abolish [ 35 S]Ca V ␤1b or [ 35 S] Ca V ␤3 subunit overlay binding to fusion proteins nor did they prevent the typical modulation of Ca V 2.3 whole cell currents by Ca V ␤3. These results suggest that the N-terminal residues of the AID helix located on the hydrophilic face of the AID helix do not play a determinant role in the interaction between Ca V ␣1 and Ca V ␤ subunits in agreement with the predictions of the crystal structure. In general, the signal intensity to the AID E peptide was found to be more robust for [ 35 S]Ca V ␤1b than for [ 35 S]Ca V ␤3 binding in agreement with previous reports showing that Ca V ␤1b displayed a significantly higher affinity for the AID A peptide than Ca V ␤3 in Ca V 2.1 channels (17,41).
Furthermore, alanine mutations of Glu 6 and Leu 7 located in the first hydrophobic ␣-helical turn of the AID E helix did not affect significantly Ca V ␤ binding and modulation of Ca V 2.3 currents. These results suggest that the binding pocket at these positions can accommodate alanine residues. The central ␣-helical turn of the AID helix was slightly less tolerant to mutations than the N-terminal residues. In fact, mutation of Gly 9 by alanine enhanced Ca V ␤-induced modulation indicating that the helical propensity conferred by the alanine residue fastened the interaction with Ca V ␤ whereas increased flexibility at this position could loosen the interaction. Mutations of Tyr 10 with either aromatic (Phe) or positively charged (Arg, Lys) residues yielded whole cell currents that responded to Ca V ␤3 essentially like the wild-type channel whereas mutations with negatively charged residues (Asp, Glu) disrupted Ca V ␤3 binding and modulation.
In contrast, mutations of the C-terminal residues Trp 13 and Ile 14 disrupted significantly Ca V ␤ binding and modulation. We had previously provided evidence that position Trp 13 could not be successfully substituted with hydrophobic (Ala, Gly), hydrophilic (Gln, Arg, Glu), or aromatic (Phe, Tyr) residues (1). Herein, we further demonstrate that modulation and binding by Ca V ␤3 was significantly weakened in I14A (neutral and hydrophobic) and I14S (neutral and polar) mutants and com-pletely eradicated in negatively charged I14D and I14E or positively charged I14R and I14K mutants. The conservative mutation I14L was found to be the only substitution that preserved the Ca V ␤3-induced modulation of Ca V 2.3 currents. Hence, in contrast to the N-terminal residues, we have compellingly established that Ca V ␤ binding as well as Ca V ␤-induced modulation of HVA Ca V currents are determined by the Cterminal pair of residues of the AID helix.
Computer Modeling and Molecular Replacements of Tyr 10 (Tyr 383 ) and Ile 14 (Ile 387 ) Positions-The analysis of the accessible surface area in the Ca V 1.2 AID-Ca V ␤2a complex showed Tyr 10 to be slightly more buried within the Ca V ␤ fold than Ile 14 suggesting that the binding pocket formed at the Tyr 10 position could impose harsher structural requirements than the binding pocket at Ile 14 (19). On the contrary, our data revealed that mutations at position Ile 14 impacted more significantly on Ca V ␤3 binding and functional modulation of Ca V 2.3 channel function than mutations at Tyr 10 . To visualize the changes in the environment caused by the AID E mutations, molecular replacements were produced in the three-dimensional model of  ). B, typical whole cell current traces obtained after the expression of the I387L (I14L) mutant in Xenopus oocytes with Ca V ␣2b␦ without exogenous Ca V ␤3 (Ϫ␤3) and after co-injection with Ca V ␤3 (ϩ␤3). Time and current scales are identical for panels B and C. C, typical whole cell current traces obtained after the expression of the I387K (I14K) mutant in Xenopus oocytes with Ca V ␣2b␦ without exogenous Ca V ␤3 (Ϫ␤3) and after co-injection with Ca V ␤3 (ϩ␤3). D, mean R300 Ϯ S.E. values of I387L were significantly decreased by the co-injection of Ca V ␤3 (p Ͻ 10 Ϫ10 at 0 Ͻ V m Ͻ 20 mV) but the modulation was milder for I387A and I387S mutants (p Ͻ 0.05 at 0 Ͻ V m Ͻ 20 mV). E, mean R300 Ϯ S.E. values of negatively and positively charged I387 mutants were unaffected by the co-injection of Ca V ␤3 at most voltages. However, the R300 for I387E and I387K were significantly decreased by Ca V ␤3 at 0 mV (p Ͻ 10 Ϫ3 ). The numbers to the right of the mutants refer to the numbers of experiments used for statistical analysis. The voltage dependence of inactivation for these mutants estimated from 5-s prepulses are provided in Table II. the wild-type AID E helix obtained for Ca V 2.3 (see "Experimental Procedures"). The resulting energy-minimized configuration was then compared with the three-dimensional model for the wild-type AID E helix.
The three-dimensional model of the AID E helix implies that Ile 14 could make van der Waals interactions with Ca V ␤2a hydrophobic residues Leu 352 , Val 241 , Met 244 , and Met 245 . From the wild-type model, Ile 14 appears to be confined to a smaller hydrophobic pocket than Tyr 10 suggesting that steric hindrance could become a key parameter at this position. As a result, mutations with hydrophobic residues of comparable size are predicted to be better tolerated than mutations with bulkier residues. Indeed, there is an excellent superposition of the C␣, C␤, and C␥ between the isoleucine and the similar leucine residues at position 14 of the AID E helix. The van der Waals interactions are also highly conserved and the I14L mutation does not appear to alter the helix conformation of the AID peptide (not shown).
The I14R mutation that introduces a bulkier and hydrophilic residue in the highly hydrophobic binding pocket produced stronger effects with a distinct deviation of the side-chain at the C␥ that pushes residues from Ca V ␤2a away from the AID E helix and results into a significantly increased r.m.s.d. value (1.53Å as compared with 0.55Å for I14L) especially for the four hydrophobic Ca V ␤2a residues (Fig. 6). Hence, the introduction of an arginine residue could result in a significantly loosening of the interaction with Ca V ␤2a residues. The I14L and I14R models thus appear to correlate well with our experimental data both in terms of in vitro binding and Ca V ␤-induced functional modulation.
In contrast with the I14R mutation, our molecular replacement experiments suggest that the introduction of an aspartate residue is not likely to impact significantly on the position of the Ca V ␤2a residues as a result of steric constraints. Indeed, the r.m.s.d. values for the AID E helix (0.12 Å) as well as for the Ca V ␤2a residues remained small throughout (0.23 Å). However, the absence of Ca V ␤-binding and modulation with the I14D mutant is likely to stem from the relatively higher electrostatic energy produced by the net negative charge and the lower hydrophobicity of the aspartate residue as compared with the arginine residue. The three-dimensional model also infers that the aspartate residue could not form additional hydrogen bonds unlike the arginine residue. Although these bonds are non-existent with the native isoleucine, they could have minimized the energy of the Ca V ␤-Ca V AID complex. The negative charge of I14D would then be associated with a high energetic cost of burial in the hydrophobic environment of the binding pocket (42,43). These properties could become noteworthy considering the hydrophobic character of the proteinprotein interface between the Ca V ␣1 and the Ca V ␤ subunits.
Since the binding pocket formed by Tyr 10 appears to be deeper than the Ile 14 site, bulky residues are expected to be better tolerated at Tyr 10 than at Ile 14 . The substitution of the tyrosine by a phenylalanine at position 10 results in the loss of two hydrogen bonds with the Ca V ␤2a residues Ser 344 and Glu 389 . Nonetheless, the aromatic rings of the tyrosine and the phenylalanine residues remain in the same orientation and the carbon atoms of the side-chains for both residues are in the same plane. The r.m.s.d. value for the AID E helix (0.18 Å) remained small suggesting that the Y10F mutation should not affect the Ca V ␤-Ca V AID interaction as we have observed experimentally. Given that the binding pocket at this position is relatively deeper than the one at position 14, the arginine residue could eventually adopt an energetically favorable elongated configuration where its carbon atoms (␣, ␤, and ␥) could espouse the same conformation as the equivalent carbon atoms from the native tyrosine residue. Indeed, the arginine residue appears to be as comfortably buried within Ca V ␤2a as the phenylalanine or the native tyrosine residues, which is substantiated by the small r.m.s.d. value observed with this mutant (0.23 Å). Finally, the arginine residue forms also a hydrogen bond with Ca V ␤2a like the native wild-type tyrosine residue does, albeit with a different residue (Asp 388 ). All these factors could account for the observation that the Y383R mutation did not affect significantly Ca V ␤ binding or Ca V ␤-induced modulation. In contrast, the aspartate residue at this position (Y10D) was too short to fully interact with Ca V ␤2a or to form any hydrogen bond with Ca V ␤2a residues. The shorter side-chain of the aspartate residue does not however appear to significantly affect the conformation of the AID E peptide since the r.m.s.d. value (0.24 Å) remained comparable to the values obtained for the arginine and the phenylalanine at this position. Hence the overall lower hydrophobicity as well as the absence of the hydrogen bond could explain in part the lack of Ca V ␤-induced modulation of the Y383D mutant in Ca V 2.3. CONCLUSION Our data proved that mutations of residues located on the N-terminal ␣-helical turn of the AID helix (Glu 6 , Leu 7 ) impact less significantly on Ca V channel modulation by Ca V ␤ than mutations on the C-terminal end of the AID helix (Trp 13 , Ile 14 ). These data agree with the three-dimensional model of Ca V 2.3/ Ca V ␤2a built using the crystal coordinates from Ca V 1.2/Ca V ␤2a. However, Ca V ␤ binding to the denatured AID peptide did not appear to completely account for the interaction between the Ca V ␣1 and the Ca V ␤ subunits. In particular, Ca V ␤ binding to the AID peptide was ablated after conserved and nonconserved mu-FIG. 6. Three-dimensional models of the AID helix of the human Ca V 2.3 obtained with INSIGHT II using the atomic coordinates for the human Ca V 1.2 AID crystal structure co-crystallized with Ca V ␤2a at a 2.0 Å resolution (1T0J.pdb) as a template (19). The backbone of the AID helix is colored in white. Key residues are shown in CPK representation. A, front view of the AID helix with the native isoleucine residue at position 14. B, front view of the AID helix with the bulkier arginine residue at position 14. The atoms of the residue at position 14 are color-coded with carbon atoms (green), oxygen (red), and nitrogen (blue). The hydrogen atoms are not shown. The residue at position 14 makes van der Waals interactions with four hydrophobic residues from Ca V ␤2a shown in CPK representation with a single color with the residue: Leu 352 (cyan), Val 241 (magenta), Met 244 (gray), and Met 245 (violet). The I14R mutation pushes Ca V ␤2a residues away from the AID E helix and results in a significantly increased r.m.s.d. value (1.53 Å as compared with 0.55 Å for I14L) for the four hydrophobic Ca V ␤2a residues. The BUILDER module was used to construct the molecular replacements onto the Ca V 2.3 model and all energy minimizations were carried out with the DISCOVER module as described under "Experimental Procedures." tations of the Tyr 10 whereas Ca V ␤-induced modulation of whole cell currents was mostly preserved with Y10F, Y10A, Y10K, and Y10R mutations. It can be speculated that the interaction is conserved when both the AID helix and Ca V ␤ are in their native conformation. It remains also possible that functional modulation of HVA Ca V currents by Ca V ␤ involves more partners than just the AID helix. Nonetheless, binding and functional modulation data were quite well correlated for the C-terminal Ile 14 residue. We have thus proposed that fewer structural constraints such as the deeper interaction site and the lower degree of hydrophobicity at Tyr 10 could account in part for these differences. Altogether, the Ca V ␤-induced modulation data are compatible with a region-specific specialization of the AID helix: the Nterminal residues playing a critical role in a hinged-lid type inactivation mechanism (44) while the C-terminal residues anchoring Ca V ␤ subunit interaction.