Structural basis for the differential effects of CaBP1 and calmodulin on Ca_V1.2 calcium-dependent inactivation

Abstract

Calcium-binding protein 1 (CaBP1), a calmodulin (CaM) homolog, endows certain voltage-gated calcium channels (Ca_Vs) with unusual properties. CaBP1 inhibits Ca_V1.2 calcium-dependent inactivation (CDI) and introduces calcium-dependent facilitation (CDF). Here, we show that the ability of CaBP1 to inhibit Ca_V1.2 CDI and induce CDF arises from interaction between the CaBP1 N-lobe and interlobe linker residue Glu94. Unlike CaM, where functional EF hands are essential for channel modulation, CDI inhibition does not require functional CaBP1 EF-hands. Furthermore, CaBP1-mediated CDF has different molecular requirements than CaM-mediated CDF. Overall, the data show that CaBP1 comprises two structural modules having separate functions: similar to CaM, the CaBP1 C-lobe serves as a high-affinity anchor that binds the Ca_V1.2 IQ domain at a site that overlaps with the Ca²⁺/CaM C-lobe site, whereas the N-lobe/linker module houses the elements required for channel modulation. Discovery of this division provides the framework for understanding how CaBP1 regulates Ca_Vs.

Introduction

Voltage-gated calcium channels (Ca_Vs) serve as a major source of calcium influx in excitable cells (Catterall, 2000). Because calcium ions are chemical messengers (Clapham, 2007), influx through Ca_Vs can directly link membrane potential charges to stimulation of intracellular signaling cascades (Catterall, 2000). Although high-voltage activated Ca_Vs consist of four essential components (Van Petegem and Minor, 2006): a Ca_V1 or Ca_V2 pore-forming Ca_Vα₁ (Catterall, 2000), a cytoplasmic Ca_Vβ (Dolphin, 2003), Ca_Vα₂δ (Davies et al., 2007), and calmodulin (CaM) (Pitt, 2007), the composition of these large protein complexes is not monolithic. In some contexts, such as cerebellar and hippocampal neurons (Lee et al., 2002; Zhou et al., 2004), photoreceptor synapses (Haeseleer et al., 2004), and auditory hair cells (Cui et al., 2007; Yang et al., 2006), members from a family of calcium binding proteins homologous to CaM, known as CaBPs (Haeseleer et al., 2000), can replace CaM. This component exchange has profound effects on how Ca_Vs respond to calcium entry and results in channels that have strikingly different functional properties than those modulated by CaM (Cui et al., 2007; Few et al., 2005; Lautermilch et al., 2005; Lee et al., 2002; Yang et al., 2006; Zhou et al., 2004; Zhou et al., 2005).

When modulated by CaM, many Ca_V1s exhibit a strong calcium-dependent inactivation (CDI) that limits calcium influx during depolarization (Dunlap, 2007). In contrast, Ca_V1s under the influence of CaBP1, a CaBP abundant in the brain and retina (Haeseleer et al., 2000), have dramatically altered functional properties. CaBP1 blocks Ca_V1.2 (Zhou et al., 2004; Zhou et al., 2005) and Ca_V1.3 (Cui et al., 2007; Yang et al., 2006) CDI and introduces an increase in Ca_V1.2 (Zhou et al., 2004) peak current upon repetitive stimulation, calcium-dependent facilitation (CDF). These effects depend on displacement of CaM from the Ca_Vα₁ C-terminal IQ domain (Yang et al., 2006; Zhou et al., 2004), a channel element that is critical for CaM-mediated CDI (Erickson et al., 2003; Zühlke et al., 1999).

As with many calcium sensor proteins (Burgoyne et al., 2004; Haeseleer et al., 2002; Weiss and Burgoyne, 2002), both CaBP1 and CaM comprise two lobes that bear a pair of EF-hand calcium binding motifs (Haeseleer et al., 2000). Beyond this common architecture, there are a number of differences. CaBP1 is myristoylated at its N-terminus, has an N-lobe that is largely insensitive to calcium (K_d > 100 μM) (Wingard et al., 2005), and a longer interlobe linker. Apart from recent NMR structures of individual CaBP1 lobes (Li et al., 2009), little is known about the high-resolution structure of complete CaBPs or the details of how they interact with Ca_Vs. Thus, it has remained uncertain which CaBP1 features contribute to the stark functional differences that CaBP1 brings to Ca_V modulation.

We set out to answer this question with respect to Ca_V1.2 and find that the key element is an interaction between the N-lobe and an interlobe linker residue that is conserved among CaBP family members. We further show that CaBP1 comprises two structural modules having separate functions. The CaBP1 C-lobe shares both structural and functional similarity with the CaM C-lobe and acts as a high-affinity anchor that binds the Ca_V1.2 IQ domain at a site that overlaps with the Ca²⁺/CaM C-lobe binding site. In contrast, the N-lobe/linker module contains the elements required for the unique functional effects CaBP1 has on Ca_V1.2. Delineation of these conserved CaBP structural elements provides the required framework for understanding how this family of calcium sensor proteins affects Ca_Vs and other target proteins in a manner distinct from CaM.

Results

CaBP1 N-lobe and interlobe linker are crucial for CDI inhibition

We used two-electrode voltage clamp recording of Xenopus oocytes to test whether CaBP1 N-terminal myristoylation contributes to CaBP1 modulation of Ca_V1.2 as this lipid anchor is important for CaBP1 modulation of Ca_V2.1 (Few et al., 2005) and is an obvious difference from CaM (Figure 1A). Examination of calcium currents produced when Ca_V1.2 was co-expressed with a CaBP1 G2A mutant (Few et al., 2005) that blocks myristoylation (Rocque et al., 1993) demonstrated that CaBP1 G2A could inhibit CDI as effectively as wild type CaBP1 (Figures 1B and C, and Table 1). The enzyme that carries out myristoylation is unable to process G2A substrates and is conserved (Rocque et al., 1993). Nevertheless, to ensure that the N-terminal modification was eliminated, we also examined the consequences of deletion of the entire CaBP1 myristoylation site, CaBP1(Δ2-15). Similar to CaBP1 G2A, CaBP1(Δ2-15) fully inhibited CDI (Figures 1B and C, and Table 1). Thus, N-terminal myristoylation is not required for CaBP1 inhibition of Ca_V1.2 CDI.

Figure 1.

CaBP1-CaM chimeras demonstrate the importance of the N-lobe and interlobe linker. A, Structure based alignment of CaBP1 and CaM. N-lobe and C-lobe are shaded green and blue, respectively. CaBP1 N-terminal myristoylation segment and interlobe linker are indicated by the yellow and orange bars, respectively. EF hand positions are indicated above the sequence according to standard nomenclature (Gifford et al., 2007). Secondary structure elements are indicated. CaBP1 residues Y56 and E94 are highlighted in red. B, CaBP1-mediated CDI inhibition is unaffected by myristoylation site mutants. Representative normalized I_Ca traces at a test potential of +20 mV for Ca_V1.2 expressed with Ca_Vβ_2a and CaBP1 myristoylation mutants G2A and Δ2-15 (grey), CaBP1 (orange), or CaM (blue). C, t_i300 values for data from ‘B’. D, Representative normalized I_Ca traces at a test potential of +20 mV for Ca_V1.2 expressed with Ca_Vβ_2a and CaBP1-CaM chimeras (grey), wild-type CaBP1 (orange), or CaM (blue). Icons show the origin of the domain elements: CaBP1 (orange), CaM (blue). Chimeras containing CaBP1 N-lobe all contain the myristoylation domain. E, t_i300 values for data from ‘D’. F, Relative current amplitude changes of I_Ca in response to 40 +20 mV pulses delivered at 3 Hz for Ca_V1.2 expressed with Ca_Vβ_2a and the indicated chimeras. First (black), last (red) and every fourth trace (grey) are shown. G, Relative current changes between the last and first pulse for Ca_V1.2 expressed with Ca_Vβ_2a and CaBP1, CaM, or the indicated chimeras. Parentheses indicate the number of oocytes tested. Error bars indicate s.d. See also Figures S1 and S2.

Table 1.

Ca_V1.2 inactivation parameters

Data are expressed as mean values ± standard deviation. All experiments correspond to the mean of at least two separate oocyte batches.

Because CaBP1 (Li et al., 2009) and CaM (Masino et al., 2000; Tsalkova and Privalov, 1985) lobes fold independently, we designed a set of CaBP1-CaM chimeras based on exchange of three elements (N-lobe, interlobe linker, and C-lobe) and tested their effects on Ca_V1.2 expressed in Xenopus oocytes. We designate the chimeras with three letter names, e.g. MMB, in which the first, second, and third positions denote N-lobe, interlobe linker, and C-lobe, respectively, and each letter denotes the element origin: ‘M’ for CaM and ‘B’ for CaBP1.

All CaM N-lobe chimeras (MMB, MBB, MBM) lacked the hallmark Ca_V1.2 CDI inhibition caused by CaBP1, and instead supported CDI similar to CaM (Figures 1D, E, and Table 1). CaM interlobe linker chimeras (MMB, BMB and BMM) also allowed CDI to proceed. Notably, the chimera having the CaM interlobe linker exchanged into CaBP1, BMB, failed to inhibit CDI. Pulldown assays showed that all chimeras retained ability to interact with the Ca_V1.2 IQ domain (Figure S1) and eliminate the possibility that the absence of CDI inhibition arose from failure of the chimeras to fold properly and bind the Ca_V1.2 IQ domain. Together, these results suggest that the inability of MMB, MBB, MBM, MMB, BMB, and BMM to inhibit CDI arises from the absence of elements in the CaBP1 N-lobe (MMB, MBB, MBM, MMB) and CaBP1 interlobe linker (MMB, BMB, and BMM). In further support of this, we found that BBM, which has CaBP1 N-lobe and interlobe linker joined to CaM C-lobe, blocks Ca_V1.2 CDI even more potently than CaBP1 (Figures 1D and E). BBM additionally causes slower Ca_V1.2 activation (Figure 1E). Taken together, the results from the chimeras strongly suggest that the key elements underlying the CaBP1 and CaM functional differences with respect to CDI reside in the N-terminal lobe and interlobe linker.

Besides CDI inhibition, CaBP1 causes Ca_V1.2 CDF (Zhou et al., 2004) (Figure 1F). We tested whether the CaBP1-CaM chimeras retained this property. Chimeras bearing either the CaM N-lobe (MMB, MBB, MBM, MMB) or CaM interlobe linker (MMB, BMB, and BMM) were unable to support Ca_V1.2 CDF. Except for BMB, which introduced a larger progressive loss in current amplitude (Figures 1F and G), channels expressed with these chimeras were indistinguishable from channels expressed with CaM. In contrast, BBM caused Ca_V1.2 CDF that was ~2-fold stronger than that of CaBP1 (Figures 1F and G). Hence, BBM embodies both main functional properties of CaBP1, the ability to inhibit Ca_V1.2 CDI and the ability to confer CDF. Together, the data indicate that the CaBP1 N-lobe and interlobe linker bear the modulatory elements unique to CaBP1, whereas CaBP1 and CaM C-terminal lobes perform similar functions.

CaBP1 interlobe linker functional properties

The CaBP1 and CaM interlobe linker lengths differ by four residues (Figure 1A), a divergence conserved among CaBPs (Haeseleer et al., 2000) (Figure S2). Given the apparent importance of the interlobe linker, we investigated whether its length, composition, or both were crucial for CaBP1 function. CaBP1 constructs having an interlobe linker composed of the first four (93-96, ‘AETA’) or last four interlobe linker residues (97-100, ‘DMIG’) failed to inhibit CDI (Figures 2A and B). Replacement of the CaBP1 interlobe linker with a duplication of the CaM interlobe linker (DTDSDTDS), octa-alanine (8A), or octa-glycine (8G) also failed to inhibit CDI (Figures 2A and B).

Figure 2.

Interlobe linker length and sequence have a crucial role in CaBP1 function. A, Representative normalized I_Ca traces at a test potential of +20 mV for Ca_V1.2 expressed with Ca_Vβ_2a and indicated CaBP1 linker mutants (grey), CaBP1 (orange), or CaM (blue). B, t_i300 values for data from ‘A’. BMB values are from Figure 1. C, Relative current amplitude changes of I_Ca and I_Ba in response to 40 +20 mV pulses delivered at 3 Hz for Ca_V1.2 expressed with Ca_Vβ_2a and CaBP1, CaM, or the DMIG linker mutant. First (black), last (red) and every fourth trace (grey) are shown. D, Averaged frequency dependent change of I_Ca current amplitude for Ca_V1.2 expressed with Ca_Vβ_2a and CaBP1, CaM, or the indicated linker mutants. E, Comparison of I_Ca recovery from inactivation for Ca_V1.2 expressed with Ca_Vβ_2a and CaM (blue) or the CaBP1 DMIG linker mutant (orange) in response to two 450 ms pulses to +20 mV separated by variable time intervals. Parentheses indicate the number of oocytes tested. Error bars indicate s.d. See also Figures S1 and S2.

Unexpectedly, the protocol used to induce CDF caused Ca_V1.2 to display a strong, calcium-dependent reduction in current amplitude in the presence of all of the CaBP1 interlobe linker mutants (Figure 2C). This phenomenon, which we term ‘CDI tachyphylaxis’, is stronger than the small current suppression seen with CaM (Figure 2D) and provides evidence that the interlobe linker manipulations did not incapacitate the CaBP1 mutants. This is corroborated by pulldown experiments that show the individual mutants retain the ability to bind the Ca_V1.2 IQ domain (Figure S1B). Further analysis of the DMIG mutant showed that the CDI tachyphylaxis arises from changes in recovery from inactivation (Figures 2E). Following a depolarization pulse, Ca_V1.2 co-expressed with CaM shows essentially full recovery after 750 ms. In contrast, ~7% of Ca_V1.2 co-expressed with DMIG fail to recover in the same period. Over longer interpulse periods, both CaM and DMIG containing channels recover fully (Figure 2E). Taken together, the data from the chimeras and interlobe linker mutants establish that both the length and composition CaBP1 interlobe linker are crucial for modulation of Ca_V1.2.

N- and C-lobes contribute to CaBP1-Ca_V1.2 IQ domain affinity

We turned to isothermal titration calorimetry (ITC) to investigate how CaBP1 interacts with the Ca_V1.2 IQ domain, the domain that is essential for CaBP1 CDI inhibition (Zhou et al., 2004). Experiments using individual CaBP1 lobes in the presence of 1 mM calcium, Ca²⁺/N-lobe_BP and Ca²⁺/C-lobe_BP, revealed that each has a single binding site on the Ca_V1.2 IQ domain (Figure 3A, B and Table 2). Ca²⁺/N-lobe_BP binding is an endothermic reaction having modest affinity (K_d = 1.11 ± 0.08 μM), whereas Ca²⁺/C-lobe_BP binds ~100-fold stronger via an exothermic reaction that has an affinity (K_d = 10.5 ± 1.9 nM) similar to Ca²⁺/C-lobe_CaM (Van Petegem et al., 2005). Competition experiments in which Ca²⁺/N-lobe_BP was titrated into a pre-formed Ca²⁺/C-lobe_BP-Ca_V1.2 IQ domain complex demonstrate that Ca²⁺/C-lobe_BP prevents Ca²⁺/N-lobe_BP binding and indicate that the binding sites overlap (Figure 3C). As expected from the affinity differences, Ca²⁺/C-lobe_BP can displace Ca²⁺/N-lobe_BP from the Ca_V1.2 IQ domain (Figure 3D). The ability of both Ca²⁺/CaBP1 lobes to bind the Ca_V1.2 IQ domain at an overlapping site is reminiscent of the behavior of individual Ca²⁺/CaM lobes (Kim et al., 2008; Van Petegem et al., 2005).

Figure 3.

Characterization of Ca²⁺/CaBP1-Ca_V1.2 IQ domain interactions. Exemplar ITC titrations for: A, 75 μM Ca²⁺/N-lobe_BP into 7.5 μM IQ domain. B, 75 μM Ca²⁺/C-lobe_BP into 7.5 μM IQ domain. C, 75 μM Ca²⁺/N-lobe_BP into 7.5 μM Ca_V1.2 IQ domain and 11.25 μM Ca²⁺/C-lobe_BP. D, 75 μM Ca²⁺/C-lobe_BP into 7.5 μM IQ domain and 50 μM Ca²⁺/N-lobe_BP. E, 75 μM Ca²⁺/CaBP1 into 5.5 μM IQ domain. F, 75 μM Ca²⁺/C-lobe_BP and 75 μM Ca²⁺/N-lobe_BP into 7.5 μM IQ domain. Red squares denote estimates of enthalpy changes using binding parameters derived from the single lobe binding.G, 75 μM Ca²⁺/CaBP1 into 7.5 μM IQ domain and 50 μM Ca²⁺/N-lobe_BP. H, Thermodyamic cycle for analysis of the binding of CaBP1 (Ka₃) to the Ca_V1.2 IQ domain. Ka₁ describes Ca²⁺/N-lobe_Bp binding. Ka₂ describes the competition of Ca²⁺/N-lobe_BP by CaBP1. Cartoons depict the observed binding reactions.

Table 2.

Titration calorimetry data

Cell	Syringe	T (°C)	N	Kd (nM)	n	ΔH (kcal mol−1)	ΔS (cal mol−1 K−1)	Ka/Ka (wt)	ΔΔG (kcal mol−1)
Cav1.2 IQ	Ca2+/N-lobeBP	15	0.82 ± 0.05	1,100 ± 80	3	2.09 ± 0.40	34.5 ± 1.5	-	-
	Ca2+/C-lobeBP	15	0.90 ± 0.07	10.5 ± 1.9	4	−2.88 ± 0.19	26.5 ± 0.6	-	-
	Ca2+/C-lobeBP	25	0.98 ± 0.11	15.0 ± 0.7	2	−8.63 ± 0.24	6.9 ± 0.9	-	-
	Ca2+/C-lobeCaM*	15	0.98 ± 0.18	2.63 ± 0.07	2	−6.77 ± 0.24	15.75 ± 0.78	-	-
Cav1.2 IQ Ca2+/N-lobe	Ca2+/C-lobeBP	15	0.96 ± 0.01	16.6 ± 0.7	2	−3.17 ± 0.05	24.1 ± 0.7	-	-
	Ca2+/CaBP1	15	0.93 ± 0.12	0.288 ± 0.070	2	−4.72 ± 0.07	26.6 ± 0.7	-	-
	Ca2+/CaBP1 E94A	15	1.04 ± 0.04	0.336 ± 0.097	2	−7.75 ± 0.31	15.9 ± 0.5	0.86 ± 0.32	−0.06 ± 0.22
Cav1.2 IQ F1628A	Ca2+/C-lobeBP	25	0.88 ± 0.03	727 ± 224	2	−4.87 ± 0.11	11.8 ± 1.0	0.021 ± 0.006	−2.25 ± 0.18
CaV1.2 IQ F1631A	Ca2+/C-lobeBP	15	0.83 ± 0.03	345 ± 107	2	−2.22 ± 0.20	21.9 ± 1.3	0.030 ± 0.011	−1.82 ± 0.21
	Ca2+/C-lobeCaM	15	0.80 ± 0.03	17.3 ± 4.3	2	−4.87 ± 0.08	18.7 ± 0.8	0.152 ± 0.038	−1.22 ± 0.14

Data are expressed as mean values ± standard deviation. ΔΔG is calculated with respect to the wild-type lobe or protein at 15°C with the exception of data for F1628A, which are calculated at 25°C.

^*indicates values from (Van Petegem et al., 2005). See also Figure S5.

Unlike the simple, individual lobe binding isotherms, titration of full-length Ca²⁺/CaBP1 into the Ca_V1.2 IQ domain showed a V-shaped isotherm that could not be attributed to a single binding event (Figure 3E). Because Ca²⁺/N-lobe_BP and Ca²⁺/C-lobe_BP bind to the Ca_V1.2 IQ domain in a competitive manner, we wondered whether the complex isotherm arose from contributions of each lobe. ITC experiments in which equimolar portions of individual Ca²⁺/CaBP1 lobes were titrated into the Ca_V1.2 IQ domain produced a binding isotherm very similar to that of full-length Ca²⁺/CaBP1 (Figure 3F). Further, when we used parameters from the single lobe experiments to simulate the isotherm in which Ca²⁺/N-lobe_BP and Ca²⁺/C-lobe_BP bind to a single, overlapping IQ domain site, we found excellent correspondence to the measured isotherm (Figure 3F, red triangles). These results indicate that the ‘V’-shaped nature of the isotherm represents a sequence of two events: (1) independent binding of Ca²⁺/N-lobe_BP and Ca²⁺/C-lobe_BP to separate Ca_V1.2 IQ domains when the IQ domain is in excess, and (2) replacement of Ca²⁺/N-lobe_BP by Ca²⁺/C-lobe_BP as the IQ domain becomes limiting. The ability to dissect the binding reaction this way set the stage for an experiment to determine the thermodynamics of the Ca²⁺/CaBP1- Ca_V1.2 IQ domain interaction and test whether the ‘V’-shaped isotherm observed with full-length Ca²⁺/CaBP1 (Figure 3E) arose from a similar course of events.

Titration of Ca²⁺/CaBP1 into pre-formed Ca²⁺/N-lobe_BP-Ca_V1.2 IQ domain complexes yielded a titration isotherm having a single transition (Figure 3G). Analysis using competition ligand binding by displacement (Sigurskjold, 2000) in the context of the Figure 3H thermodynamic cycle reveals that Ca²⁺/CaBP1 binds the Ca_V1.2 IQ domain ~40-fold stronger than measured for Ca²⁺/C-lobe_BP alone (K_d= 296 ± 70 pM)(Table 2). This increased affinity is accompanied by a binding enthalpy increase that indicates that Ca²⁺/N-lobe_BP, the interlobe linker, or both contribute to the binding reaction by interacting with the Ca_V1.2 IQ domain at sites separate from the Ca²⁺/C-lobe_BP binding site. Taken together, the ITC experiments establish that CaBP Ca²⁺/C-lobe interacts with the Ca_V1.2 IQ domain in a manner similar to Ca²⁺/CaM C-lobe, and show that elements from the entire CaBP1 participate Ca_V1.2 IQ domain binding.

Functional EF-hands not required for CDI inhibition

CaBP1 has four EF hands; however, the importance of metal binding to N-lobe EF-hands is unclear. EF1 has weak Ca²⁺ affinity (Wingard et al., 2005) and EF2 is non-functional due to the lack of a canonical residue at the ‘z’ position (Figure 1A)(Gifford et al., 2007; Haeseleer et al., 2000). To test whether CaBP1 inhibition of Ca_V1.2 CDI requires the ability of the CaBP1 EF-hands to bind metal ions, we examined the consequences of introduction of a D➔A mutation at the ‘x’ position of each functional EF hand. This mutation is analogous to those used to dissect CaM EF hand function (Peterson et al., 1999) and should reduce metal-binding ability substantially and. CaBP1 bearing a disrupted EF1 was functionally indistinguishable from wild-type (Figures 4A and B, and Table 1). In contrast, EF3, EF4, and EF34 mutations diminished but did not eliminate the ability of CaBP1 to inhibit Ca_V1.2 CDI. Thus, the capacity of CaBP1 C-lobe EF hands to bind metal ions is important but not essential for CDI inhibition. This relative insensitivity to EF hand disruption stands in contrast to CaM where functional C-lobe EF-hands are required for CDI (Alseikhan et al., 2002; Peterson et al., 1999). The effects of CaBP1 EF34 are reminiscent of the ability of the CaM EF34 mutant to block CDI (Peterson et al., 1999) and suggest that part of the CaBP1 mechanism may be competition with apo-CaM. In contrast to the minor effects on CDI inhibition, the EF3 and EF4 mutants greatly diminished Ca_V1.2 CDF (Figure 4C and D) and indicate that CaBP1-mediated CDF requires Ca²⁺ binding to the C-lobe.

Figure 4.

C-terminal EF-hands are required for CaBP1-mediated CDF but not for inhibition of CDI. A, Representative normalized I_Ca traces at a test potential of +20 mV for Ca_V1.2 expressed with Ca_Vβ_2a and indicated CaBP1 EF hand mutants (grey), CaBP1 (orange), or CaM (blue). B, t_i300 values for data from ‘A’. C, Relative current amplitude changes of I_Ca in response to 40 +20 mV pulses delivered at 3 Hz for Ca_V1.2 expressed with Ca_Vβ_2a and the indicated chimeras indicated CaBP1 EF hand mutants, CaBP1, or CaM. D, Averaged frequency dependent change of I_Ca current amplitude for Ca_V1.2 expressed with Ca_Vβ_2a and the indicated chimeras indicated CaBP1 EF hand mutants, CaBP1, or CaM. Parentheses indicate the number of oocytes tested. Error bars indicate s.d. P-values with respect to wild-type were calculated using student’s t-Test: N.S., not significant, P > 0.05; ***, P < 0.005.

CaBP1 crystal structure

To understand how the CaBP1 N-lobe and interlobe linker contribute to function, we crystallized and determined the structure of the CaBP1 functional core, CaBP1(Δ2-15). CaBP1(Δ2-15) crystallized in the I23 space group and diffracted X-rays to 2.9Å (Table 3). Surface entropy reduction (Derewenda and Vekilov, 2006) identified a mutant, CaBP1(Δ2-15) K130A, that did not alter function (Table 1), gave crystals having a different space group, P3₁21 and improved resolution, 2.4Å, and that enabled solution by MAD (Hendrickson and Ogata, 1997) using selenomethionine-substituted protein. The 2.4Å structure was used for molecular replacement of the I23 crystal form. As there were no major differences between the structures, we used chain A from the 2.4Å structure for analysis.

Table 3.

Crystallographic data collection and refinement statistics

	CaBP1	CaBP1 K130A
Data Collection		Native	SeMet
Space group	I23	P3121	P3121
Cell dimensions a/b/c (Å)	206, 206, 206	69, 69, 344	69, 69, 345
α/β/γ (°)	90, 90, 90	90, 90, 120	90, 90, 120
Resolution (Å)	50-2.9 (3.06 – 2.90)	50 - 2.4 (2.53 - 2.40)	50-2.9 (3.06 - 2.90)
			Peak	Inflection	Remote
Wavelength	1.11587	1.11587	0.97958	0.97972	1.01987
Rsym (%)	9.5 (45.1)	7.2 (43.2)	9.0 (38.2)	8.5 (34.9)	7.9 (28.6)
I / σI	6.4 (1.6)	7.1 (1.6)	6.7 (1.8)	6.9 (2.0)	7.2 (2.4)
Completeness (%)	97.5 (96.2)	99.4 (99.0)	100.0 (100.0)	100.0 (100.0)	99.9 (100.0)
Redundancy	4.0 (2.1)	3.9 (2.8)	6.9 (7.2)	6.9 (7.2)	6.9 (7.2)
Unique reflections	32321	43362
Average B-factor	99.7	54.4
Refinement
Rwork/Rfree (%)	21.3 / 25.0	25.2 / 29.5
No. of chains in AU	6	6
No. of protein atoms	7093	6835
No. of ligand atoms	12	42
RMSD bond lengths (Å)	0.020	0.016
RMSD angles (°)	1.697	1.556
Ramachandran best /disallowed regions (%)	94.9/0.0	96.1/0.0

One crystal was used for each dataset. Values in parentheses are for highest-resolution shell.

CaBP1 has four EF-hands arranged into two lobes. Unexpectedly, a well-ordered interlobe linker (residues 93-100) connects the lobes (Figure 5A). N-lobe EF hands, EF1 and EF2, lack ligands; whereas C-lobe EF hands, EF3 and EF4, each contain a calcium ion. The compact structure is different from what has been observed by NMR, which found independent lobes and a disordered interlobe linker (Li et al., 2009). Comparison with crystal structures of compact (Fallon and Quiocho, 2003) and extended (Wilson and Brunger, 2000) Ca²⁺/CaM conformations reveals that the CaBP1 lobe orientation is different from both (Figure S3A). Despite the compact conformation, the CaBP1 lobes make few contacts with each other and bury only 270 Ų.

Figure 5.

Structure of CaBP1. A, Cartoon representation of CaBP1. N-lobe (green), linker (red), and C-lobe (blue) are indicated. Calcium ions are depicted as yellow spheres. Secondary structure elements are labeled. B, Superposition of CaBP1 apo-N-lobe structure determined by X-ray (green) and Mg²⁺-bound form determined by NMR (2K7B)(Li et al., 2009) (light green). Mg²⁺ ion is a yellow sphere. B-hairpin from the Mg²⁺-bound form is shown (salmon) and labeled. Equivalent residues in the apo-lobe structure are highlighted in red. C, Superposition of CaBP1 Ca²⁺/C-lobe structure determined by X-ray (blue) and NMR (grey). D, Superposition of CaBP1 Ca²⁺/C-lobe (dark blue) and CaM Ca²⁺/C-lobe (marine) (1EXR)(Van Petegem et al., 2005). Hydrophobic residues involved in CaM Ca²⁺/C-lobe-IQ domain binding and their CaBP1 Ca²⁺/C-lobe counterparts are shown as sticks. E, Details of the Glu94/N-lobe cleft interaction. Glu94 and residues from the N-lobe are shown as sticks. N-lobe is colored green and shown in surface representation. Linker is indicated in red. E94 is shown in salmon. Hydrogen bond interaction between E94 and the backbone carbonyls of Met54 is shown as black dashes. See also Figure S3.

N-lobe is in the calcium-free apo-form. Neither EF1 nor EF2 have electron density indicative of a metal ion, despite crystallization conditions containing millimolar Ca²⁺. Apo-N-lobe shows a number of differences from the Mg²⁺-bound N-lobe NMR structure (Li et al., 2009) (RMSD_Cα=2.5Å). The most notable is in the E1 helix (Figure 5B). In the apo-form, E1 extends by an extra turn that encompasses the ‘x’ and ‘y’ ligand-binding residues, Asp35 and Asp37, and clashes with the metal ion position in the Mg²⁺-bound structure (Li et al., 2009) (Figure 5B). Additionally, apo-N-lobe EF1 and EF2 are further away from each other and lack the short β-hairpin that connects them in the Mg²⁺-bound structure (Figure 5B). These differences indicate that conformational changes, particularly in EF1, accompany metal binding and are consistent with divalent-ion induced chemical shift changes (Wingard et al., 2005). Although there are some differences in packing, the N-lobe hydrophobic core is buried in both forms (Figure S3B) and does not undergo large changes in exposure seen in the classic apo➔Ca²⁺/CaM transition (Gifford et al., 2007; Grabarek, 2006) or in the CaBP1 apo➔Ca²⁺/C-lobe transition (Li et al., 2009). The lack of large divalent ion induced conformational changes agrees with the absence of EF1 D➔A functional effects (Figures 4A and B) and suggests that N-lobe metal binding is unnecessary for CaBP1-mediated Ca_V1.2 modulation.

CaBP1 Ca²⁺/C-lobe is similar to the NMR structure (RMSD_Cα = 1.3Å over 60 Cα positions) (Figure 5C) and resembles the Ca²⁺/CaM C-lobe ‘open’ conformation. The structural similarity is even better when compared to Ca²⁺/CaM crystal structures regardless of whether the comparison is made with Ca²⁺/CaM C-lobe alone (Wilson and Brunger, 2000) or bound to a target such as the Ca_V1.2 IQ domain (Van Petegem et al., 2005), (RMSD_Cα =1.0 Å over 60 Cα positions for each) (Figure 5D). This similarity extends to many of the hydrophobic pocket sidechains. Consequently, the CaBP1 Ca²⁺/C-lobe hydrophobic pocket appears ready to engage the Ca_V1.2 IQ domain in a manner similar to CaM Ca²⁺/C-lobe (Van Petegem et al., 2005).

The most striking aspect of the structure is the well-ordered interlobe linker (Figures 5A and S3C), which is clamped to the N-lobe by insertion of Glu94 into a largely hydrophobic N-lobe cleft lined by Tyr56 and Leu91 (Figure 5E). This interaction is absent from the NMR structure (Li et al., 2009). Within the N-lobe cleft, the Glu94 sidechain is within hydrogen bonding distance of the Met54 backbone carbonyl oxygen. Comparison of non-crystallographically related asymmetric unit monomers reveals that the single helical turn in the C-terminal half of the linker (residues 95-99) forms a hinge point for a 5° rotation between the lobes (Figure S3D) and indicates that even in the compact state there is flexibility between the lobes.

Ca²⁺/C-lobe_BP and Ca²⁺/C-lobe_CaM-IQ domain binding sites overlap

MMB and BBM chimeras having swapped CaBP1 and CaM C-lobes retain the properties of the parent, suggesting that CaBP1 and CaM C-lobes are functionally interchangeable (Figure 1). This, together with the strong structural similarity between Ca²⁺/C-lobe_BP and Ca²⁺/C-lobe_CaM bound to the Ca_V1.2 IQ domain (Figure 5D) and the prior evidence that CaBP1 acts by competing with CaM for the IQ domain (Zhou et al., 2004) led us to wonder whether CaBP1 and CaM Ca²⁺/C-lobes interact with the Ca_V1.2 IQ domain at a common site. Titration of Ca²⁺/C-lobe_BP into a pre-formed Ca²⁺/C-lobe_CaM-Ca_V1.2 IQ domain complex shows that Ca²⁺/C-lobe_CaM blocks binding of Ca²⁺/C-lobe_BP (Figure 6A) and supports the idea that the high affinity Ca_V1.2 IQ domain binding sites for each overlap. To explore this possibility further, we examined the effects of alanine mutations at F1628 and F1631, two residues that serve as anchors for Ca²⁺/C-lobe_CaM-Ca_V1.2 IQ domain interactions (Van Petegem et al., 2005).

Figure 6.

Ca²⁺/C-lobe_BP and Ca²⁺/C-lobe_CaM binding sites overlap on the Ca_V1.2 IQ domain. Exemplar ITC titrations for: A, 75 μM Ca²⁺/C-lobe_BP into a solution of 7.5 μM IQ domain and 11.25 μM Ca²⁺/C-lobe_CaM. B, 75 μM Ca²⁺/C-lobe_BP into 7.5 μM IQ domain F1628A mutant, 15°C. C, 75 μM Ca²⁺/C-lobe_BP into 7.5 μM IQ domain F1628A mutant, 25°C. D, 75 μM Ca²⁺/C-lobe_BP into wild-type IQ domain, 25°C. E, 75 μM Ca²⁺/C-lobe_BP into 7.5 μM IQ domain F1631A mutant. F, 75 μM Ca²⁺/C-lobe_CaM into 7.5 μM IQ domain F1631A mutant. Cartoons depict the observed binding modes for Ca²⁺/C-lobe_BP (dark blue) and Ca²⁺/C-lobe_CaM (light blue). Stars indicate mutant IQ domains.

ITC experiments with Ca²⁺/C-lobe_BP and the F1628A peptide at 15°C, the temperature of our other ITC analyses, gave a small, endothermic signal (Figure 6B). Therefore, we evaluated Ca²⁺/C-lobe_BP binding to the F1628A peptide at 25°C, where the signal was robust (Figure 6C) and compared the results to titration of Ca²⁺/C-lobe_BP into the wild-type peptide at the same temperature (Figure 6D). The data show that Ca_V1.2 F1628A reduced Ca²⁺/C-lobe_BP affinity and binding enthalpy by ~48-fold and ~4 kcal mol⁻¹, respectively. The enthalpy reduction magnitude is similar to that caused by F1628A for Ca²⁺/C-lobe_CaM (Van Petegem et al., 2005) and is consistent with a loss of hydrophobic interactions. Measurements with the F1631A peptide (Figure 6E) showed a ~33-fold and ~0.6 kcal mol⁻¹ reduction in affinity and binding enthalpy, respectively. The F1631A change also reduces binding affinity and enthalpy for Ca²⁺/C-lobe_CaM (Figure 6F). Taken together, these experiments strongly support the idea that CaBP1 Ca²⁺/C-lobe interacts with the Ca_V1.2 IQ domain in a manner akin to Ca²⁺/CaM C-lobe (Fallon et al., 2005; Van Petegem et al., 2005).

Interlobe linker Glu94 is essential for CDI inhibition but not Ca_V1.2 IQ domain binding

Given that the CaBP1 interlobe linker is well ordered, makes specific interactions with the N-lobe, and is crucial to CDI inhibition, we used an alanine scan to investigate the functional contribution of each residue. Each alanine mutation diminished CDI inhibition (Figure 7A and B); however, E94A was the only change that completely eliminated CDI inhibition (Figure 7A and B). E94A also caused the appearance of CDI tachyphylaxis (Figure 7C and D) resulting from changes in recovery from inactivation similar to the DMIG linker mutant (Figure S4). To explore the importance of the Glu94 position further, we tested the effects of E94Q and E94D. E94Q almost completely eliminated CDI inhibition (Fig 7A and B) whereas, E94D allowed partial CDI inhibition. These results indicate that the sidechain carboxylate and precise sidechain geometry are essential. To probe the N-lobe/Glu94 interaction further, we replaced Tyr56, which contributes many N-lobe pocket/Glu94 contacts, with alanine. This change substantially diminishes CDI inhibition (Figure 7E and Table 1) and further supports the importance of the N-lobe/Glu94 interaction. The paramount functional role for Glu94 agrees well with the structurally defined N-lobe/Glu94 interaction (Figure 5E).

Figure 7.

Functional examination of the CaBP1 interlobe linker and E94/N-lobe interaction. A, Representative normalized I_Ca traces at a test potential of +20 mV for Ca_V1.2 expressed with Ca_Vβ_2a and indicated CaBP1 linker mutants (grey), CaBP1 (orange), or CaM (blue). B, t_i300 values for data from ‘A’. C, Relative current changes between the last and first pulse for Ca_V1.2 expressed with Ca_Vβ_2a and CaBP1, CaM, or the indicated mutants. D, Averaged frequency dependent change of I_Ca amplitude for ‘C’. E, and F, Representative normalized I_Ca traces at a test potential of +20 mV for Ca_V1.2 expressed with Ca_Vβ_2a and indicated CaBP1 mutants (grey), CaBP1 (orange), or CaM (blue). Graphs compare t_i300 values. Parentheses indicate the number of oocytes tested. Error bars indicate s.d. See also Figure S4 and Figure S6.

Despite the effects that E94A had on function, ITC experiments revealed that the loss of the N-lobe/Glu94 interaction caused by E94A altered ΔH and ΔS but spared the affinity for the Ca_V1.2 IQ domain (K_d = 0.336 ± 0.097 nM, Table 2, Figure S5). This result prompted us to test whether the ordered nature of the linker was a key element of CaBP1 function. We made a mutant (4G) that maintained the N-lobe/Glu94 interaction but that converted the C-terminal half of the linker (residues 97-100) to polyglycine. In contrast to the devastating effect of E94A, 4G retained an ability to inhibit CDI that was on par with the single alanine mutants (Figure 7F). Thus, although both the Glu94/N-lobe interaction and interlobe linker length (Figure 2) are essential for CaBP1 function, the order seen in the C-terminal half is not.

CaBP1 and CaM mediated CDF are two distinct processes

CaM-mediated Ca_V1.2 CDF requires Ca_Vβ (Findeisen and Minor, 2009; Grueter et al., 2006; Hudmon et al., 2005) and CaMKII (Anderson et al., 1994; Grueter et al., 2006; Hudmon et al., 2005; Yuan and Bers, 1994). Although CaMKII activation is not necessary for CaBP1-mediated CDF (Zhou et al., 2004), the extent to which Ca_V1.2 CaM-mediated CDF (Van Petegem et al., 2005; Zühlke et al., 1999; Zühlke et al., 2000) and CaBP1-mediated CDF (Zhou et al., 2004) share molecular requirements has remained unclear. To test whether CaBP1-mediated CDF necessitates the presence of Ca_Vβ, we used a Ca_V1.2 mutant, ‘HotA’, that cannot bind Ca_Vβ (Van Petegem et al., 2008) and that eliminates CaM-mediated CDF (Findeisen and Minor, 2009). Unlike the case with CaM, CaBP1 supports CDF when co-expressed with Ca_V1.2 HotA or wild-type Ca_V1.2 in the absence of Ca_Vβ_2a (Figure 8A and B). Thus, CaBP1-mediated CDF does not require Ca_Vβ.

Figure 8.

Comparison of CaBP1-mediated and CaM-mediated Ca_V1.2 CDF. A, Relative current changes between the last and first pulse in response to 40 +20 mV pulses delivered at 3 Hz for Ca_V1.2 and Ca_V1.2 mutants expressed with indicated Ca_Vβ_2a, CaBP1, and CaM combinations. B, Averaged frequency dependent change of I_Ca amplitude for ‘A’. C, Representative normalized I_Ca traces at a test potential of +20 mV for Ca_V1.2 with Ca_Vβ_2a and CaBP1 (orange) or CaM (blue) and Ca_V1.2 TripleA (grey) expressed with Ca_Vβ_2a and CaBP1. Parentheses indicate the number of oocytes tested. Error bars indicate s.d.

Ca_V1.2 CaM-mediated CDF is unmasked by the Ca_V1.2 IQ domain mutation, I1624A (Van Petegem et al., 2005; Zühlke et al., 1999; Zühlke et al., 2000) (Figure 8A and B). If CaBP1-mediated CDF were similar to CaM-mediated CDF, one might expect that I1624A would enhance CaBP1-mediated CDF. Contrary to this expectation, co-expression of Ca_V1.2 I1624A with CaBP1 produces CDF having a magnitude indistinguishable from that seen with CaBP1 and Ca_V1.2 (Figure 8A and B). Ca_V1.2 IQ domain residues F1618, Y1619, and F1622 are involved in Ca²⁺/CaM N-lobe-IQ domain interactions that play a role in Ca_V1.2 CaM-mediated CDF (Hudmon et al., 2005; Van Petegem et al., 2008). The triple alanine mutant, F1618A/Y1619A/F1622A, (‘TripleA’), eliminates CaM-mediated CDF (Van Petegem et al., 2008). In contrast, TripleA had no effect on CaBP1 mediated CDF (Figure 8A and B) or on CaBP1 CDI inhibition (Figure 8C). The insensitivity of CaBP1-medated CDF to manipulations that affect CaM-mediated CDF demonstrates that Ca_V1.2 CaM-mediated CDF and Ca_V1.2 CaBP1-mediated CDF are different and indicates that their underlying molecular mechanisms are different.

Discussion

CaBPs belong to a large calcium sensor family found throughout the nervous system (Burgoyne et al., 2004; Haeseleer et al., 2002; Weiss and Burgoyne, 2002) and closely resemble CaM (Haeseleer et al., 2002; Weiss and Burgoyne, 2002). Accordingly, CaBPs interact with diverse classes of CaM-modulated proteins (Haynes et al., 2006): Ca_Vs (Lee et al., 2002; Yang et al., 2002; Zhou et al., 2004), IP3 Receptors (Kasri et al., 2004; White et al., 2006; Yang et al., 2002), TRP channels (Kinoshita-Kawada et al., 2005), and myosin 1c (Tang et al., 2007). CaBP-target interactions impart functional changes distinct from those caused by CaM (Kasri et al., 2004; Lee et al., 2002; Yang et al., 2002; Zhou et al., 2004) and may diversify neuronal responses to calcium signals. CaBPs interact with and reshape the functional properties of certain Ca_Vs (Cui et al., 2007; Haeseleer et al., 2004; Lee et al., 2002; Tippens and Lee, 2007; Yang et al., 2006; Zhou et al., 2004). Because Ca_Vs are prominent in cellular calcium signaling pathways (Clapham, 2007), CaBP remodeling of Ca_V activity should reach well beyond electrical excitation properties.

CDI is an important type of Ca_V feedback modulation that limits cellular calcium entry in response to electrical activity. There are a number of situations in which CDI is overridden (Striessnig, 2007). One mechanism is CaBP1 substitution for CaM within Ca_V multiprotein complexes in retinal (Haeseleer et al., 2004) and auditory (Cui et al., 2007; Yang et al., 2006) neurons. This component change also happens in the brain (Zhou et al., 2004), blocks CDI in Ca_V1.2 and Ca_V1.3 (Cui et al., 2007; Yang et al., 2006; Zhou et al., 2004; Zhou et al., 2005), and introduces CDF to Ca_V1.2 (Zhou et al., 2004). CaBP1 and CaM have two independently-folded EF-hand containing lobes separated by a linker (Li et al., 2009; Masino et al., 2000; Tsalkova and Privalov, 1985). This likeness presents the question of which elements endow CaBP1 with the ability to alter Ca_V1.2 behavior differently from CaM. Our chimera-based analysis eliminated a role for CaBP1 N-terminal myristoylation, which is important for Ca_V2.1 modulation (Few et al., 2005), and identified the N-lobe and interlobe linker as the elements that allow CaBP1 to inhibit CDI and enable CDF in Ca_V1.2 (Figure 1).

The ability of CaBP1 C-lobe to bind calcium was important but not integral to CDI inhibition. This lack of sensitivity would seem to conflict with the reported calcium-dependency of the CaBP1-Ca_V1.2 IQ domain interaction (Zhou et al., 2004). However, the ability of CaBP1-EF3, EF4, and EF34 mutants to block CDI is reminiscent of the ability of the CaM EF34 mutant to block CDI (Peterson et al., 1999) and suggests that part of the CaBP1 CDI inhibition mechanism could arise from simple competition with apo-CaM binding to Ca_V1.2. Nevertheless, this effect cannot embody the entire mechanism. Our data show that only the BBM chimera but not BMM or MBM chimeras or the E94A mutant block CDI. If competition were the only effect, these mutants would be potent CDI inhibitors. Thus, the data strongly suggest that besides CaM competition, there must be an active role for the CaBP1 N-lobe/Glu94 module in Ca_V1.2 CDI inhibition and this role is the dominant contributor.

The CaBP1 crystal structure revealed an unanticipated N-lobe/Glu94 interaction (Figure 5E) that is indispensible for CaBP1 CDI inhibition and CaBP1-mediated CDF of Ca_V1.2 (Figure 7). It is notable that this position is conserved CaBP2, CaBP4, and CaBP5 (Figure S2), particularly as CaBP4 inhibits Ca_V1.3 CDI (Cui et al., 2007; Yang et al., 2006). Thus, in addition to serving as part of the CaBP1 module required for Ca_V1.2 CDI inhibition, the N-lobe/Glu94 interaction may have a role in other CaBPs.

The Ca_V1.2 IQ domain is the essential CaM modulation site (Peterson et al., 1999; Pitt et al., 2001) and main CaBP1 binding site (Zhou et al., 2004). Our data uncover a strong structural similarity between the CaM and CaBP1 Ca²⁺/C-lobes (Figure 5D). This structural mimicry suggests that both bind the Ca_V1.2 IQ domain in a similar manner. This idea is supported by the interchangeability of CaBP1 and CaM C-lobes with respect to parent molecule Ca_V1.2 modulation (Figures 1D-G) and is consistent with the similar functional sensitivity of both proteins to Ca_V1.2 IQ domain mutations (Zhou et al., 2004). Further, our ITC studies reveal that Ca²⁺/C-lobe_BP and Ca²⁺/C-lobe_CaM bind with similar affinities to a mutually exclusive Ca_V1.2 IQ domain site (Figure 6A, C and D, Table 2) and that mutations of residues that form anchors for Ca²⁺/C-lobe_CaM greatly diminish the ability of Ca²⁺/C-lobe_BP to bind (Figure 6C-F). Together, these results suggest that CaBP1 and CaM C-terminal lobes have the same main function, formation of high affinity attachment to the Ca_V1.2 IQ domain and that this attachment site overlaps.

ITC experiments with full-length CaBP1 show that it binds the Ca_V1.2 IQ domain ~40-fold stronger than Ca²⁺/C-lobe_BP and that the CaBP1 E94A mutation, which eliminates CDI inhibition, has no effect on binding affinity. Superposition of the Ca²⁺/CaBP1 and Ca²⁺/CaM C-lobes in the context of the Ca²⁺/CaM-Ca_V1.2 IQ domain complex (Van Petegem et al., 2005) indicates that the compact CaBP1 conformation observed in the crystal structure is incompatible with Ca_V1.2 IQ domain binding. Many CaBP1 N-lobe/Glu94 atoms clash with the Ca_V1.2 IQ helix (Figure S6). Therefore, although there is good evidence that the Ca²⁺/CaBP1 C-lobe binds the Ca_V1.2 IQ domain in a mode similar to Ca²⁺/CaM C-lobe, CaBP1 binding must involve a conformation in which the CaBP1 N-lobe is in different position than seen in the CaBP1 structure.

Our ITC experiments indicate that the CaBP1 N-lobe participates in Ca_V1.2 IQ domain binding. If CaBP1 Ca²⁺/C-lobe and CaM Ca²⁺/C-lobe bind the Ca_V1.2 IQ domain similarly, one candidate CaBP1 N-lobe interaction site would be the Ca_V1.2 IQ domain residues that comprise the CaM Ca²⁺/N-lobe binding site (F1618, Y1619, and F1622) (Fallon et al., 2005; Van Petegem et al., 2005). Nevertheless, functional analysis of the Ca_V1.2 TripleA mutant, which disrupts these anchors and eliminates CaM-mediated CDF (Van Petegem et al., 2005), reveals that neither CaBP1-mediated CDF or CDI inhibition depend on these residues. Therefore, CaBP1 N-lobe must interact with Ca_V1.2 differently than Ca²⁺/CaM N-lobe. Conformational differences between the CaBP1 crystal and solution structures (Li et al., 2009) show that CaBP1 can adopt various interlobe conformations. Further, CaBP1 modulation of Ca_V1.2 tolerates substitutions that render the C-terminal part of the interlobe linker maximally flexible (Figure 7F). Thus, we propose that the CaBP1 binds the Ca_V1.2 IQ domain using a similar C-lobe position to CaM but an alternate N-lobe pose such as an open structure that would maintain the CaBP1 N-lobe/Glu94 interaction but flex in the C-terminal portion of the interlobe linker. This could allow CaBP1 to interact with the Ca_V1.2 cytoplasmic N-terminal domain, a channel element that has the capacity to bind CaBP1 and that is required for CaBP1 modulation of Ca_V1.2 (Zhou et al., 2005).

CaBP1 affects Ca_V1s and Ca_V2s differently. CaBP1 inhibits Ca_V1.2 and Ca_V1.3 CDI (Cui et al., 2007; Yang et al., 2006; Zhou et al., 2004; Zhou et al., 2005), whereas it accelerates Ca_V2.1 CDI and changes activation (Few et al., 2005; Lee et al., 2002). Additionally, CaBP1 blocks Ca_V2.1 CDF (Few et al., 2005; Lee et al., 2002). These Ca_V2.1 functional changes require CaBP1 N-terminal myristoylation (Few et al., 2005), a modification that we find irrelevant for CaBP1 modulation of Ca_V1.2. This result is corroborated the ability of a non-myristyolated CaBP1 splice variant, caldendrin, to block Ca_V1.2 CDI (Tippens and Lee, 2007). The mechanistic basis for the difference between Ca_V1.2 and Ca_V2.1 modulation by CaBP1 remains unresolved. CDI and CDF in Ca_V1s and Ca_V2s arise from the actions of different CaM lobes (Dunlap, 2007) that are correlated with different Ca²⁺/CaM binding modes on the Ca_V1 (Van Petegem et al., 2005) and Ca_V2 (Kim et al., 2008) IQ domains (however, cf., (Mori et al., 2008)). Similar to Ca²⁺/CaM (Kim et al., 2008; Van Petegem et al., 2005), we find that both CaBP1 lobes can bind the IQ domain. It is tempting to speculate that the differential effects of CaBP1 on Ca_V1 and Ca_V2 CDI and CDF and the differential effects of the CaM lobes with respect to the same processes are related. CaBP1 also binds other CaM binding channel segments (Dick et al., 2008; Ivanina et al., 2000; Lee et al., 1999). It is possible that the differential effects also require interactions with these or other channel domains. Delineating how CaBPs act should not only unravel CaBP-specific mechanisms, but may help dissect the still imperfectly understood action of CaM with respect to CDI and CDF. Our data provide the framework for investigation of these issues.

We have identified the features that allow CaBP1 to modulate Ca_V1.2 differently from CaM. Our studies reveal an unanticipated essential role for the interlobe linker and show that CaBP1 has two structural modules that encode two separate functions. CaBP1 C-lobe serves as the IQ domain anchor, whereas the N-lobe/linker module contains the elements required for channel modulation. The elements comprising the N-lobe/Glu94 interaction are conserved in many CaBPs that interact with diverse targets. Thus, the structural underpinnings outlined here are likely to aid understanding how CaBPs act on a wide range of targets that have dual modulation by CaM and CaBPs.

Experimental procedures

Expression and Purification

Individual human CaBP1-S (AF169148) lobes were expressed as fusion proteins from a pET28 vector denoted ‘HMT’ (Van Petegem et al., 2004) in E. coli BL21(DE3)pLysS. CaBP1(Δ2-15) and CaBP1(Δ2-15) K130A, and selenium methionine substituted CaBP1(Δ2-15) K130A were expressed a pEGST vector in E. coli BL21(DE3)pLysS. Ca_V1.2 IQ domains were made as described previously (Van Petegem et al., 2005). Expression and purification details can be found in Supplementary Experimental Procedures.

Crystallization and Refinement

CaBP1(Δ2-15) was crystallized by hanging drop vapor diffusion at 20°C from equal volumes of protein in buffer A and well solution containing 1.9 M (NH₄)₂SO₄, 2-4 % 1,2-propanediol and 0.1 M citrate/sodium citrate, pH 5.5. Native and selenomethionine-substituted CaBP1(Δ2-15) K130A were crystallized similarly except that 1,2-propanediol was replaced by 5-7 % 1,6-hexanediol. Crystals were flash-frozen in well solution plus 20% glycerol. Diffraction data were collected at Beamline 8.3.1 (Advanced Light Source, Lawrence Berkeley National Laboratories) and processed using MOSFLM 7.0.4 (Leslie, 1992). A three-wavelength MAD experiment was performed on crystals of SeMet-substituted CaBP1(Δ2-15) K130A. 64 initial selenium positions were located using SHELXD (Uson and Sheldrick, 1999). After phasing and density modification using DM, the resulting electron density allowed building of an initial model that was improved by iterative cycles of manual building and refinement against native data using Refmac5 (Collaborative Computational Project, 1994). TLS-tensors were used throughout refinement. Tight NCS restraints were employed initially and relaxed in later refinement cycles. Each final model consists of six CaBP1 monomers.

Isothermal titration calorimetry

Titrations were performed at 15°C or 25°C using a VP-ITC Microcalorimeter (MicroCal). Samples were dialyzed overnight at 4°C (Slide-A-Lyzer, 3.5 kDa molecular weight cut-off, Thermo Scientific) against 5 mM KCl, 1 mM CaCl₂, 10 mM Hepes, pH 7.4. After centrifugation (30′, 70,000 × g, 4 °C) protein concentrations were determined by absorbance at 280 nm (Edelhoch, 1967). All samples were degassed for 5− prior to loading into the VP-ITC. Ca_V1.2 IQ domain peptide at concentrations of 5.5 or 7.5 μM were titrated with 75 μM CaBP1 constructs using a schedule of one 4 μl titrant injection followed by 29 injections of 10 μl. For competition experiments, 7.5 μM Ca_V1.2 IQ domain peptide was preincubated with 50 μM of Ca²⁺/N-lobe_BP and titrated with a solution of 75 μM of CaBP1. To correct the baseline, either heat of dilution from titrations of injectant into buffer was subtracted (Figures 3A and 3D) or the final titration points were used to estimate the baseline (Figures 3B, 3F-H). Data were processed with MicroCal Origin 7.0 using a single site binding model or a competitive binding model, where appropriate.

For binding isotherm analysis of the simultaneous addition of Ca²⁺/N-lobe_BP and Ca²⁺/C-lobe_BP to the Ca_V1.2 IQ domain, we used a modified version of competition ligand binding by displacement ITC (Sigurskjold, 2000) that takes into account both ligands are added simultaneously. Thus, the stoichiometric concentration of B in the calorimeter cell, equation [26] from (Sigurskjold, 2000), [B]_0,i = f_t,i[B]₀,, is replaced by [B]_0,i = [B]₂ (1-f_t,i), where [B]_s is concentration of Ca²⁺/N-lobe_BP in the injection syringe and f_t,i is the total dilution factor after the ith injection, respectively.

Electrophysiology

Human Ca_V1.2 (α₁C77, CAA84346) in pcDNA3.1 (Invitrogen), rat Ca_Vβ_2a (NP 446303) in pGEMHE, and Ca_Vα₂δ-1 (NM_00182276) in pcDNA3.1 were co-expressed with either Homo sapiens CaM (NM_006888) or the short isoform of human CaBP1-S (AF169148) in pGEMHE for Xenopus oocyte two-electrode voltage clamp experiments. Point mutations were introduced using QuikChange (Stratagene). For chimera construction, CaM and CaBP1 were divided into three elements: N-lobe (CaM 1-78 and CaBP1 1-92), linker (CaM 79-82 and CaBP1 93-100 for CaBP1), and C-lobe (CaM 83-149 and CaBP1 101-167). All chimeras having CaBP1 N-lobe include the N-terminal myristoylation domain. Inter-element junctions were framed by introduction of EcoRV sites generated using QuikChange. Elements were excised using EcoRV and either HindIII or XhoI present in the parent vector and combined to produce the chimeras. Following chimera generation, junction sites were returned to the native sequence using QuikChange.

After overnight linearization (XhoI for pcDNA3.1 and NheI for pGEMHE), capped mRNAs were synthesized (T7 mMessenger kit (Ambion)) according to the manufacturer’s protocols. RNA concentrations were determined by A_260nm. 50 nl of Ca_Vα₁, Ca_Vβ, Ca_Vα₂δ-1, and CaBP1 or CaM mRNA at a molar ratio of 1:1:1:20 were injected into defolliculated stage VI Xenopus oocytes using Nanoject II injector (Drummond Scientific). Oocytes were kept at 18 °C in ND96 medium containing penicillin and streptomycin. Two-electrode voltage-clamp experiments were performed 2 to 4 days post-injection using a GeneClamp 500B (Axon Instruments) amplifier controlled by a 1,200 MHz processor computer (Celeron, Gateway) running CLAMPEX 8.2.0.244, and digitized at 1kHz with a Digidata 1332A (Axon Instruments). Immediately prior to recording, oocytes were injected with 47 nl of 100 mM BAPTA to minimize Ca²⁺-activated Cl⁻ current. Recording solutions contained either 40 mM Ba(OH)₂ or 40 mM Ca(NO₃)₂, 50 mM NaOH, 1 mM KOH, and 10 mM HEPES, adjusted to pH 7.4 using HNO₃. Electrodes were filled with 3M KCl and had 0.3-2.0 MΩ resistances. Leak currents were subtracted using a P/4 protocol. Currents were analyzed with Clampfit 8.2 (Axon Instruments). Oocytes were superfused during recording using a Valvelink 16 controller (Automate Scientific). Holding potential for all experiments was −90mV. Results are from at least two independent oocyte batches. ti₃₀₀ values were calculated from normalized currents at +20 mV and represent the percentage inactivation after 300 milliseconds. Inactivation τ values were calculated at a test potential of +20 mV as described (Findeisen and Minor, 2009). CDF was elicited by a train of 40, 50 ms pulses to +20 mV at 3Hz and calculated as the ratio of the peak current from the last pulse divided by the first. Recovery from inactivation was measured by a protocol with two 450 ms pulses to +20 mV separated by variable time intervals. Consecutive sweeps were separated by 30 s.

Pull-down experiments

Bacterial pellets from a 50 ml culture of CaBP1, CaBP1 mutants, CaM, or CaBP1/CaM chimeras co-expressed with HMT-tagged Ca_V1.2 IQ domain were resuspended in 1.6 ml lysis buffer (10% sucrose, 150 mM KCl, 5 mM MgCl₂, 1 mM CaCl₂, 25 μg/ml DNAaseI, 1 mM PMSF, 100 mM Tris, pH 8.8) and lysed by sonication. 100 μl amylose resin in buffer AmyA (250 mM KCl, 1 mM CaCl₂, 10 mM Hepes/KOH, pH 7.4) was incubated with the bacterial lysate for 30 minutes at 4 °C with gentle agitation. After four washes with 500 μl AmyA, bound material was eluted in 500 μl AmyA containing 10 mM maltose. Input and eluate fractions were analyzed by SDS-PAGE.