Calmodulin regulates Cav3 T-type channels at their gating brake

Calcium (Cav1 and Cav2) and sodium channels possess homologous CaM-binding motifs, known as IQ motifs in their C termini, which associate with calmodulin (CaM), a universal calcium sensor. Cav3 T-type channels, which serve as pacemakers of the mammalian brain and heart, lack a C-terminal IQ motif. We illustrate that T-type channels associate with CaM using co-immunoprecipitation experiments and single particle cryo-electron microscopy. We demonstrate that protostome invertebrate (LCav3) and human Cav3.1, Cav3.2, and Cav3.3 T-type channels specifically associate with CaM at helix 2 of the gating brake in the I–II linker of the channels. Isothermal titration calorimetry results revealed that the gating brake and CaM bind each other with high-nanomolar affinity. We show that the gating brake assumes a helical conformation upon binding CaM, with associated conformational changes to both CaM lobes as indicated by amide chemical shifts of the amino acids of CaM in 1H-15N HSQC NMR spectra. Intact Ca2+-binding sites on CaM and an intact gating brake sequence (first 39 amino acids of the I–II linker) were required in Cav3.2 channels to prevent the runaway gating phenotype, a hyperpolarizing shift in voltage sensitivities and faster gating kinetics. We conclude that the presence of high-nanomolar affinity binding sites for CaM at its universal gating brake and its unique form of regulation via the tuning of the voltage range of activity could influence the participation of Cav3 T-type channels in heart and brain rhythms. Our findings may have implications for arrhythmia disorders arising from mutations in the gating brake or CaM.

before the split of animals and fungi, such as apusozoan, Thecamonas trahens, and Na v 1 channels (18), including cnidarian jellyfish, which are the extant relatives of the likely ancestors containing the first metazoan nervous systems (19,20). The only member of the calcium and sodium channel superfamily to lack the reported CaM-binding elements (the C-terminal IQ motif and the N-terminal NSCaTE) are Ca v 3 T-type channels (1,2,21).
Here, we show that Cav3 T-type channels possess high-affinity calmodulin binding at the "gating brake," a helix-loop-helix motif located in the proximal I-II linker of known Ca v 3 T-type channels in the analogous position where accessory Cav␤ subunits regulate Ca v 1 and Ca v 2 channel complexes (22). This gating brake governs the low-voltage dependence of T-type channels, and its absence generates a "runaway gating" phenotype (23)(24)(25)(26). Mutations that alter the function of the Ca v 3.2 gating brake have been found in human patients with absence epilepsy (27,28).
CaM-binding sequences are variable, which can range from a striking likeness to the C-terminal IQ motif from L-type channels, as in the gating brake of the basal metazoan Trichoplax adherens, to diverse sequences that generate a nanomolar affinity for CaM binding in protostome invertebrates such as pond snail LCa v 3 and all human (Ca v 3.1, Ca v 3.2 , and Ca v 3.3) channel isoforms. We demonstrate that CaM facilitates the formation of ␣-helices in gating brake sequences, can pre-associate with Ca v 3 T-type channels without calcium ions, and its binding involves structural conformational changes in both N-and C-terminal pairs of EF hands in CaM and the gating brake. Dialysis of CaM binding (CaMB) peptides, or co-expression of apo-CaM (CaM 1234 ) generates a significant hyper-polarizing shift in voltage sensitivities and faster gating kinetics, consistent with the mutant phenotype of Ca v 3 T-type channels lacking a gating brake in the I-II linker.
Cav3 T-type channels contribute to pacemaker rhythms, such as the conducting system of the heart (29) and low-threshold calcium potentials (also known as low-threshold spikes), that trigger rhythmic burst firing classically associated with thalamic neurons during non-rapid eye movement sleep and to the spike wave discharge during absence seizures (30,31). There is also significant evidence for T-type currents participating in "low threshold" neurotransmitter release (32) and in the maintenance of vascular tone (33). Ca v 3 T-type channels possess a significant "window current" that provides a small but continuous stream of calcium influx through a population of Ca v 3 T-type channels open at rest (34). This calcium, available through Ca v 3 T-type channels at rest, is modeled to contribute to cellular proliferation during organ development, to the aberrant proliferation in many cancers (35), and to the hypertrophied condition of the mammalian heart (36). Classically, Ca v 3 T-type channels are mostly inactivated at rest, and their participation rate is steeply voltage-dependent (37). CaM's regulation of the voltage sensitivities of Ca v 3 T-type channels at the gating brake thus has dramatic consequences to the participation of Ca v 3 T-type channels in normal functions, as well as during development and disease.

CaM complexes with full-length Ca v 3 T-type channels
Full-length mammalian Ca v 3.1 channels were purified from Sf9 insect cells (38) and co-incubated with biotin-CaM com-plexed to a streptavidin-gold conjugate. Fig. 1A (left panel) shows a field of purified Ca v 3.1 (protein appears white) coupled to an electron dense (black) gold particle, indicating CaM is bound. Individual images illustrating different orientations of the single particles of Ca v 3.1 channels alone or calmodulingold alone and Ca v 3.1 channels in complex with calmodulingold are shown in montages of viewpoints on top, bottom, and middle panels, respectively, in Fig. 1A. The orientation of the channel in some images allows identification of the C terminus, a protruding, finger-like projection from the transmembrane domain of Ca v 3.1 as highlighted by an asterisk; a feature previously identified through single-particle electron microscopy 3-D reconstruction of Ca v 3.1 (38). We confirmed the complexing of calmodulin with Ca v 3 T-type channels as co-immunoprecipitants in Western blottings (see Fig. 1B). CaM-GFP fusion proteins transfected and purified from HEK-293T cells and isolated on anti-GFP Sepharose beads were identified in the complex with HA-tagged Ca v 3.2 channels in Western blottings, labeled with anti-HA antibody, as an ϳ259-kDa band (

Ca v 3 T-type channels possess a predicted high-affinity CaM-binding site at their gating brake in the I-II linker
Ca v 3 T-type channels lack the C-terminal IQ motif shared among other calcium channels (Ca v 1 and Ca v 2) and sodium (Na v 1 and Na v 2) channels (1,2,21). Because both EM and coimmunoprecipitation experiments indicated that CaM binds to full-length mammalian Ca v 3.1 and Ca v 3.2 channels, we therefore next sought to delineate the CaM-binding domain. A unique but ubiquitous feature in Ca v 3 T-type channels is a helix-loop-helix, gating brake motif (similar in structure to a region in fumarase enzyme) in the proximal I-II linker that is in the analogous position of ␤ subunit binding to Ca v 1 and Ca v 2 calcium channels (25). It is within the second helix of the gating brake of the I-II cytoplasmic linker where we have found an analogous region to the C-terminal IQ motif, with nanomolar affinity binding to CaM ( Fig. 2A). Sequence alignments of Ca v 3 T-type channels (Fig. 2B) indicated a predicted CaM-binding site (illustrated by red color gradient) in helix-2 of the gating brake of representative metazoan species from cnidarians to the three human Ca v 3.1, Ca v 3.2, and Ca v 3.3 channels. A CaM-binding site is also predicted in Ca v 3 T-type channels by the on-line tool, CaM Target Database (39). Predictions suggest that the gating brake sequence is cytoplasmic, and a helical wheel analysis indicates its amphipathic nature.

Ca v 3 T-type channel from most basal multicellular animals resembles the C-terminal IQ motif of Ca v 1, Ca v 2, Na v 2, and Na v 1 channels
The most basal multicellular organism with a Ca v 3 T-type channel is T. adherens (a placozoan) (40), which has a region spanning the gating brake that is unlike other gating brake sequences (Fig. 2C, blue bar) with a core 17 amino acids that Calmodulin at the T-type channel gating brake resembles the C-terminal CaM-binding IQ motif of Ca v 1, Ca v 2, and Na v 2 channels (Fig. 2, C and D, blue bars). The Ca v 3 T-type channel in the basal multicellular placozoan may represent a structural intermediate involving the positioning of the C-terminal IQ motif in the common ancestor to calcium channels and sodium channels to the proximal I-II linker, before a divergence and appearance of a gating brake helix-loop-helix motif shared among all other metazoan Ca v 3 T-type channels. Comparatively, the IQ motif of Ca v 1 L-type channels is mostly identical from microbial eukaryotes to human homologs (9), where the CaM-binding gating brake motif in Ca v 3 T-type channels is more divergent. The protein similarity of gating brake sequences between protostome invertebrate (e.g. pond snail LCa v 3) and among the human Ca v 3 isoforms is 71-76% (Fig.  2C, red bars). Differences in gating brake sequences may reflect a local adaptability of the gating brake among Ca v 3 T-type channels in different animals.

Gating brake peptides from snail LCa v 3 and the three human Ca v 3.1, Ca v 3.2, and Ca v 3.3 T-type channels associate with Ca 2؉ -CaM
26-mer peptide sequences spanning the gating brake region were synthesized for snail LCa v 3 and human Ca v 3.1, Ca v 3.2, and Ca v 3.3 channels dubbed "Ca v 3 CaMB" peptides. In CaM associates with full-length Ca v 3 T-type channels illustrated by nano-particle cryoelectron microscopy (A) and co-immunoprecipitation (B). A, field of negatively stained (2% w/v uranyl acetate) full-length Ca v 3.1 in complex with CaM in the presence of 1.5 mM Ca 2ϩ . Scale bar is 50 nm. Left panel, CaM has a biotin label that binds streptavidin gold (5 nm). Protein appears white, and the gold that is electron-dense is black. Right panel, top row, montage of purified Ca v 3.1 particles (white density) presenting different views of the channel. Asterisk indicates "tail" domain that we have previously determined corresponds to the C terminus (38). Middle row, montage of Ca v 3.1 particles with CaM-biotin bond indicated by the presence of streptavidin gold (black sphere). Bottom row, control images of CaM-biotin-streptavidin gold particle without Ca v 3.1. Observed relative differences in Ca v 3.1 and streptavidin gold particle sizes relate to the orientation and contrast of each Ca v 3.1 channel particle adhered to the EM support film and the size variation in the streptavidin gold particles (3-6 nm with nominal size of 5 nm, according to the manufacturer). B, CaM-GFP bound to anti-GFP magnetic beads associates with hemagglutinin (HA)-tagged Cav3.2 channel (top panel, middle lane) as illustrated by HA antibody labeling (259-kDa band) of the Ca v 3.2 channel co-immunoprecipitant (IP) bound to beads. 259-kDa HA-tagged Cav3.2 channel band does not appear as a co-immunoprecipitant in the Western blot without co-expression of pCDNA3.1 plasmid inserts containing HA-tagged Cav3.  CaM is predicted to associate with Ca v 3 T-type channels in helix-2 of the gating brake in the proximal I-II linker. A, illustration of the four domain (DI, DII, DIII, and DIV) ϫ 6 transmembrane helices structure common to sodium channels, calcium channels, and NALCN. Ca v 1, Ca v 2, and Na v channels contain a canonical proximal C-terminal IQ motif that possess high-affinity CaM-binding site. A proposed equivalence of the CaM-binding C-terminal IQ motif in Ca v 3 T-type channels is a helix-loop-helix gating brake motif in the proximal I-II linker that is in the analogous position of ␤-subunit binding to Ca v 1 and Ca v 2 calcium channels. B, sequence alignment of T-type channels illustrates a predicted CaM-binding site in helix-2 of the gating brake (illustrated by red color) in representative species from cnidarians to the three human genes. CaM-binding site prediction from CaM Target Database (39). Predictions suggested the putative CaMB peptide sequence is cytoplasmic, and helical wheel analysis indicates its amphipathic nature. In the most primitive metazoan with a T-type channel, Trichoplax (placozoan), the most basal extant T-type channel in multicellular organisms known possesses a gating brake motif resembling the C-terminal IQ motif shared with other calcium (Ca v 1 and Ca v 2) and sodium (Na v 2 and Na v 1) channels (illustrated by blue color). Yellow outlined residues are the sequences for the synthetic CaMB peptides. 39 and 111 amino acid sequence deleted in gating brake deletion mutant, Ca v 3.2(D 453-491 ) and Ca v 3.2(D 429 -539 ) are indicated. C and D, protein similarities among CaMB peptide sequences in the I-II linker, illustrating the high similarity among invertebrate (snail) LCa v 3 and the human Ca v 3.x homologs (C) and the greater similarity of Trichoplax Ca v 3 CaMB peptide sequences to the C-terminal IQ motif of Ca v 1.2 channels (D), rather than other gating brake sequences (C).

Calmodulin at the T-type channel gating brake
addition, a PGPGPG substituted Ca v 3.2 channel peptide (Ca v 3.2mut) was synthesized, which served as a negative control peptide in experiments (see yellow highlighted sequence in Fig. 2B). CaMB peptides promoted a gel-mobility shift with CaM in the presence of 0.1 mM CaCl 2 -containing solution and increasing molar ratios (0.5 to 4ϫ) of snail LCa v 3 or human Ca v 3.1, Ca v 3.2, and Ca v 3.3 channel CaMB peptides (Fig. 3). Mutated Ca v 3.2 CaMB peptide (Ca v 3.2mut) with the PGPGPG substitution did not promote a gel-mobility shift with CaM ( Fig.  3). Ca v 3.3 peptide has the weakest apparent interaction compared with other Ca v 3.1 and Ca v 3.2 peptides, which correlates with its weaker predicted binding affinity (Fig. 2B). Presence of urea attenuated the weakest interacting human Ca v 3.3 channels with CaM without affecting the higher affinity LCa v 3 gating brake peptide with CaM.

Gating brake peptides assume a helical conformation upon association with Ca 2؉ -CaM
Differential circular dichroism (CD) spectroscopy of snail LCa v 3, human Ca v 3.2, and Ca v 3.3 gating brake peptides suggest that the peptides assume a more helical secondary structure upon co-incubation with CaM or with helix-stabilizing agent trifluoroethanol (TFE) (Fig. 4A). Helical propensity of peptides increases after titrating higher TFE concentrations from 10 to 50% (41,42). The apparent ␣-helical formation of LCa v 3 and Ca v 3.2 resembles the CD spectral signature with 10% TFE, whereas Ca v 3.3 resembles more the spectral signature with 25% TFE. Human Ca v 3.1 is unique among the gating brake peptides in appearing to assume an ␣-helical conformation when free in solution (without CaM or TFE) (Fig. 4A), and this can be explained by its high-native alanine content (Fig.   4B). Although the circular dichroism results are suggestive of induction of helical secondary structure in the peptide upon binding CaM, some of the observed ellipticity changes may be induced by tertiary structural changes, not secondary structural ones (43). Mutated peptide (Ca v 3.2mut) has no ␣-helical propensity as predicted due to its proline substitutions (Fig. 4, A and B).

Gating brake peptides bind with a high nanomolar affinity to Ca 2؉ -CaM
CaM binds Ca v 3 CaMB peptides in 0.5 mM CaCl 2 solution with a 1:1 stoichiometry, at nanomolar affinities that vary from 12, 43, 187, and 383 nM for snail LCa v 3, Ca v 3.1, Ca v 3.2, and Ca v 3.3 CaMB peptides, respectively (representative data in Fig.  5 and table of parameters in Fig. 6A). The rank order of measured binding affinities determined in the isothermal titration calorimetry (ITC) (Fig. 6A) is consistent with the rank order of concentration of CaMB peptide to CaM ratio required for a saturating gel-mobility shift, 1.0, 1.5, 1.5, and Ͼ4.0ϫ for snail LCa v 3, Ca v 3.1, Ca v 3.2, and Ca v 3.3 CaMB peptides, respectively (Fig. 3). The Ca v 3.3 CaMB peptide possessed the weakest CaMbinding site of all isoforms as evidenced by the lowest measured affinity with ITC and incomplete gel shift of CaM in native PAGE (Fig. 3). Besides their variable sequence and binding affinity differences for CaM, the Ca v 3 CaMB peptides vary in their thermodynamic properties during their association with CaM. The association appears to be entropy-driven for LCa v 3, LCa v 3.1, and Ca v 3.3 peptides and enthalpy-driven for Ca v 3.2 peptide, as illustrated by the latter's negative enthalpy (exothermic) (Figs. 5 and 6A).

Gating brake peptides bind and induce structural changes to both N-and C-terminal lobes of Ca 2؉ -CaM and will likely associate with CaM in the absence of Ca 2؉
Mutant CaM 1234 possess critical Asp 3 Ala mutations in its EF hands that disable Ca 2ϩ binding, creating a mutant that serves as a proxy for apo-CaM (44). The gating brake may serve as a pre-association site for binding CaM in the absence of calcium ions (apo-CaM), with binding of snail LCa v 3 and human Ca v 3.1 CaMB peptides with Ca 2ϩ -binding deficient CaM (CaM 1234 ), albeit at an ϳ12and ϳ349-fold lower affinity, respectively, compared with wild-type CaM (Figs. 5 and 6A). The interactions with CaMB peptides likely involve both lobes of EF hand pairs of CaM, because both N-and C-terminal CaM (74-aa constructs spanning exactly half of the CaM molecule) were able to bind Ca v 3.2 CaMB peptide, with the N-lobe having a significantly higher affinity (Figs. 5 and 6A). LCa v 3 CaMB peptide will bind to CaM with mutated N-terminal Ca 2ϩ -binding sites (CaM 12 ) but not to CaM with mutated C-terminal Ca 2ϩ -binding sites (CaM 34 ) (Fig. 6A). Chemical shift changes observed in the 1 H-15 N HSQC NMR spectra of CaM in the presence of both snail LCa v 3 (red to cyan color shift, Fig. 6B) and human Ca v 3.2 (red to green color shift, Fig. 6B) indicate a dramatic conformational rearrangement of both N-and C-terminal lobes of CaM in response to high-affinity binding to gating brake sequences. The widespread and dramatic chemical shift differences in Ca 2ϩ -CaM alone compared  5). Note that Ca v 3.1 and Ca v 3.2 peptides both completely displace Ca 2ϩ -CaM at 1.5ϫ peptide to CaM molar ratio, but Ca v 3.1 appears to possess a higher affinity based on sub-saturation levels of peptide. Urea attenuates the weaker interacting human Ca v 3.x channels but does not attenuate the higher affinity LCa v 3 CaMB peptide.

Calmodulin at the T-type channel gating brake
with a CaM-Ca v 3.2 or CaM-LCa v 3 CaMB peptide complex is illustrated in Fig. 6C.

Intracellular dialysis of gating brake peptides, but not mutated gating brake peptides, causes a hyperpolarizing shift in voltage dependence and faster gating kinetics in Ca v 3.2 channels
We tested whether the observed nanomolar affinity of CaM for the gating brake governs the unique gating properties of Ca v 3 T-type channels such as its low voltage of activation. We challenged cells expressing Ca v 3.2 T-type channels with 5 M Ca v 3.2 CaMB peptides pre-loaded in the intracellular pipette, and we evaluated the change of gating behavior in whole-cell patch-clamp recording of transfected Ca v 3.2 T-type channels in HEK-293T cells, before and after equilibration of the intracellular dialysis of gating brake peptide (after 15-20 min) (Fig.  7). We conducted parallel control experiments with the intracellular dialysis of mutant gating brake peptide (Ca v 3.2 CaMBmut) (mutant sequence in Figs. 2B and 4B). The Ca v 3 CaMB peptide challenge to Ca v 3.2 channels causes a phenotype change, which resembles the 33-or 111-amino acid deletion mutant phenotype of Ca v 3 channels lacking the gating brake, Ca v 3.2(D 453-491 ) and Ca v 3.2(D 429 -539 ) (23), respectively (Fig.  7). The gating brake is so named because it is considered a unique regulator of Ca v 3 T-type channels, maintaining them in the closed state at resting membrane potentials (23)(24)(25)(26), where disruptions of the gating brake cause the voltage threshold for activation and inactivation gating to be shifted 20 -30 mV in the hyperpolarizing direction (e.g. see Figs. 8 and 9B) (23,25,26,45), with dramatically accelerated activation and inactivation kinetics (23)(24)(25)(26). Activation curves are shifted to negative potentials by an average of 8 mV (a 50% activation voltage change from Ϫ55.28 Ϯ 1.33 to Ϫ63.10 Ϯ 1.29 mV) after dialysis of Ca v 3.2 CaMB peptide, whereas no corresponding voltage shifts are observed with mutated CaMB peptides (Ca v 3.2 CaMBmut) (Fig. 7A). Steady-state inactivation curves of HEK-293T cell-transfected Ca v 3.2 channels are also shifted toward negative potentials by an average of 16 mV (a 50% inactivation voltage change from Ϫ83.19 Ϯ 2.35 mV to Ϫ99.62 Ϯ 1.37 mV) after 20 min of dialysis with Ca v 3.2 CaMB peptide but not after dialysis with mutated peptide (Ca v 3.2 CaMBmut)-treated cells (Figs. 7B and 8B). Intracellular dialysis of Ca v 3.2 CaMB peptide also confers a significant speeding up of channel activation and inactivation kinetics, compared with Ca v 3.2 channels dialyzed with mutated peptide (Ca v 3.2 CaMBmut). This is quantified as significant reduction in the for exponential fits of the kinetics of activation (Fig. 7A) and inactivation (Fig. 7B) in the presence of Ca v 3.2 CaMB peptide.

Dialyzed gating brake peptides specifically target the gating brake of Ca v 3.2 channels
We observed that the intracellular dialysis of Ca v 3.2GB peptide does not confer additional hyperpolarizing shifts in voltage sensitivity of activation (Fig. 8A) or steady-state inactivation x T-type channels assume an ␣-helical structure when bound to Ca 2؉ -CaM. A, differential circular dichroism (CD) spectra of CaMB peptides co-incubated in the presence of helix-stabilizing agent, TFE (gray dashed line), 20 M CaM (red line), or without CaM (blue line). 26-mer snail LCa v 3, human Ca v 3.2, and Ca v 3.3 CaMB peptides (sequence, bottom right) assume a more helical secondary structure upon addition of CaM and TFE. TFE has a much greater effect on the highest binding affinity peptide, snail LCa v 3, than the others. Ca v 3.1 peptide is the only isoform that is ␣-helical when free in solution, as evidenced by the characteristic negative peaks at 208 and 222 nm. Ca v 3.2-Gbmut peptide was incapable of adopting an ␣-helix even at 50% TFE, nor did it appear to interact with CaM. CaM-alone curve was subtracted from each spectra, converted to mean residue ellipticity (), and smoothed. For TFE experiments, 50 M peptide in PBS was used, whereas the baseline was corrected against PBS as background. B, gating brake peptide sequences highlighting the alanine-rich Ca v 3.1 sequence in blue outline, and the "PGPGPG" substitution in red outline that serves as the Ca v 3.2 gating brake mutant. The 24 residues of the 26-mer peptides with a yellow-colored background are predicted to contain ␣-helices according to PSIPRED version 3.3 (Bioinformatics Group at University College London) (91).

Results of overexpression experiments with mutated CaM (CaM 1234 ) strongly suggest that CaM regulates Ca v 3.2 channels at the gating brake
ITC measurements suggest that the gating brake associates with both N-and C-terminal Ca 2ϩ -binding EF hand pairs. CaM binds to the gating brake in the absence of viable calcium-bind-ing sites (CaM 1234 ). This CaM association for Ca v 3.1 and LCa v 3 CaMB peptide in the absence of calcium is at an ϳ12or ϳ349fold reduced affinity, respectively, compared with wild-type CaM (Fig. 6A). We tested whether calcium occupancy in CaM lobes is required to prevent the gating brake phenotype. We overexpressed a plasmid construct coding for CaM 1234 to com- Representative current traces after 20-min dialyses are shown. Calcium currents were measured for their current-voltage relationships and mono-exponential fits for the kinetics of activation (A) and inactivation (B). Activation and steady-state inactivation curves were created with peak currents generated from a step depolarization from Ϫ110 to Ϫ80 to Ϫ10 mV and to Ϫ30 mV from holding potentials ranging from Ϫ140 to Ϫ40 mV, respectively. Statistical significance (p Ͻ 0.05) using a non-parametric Wilcoxon matched-pairs signed rank test for the kinetic data is shown by * measured before and after equilibration of the intracellular dialysis of Ca v 3 CaMB peptide or mutated CaM binding (CaMBmut) peptide. n values for current-voltage relationship Ca v 3 CaMB peptide (n ϭ 7) and CaMBmut (n ϭ 6). n values for inactivation Ca v 3 CaMB peptide (n ϭ 5) and CaMBmut (n ϭ 6).

Calmodulin at the T-type channel gating brake
pete with native CaM (46). The CaM 1234 overexpression construct included the bicistronic marker mRFP (red color fluorescent protein) (46) to be used as a visual estimate of the expected level of CaM 1234 overexpression, after mercury lamp excitation with the epifluorescence-inverted microscope used for patch-clamp recording. The CaM 1234 overexpression construct also contained two shRNA sequences to knock down CaM (CALM 1,2,3 genes) (46 -48). The shRNA knockdown sequences is reported to be 70% complete in disabling native CaM expression (46 -48). However, knockdown was not effective under our experimental conditions, as we observed no significant physiological consequence after transfection of the shRNA and steady-state inactivation (B) taken from Boltzmann fits of activation and steady-state inactivation curves for wild-type Ca v 3.2 channels and Ca v 3.2 channels with a deleted gating brake (deleted sequence from 429 to 539, see Fig. 2A). Representative current traces at time 0 min and after 20-min dialysis of Ca v 3.2 CaMB peptide are shown. Dialysis of CaMB peptide (after 20 min) generates a hyperpolarizing shift in voltage dependence of wild-type Ca v 3.2 channels, but it does not promote additional hyperpolarizing shifts in voltage dependence on the Ca v 3.2 channels that are highly hyperpolarizing shifted after gating brake deletion. The statistical significance after dialyses of CaMB peptide in Ca v 3.2 channels containing an intact gating brake is p ϭ 0.0156 (*) and p ϭ 0.0313 (*) for V 50% activation and V 50% inactivation, respectively, in a Wilcoxon matched-pairs signed rank test.

Calmodulin at the T-type channel gating brake
CaM knockdown alone. The combined construct of shRNA knockdown of wild-type CaM and overexpression of CaM 1234 had a dominant effect, compared with the overexpression of a similar construct with wild-type CaM (Fig. 9). CaM 1234 causes a hyperpolarizing shift in voltage dependence for activation and inactivation and more rapid gating kinetics (Fig. 9), resembling the phenotype after saturating CaM-binding sites with dialyzed gating brake peptides (Figs. 7 and 8) and resembling the runaway gating phenotype of T-type channels after deletion or mutation of their gating brake sequences (Fig. 8, A and B) (23)(24)(25)(26). The hyperpolarizing voltage shift conferred by CaM 1234 appear to require association with an intact gating brake, because additional hyper-polarizing shifts in voltage dependence are lacking in Ca v 3 channels containing deletion of 39 and 111 amino acids spanning the gating brake, Ca v 3.2(D 453-491 ) and Ca v 3.2(D 429 -539 ), respectively (Fig. 9B) (23).

Ca 2؉ -CaM readily complexes with full-length Ca v 3 T-type channels in vitro
We show here that the two most common isoforms of Ca v 3 T-type channels expressed in both the mammalian brain and heart (Ca v 3.1 and Ca v 3.2) (49), in their full-length configurations, complex with Ca 2ϩ -CaM in vitro. We illustrate using electron microscopy that purified full-length Ca v 3.1 binds calcium-CaM, data supported by co-immunoprecipitation experiments. A recent study by Lee et al. reported that a rat Ca v 3.3 splice variant co-immunoprecipitates with CaM (50), adding support for a universality in CaM associating with all Ca v 3 T-type channels, in the same manner as CaM serves as a ubiquitous resident calcium sensor at the C-terminal IQ tails of Ca v 1 and Ca v 2 calcium channels (1-5) and Na v 1 sodium channels (6). Ca v 3 T-type channels (like all eukaryotic calcium and sodium channels), are large proteins (ϳ250 kDa) that do not readily express well enough for the purposes of extensive in vitro biochemical analyses. We chose candidate peptide fragments based on in silico prediction for resolving finer details of the CaM regulation of Ca v 3 T-type channels. Examination of candidate peptides is the standard approach used for other CaM targets (51).

All known Ca v 3 T-type channels have a predicted high-affinity CaM-binding site in helix2 of the helix-loop-helix of the gating brake in the I-II cytoplasmic linker
Ca v 3 channels lack the canonical C-terminal IQ motif (FILV)QXXX(RK)GXXX(RK)XX(FILVWY) of Ca v 1 L-type Ca 2ϩ channels or (FILV)QXXX(RK)XXXXXXXX Na v 1 sodium channels for CaM association (51). One of the criteria for a potential CaM-binding motif is its omnipresence from microbial eukaryotes to humans such as the IQ motif within different Ca v 1 and Na v channels (9). This criterion was satisfied with the predicted CaM association in the second helix of the helixloop-helix motif forming the Ca v 3 gating brake in the I-II linker of almost all known Ca v 3 channels. All 26-mer gating brake sequences are similar (e.g. 71-76% between pond snail and human Ca v 3 isoforms) but are also completely lacking the string of identical and canonical amino acids conserved in the IQ motif in different Ca v 1 and Na v channels. Interestingly, the most basal of multicellular organisms, the placozoan species T. adherens, possess a Ca v 3 T-type channel (40) but lack an identifiable gating brake motif (only 10% similarity to pond snail and human Ca v 3 isoforms). In its place, the Trichoplax channel sequence is similar to the IQ motifs of L-type calcium channels (48% to the human Ca v 1.2). It is highly suggestive that the earliest eukaryotic Ca v 3 T-type channels had a CaM-binding IQ motif shared with other Ca v and Na v channels, before its divergence and establishment of its unique identity in Ca v 3 T-type channels, as the gating brake.

Conformational changes in both Ca 2؉ -CaM and CaMB peptides are observed upon their high nanomolar affinity association
Nanomolar and specific CaM binding to Ca v 3 CaMB peptides are corroborated in the ability of Ca 2ϩ -CaM to increasingly associate with Ca v 3 CaMB peptides corresponding to increasing concentrations of the Ca v 3 CaMB peptides illustrated by gel-mobility-shift assays. Ca v 3 CaMB peptides appear to undergo dramatic conformational changes (i.e. increasing ␣-felicity) on association with Ca 2ϩ -CaM illustrated by circular dichroism. Although ellipticity changes in the circular dichroism likely involve the induction of ␣-helices in the peptide upon binding CaM, some of the changes may be induced by tertiary structural changes in the proteins themselves (43). The association of the Ca v 3 CaMB peptides with CaM generates dramatic conformational changes in Ca 2ϩ -CaM, illustrated by amide chemical shifts of CaM's amino acids in 1 H-15 N HSQC NMR spectra spanning both N-and C-terminal lobes of CaM. Thermodynamic properties measured within the isothermal cell of the Ca 2ϩ -CaM complexing with Ca v 3 CaMB peptides illustrate that Ca 2ϩ -CaM and Ca v 3 CaMB peptides associate in a 1:1 stoichiometry (N value) at a high nanomolar association constant (K a ).  Fig. 2B. The statistical significance of CaM 1234 overexpression compared with wild-type conditions was p ϭ 0.0028 (***) and p ϭ 0.0308 (*) for V 50% activation and V 50% inactivation, respectively, using a Kruskal-Wallis test followed by a Dunn's multiple comparisons test. The effect of CaM 1234 overexpression compared with wild-type conditions for Ca v 3.2 deletion mutants (D 429 -539 or D 453-491 ) were non-significant using a Mann-Whitney test.

Calmodulin at the T-type channel gating brake Highest affinity association of Ca 2؉ -CaM CaMB peptides requires both lobes of CaM and their Ca 2؉ -binding sites
Further evidence of the specificity of the Ca v 3 CaMB peptides to Ca 2ϩ -CaM is in the requirement of intact Ca 2ϩ -CaM for high nanomolar affinity binding. We show that the N-lobes and C-lobes of CaM alone will associate with Ca v 3 CaMB peptides, but at a lower measurable affinity compared with both CaM lobes. The high-affinity of CaM for Ca v 3 CaMB peptides requires intact Ca 2ϩ -binding sites for CaM, as evidenced by the greatly weakened affinity of CaM in the CaM 1234 configuration for the Ca v 3 CaMB peptides.

Different sequences for CaMB peptides from sample invertebrate LCa v 3 and mammalian Ca v 3 channels possess a high nanomolar affinity and similar binding properties in complexing with Ca 2؉ -CaM
Differing sequences of CaMB peptides for snail LCa v 3 and the three human Ca v 3 gene isoforms are 71-76% similar to one another and possess similar binding properties with Ca 2ϩ -CaM. Snail LCa v 3 and the different human Ca v 3 isoforms of the CaMB peptides associate with a nanomolar affinity and in a 1:1 stoichiometry with Ca 2ϩ -CaM. All CaMB peptides assume a greater ␣-helicity upon association of Ca 2ϩ -CaM, except the alanine-rich Ca v 3.1 isoform, and both snail LCa v 3 and mammalian Ca v 3.2 CaMB peptides generate similar and large conformational changes to both N-and C-lobes of CaM as measured by the amide chemical shifts of CaM's amino acids in 1 H-15 N HSQC NMR spectra. Both snail LCa v 3 and mammalian Ca v 3.2 CaMB peptides also associate with Ca 2ϩ -CaM at a much weakened affinity when Ca 2ϩ -binding sites are mutated in CaM.

Differing sequences of CaMB peptides differ in their relative affinities for Ca 2؉ -CaM
The sequence differences between the Ca v 3 CaMB peptides contributed to significant differences in binding affinities, which varied from rank order of highest to lowest binding affinities of LCa v 3 Ͼ Ca v 3.1 Ͼ Ca v 3.2 Ͼ Ca v 3.3 of 12, 43, 187, and 383 nM, respectively. The rank order of binding affinities measured by the thermodynamic changes resulting from the 1:1 complexing of Ca v 3 CaMB peptides and Ca 2ϩ -CaM measured in the isothermal cell is also the same observed rank order of the fold concentration of Ca v 3 CaMB peptides required for saturation of Ca 2ϩ -CaM with Ca v 3 CaMB peptides for a complete gel-mobility shift.

Nanomolar affinities of Ca v 3 CaMB peptides are approximately equivalent to the measured high affinities of the IQ motif of Ca v 1.2 and Ca v 2.1 channels and greater than the affinity of Ca 2؉ -CaM for Na v 1.5 channels
Our measured binding affinities of Ca v 3 CaMB peptides for N-and C-lobes of Ca 2ϩ -CaM (in nM) of 31 Ϯ 1.9 and 690 Ϯ 42, respectively, is in the relative range of nanomolar affinities for CaM as the C-terminal IQ domain of Ca v 1.2 of 57.6 Ϯ 35.5 and 2.63 Ϯ 0.07 (52) and the C-terminal IQ domain of Ca v 2.1 of 51 Ϯ 20 and 4.32 Ϯ 0.39 (53). The higher nanomolar binding affinity of the Ca v 3.2 channel CaMB peptides for N-and C-lobes of Ca 2ϩ -CaM is greater than NSCaTE, a secondary CaM-binding site of Ca v 1.2 channels (in nM): 650 Ϯ 60 and 1330 Ϯ 300 (16), respectively. Ca v 3.2 channel CaMB peptides are also much higher affinity than the micromolar binding affinity range for the N-and C-lobes of Ca 2ϩ -CaM of Na v 1.5 sodium channel fragments (in nM), such as its C-terminal IQ domain of 2170 Ϯ 400 and 6130 Ϯ 900 (6) and the III-IV linker of 19,270 Ϯ 940 and Ͼ500,000, respectively (54).

Phenotype after dialysis of Ca v 3 CaMB peptides corresponds to immobilizing the gating brake in the I-II linker
Our biochemical studies reflect that Ca 2ϩ -CaM associates with helix-2 of the gating brake in both invertebrate and all mammalian isoforms of Ca v 3 channels, in a high nanomolar affinity with binding characteristics that is reminiscent of the canonical binding of Ca 2ϩ -CaM to the proximal C-terminal IQ motif of Ca v 1 and Ca v 2 channels. We then tracked the consequences of introducing the high-affinity CaMB peptide in expressed Ca v 3.2 channels in real time, by comparing the biophysical changes to the Ca v 3.2 channels from the start of a stabilized patch-clamp recording (0 min) and after complete dialysis of 5 M CaMB peptides from our patch pipette (15-20 min). Biochemically, we observed that equivalent mutated CaMB peptide did not associate with Ca 2ϩ -CaM, so it served as a control to compare the effects of the CaMB peptide. We observed ϳ8and ϳ16-mV hyperpolarization shifts in activation and inactivation curves, respectively, as a result of CaMB peptide dialysis. Faster channel kinetics were also observed, which are a likely a consequence of the hyperpolarizing shift, because corresponding kinetic changes are usually associated with shifts in voltage sensitivities within Cav3 T-type channels (37). The target for the CaMB peptide effect was the gating brake, as we observed no additional hyperpolarizing shift in voltage sensitivities after CaMB peptide dialysis in Ca v 3.2 channel mutants lacking the gating brake. We carried out the complete set of electrophysiology experiments using mammalian Ca v 3.2 because of the availability of gating brake mutants. We will carry out future analyses to confirm the hyperpolarizing shifts in voltage sensitivities in response to CaMB peptide dialyses are shared among other Ca v 3 T-type channels, including invertebrate LCa v 3, all of which have a common nanomolar affinity for CaM binding to the gating brake.

Overexpression of Ca 2؉ -binding deficient CaM (i.e. CaM 1234 ) reveals that CaMB peptides likely compete for CaM binding at the gating brake in the I-II linker of Ca v 3 channels
Involvement of CaM regulation of ion channels has been supported by introducing a Ca 2ϩ -insensitive mutant form of CaM (CaM 1234 ), which possess point mutations within both pairs of EF hands in the N-and C-lobes of CaM to eliminate Ca 2ϩ binding (44). It is believed that CaM pre-associates onto its target in the apo-CaM form (lacking Ca 2ϩ occupied in its EF hands) (55)(56)(57). CaMB peptides have a measurable affinity with CaM 1234 , albeit at levels that are ϳ12and ϳ349-fold lower compared with wild-type CaM, for Ca v 3.1 and LCa v 3 CaMB peptides. Apo-CaM bound to its channel target is considered a necessary antecedent to the subsequent regulation of CaM by Ca 2ϩ . Overexpression of the mutant apo-CaM, which is incapable of binding Ca 2ϩ , is thus considered a competitive inhib-

Calmodulin at the T-type channel gating brake
itor to silence the effects of Ca 2ϩ -CaM. For example, the dominant-negative effects of CaM 1234 overexpression have been used to illustrate the importance of Ca 2ϩ -CaM in the gating of small-conductance Ca 2ϩ -activated K ϩ channels (44), and the necessity of Ca 2ϩ -CaM in the Ca 2ϩ -dependent inactivation of Ca v 1 L-type channels (58,59) as well as the calcium-dependent facilitation of Ca v 2 channels (60). We observe that CaM 1234 causes a modest (6.7 and 7.2 mV) hyperpolarizing shift in activation and inactivation curves of Ca v 3.2 channels, respectively, compared with wild-type CaM overexpression. The likely target for the CaM 1234 effect is the gating brake, as we observed no additional CaM 1234 causing hyperpolarizing shifts on Ca v 3.2 channels lacking the gating brake.

Combined biochemistry and electrophysiology data highly suggest that Ca 2؉ -CaM is a high-affinity resident sensor at the gating brake of Ca v 3 channels
To summarize our data, we illustrate that CaM 1234 , considered as a substitute for apo-CaM (CaM in the absence of calcium), associates with helix-2 of the gating brake at a low micromolar affinity. Upon binding Ca 2ϩ , CaMB peptides and CaM both undergo dramatic conformational changes, leading to CaM's tighter association at the gating brake at a higher nanomolar affinity. Our data suggest that Ca 2ϩ -CaM is behaving as a similar high-affinity and resident sensor at the gating brake in the I-II linker of Ca v 3 T-type channels, in both sample invertebrate and the three human gene isoforms, in a universal manner similar to the calcium-CaM regulation on the IQ motif containing C-terminal tails of Ca v 1 and Ca v 2 calcium channels (1-5) and Na v 1 sodium channels (6). Introducing CaMB peptides or CaM 1234 competitively inhibits Ca 2ϩ -CaM at the gating brake of Ca v 3.2 channels, generating a runaway gating phenotype, with hyperpolarized shifts in voltage sensitivities that are even greater than observed after deletion of the brake from Ca v 3 channels. Because the introduced CaMB peptides or CaM 1234 have no effect on Ca v 3.2 channels with a missing gating brake, it suggests that the primary target for Ca 2ϩ -CaM is the gating brake of Ca v 3 channels. We did not observe the complexing of full-length Ca v 3.1 and Ca v 3.2 channels with CaM in vitro in the absence of calcium ions, but we suspect that it may relate to the structural requirements for calcium ions in recapitulating the native behavior of full-sized, Ca v 3 channels in vitro, the sensitivities of our assays, and the technical challenges in working with whole ion channel proteins of such large sizes (ϳ259 kDa) in vitro.  (50) illustrate that rat Ca v 3.3 channels possess a unique C-terminal sequence that associates with Ca 2ϩ -CaM, conferring the unique CaM regulation observed for Ca v 3.3 channels.

Observed variability of the Ca 2؉ -dependent regulation within Ca v 3 channels is consistent with the variability in Ca 2؉ -CaM binding within the Ca v 1 L-type channels
A suggested hypothesis is that all Ca v 3 channels possess a central, high-affinity Ca 2ϩ -CaM binding at the gating brake in the I-II linker, with additional modulatory, intracellular Ca 2ϩ -CaM-binding sites elsewhere, which can confer novel attributes, such as the Ca 2ϩ -CaM regulation at the C terminus for Ca v 3.3 channels (50). A highly customizable form of Ca 2ϩ -CaM regulation is observed within the "IQ motif centric" calcium (Ca v 1 and Ca v 2) and sodium (Na v 1) channels. The preoccupancy of CaM at the primary IQ motif in the absence of Ca 2ϩ ions is one form of regulation. The specialized Ca v 1.4 homolog localized in vertebrate photoreceptors possesses a highly conserved IQ motif but exhibits almost no calcium-dependent inactivation compared with closely-related Ca v 1.2 channels. Ca v 1.4 possess a distal C-terminal auto-inhibitory domain, which effectively competes for apo-CaM for binding, thus preventing Ca 2ϩ -CaM regulation at Ca v 1.4's conserved proximal C-terminal IQ motif (62). Ca v 1.3 possess a similar auto-inhibitory domain in the distal C terminus generated through alternative splicing, which competes with apo-CaM (63,64).
Another source of variability is in the semi-autonomy and cooperativity among CaM's pairs of Ca 2ϩ -binding EF hands in the N-and C-lobes, each of which is capable of conferring unique attributes. For example, the Ca 2ϩ -CaM-binding site NSCaTE shared within the N terminus of invertebrate Ca v 1 channels (9) and mammalian Ca v 1.2 and Ca v 1.3 channels (16,17) interacts specifically with CaM's N-lobes, conferring a capacity for more global Ca 2ϩ -sensing and buffer-resistant calcium-dependent inactivation in classical L-type channels. The C-lobe of CaM is occupied with the IQ motif, imparting a local, more rapid, and spatially restricted calcium-sensing of Ca 2ϩ ions (16,17).

Observed variability of the Ca 2؉ -dependent regulation within Ca v 3 channels is consistent with the variability in Ca 2؉ -CaM binding within the Ca v 2 synaptic Ca 2؉ channels and Na ؉ channels
The bipartite nature of CaM's N-and C-lobes is also evident in the synaptic Ca v 2 channel family, where Ca 2ϩ regulation manifests as a rapid, calcium-dependent facilitation of the calcium current triggered at CaM's C-lobe and a slower-developing calcium-dependent inactivation imparted by CaM's N-lobe (60,65).
One confounding issue is a potential interplay between calcium-and voltage-dependent properties, such as the consequences of a Ca 2ϩ -CaM-binding site identified in the III-IV linker of Na v 1.5 sodium channels, which is also the region responsible for the fast voltage-dependent inactivation within sodium channels (6, 54, 66). The contributions of both calcium-

Calmodulin at the T-type channel gating brake
and voltage-dependent properties can obscure the typical U-shaped calcium-dependent inactivation (39).

Technical challenges in measuring the Ca 2؉ occupancy of CaM in vivo
Observations in vitro often do not correlate to those gathered within different cell types too, where different factors influence CaM's effects within the cellular environment. For example, when CaM has many high-affinity targets within a cell, it has been reported that CaM availability is likely to be limiting in many cell types (67,68). In such cases, CaM regulation may be unavailable because individual CaMs may be preoccupied at different high-affinity cellular targets. There is also likely competition among differing calcium-sensing proteins. A good example is calcium-binding protein expressed in the vertebrate nervous system, whose expression can antagonize CaM-dependent regulation of Ca v 1 L-type channels (69).
The last confounding issue with studies involving CaM is in the technical challenge in modeling calcium availability for CaM regulation. It is challenging to precisely measure native calcium-buffering conditions because they are variable within a particular cell or among different cell types. Ion channel studies usually involve intracellular dialysis in the whole-cell patchclamp configuration of intracellular calcium chelators (e.g. EGTA and BAPTA), which impose artificial calcium-buffering conditions, whose calcium affinity and on-off rates are not known in comparison with the native buffering conditions in vivo.

Universal "calmodulation" among sodium and Ca 2؉ channels also involves Ca v 3 channels
Ca v 3 channels are the least well studied of the calcium or sodium channels, which means that Ca v 3 channels are interpreted through the lens of what is understood for other calcium and sodium channels (37). A universal IQ centric calmodulation of sodium and calcium channels has more variations than commonality by the many interactions that fine-tune the Ca 2ϩ -CaM regulation outside of the primary C-terminal IQ motif. So it should not be surprising that a functional Ca 2ϩ -CaM regulation is only evident within one of three mammalian Ca v 3 channels in preliminary analyses, and lacking evidence so far within Ca v 3.1 and Ca v 3.2 channels (50). Despite possessing proximal C-terminal Ca 2ϩ -CaM binding in the IQ motif and the III-IV linker, cardiac muscle Na v 1.5 channels possess no obvious Ca 2ϩ -CaM regulation, whereas skeletal muscle Na v 1.4 has a robust calcium-dependent inactivation (2,70,71). A deeper molecular analysis uncovered, for example, a Ca 2ϩ -CaM regulation of classical L-type channels within Ca v 1.4 channels, after deletion of its C-terminal auto-inhibitory domain (62).

Potential implications for Ca v 3 channels possessing Ca 2؉ -CaM regulation at the gating brake
Ca v 3 channels are equally distant structurally to both calcium and sodium channels (37), and it should come as no surprise that its Ca 2ϩ -CaM regulation is also strikingly different, where the equivalence to the C-terminally positioned IQ motif is transposed onto a different cytoplasmic region. The gating brake is a unique element within all Ca v 3 T-type channels, which tunes the unique low voltage sensitivity of this channel class (72). The gating brake is a position that overlaps with the Alpha1 Interaction Site of Ca v 1 and Ca v 2 calcium channels where accessory Ca v ␤ subunits regulate their expression levels and biophysical properties (25). Ca v 3 T-type channels lack equivalency of the high-nanomolar affinity Ca v ␤ subunit-binding site within Ca v 1 and Ca v 2 calcium channels (73). It is possible to confer Ca v ␤ subunit regulation onto Ca v 3 T-type channels after transplanting the I-II linker from Ca v 2 channels (45). But in their native state, Ca v 3 T-type channels possess remarkably similar properties as their in vivo counterparts without requirement of co-expressed accessory subunits in invertebrates (74) and mammals (45,75). Ca v ␤ subunits will boost the level of expression of invertebrate (76) and mammalian (75) Ca v 3 T-type channels, but this is considered to be the result of indirect and generalized consequences of overexpressing a multifunctional, scaffolding, and regulatory protein, such as the Ca v ␤ subunit in vitro (77).
With a lack of equivalent accessory subunit interactions within Ca v 3 channels, opportunities are created for a Ca 2ϩ -CaM regulation within the gating brake of the I-II linker. The gating brake is placed in the I-II loop, just downstream of a stiff ␣-helical linker extending from the segment 6 of domain II, which contributes to pore opening (72). Affixing Ca 2ϩ -CaM to the I-II loop may have structural implications because the I-II loop is immobilized at both ends compared with the C-terminal tails containing the IQ motif of sodium and calcium channels, which is linked only a single end. A more proximate location of the gating brake to the intracellular pore mouth may also have implications, contributing to a unique Ca 2ϩ -CaM regulation within Ca v 3 channels.

Potential importance of Ca 2؉ -CaM regulation of Ca v 3 channels for human physiology and disease
The unique high-affinity binding and regulation of CaM at the gating brake of Ca v 3 T-type channels has dramatic potential consequences, because of the gating brake's role as a primary regulator of the voltage sensitivity of Ca v 3 T-type channels. Ca v 3 T-type channels are unique among sodium and calcium channels by the low voltage range of activation that is close to typical resting membrane potentials. CaM at the gating brake is perfectly positioned in Ca v 3 channels for fine-tuning the channels' responsiveness to voltage, providing a means to govern the capacity of Ca v 3 T-type channels to facilitate pacemaker rhythms where T-type channels have established roles, such as the thalamo-cortical circuits of the brain (where Ca v 3 T-type channel expression is highest) and the conducting system in the heart (49). Human mutations in the gating brake are associated with arrhythmic vulnerabilities such as epilepsy (27,28) or autism spectrum disorder (78), and it is the platform where Ca 2ϩ and CaM can modulate the low-voltage gating behavior in T-type channels, in a manner that is unique from the regulation of CaM universally at the IQ motif in the C-terminal tails of Ca v 1 and Ca v 2 calcium channels (1)(2)(3)(4)(5) and Na v 1 sodium channels (6).

Outside of Ca 2؉ -CaM regulation, Ca v 3 channels are very different from other calcium and sodium channels
Ca v 3 channels uniquely different Ca 2ϩ -CaM regulation goes alongside its many observed differences from other sodium and calcium channels. This includes their ion selectivity, which can range from mostly calcium-conducting (Ca v 3.1 and Ca v 3.2), equally sodium-and calcium-conducting (Ca v 3.3) and as sodium-conducting as Na v 1 sodium channels (protostome invertebrate Ca v 3 channel isoforms).
We have shown that most non-vertebrate Ca v 3 T-type channels have alternative splicing of the L5 extracellular turret in domain II, which converts the mostly calcium-conducting T-type channel into a highly sodium-conducting isoform, whose expression in the invertebrate heart serves as the primary sodium current (74). The plasticity of invertebrate Ca v 3 channels in assuming calcium-or sodium-conducting pore configurations is to expand their niche in the absence of Na v 1 channel genes expressed outside of the nervous systems of many invertebrates, and in the many species of non-vertebrates completely lacking any Na v 1 channel in their genomes (nematodes, echinoderms, hemichordates, and parasitic flatworms) (74). It is an interesting question to address how Ca 2ϩ -CaM regulation varies when Na ϩ is the primary conducting ion rather than Ca 2ϩ through differing invertebrate Ca v 3 channel isoforms.

Potential role for calcium sensing at the gating brake is in influencing calcium ion selectivity within Ca v 3 channels
We have shown that the Ca v 3 channels are endowed with ϳ2-fold increase in the number of cysteines in L5 and L6 extracellular turrets, contributing to uniquely folded extracellular scaffolds (79). In particular, alternative turrets with unique cysteine configurations in both DII L5 and DIV L6 regulate the degree of sodium and calcium conductances through invertebrate Ca v 3 channels. 4 Deletion of these unique L5 cysteines in domain II, alters drug affinity, divalent cation block with Ni 2ϩ and Zn 2ϩ , and also increasing the apparent pore size, allowing for the accommodation of a simultaneous flow of both monovalent and divalent cations. 5 The simultaneous flow of differing divalent and monovalent ions may only be possible within Ca v 3 T-type channels, given its likely wider, more accommodating EEDD selectivity filter compared with other Ca v 1 and Ca v 2 calcium channels with an EEEE selectivity filter (49). One of the possible implications of a high-affinity calcium sensor, such as CaM near the internal mouth of the pore, is in its potential in influencing calcium ion conductance. We observe an apparent simultaneous flow of monovalent and divalent ions through Ca v 3 channels containing mutated cysteines in the domain II, L5 extracellular loop, which is only possible if the divalent cation alongside the monovalent (e.g. Na ϩ ) is either Sr 2ϩ or Ba 2ϩ but not Ca 2ϩ . 5 A high-affinity calcium sensor, like CaM at the pore mouth, may contribute to the explanation for the observations that Ca 2ϩ appears as the dominant and pre-ferred divalent ion carrier through the Ca v 3 channel pore, even when the channel is highly sodium-conducting. 5

Expression of Ca v 3 constructs
The cDNAs encoding for Ca v 3.2 were previously characterized (61) as well as hemagglutinin (HA) epitope-tagged Ca v 3 subunits expressed using pcDNA 3.1 vector (23,75). We illustrate how to maintain HEK-293T cell lines and introduce calcium channel expressible plasmids for transient transfection, expression, and whole-cell patch-clamp recording in an on-line video (JoVE) format (82). HEK-293T cells were cultured as described previously (23,75). For optimal transfection, cells were plated at 50 -70% confluence. Cells were transfected using the JetPEI transfection reagent (QBiogene) according to the manufacturer's protocol. In all conditions, corresponding empty plasmids were used to adjust the quantity of the transfected material (1 g of each plasmid/35-mm dish). Gating brake deletion mutants for Ca v 3.2, D 429 -539 , and D 453-491 were used previously (23).

Expression of CaM, CaM mutant, and CaM knockdown constructs
We have used human CaM and associated CaM mutants that we have described before (9), CaM(1-44) and partial CaMs: N-CaM(1-74), C-CaM , and CaMs that simulate CaM binding in the absence of calcium, by substitution of glutamate residues in the two calcium-binding sites in the N-terminal lobe (CaM 12 ) or C-terminal lobe (CaM 34 ), or mutation of all four calcium-binding sites (CaM 1234 ) of CaM. CaM-GFP used for Western blotting is a construct consisting of GFP fused to the C terminus of CaM expressed in pcDNA3.1 (Cell Signaling Technology) (83). EGFP expression vector was obtained by subcloning EGFP from pEGFP-C1 (BD Biosciences) into the multiple cloning site of pcDNA3.1. The CaM 1234 overexpression construct contained bicistronic marker mRFP (to generate a red color) (46) to distinguish from the channel expression with bicistronic EGFP (to generate a green color). The CaM 1234 overexpression construct also containing CaM shRNA knockdown includes two short hairpin sequences described above to knock down CaM (Calm 1,2,3 genes) (46 -48). The shRNA knockdown sequences is reported to be 70% complete in disabling of native CaM expression (46 -48), but this knockdown was insufficient for our purposes, as we observed no significant physiological consequence after transfection of the shRNA CaM knockdown alone.

Synthetic peptide experiments
We purified bacterially expressed His-tagged CaM (CaM) and measured in vitro association with 26-mer synthetic peptides (BioBasic) spanning the gating brake of human Ca v 3.x, Ca v 3.2GBmut, and snail LCa v 3 in gel-shift mobility assays, differential circular dichroism, and isothermal titration calorimetry. We followed protocols that we have described previously in an analysis of CaM binding with snail or human L-type calcium channel peptides in the presence of 0.1 or 0.5 mM CaCl 2 -containing solution (9).

CaM protein expression for biochemistry experiments
Rat CaM was purified using phenyl-Sepharose 6 Fast-Flow (High Sub) (GE Healthcare), packed to ϳ10 ml volume, by gravity flow at 4°C. Codon-optimized CaM was cloned in pET9d, a T7 Escherichia coli expression vector. CaM was transformed into BL21 DE3 E. coli and grown at 37°C and 200 rpm to OD of 0.4 -0.6 (for Luria Broth, LB) or 0.5-0.9 (for SuperBroth, SB), at which point samples were induced by 1 mM added isopropyl ␤-D-1-thiogalactopyranoside and grown for another 4 (for SB) to 6 (for LB) h. Robust expression of soluble CaM was confirmed by SDS-PAGE. Cells were collected by centrifugation at ϳ4000 ϫ g, then lysed in high-salt buffer (0.1 M Tris-Cl, pH 7.5, 0.5 M NaCl, 0.5 M ammonium sulfate, 1 mM DTT) using an Avestin Emulsiflex homogenizer at 15,000 -20,000 p.s.i. with jacketed water cooling, through which the cells were passed ϳ3ϫ on average. The cell lysate was clarified by centrifugation at 20,000 rpm in an SS-34 rotor (ϳ50,000 ϫ g) at 4°C for 30 min. To the cleared lysate, CaCl 2 was added to a final concentration of 10 mM. The lysate was then applied to the gravity phenyl-Sepharose column equilibrated with the same high salt buffer (and 1 mM CaCl 2 ). The column was washed with highsalt buffer. CaM was then eluted using 10 mM Tris-Cl and EDTA solution, pH 7.5. The highest fractions were pooled, concentrated, and applied to the preparation grade Superdex 75 16/60 column (GE Healthcare) on an AKTA FPLC system. Typical injection volume was 1 ml, and the gel filtration buffer (50 mM Tris-Cl, pH 7.5, 150 mM NaCl, 1 mM CaCl 2 ) was set to flow at 1 ml/min. Wild-type CaM reproducibly elutes at ϳ68 ml on this column. Elution peak width is dependent on the amount of CaM present; routinely up to 10 1-ml fractions were obtained of concentrations 200 -400 M; these were pooled and concentrated using the spin concentrators from Sartorius or Amicon (10,000 molecular weight cutoff). Extinction coefficient of 3006 M Ϫ1 cm Ϫ1 at A 276 was used (66), and the final protein stock was aliquoted to 500 l and stored at Ϫ80°C.

Purification of Ca 2؉ -deficient CaM proteins
The interaction of CaM with hydrophobic interaction chromatography media depends on its exposure of the hydrophobic residues upon Ca 2ϩ binding. The phenyl-Sepharose method is not nearly as efficient for the purification of CaM 12 , CaM 34 , or CaM 1234 . So for Ca 2ϩ -deficient CaM proteins, we used a combination of ion-exchange chromatography and subsequent gel filtration for purification. A Q-Sepharose (quaternary ammonium) 6-ml column (GE Healthcare) was used for ion exchange. Lysis and low-salt buffer consisted of 5 mM BisTris, pH 6.0, with no Ca 2ϩ or EDTA. Lysis buffer also had DTT added to 1 mM final concentration to maintain CaM's methionines in the reduced state. CaM 12 , CaM 34 , and CaM 1234 (in pET28a vector) were expressed and harvested in same manner as wild-type CaM. Cells were lysed using the Avestin Emulsiflex homogenizer (Ͼ15,000 p.s.i.), clarified by centrifugation at ϳ 50,000 ϫ g, and applied to the column (ion exchange was done using the AKTA FPLC system, same as for gel-filtration chromatography). The column was then briefly washed with low-salt buffer (5-10 column volumes), and a shallow salt gradient was applied (100 mM NaCl over 10 column volumes). Subsequently, the column proteins were eluted with a steep gradient (1.6 M NaCl over 2 column volumes, held for 2 more column volumes), after which the column was re-equilibrated or cleaned. The resulting fractions were analyzed by SDS-PAGE and read for absorbance levels at 278 nm. Best candidates were pooled and concentrated using YM-10 spin columns (Millipore) and further purified by gel-filtration chromatography. Purity of the CaM proteins was confirmed by SDS-PAGE and ESI-MS (electrospray ionizationmass spectrometry), and their concentrations were determined using the modified Bradford assay (84) (with wild-type CaM as the standard), aliquoted, and stored in gel filtration buffer at Ϫ80°C.

Bioinformatic analyses of peptide sequences
Protein parameters (pI and molecular weight) were determined using the prediction tool in ExPaSy (http://web.expasy. org/protparam/), 6 and the aggregation propensity was determined using the on-line software at Centre for Genomic Regulation (http://tango.crg.es/). 6 All phylogenetic alignments were done using variations of the BLAST tool on the GenBank TM web site. The CaM-binding site prediction database (http:// calcium.uhnres.utoronto.ca/ctdb/contacts/index.htm) 6 was used to search for other possible "hits" in the channel ORFs (39) and to design optimal peptide sequences for synthesis.

Native Trp fluorescence
Steady-state measurements were obtained using a Photon Technologies International Quantamaster Fluorimeter (London, Ontario, Canada) in a 50-l 1-mm fluorescence cuvette (Hellma) at room temperature. The same buffer was used as for gel filtration and ITC experiments (50 mM Tris-Cl, pH 7.5, 150 mM NaCl) with either 1 mM CaCl 2 or 10 mM EDTA; buffer was used as a baseline for subtraction. To the cuvette containing the starting peptide solution (25 M), increasing amounts of CaM were added (in increments corresponding to ϳ2.5 M).
Stepwise mission scans from 300 to 400 nm using a 280-nm excitation wavelength and 1-nm slit width (averaging over 1s each) were performed.

Dansyl-CaM fluorescence
Dansyl (5-(dimethylamino) naphthalene-1-sulfonyl chloride) is an amine-reactive fluorescent dye. Dansyl-CaM was prepared as described previously (85). CaM (1 mg/ml) was transferred into 10 mM NaHCO 3 , 1 mM EDTA, pH 10.0, at 4°C. 30 l of 6 mM dansyl-chloride (1.5 mol/mol of CaM) in DMSO was added to 2 ml of CaM, with stirring. After incubation for 12 h at 4°C, the mixture was first dialyzed against 500 volumes of 150 mM NaCl, 1 mM EDTA, 20 mM Tris-HCl, pH 7.5, at 4°C, and then dialyzed against 500 volumes of water. Labeling yields were determined from absorbance spectra using the ⑀320 of 3,400 M Ϫ1 cm Ϫ1 and were compared with protein concentrations determined using the Bradford method with wild-type CaM used as the protein standard (86) ESI-MS was used to confirm successful dansyl-labeling of CaM protein. The concentration of dansyl-CaM in all experiments was 2 M. Steady-state fluorescence was performed in a similar manner as the Trp experiments, except that a newer Fluorolog 3-22 (Horiba Scientific, Ltd.) fluorimeter was used, and the buffer used was 10 mM HEPES, pH 7.0, with supplemented 0.1 mM CaCl 2 .

Differential circular dichroism
CD measurements were performed using a Jasco-715 spectropolarimeter (Jasco Instruments, Easton, MD), using the following parameters: 250 -190-nm range, 50 nm/min speed, with 1-s response time, 1-nm bandwidth, and 100-millidegree sensitivity at room temperature. 16 -25 accumulations were gathered depending on signal quality. Recordings were made in a 1-mm cuvette, in 10 mM sodium phosphate buffer with 0.1 mM added CaCl 2 , and used as baseline, which was subtracted from each recording. For CaM and CaMB peptide spectra, the same solution of 10 M CaM was scanned before stock (1 mM) peptide was added in incremental amounts before being scanned again. The resulting spectra were subtracted (e.g. CaM and CaMB peptide-CaM alone) to extract the spectrum corresponding to the change associated with adding the peptide. Every recording was an average of at least 16 accumulations and subsequently corrected for protein concentration (converted to units of mean residue ellipticity). These were smoothed and exported into ASCII format for analysis with Excel and either SOMCD or Dichroweb K2d algorithm (http://dichroweb.cryst. bbk.ac.uk) 6 (92,93). TFE recordings were performed using pure 2,2,2-trifluoroethanol (Sigma) at 10, 25, and 50% after thorough pre-mixing and incubation with peptides in the phosphate buffer, baseline-corrected against the same TFE/buffer solution without peptide. LCa v 3 and Ca v 3.2 CaMB peptides were contained in PBS buffer (10 mM sodium phosphate, pH 7.3, 137 mM NaCl, 2 mM KH 2 PO 4 , and 2.7 mM KCl). Ca v 3.1, Ca v 3.3, CaMB peptides were characterized in low ionic strength buffers due to severe aggregation issues in PBS. Low ionic strength buffer for Ca v 3.1 and Ca v 3.3 CaMB peptides was composed of 10 mM HEPES, pH 7. A 0.1-cm quartz cuvette (Hellma) was used for CD measurements. CaM-binding experiments were performed by adding increments of stock (1 mM) CaMB peptide to the same cuvette containing 10 M CaM (300 l). Subsequent scans would then have the CaM-only trace subtracted and corrected for total peptide concentration, smoothed, and converted to mean residue ellipticity units.

Gel mobility-shift assays
Gel assays were run in native conditions with 4 M urea, containing 15% acrylamide separating and 4% stacking gels. Running buffer contained 192 mM glycine, 50 mM Tris-Cl, 0.1 mM CaCl 2 , pH 8.3. Separating gel buffer (2ϫ) contained 0.7 M Tris-Cl and 1 mM CaCl 2 , pH 8.8. Stacking buffer (2ϫ) contained 0.5 M Tris-Cl and 1 mM CaCl 2 , pH 6.7. Peptides were incubated with CaM for Ն1 h at 4°C in gel filtration buffer (150 mM NaCl, 50 mM Tris-Cl, pH 7.5, 1 mM CaCl 2 ) prior to loading. 50% glycerol was used in the gel filtration buffer, and Ͻ0.1% bromphenol blue (as tracker dye) was used at 1:2 sample volume just prior to loading for each sample. Gels were run at 100 V in 4°C jacketed/ ice-bath conditions for 6 -8 h as needed. All gel-mobility-shift assays were confirmed in three replicate experiments.

Isothermal titration calorimetry
Heat gain or heat requirements in the sample cell (for exothermic binding or endothermic binding, respectively) correspond to the direction of the injection peaks measured in the isothermal titration calorimeter (MicroCal iTC200, Malvern Instruments, UK). The peaks were integrated and plotted as a sigmoidal plot of cal (concentration of the ligand in the cell) versus mol of injectant (protein concentration in the syringe). Origin 2017 software (OriginLab) was used to fit the resulting data through iterative least-squares-type regression. Due to the tendency of CaMB peptides to aggregate in any salt, we reconstituted the CaMB peptides for ITC experiments in a low ionic strength (10 mM HEPES, pH 7.0, 0.1 mM CaCl 2 ) buffer for all CaMB peptides. All experiments were performed at 25°C, with 1000 rpm stirring, 5 cal/s reference power, and high-feedback gain. Injection conditions were typically 1.2-1.8 l each (22-28 injections), and cell/syringe concentrations were optimized for individual experiments (c value between 10 and 1000) (87)(88)(89) but typically between 10 and 50 M for the cell and 100 to 500 M for the syringe. Periodically, water-water and buffer-buffer controls were conducted to ensure the syringe and cell were clean and in good working order. Most of the endothermic titrations had to be baseline-subtracted against the final injection heats (end-weighted auto-baseline or manual baseline for Ca v 3.3 correct for the heat of dilution) due to fitting artifacts from the isotherm passing through the x axis. Baseline-corrected curves were all fitted to a one-set-of-sites model in Origin 2017 (OriginLab) software with satisfactory results.

Electron microscopy
Full-length mammalian Ca v 3.1 was expressed in Sf9 insect cells and purified as described previously (38). Recombinant human Ca v 3.1 was incubated (1:1 molar ratio) with biotin-CaM (Calbiochem) in 1.5 mM CaCl 2 after which a streptavidingold conjugate (5 nm) was added (38). Samples were examined using standard negative staining techniques with 2% (w/v) uranyl acetate. Images of the Ca v 3.1-CaM complexes were recorded on a Tecnai 12 transmission electron microscope operated at 100 kV.

NMR
Complex formation of gating brake peptide with CaM in a 1:1 ratio was monitored by acquiring a 1 H-15 N heteronuclear single-quantum coherence (HSQC) spectrum in a manner previously published with CaM and peptides (90). NMR spectra were recorded at 25°C on Bruker 600 and 700 MHz DRX spectrometers equipped with XYZ-gradient triple-resonance probes (Bruker, Billerica, MA), and analyzed with CARA (computeraided resonance assignment). CaM for NMR experiments was expressed in M9 minimal media (11.03 g/liter Na 2 HPO 4 ⅐7H 2 O, 3.0 g/liter KH 2 PO 4 , 0.5 g/liter NaCl, 2 mM MgSO 4 , 0.1 mM CaCl 2 , 5 mg/ml thiamine, 100 g/ml kanamycin) containing 2 g/liter [ 13 C]glucose and 1 g/liter 15 NH 4 Cl; 13 C-15 N CaM was purified. The samples were prepared for NMR experiments via a buffer exchange into NMR solution (100 mM KCl, 10 mM CaCl 2 , 0.2 mM NaN 3 , 90% H 2 O, 10% 2H 2 O) at pH 6.0 using a YM10 centrifugal filter device (Millipore, Billerica). All NMR samples contained 300 M CaM in a total volume of 500 l. The Calmodulin at the T-type channel gating brake samples were transferred into 5-mm NMR sample tubes and stored at 4°C until required for NMR experiments. NMR experiments on the complexes were conducted on CaM samples titrated with each peptide to saturation in a 1:1 CaM/peptide ratio. Complex formation was monitored after each addition by acquisition of a 1 H-15 N HSQC pulse sequence. NMR spectra were recorded at 25°C on Bruker 600 and 700 MHz DRX spectrometers equipped with XYZ-gradients triple-resonance probes (Bruker, Billerica, MA.). Spectra were analyzed using the program CARA.

Electrophysiological recordings
Macroscopic currents were recorded at room temperature using an Axopatch 200B amplifier (Molecular Devices). Borosilicate glass pipettes had a resistance of 1.5-2.5 megaohms when filled with an internal solution containing (in mM): 140 CsCl, 10 EGTA, 10 HEPES, 3 Mg-ATP, 0.6 GTP-Na (pH adjusted to 7.25 with CsOH, ϳ315 mosM). The extracellular solution contained (in mM): 145 tetraethylammonium-Cl, 10 HEPES, and 2 CaCl 2 (pH adjusted to 7.25 with tetraethylammonium-OH, ϳ330 mosM). Recordings were filtered at 2 kHz. Data were analyzed using pCLAMP9 (Molecular Devices), GraphPad Prism (GraphPad), and Origin 2017 (OriginLab) software. Current-voltage relationships were generated in 5-or 10-mV voltage steps from a holding potential of Ϫ100 mV. Steady-state inactivation curves were constructed by plotting normalized peak currents (peak test pulse current/peak prepulse current) as a function of inactivating potentials. Kinetics of activation and inactivation were determined by fitting monoexponential functions over the growing or decaying phases of each current trace.

Statistical analyses
Results are presented as mean Ϯ S.E., and n ϭ the number of cells. Statistical analysis was first performed with the Student's t test or with one-way ANOVA combined with a Tukey posttest for multiple comparisons (*, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001). For the electrophysiology data sets as paired data of treatment values compared with control values at time 0 and after 20 min of CaMB peptide dialyses in Figs. 7 and 8, we also used a non-parametric Wilcoxon matched-pairs signed rank test. In Fig. 9, there were three sets of data to compare between wild type, CaM overexpression, and CaM 1234 overexpression, so we used a Kruskal-Wallis test here followed by a Dunn's multiple comparison test or a Mann-Whitney test. The statistical analyses using parametric and non-parametric analyses produced similar results.

Co-immunoprecipitation experiments
Forty eight hours after transfection, cells cultured in 35-mm dishes were lysed on ice for 20 min with Nonidet P-40 buffer containing 50 mM Tris-HCl, pH 7.4, 125 mM NaCl, 25 mM MgCl 2 , 33.3 M CaCl 2 , 5% glycerol, 1% Nonidet P-40. Cell lysates were centrifuged at 10,000 ϫ g for 30 min at 4°C. Clarified cell lysates were then incubated with anti-GFP-conjugated magnetic beads (Thermo Fisher Scientific) for 4 h at 4°C. The beads were washed four times with lysis buffer and resuspended in 2ϫ loading buffer. The immunoprecipitates were then analyzed by SDS-PAGE and Western blotting.

Western blotting
Samples were loaded on SDS-PAGE. Proteins were then transferred onto nitrocellulose membranes and blocked with 5% powdered nonfat milk in PBS-T (Tween 0.1%). Primary antibody was incubated overnight at 4°C in 5% powdered nonfat milk in PBS-T (Tween 0.1%). After three washes in 5% powdered nonfat milk in PBS-T, the secondary HRP-coupled antibodies were incubated for 1 h at room temperature in the same buffer. The signal was detected using the Super Signal West Pico chemiluminescent system (Thermo Fisher Scientific).
Author contributions-The CaV3.1-baculovirus construct was made and Sf9 cells harvested in the laboratory of A. C. D., which were used for the single particular cryo-EM carried out in the laboratory of A. K. The co-immunoprecipitation experiments were carried out by A. M. in his laboratory. The gel-mobility-shift assays, circular-dichroism spectroscopy, and isothermal-titration calorimetry experiments were carried out by V. T. in the laboratory of J. G. G. Most electrophysiology experiments of mammalian Cav3.2 channels were conducted by J. C. in his laboratory, based on preliminary discoveries by the electrophysiology work of R. F. S. and W. G. in the laboratory of J. D. S. Cav3.2 channel mutants were provided by E. P. R., who was especially helpful in the design of electrophysiology experiments involving the gating brake. The NMR experiments were conducted by M. P. in the laboratory of T. D. Vectors for CaM1234 overexpression and shRNA knockdown of Calm1,2,3 genes were provided by Z. P. P. All authors contributed to the writing of the manuscript. Team management and the bulk of the writing of the manuscript was conducted by J. D. S.