Calcium can mobilize and activate myosin-VI

Significance

The timing of motor protein activation is central to a broad range of cellular motile processes including endocytosis, cell division, and cancer cell migration. The cytoskeletal motor myosin-VI is involved in these processes and is the only myosin in the human genome shown to move toward the minus end of actin filaments. Using electron microscopy, fluorescence spectroscopy, and motility assays, we demonstrate that calcium is the cellular switch that directs the rearrangement of the motor from a dormant, inactive state at low calcium to a cargo-binding nonmotile state at high calcium. The return to low calcium generates either cargo-bound active motors that translocate to the center of the cell or refolded inactive motors ready for the next cellular calcium flux.

Abstract

The ability to coordinate the timing of motor protein activation lies at the center of a wide range of cellular motile processes including endocytosis, cell division, and cancer cell migration. We show that calcium dramatically alters the conformation and activity of the myosin-VI motor implicated in pivotal steps of these processes. We resolved the change in motor conformation and in structural flexibility using single particle analysis of electron microscopic data and identified interacting domains using fluorescence spectroscopy. We discovered that calcium binding to calmodulin increases the binding affinity by a factor of 2,500 for a bipartite binding site on myosin-VI. The ability of calcium-calmodulin to seek out and bridge between binding site components directs a major rearrangement of the motor from a compact dormant state into a cargo binding primed state that is nonmotile. The lack of motility at high calcium is due to calmodulin switching to a higher affinity binding site, which leaves the original IQ-motif exposed, thereby destabilizing the lever arm. The return to low calcium can either restabilize the lever arm, required for translocating the cargo-bound motors toward the center of the cell, or refold the cargo-free motors into an inactive state ready for the next cellular calcium flux.

In human cells, cytoskeletal motor proteins move along microtubules and actin filaments to generate complex cellular functions that require a precise timing of motor activation and inactivation. Myosin-VI is thought to have unique properties because it is the only myosin in the human genome shown to move toward the minus end of actin filaments (1). Apart from its roles in the formation of stereocilia in cells of the auditory system (2, 3), membrane internalization (4–6), and delivery of membrane to the leading edge in migratory cells (7), myosin-VI is an early marker of cancer development, aggressiveness, and cancer–cell invasion because of its dramatically up-regulated expression in breast, lung, prostate, ovary, and gastresophagus carcinoma cells (7–11). How this motor might promote cancer–cell migration, proliferation, and survival is unknown.

In migrating cells, localized calcium transients (∼50 nM to ∼10 μM) (12, 13) have been reported to play a multifunctional role in steering directional movement (14), cytoskeleton redistribution, and relocation of focal adhesions (15). The effect of calcium transients on the mobilization and cargo binding of myosin-VI and on its mechanical activation, however, are not understood. In the current model, the catalytic head domain hydrolyzes ATP, whereas the tail domain anchors the motor to specific compartments. In vitro studies have shown that calcium affects myosin-VI binding to phospholipids (6), as well as the kinetics and motility rate of the motor (16, 17). The underlying molecular mechanisms, however, are unknown. It has also been discussed that myosin-VI might be able to adopt an inactive folded state (18, 19), perhaps similar to nonmuscle myosin II and myosin-V (20–22), with folding and unfolding regulated by some unknown mechanism. When activated, the myosin-VI head domain binds to actin, generating conformational changes that are transduced by the converter to the lever arm or neck domain and amplified to nanometer displacements. The neck consists of an extended α-helix stabilized by the binding of calmodulin (23), which pointed to the intriguing possibility that the calcium sensor calmodulin bound to the myosin-VI neck domain might constitute a molecular mechanism to control both the cellular mobilization and activation of myosin-VI in migrating cells. We therefore set out to investigate the effect of calcium on the structural conformation, mechanical properties, and activation of single myosin-VI motor molecules using electron microscopy (EM), spectroscopic, and mechanical experiments.

Results

Calcium Binding to Calmodulin Induces a Structural Rearrangement of Myosin-VI.

We characterized the effect of calcium on the structure and function of full-length myosin-VI using single particle analysis of negatively stained EM images (24, 25). We chose calcium concentrations close to the physiological range, which ensured homogeneous populations of molecules (SI Text). The nucleotide-free motor molecules adhered to the carbon-coated grids in two main orientations (Fig. 1 and Fig. S1), providing a front view with the neck along the long axis of the head domain (=straight conformation) and a side view with the head and neck at an angle of ∼53° (=bent conformation). The images in front view did not fit to any crystal structures and were therefore not analyzed in further detail (Fig. S1 and SI Text). The images in side view revealed detailed information on the conformation of the calmodulins and the tail domain. To interpret the EM data, we modeled a structure that combined the crystal structures of the myosin-VI head domain (23) and the neck region comprising two calmodulins and the subsequent three-helix bundle of the tail domain (26). At low calcium, the modeled structure in the optimized spatial orientation could account for the contour of the EM image (Fig. 1B, class average of n = 2,998, and Figs. S1 and S2). Zooming into the neck region (Fig. 1C) showed an excellent agreement between the crystal structure of both calmodulins with a cross-correlation coefficient (CCC) of 0.99 and 0.96, respectively, plus the helix bundle (represented here by the first helix in dark green; CCC, 0.89) and the EM outline (solid white line). At high calcium (Fig. 1 D–F), in a complete reversal to the low calcium condition, most molecules adhered to the grid in a side view, indicating a major change in surface charge and structure (Fig. S3). The crystal structure of the second calmodulin (CCC, 0.76) and helix bundle (CCC, 0.68; Fig. 1 E and F) were no longer in agreement with the EM contour (dashed white line, n = 2,392). The data indicated that the second calmodulin had rotated clockwise by ∼30° in the projection plane, whereas the helix bundle had rotated anticlockwise by ∼30° (black arrows).

Fig. 1.

Calcium-induced conformational change of calmodulin revealed using negative stain EM of myosin-VI. (A) The nucleotide-free molecules adhered to the carbon-coated grids in a straight conformation (class average of n = 13,583) or a bent conformation (class average of n = 2,998), pCa 8. (B) The optimized projection of a modeled structure (catalytic domain blue, converter gray, first calmodulin red, second calmodulin yellow, first helix of the three-helix bundle dark green) was overlaid onto the inverted image of the bent conformation shown in B. (C) The EM image was color-coded with high intensities in blue and low intensities in green and yellow. The contour lines were created using an intensity threshold. (D–F) Class averages at pCa 4 in the straight (n = 1,405) and bent conformations (n = 2,392). Note the differences (marked by arrows) between the EM image at high calcium and the overlaid modeled structure in a projection optimized for the catalytic domain.

Fig. S1.

The two main orientations of myosin-VI on the EM grid. (A) The modeled structure (myosin-VI in rigor) was based on 2BKI (23) and 3GN4 (26), as described in SI Text. Molecules bound to the EM grid with the catalytic domain and the calmodulin-binding lever arm at an angle of ∼53° (=bent conformation) or ∼180° (=straight conformation) in the projection plane. To fit the model in the optimized orientation for the molecules in the bent conformation to the catalytic domain of the straight molecules, the model had to be rotated by 89° around the y axis and tilted in the y/z plane (vector r). In this projection, the calmodulin-binding lever arm lies underneath the catalytic domain. (B) A total of 799 low pass-filtered 2D projections of the model in different 3D orientations were cross-correlated with the real EM averages. The CCC values calculated for the bent molecules at low calcium (black) were ranked and the projection with the highest CCC value labeled X1. For comparison, the CCC values for the same projections were also calculated for the bent molecules at high calcium (red). (C) The CCC values of the model projections, fitted to the catalytic domain of the straight molecules at low calcium, were ranked and the best fitting projection labeled X1. (D) The CCC values for the projections of the myosin-VI–ADP structure, fitted to the catalytic domain of the straight molecules at low calcium, were also ranked. In B–D, three simulated projections of the structures (X1–X3) show the best, an intermediate, and the worst fitting projection in each case. (E) (Upper) Best low pass-filtered projection of the model fitted to the bent conformation (X1). (Lower) Model in this particular orientation, overlaid onto the real EM class average. (F) (Upper) Best low pass-filtered projection of the model fitted to the catalytic domain of the straight conformation. (Lower) Structure overlaid onto the real EM class average. Note, the lever arm of the model lies underneath the catalytic domain. The structure was inconsistent with the real EM of straight molecules stretched out along the long axis of the catalytic domain. (G) (Upper) Best low pass-filtered projection of the myosin-VI–ADP structure fitted to the catalytic domain of the straight conformation. (Lower) Structure overlaid onto the real EM class average. The position of the lever arm of the structure is inconsistent with this EM dataset.

Fig. S2.

Calcium-sensitive structures in the electron micrographs. To evaluate the calcium-dependent structural differences in the EM images, the CCC values for the optimized projections of different parts of the model structure at high and low calcium were determined. The lowest CCC values, indicating differences between the model structure and the EM data, were found for the neck-tail and the second calmodulin in high calcium. The highest CCC values were obtained for the catalytic domain both at high and low calcium, indicating that this region is not much affected by calcium. Note, however, that at high calcium, the correlation between the catalytic domain of the model and the EM average is still better than at low calcium (indicated by the negative value at ΔCCC). This indicates a structural difference and is consistent with the presence of a backfolded tail at low calcium, in contrast to high calcium.

Fig. S3.

Statistics on electron micrographs and data analysis. (A) Electron micrographs of myosin-VI in rigor were taken in four different conditions with respect to calcium and calmodulin concentrations. (B) Apart from the parameter straight or bent, which refers to the angle between catalytic domain and lever arm of the molecules on the EM grid, the molecules were also analyzed with respect to the length of the lever arm. For the long molecules, two calmodulins and parts of the tail could be resolved, whereas short identifies molecules for which only one calmodulin could be resolved. The parameter left or right refers to the orientation of the catalytic domain with respect to the mirror symmetry along the long axis of the catalytic domain. The statistics show the effect of calcium on molecule binding in the straight or bent conformation. (C) The gallery of the molecule classes shows the averaged conformations and statistics of the classes that were found for the four different conditions.

Ca²⁺-Calmodulin Binds to a Bipartite Binding Site.

To investigate the conformational change induced by calcium, we performed titration experiments (Fig. 2, Table 1, and Fig. S4) using apo- and Ca²⁺-calmodulin and calmodulin mutants with either the N- or C-terminal calcium binding site eliminated (16), against 16 different peptides containing potential calmodulin binding sites (Table S1) (27, 28). We used the intrinsic tryptophan fluorescence in these target peptides to detect calmodulin binding (SI Text and Fig. S4) (29). We found that apo-calmodulin bound to the target peptide P2 (Fig. 1B, yellow) with a stoichiometry of 1:1 (K_d ∼ 290 nM). For Ca²⁺-calmodulin, however, the stoichiometry was 0.5:1, which indicated that the coordination between the N- and C-terminal lobes was lost in the Ca²⁺ state (Fig. 2A), i.e., that only one calmodulin lobe could bind. Peptide P3 had a very weak affinity for Ca²⁺-calmodulin and did not bind apo-calmodulin at all (Table 1). Intriguingly, the binding stoichiometry for the double peptide P2–3 was 1:1, for both apo- and Ca²⁺-calmodulin (Fig. 2B), with a high affinity for the Ca²⁺ state (K_d ∼ 83 nM). This result revealed that the double peptide must contain a bipartite binding site available only for Ca²⁺-calmodulin (Fig. 2E). Using peptides omitting the N-terminal half of P2, the C-terminal half of P3, and C-terminal mutations in the peptide P2-3‡, we localized the two nonadjacent halves of this composite holo-calmodulin binding site (contained in the peptide P½ 2–3, K_d ∼ 38 nM; Fig. 2 D and E). Interestingly, comparison with the structure of the three-helix bundle (26) revealed that the distal half of this binding site was located at the extended loop between the first and the second helix of the bundle. This structural comparison indicated that, in the presence of calcium, the partly detached Ca²⁺-calmodulin can seek out and form a bridge between P2 and the loop in the helix bundle, with a change in affinity for the bipartite binding site by a factor of 2,500 (Fig. 2D and Table 1, for P½ 2–3 K_d ≥ 100 μM at pCa 8 and 38 nM at pCa 4). This binding pattern is consistent with the second calmodulin rotating clockwise and the helix bundle being pulled anticlockwise in the EM images (Fig. 1 E and F). The calcium-dependent transitions between the different binding modes of apo- and Ca²⁺-calmodulin to the peptide P2–3 were fully reversible with a rearrangement, but no detachment, of calmodulin observed in either condition (Fig. 2C). We did not observe the binding of a third calmodulin to this region, as described for artificially zippered myosin-VI dimers (30), where modified tertiary structures may expose extra binding sites. We found the peptide P1 binding to calmodulin to be extremely calcium sensitive at pCa 4 (K_d = 11 nM), with no binding observed at pCa 8 (Fig. S4). Our EM data, however, support the report by Bahloul et al. (31), who showed that when additional parts of the converter were included, calmodulin was locked into position regardless of the calcium concentration. To locate any further C-terminal calmodulin binding sites, we studied six further peptides covering the entire myosin-VI tail up to the cargo binding domain (Table S1). No binding was observed for any of the peptides with either apo- or Ca²⁺-calmodulin (all K_d ≥ 100 μM).

Fig. 2.

Ca²⁺-calmodulin binds to a bipartite site, bridging between sites P2 and P3 and rearranging the conformation of the tail domain. Tryptophan fluorescence (λ_ex = 290 nm, λ_em = 323 nm) for (A) 2 μM peptide P2, (B) 2 μM P2-3, or (D) 2 μM P½ 2–3 and increasing calmodulin concentrations was measured at pCa 4 and pCa 8. (C) Tryptophan fluorescence in response to a change in calcium concentration was measured with 2 μM P2-3 and 2 μM calmodulin. (E) Predicted calmodulin binding sites P1–P4, and Helix 1, Helix 2 of the three-helix bundle (26), and single α helical domain (SAH) (44). The apo-calmodulin binding site identified on P2 and the bipartite Ca²⁺-calmodulin binding site on P2–P3 are indicated. For P2-3‡, the C-terminal serine and valine were deleted. For HB1‡, the residues on HB1 labeled by * were replaced by alanine or glycine.

Table 1.

Calmodulin binding to myosin-VI target peptides

Peptide	pCa 4.0			pCa 8.0
	Kd (nM)	Stoichiometry CaM:peptide		Kd (nM)	Stoichiometry CaM:peptide
	CaM	CaM		CaM	CaM
	WT	WT	Mutants	WT	WT
P1	11 ± 0.7	1:1	2:1	NA	NA
P2	170 ± 6	0.5:1	1:1	290 ± 5	1:1
P2-3	83 ± 4	1:1	1:1	142 ± 11	1:1
P2-3‡	278 ± 13	0.5:1	NA	143 ± 10	1:1
P ½ 2–3	38 ± 17	1:1	2:1	≥100,000	NA
P 2–½ 3	167 ± 19	0.5:1	2:1	165 ± 9	1:1
P3	≥6,000	NA	NA	NA	NA
P4	NA	NA	NA	NA	NA
HB1	NA	NA	NA	≥100,000	NA
HB1‡	NA	NA	NA	≥500,000	NA

The titrations of target peptide sequences (Fig. 2E) with calmodulin were performed at 20 °C. The dissociation constants K_d in Table 1 for the Trp-containing peptides were determined by direct titration, and the data were analyzed as previously described (29). NA, K_d was bigger than 500 μM. The binding stoichiometry for calmodulin binding to the peptide at pCa 8 was 1:1 for all peptides. At pCa4, the binding stoichiometry is specified for each peptide. For each experiment, four independent titrations were performed, and the average K_d value is reported with its SD.

Fig. S4.

Calmodulin binding to the binding motif P1. The change in tryptophan fluorescence of the target peptide in the presence and absence of calmodulin was used to study calmodulin binding to the synthetic target sequence. (A) Tryptophan fluorescence (λ_ex 290 nm, λ_em 323 nm) for 2 μM target peptide P1 and increasing concentrations of calmodulin measured at low and high calcium concentration. P1 bound to calmodulin with a 1:1 stoichiometry and a K_d of 11 nM at pCa 4; at pCa 8 no binding was observed. (B) Calmodulin binding to P1 was fully reversible and calcium dependent. Black line: pCa ≥8, red line: pCa 4. (C) 2 μM peptide P1 was mixed with 2 μM calmodulin at different calcium concentrations. No difference in the fluorescence signal between pCa 5.1 (∼8 μM) and pCa 4.0 (∼100 μM) was found, indicating that at 8 μM free calcium, calmodulin was already saturated. There was also no difference in fluorescence between pCa 7.03 (∼93 nM) and pCa 8.0 (∼10 nM). This indicated that the differences seen between pCa 4 and pCa 8 in the titrations and the EM experiments in our study also apply to the cellular variations in free calcium between pCa 5 and pCa 7. (D) Although at pCa 4 WT calmodulin (WT-Cam) bound to the peptide P1 with a 1:1 stoichiometry, the N-terminal and the C-terminal calmodulin mutants bound with a 2:1 stoichiometry, indicating that the mutants attached to the peptide with only one lobe. (E) The cartoon illustrates the effect of calcium on calmodulin binding to the peptide P1.

Table S1.

Peptide sequences of predicted calmodulin binding sites

Peptide	Amino acid position	Amino acid sequence
P1	788–810	LTCSRWKKVQWCSLSVIKLKNKI
P2	812–834	YRAEACIKMQKTIRMWLCKRRHK
P3	832–856	RHKPRIDGLVKVGTLKKRLDKFNEV
P4	855–877	EVVSVLKDGKPEMNKQIKNLEIS
P5	872–895	KNLEISIDTLMAKIKSTMMTQEQI
P6	889–912	MMTQEQIQKEYDALVKSSEELLSA
P7	925–944	RLRRIQEEMEKERKRREEDE
P8	950–974	EEEERRMKLEMEAKRKQEEEERKKR
P9	975–944	EDDEKRIQAEVEAQLARQKE
P10	1,016–1,036	IAQSEAELISDEAQADLALRR
P2–3	812–860	YRAEACIKMQKTIRMWLCKRRHKPRIDGLVKVGTLKKRLDKFNEVVSVL
P2–3‡	812–858	YRAEACIKMQKTIRMWLCKRRHKPRIDGLVKVGTLKKRLDKFNEVVS
P2–1/23	812–847	YRAEACIKMQKTIRMWLCKRRHKPRIDGLVKVGTLK
P1/22–3	825–860	RMWLCKRRHKPRIDGLVKVGTLKKRLDKFNEVVSVL
HB1	833–874	HKPRIDGLVKVGTLKKRLDKFNEVVSVLKDGKPEMNKQIKNL
HB1‡	833–874	HKPRIDGLVKVGTLKKRLDKGNEGGSAGKDGKPEMNKQIKNL

The peptides of potential calmodulin binding sites, as determined for human myosin-VI, were used in the tryptophan fluorescence assays to measure the calmodulin binding affinity. The numbers specify the amino acid position in the human myosin-VI sequence.

Apo-Calmodulin Binds the Tail Segment aa1060–1125 and Causes the Motor to Backfold.

Closer inspection of the head domain at low calcium (Fig. 1 B and C) also revealed some extra mass in the averaged EM between the converter (gray), the catalytic domain (blue), and the first calmodulin (red), which was not accounted for by the modeled structure and indicated the presence of a backfolded myosin-VI tail. Intriguingly, at high calcium this extra mass was not observed (Fig. 1 E and F), implying that in these conditions the tail was more flexible. To test the idea of calcium regulating the backfolding of the myosin-VI tail, we performed pull-down and microscale thermophoresis (MST) experiments with truncated myosin-VI head and tail constructs. At low calcium, the Head 913 (aa1–913) was pulled down together with F-actin (pellet P), whereas the Tail 1125 (a1125–1276) remained in the supernatant (S), indicating that this part of the tail did not bind to the head (Fig. 3 A and C). In contrast, the Tail 1060 (aa1060–1276, GST-tagged) bound to the Head 913 (K_d = 8.21 ± 0.51 μM; Fig. 3 B and C). These experiments identified the segment aa1060–1125 as the part of the myosin-VI tail that bound to the catalytic domain and/or the neck region, enabling the molecule to backfold. The Western blot-based binding studies with peptides and calmodulin confirmed that it was in fact calmodulin, and not the calmodulin target sites P1 or P2, that bound the tail segment Tail 1060 (K_d = 17.79 ± 10.5 nM; Fig. 3 C and D). To investigate whether calmodulin with the peptide bound, as found in the physiological complex, would affect the binding of the tail we probed increasing ratios of the complex formed by calmodulin and peptide P2 in the presence and absence of calcium. Intriguingly, the experiments showed that the tail segment only bound to the calmodulin-peptide complex in the apo-calmodulin state and not the Ca²⁺ state, supporting the idea that backfolding is regulated by calcium (Fig. 3E).

Fig. 3.

Backfolding of the tail segment aa1060–1125 onto apo-calmodulin. (A) Cosedimentation of the myosin-VI construct Head 913 (aa1–913) with F-actin in the presence of the Tail 1125 (aa1125–1276), analyzed using SDS/PAGE. The Head 913 was pulled down with actin into the pellet P, whereas the Tail 1125 remained in the supernatant (S). Controls of the proteins alone are labeled in red. (B) In contrast, the Head 913 was pulled down together with the GST-tagged Tail 1060 (aa1060–1276), indicating that the tail segment aa1060–1125 was binding to the Head 913. Controls of the proteins alone labeled in red. (C) MST. Fluorescently labeled proteins indicated in red, and the binding partner in black. Values are mean values ± SD for three separate experiments. Squares: K_d 17.79 ± 10.5 nM; triangles: 8.21 ± 0.51 μM; circles: 554.1 ± 165.6 μM. (D) In the dot far Western blot, the Tail 1125 did not interact with calmodulin, whereas the Tail 1060 did. The Tail 1060 did not interact directly with the calmodulin-binding peptides P1 or P2. The Tail 1125 and Tail 1060 were detected when spotted directly onto the membrane (positive controls), whereas no reaction was detected when the tails were not applied (negative control). All experiments in A–D were performed at pCa 8. (E) For the dot far Western blot, the calmodulin-P2 complex was spotted onto the membrane at high and low calcium. Binding of the GST-tagged Tail 1060 was probed using an anti-GST antibody. The cartoon illustrates that the tail section aa1060–1125 binds to calmodulin at low, but not high, calcium.

Calcium Induces the Release of the Myosin-VI Tail and Increases Lipid Cargo Binding.

We then investigated whether the release of the backfolded tail, induced by calcium binding to calmodulin, also increased the binding affinity of the myosin-VI tail for lipid cargo. We performed pull-down experiments with myosin-VI and liposomes prepared from mixed bovine brain lipids (Folch fraction 1) in the presence and absence of calcium (Fig. 4A). At low calcium, ∼30% of myosin-VI remained in the supernatant. At high calcium, essentially the entire myosin-VI sample was pulled down together with the liposomes, consistent with an increased availability of the lipid-binding domains (6) on the myosin-VI tail. The fat blots (Fig. 4B) confirmed that the lipid binding of the recombinant full-length myosin-VI used in this study was calcium dependent and specific. Myosin-VI only bound to the mixed brain lipids containing phospholipids as described previously (6) but not to the main constituents of Folch fraction 1, i.e., phosphatidylcholine (PC) or -ethanolamine (PE). These experiments showed that the calcium-induced release of the myosin-VI tail can mobilize the motor and target it to bind to cargo, which led to the question of how calcium binding would affect the mechanical properties of the motor.

Fig. 4.

Calcium regulates the binding to lipid cargo and mechanical activity of myosin-VI. (A) Pull-down of full-length myosin-VI (FL-MyoVI) with Folch liposomes at low and high calcium, analyzed using SDS/PAGE. At low calcium, ∼70% of myosin-VI was pulled down, whereas at high calcium, the pull-down was virtually complete. (B) The protein-lipid-overlay (PLO) showed calmodulin binding to Folch mixed lipids in the presence and absence of calcium, but not to the pure main Folch constituents, phosphatidylcholine (PC) or -ethanolamine (PE). (C) Actin filament gliding observed with myosin-VI immobilized via a globular tail antibody (green), via a biotin-streptavidin linker (blue), and when attached to a surface coated with Folch liposomes (black). (D) To study the structural flexibility of myosin-VI, EM images at high and low calcium were aligned and classified using a mask covering the catalytic domain and grouped according to the orientations of the neck and tail domains. Extreme orientations of the neck are shown in D and E and H and I for low and high calcium, respectively. The outlines of the classes are overlaid in F and J, catalytic domain blue, converter gray, first calmodulin red, second calmodulin yellow. The mechanical pivot points at low and high calcium are indicated by a black dot. Note, that the rotations also include movement in and out of the projection plane. (G) Maximum intensity projection showing the tracks of actin filaments gliding over myosin-VI in red, with the first frame shown in yellow, pCa 8. Myosin-VI was immobilized on the surface via a biotin-streptavidin linker. (K) As in G, but at pCa 4; the filaments remained attached to the myosin-coated surface but did not move.

Ca²⁺-Calmodulin Destabilizes the Myosin-VI Lever Arm.

To investigate the effect of calcium on the mechanical stiffness of the myosin-VI neck, we reexamined the two calmodulins bound to the extended α-helix in the EM averages at high and low calcium (Fig. 1). In contrast to their close proximity at low calcium (Fig. 1 A and C), the two calmodulins in the EM images at high calcium seemed to have lost contact as a consequence of the rotation of the second calmodulin (Fig. 1 D and F). This loss of contact indicated a possible loss of mechanical stability of the neck domain, which serves as a mechanical lever arm. To further explore this possibility, we realigned and classified the EM images using a mask covering only the catalytic domain (Fig. 1B, blue). The outlines of the extreme orientations of the neck region (Fig. 4 D and E and H and I) at low and high calcium, respectively, were overlaid in Fig. 4 F and J. At low calcium, the overlay revealed some flexibility arising from a single pivot within the converter domain (Movie S1), whereas the calmodulin binding neck domain itself appeared inflexible. At high calcium, a second pivot point emerged between the two calmodulins. This second pivot indicated that the rearrangement due to calmodulin binding to the bipartite site caused a break in the structure of the lever arm (Movie S2). To test whether such a lever arm was still mechanically stable and functional, we performed gliding filament assays (32, 33). In these experiments, fluorescently labeled actin filaments were imaged while gliding over a lawn of myosin-VI motors attached to the surface of the experimental chamber in the presence of ATP. The maximum intensity projections in Fig. 4 G and K illustrate the motility. The tracks of actin filaments gliding over myosin-VI in the absence of calcium (Fig. 4G) are shown in red, whereas the actin filaments in the first frame of the movie are shown in yellow. In the absence of calcium, the maximum gliding velocity was ∼60–80 nm/s and largely independent of the mode of motor attachment to the surface of the experimental chamber. We tested three conditions of myosin attachment (Fig. 4C), namely an antibody against the myosin-VI globular tail domain (green), a C-terminal biotin-streptavidin-linker (blue), and Folch fraction 1 liposomes deposited onto the surface (black). When calcium was added, the motility stopped completely, consistent with a mechanically unstable lever arm. This calcium-regulated motility is illustrated in the experiment in Fig. 4K, where myosin-VI attached to streptavidin via a C-terminal biotin at high calcium was able to bind actin filaments. As long as calcium was high, the motor remained target bound but mechanically inactive. The position of the actin filaments in the first (yellow) and last (red) frame of the 600-s movie overlapped. Only when calcium was lowered did the lever arm gain the mechanical rigidity required to allow the generation of force and movement as seen in Fig. 4G. The structural flexibility of the calmodulin-binding lever arm at high calcium in the EM images (Fig. 4J) indicated that it was not the detachment, but in fact a conformational change of the second calmodulin, including a loss of contact between the calmodulins, that caused the loss in rigidity of the lever arm structure and thus the loss of mechanical functionality at high calcium.

Discussion

A central feature of cellular motile processes lies in the ability to coordinate the timing of motor protein activation and inactivation. The molecular mechanisms of chemo-mechanical energy transduction of several cytoskeletal motors have been investigated at the single molecule level (34). However, the mechanisms and intracellular signals coordinating motor protein mobilization, targeting, and activation in the cell remain unclear. Here we found that the second messenger calcium fundamentally changes the conformation and structural flexibility of the myosin-VI motor. We discovered that this is due to calmodulin changing affinity for binding sites on calcium binding. The rotation of calmodulin from the IQ-motif P2 to the bipartite binding site directs a major rearrangement of the motor from a compact dormant state into a primed, cargo binding state.

To interpret the negatively stained electron micrographs, we modeled a structure that combined the crystal structures of the myosin-VI head domain (23) and the neck region comprising two calmodulins and the subsequent three-helix bundle of the tail domain (26). At low calcium, we found an excellent agreement between the crystal structure of both calmodulins plus helix bundle and the EM image. The extra mass observed in the EM at low calcium between the catalytic domain, the converter, and the first calmodulin indicated the presence of a backfolded myosin-VI tail, as speculated previously based on cell biological and small angle X-ray scattering studies (19, 35). Intriguingly, instead of interactions between the C-terminal cargo-binding domain and the catalytic head domain, as observed in EM studies on myosin-V (21, 22), our spectroscopic studies revealed that in the case of myosin-VI a C-terminal tail segment (aa1060–1125) interacts with apo-calmodulin at the neck region of the motor. This direct interaction of the cargo-binding tail with the calcium sensor on the motor provides a molecular mechanism to couple the folding/unfolding and targeting of the motor to the intracellular calcium signaling pathways.

The majority of our EM data at high calcium showed that the second calmodulin had rotated clockwise by ∼30° in the projection plane, whereas the helix bundle had rotated anticlockwise by ∼30°. This rotation indicated a major structural rearrangement of the motor without the detachment of calmodulin, in contrast to a previous report (36). Consistent with the previous study, we also obtained classes of short molecules that seemed to have lost the second calmodulin at high calcium (Fig. S3C). However, we found that the loss of calmodulin was strongly dependent on free calmodulin supplemented to the buffer solution. In the presence of physiological levels of ∼10 μM calmodulin (37), ∼65% of the myosin-VI molecules retained the second calmodulin at high calcium and underwent the conformational changes described above. In contrast, at low calcium, supplementation with free calmodulin had little effect, with the second calmodulin remaining bound in the majority of molecules. These results indicated that the second calmodulin detaches only partly at the higher, more physiological calmodulin concentrations and rebinds to the bipartite binding site, thus rearranging the fold of the neck and tail region. This rearrangement leads to a release of the tail and primes the motor for mechanical activity. In this way, calcium activates the motor from a dormant, backfolded state into an unfolded, primed state with a tail free to attach to binding partners and cargo.

Using tryptophan fluorescence, we identified the sites and calcium dependence of calmodulin binding to the myosin-VI neck and tail segments. Calmodulin N- and C-terminal lobes interact with amino acids in IQ motifs both independently and in conjunction with each other, giving rise to complex binding behavior. Using calmodulin and calmodulin mutants, as well as a comprehensive library of peptides several of which overlap, we revealed a previously undiscovered bipartite Ca²⁺-calmodulin binding site with two, nonadjacent halves contained in the peptide P½ 2–3. This composite binding site, comprising parts of P2 and the extended loop between the first two helices in the subsequent helix bundle (26), enables Ca²⁺-calmodulin to bridge between the neck and the proximal tail of the motor, probably keeping the structure of the helix bundle intact. This mechanism makes this calmodulin binding site a very likely candidate for the structural rearrangement triggering the release of the backfolded tail at high calcium as observed in the EM.

The lipid-binding studies showed that calcium increased myosin-VI binding to the mixed brain Folch liposomes, consistent with a previous study (6). At high calcium, no actin gliding could be obtained with myosin-VI attached via a C-terminal biotin to a streptavidin surface. In contrast, at low calcium, the actin gliding velocities of 60–80 nm/s at saturating ATP concentrations were similar to those obtained under control conditions with the motor molecules attached at the C terminus via an antibody against the globular tail (18), via Folch liposomes or a biotin-strepavidin linker, and similar to those previously reported for myosin-VI constructs (16, 17). An explanation for the calcium effect is provided by the EM image analysis showing that, at low calcium, the calmodulin binding lever arm pivoted around a single fulcrum near the converter and the lever arm itself appeared largely inflexible, consistent with a mechanically functional lever arm. Intringuingly the emergence of a second fulcrum in between the two calmodulins at high calcium indicated that the lever arm had lost stiffness and had become largely flexible and therefore unable to transmit force. The inhibition of motility at high calcium is consistent with either the binding of the second calmodulin to the neck domain at the bipartite site or the detachment of calmodulin (Fig. S3). Both scenarios are expected to destabilize the mechanical lever arm and would provide a simple explanation for the lack of motility at high calcium, seen here and reported previously (17), and also the uncoupling of the heads in enforced myosin-VI dimers (16).

We propose the following model how calcium can mobilize myosin-VI to attach to cargo and empower the motor mechanically to generate motility (Fig. 5): (i) at low calcium, myosin-VI adopts a backfolded, dormant state, with the tail domain aa1060–1125 binding to apo-calmodulin; (ii) at high calcium, the second calmodulin detaches partly and seeks out an alternate, high-affinity bipartite binding site, coupled with the strong reduction in affinity of the tail for Ca²⁺-calmodulin, myosin-VI unfolds and is now primed to bind cargo; (iii) the conformational change of the second calmodulin leads to a destabilization of the lever arm that prevents myosin-VI from translocating before binding to (lipid) cargo; and (iv) once calcium is lowered again the second calmodulin rebinds to P2 alone, restabilizing the lever arm; the motor, attached to cargo, is now mechanically active; on the release of cargo, or in the event that no cargo is bound, the tail folds back onto calmodulin, switching the motor off. It is conceivable in this condition that cargo could directly compete with the tail–calmodulin interaction if the K_d for the cargo is strong enough. However, increasing calcium concentrations makes this much more likely, by reducing the tail-calmodulin affinity in the backfolded conformation. Our simple model opens up new perspectives of how transients in the ubiquitous second messenger calcium can orchestrate the timing of localized motor activation in this superfamily of molecular motors. The proposed mechanism uncouples the process of target/cargo binding from the mechanical activation. Such a two-step mechanism might represent a paradigm for the problem of recruitment and timing of activation and inactivation of cytoskeletal motors in general.

Fig. 5.

Model. The model illustrates the effect of calcium on the back-folding and release of the tail, conformation and target binding of the second calmodulin, and mechanical stability of the lever arm.

Materials and Methods

Molecular Biology.

Myosin-VI from chicken brush border cells containing the large insert (residues 1–1276) (18) was cloned into pFastbacHtB (Invitrogen). The myosin-VI heavy chain was coexpressed with human calmodulin and purified as previously described (18, 32). The expression of excess calmodulin compared with the myosin-VI heavy chain was confirmed by the presence of unbound calmodulin in the flow-through on loading the cell extract onto the HisTrap-myosin-VI-purification column (GE Healthcare). To generate calmodulin with either the N- or C-terminal calcium binding sites eliminated, the cDNA of calmodulin was mutated to make E104A, E140A for the C-terminal mutant, and E31A, E67A for the N-terminal mutant. The mutant cDNA constructs were then cloned into pET28a bacterial vector and expressed and purified as described previously for WT calmodulin (32).

Tryptophan Fluorescence.

All tryptophan fluorescence studies were performed with target peptide sequences from human myosin-VI (National Center of Biotechnology Gene ID 92859701; amino acids 788–1036, synthesized by GenScript) and human calmodulin. The titrations of predicted target peptides (27, 28) with calmodulin were performed and analyzed at 20 °C in the following buffer (in mM): 25 Tris (pH 8.0), 100 KCl, and 1 DTT, supplemented with either 1 CaCl₂ or 0.2 EDTA, using a Varian Cary Eclipse fluorescence spectrophotometer: λ_ex = 290 nm and λ_em = 323 nm, as previously described (29, 32). The dissociation constants K_d for the tryptophan-containing peptides were determined by direct titration, whereas peptides without tryptophans were titrated against preformed calmodulin–peptide complexes with a known K_d as previously described (29). The equations used to fit the data are described in SI Text.

Motility Assay.

Procedures were adapted from assays as previously described (18, 32, 33) and as detailed in SI Text. Images of the fluorescently labeled actin filaments (90× magnification) were recorded every 10 s for a total period of 600 s. Only filaments moving continuously for at least 20 frames were included in the data analysis. For the assays with biotinylated myosin-VI, myosin was bound to a streptavidin-coated nitrocellulose surface. For the lipid-based assays, before adding myosin, Folch liposomes were introduced into the chamber, as previously described (38). The gliding velocity of filaments was calculated using the analysis software GMimPro (www.mashanov.uk). All assays were carried out at 37 °C. A composite of the first frame (Fig. 4 G and K in yellow) overlaid on a maximum intensity projection of the following 40 frames shows the tracks of the actin filaments.

Cosedimentation of Myosin-VI Head Constructs with F-Actin.

Interactions between the myosin-VI head and tail constructs were probed with pull-down experiments in the presence of F-actin. The Head 913 (aa1–913; 1 μM) and Tail 1060 (aa1060–1276; 3 μM) were mixed with F-actin (5 μM) in a pull-down buffer (in mM: 20 Hepes, pH 7.4, 150 NaCl, and 1 DTT) and incubated at 23 °C for 10 min before centrifugation at 160,000 × g for 15 min at 4 °C. The pellet (resuspended in pull-down buffer) and supernatant were run on SDS/PAGE and stained with Coomassie.

Cosedimentation of GST-Tagged Constructs, Using MagneGST Beads.

Interactions between GST-tagged and His-tagged or nontagged protein were studied using a MagneGST beads system (Promega). The beads with high affinity for GST were incubated with GST-fusion protein in PBS plus 1 mM DTT for 10 min, including 1% BSA to reduce nonspecific binding to the beads. Washing using PBS and incubation with the His-tagged or nontagged binding partner plus 10% BSA for 60 min was followed by further washing steps in PBS. Finally SDS sample buffer was added, the sample was boiled for 5 min, the beads were removed, and the sample was analyzed using SDS/PAGE. The experiments were performed at 23 °C.

Dot Far Western Blot.

To detect protein–protein interactions, unlabeled prey-peptides were titrated against calmodulin, and samples were drawn at predefined ratios. One microliter of a 2 μM solution was spotted onto a nitrocellulose membrane and left to dry for 30 min before the membrane was incubated for 60 min in blocking buffer (in mM: 50 Tris HCl, pH 7.5, 150 NaCl, 0.1% Tween 20 plus 20 mg/mL fatty acid-free BSA; Sigma). Following incubation with the bait protein (15 nM) in blocking buffer for 60 min and five washing steps in TBST, the membrane was incubated for 60 min with an antibait antibody, which was detected using an ECL kit (BioRad) following the manufacturer’s instructions and imaged using a BioRad Geldoc system.

MST.

The method as described previously (39) was adapted for probing the interaction of myosin-VI tail fragments with myosin-VI motor domains and calmodulin (SI Text).

Liposome Cosedimentation.

Liposomes were prepared from bovine brain extract, type I, Folch fraction I (Sigma), and the cosedimentations with myosin-VI were performed as described previously (6, 32).

Negative Stain EM.

Nucleotide-free myosin-VI was diluted to 200 nM in a buffer containing (in mM) 25 NaCl, 20 Tris⋅HCl, pH 7.5, 20 imidazole, pH 7.5, 5 MgCl₂, 1 EGTA, and 10 DTT. For studies at high calcium, instead of EGTA, CaCl₂ (pH 7.5) was added to obtain 100 μM free calcium to ensure saturation of the myosin-VI–calmodulin complex, whereas for the experiments at high calmodulin, the buffer was supplied with 10 μM calmodulin, which yielded four different conditions in terms of calcium and calmodulin concentrations. For each condition, the protein samples were applied to hydrophilized (glow discharge) carbon-coated copper grids (Science Services) and negatively stained with 2% uranyl formate as previously described (32, 40). The grids were examined using a Philips CM 100 electron microscope (Hendrick Dietz, Laboratory for Biomolecular Nanotechnology, Technische Universität Munich) operating at 100 kV, and the micrographs were recorded using a CCD camera at a resolution of 0.33 nm/pixel.

Single Particle Analysis.

The micrographs were processed using EMAN2 (41) and SPIDER (42) software for particle picking, alignment, and classification. If not stated otherwise, alignments were done reference free, and classifications were obtained using K-means clustering. To reduce the influence of background noise and to focus the classification onto specific parts of the molecules, alignments and classifications were preceded by the application of a mask, as previously described (43). Stacks of 40,000–100,000 images, processed in a first round of reference-free alignment and K-means classification, led to stacks of 12,000–45,000 usable images of myosin-VI molecules for each of the four conditions. As described further in SI Text, each stack was further classified into four classes representing major structural differences, called bent and straight, each in two different orientations, related by mirror symmetry with respect to the long axis of the catalytic domain of myosin.

SI Text

Single Particle Analysis.

Stacks of 40,000–100,000 images, processed in a first round of reference-free alignment and K-means classification, led to stacks of 12,000–45,000 usable images of myosin-VI molecules for each of the four conditions in terms of calcium and calmodulin concentration. Each stack was further classified into four classes representing major structural differences, called bent and straight, each in two different orientations, related by mirror symmetry with respect to the long axis of the catalytic domain of myosin. Further alignment and classification of these classes revealed more structural differences and resulted in new classes, for which class averages containing ∼20–60 images were created. Using these class averages as references, a multireference-based alignment of the entire image set could be applied, followed by the creation of classes based on the best fit to one of the references. Then each class was reference-free aligned and classified to detect misclassified images. After reassigning misaligned images to the correct class, the final classes were treated to various rounds of reference-free alignment and classification by K-means.

Projections of Crystal Structures Fitted to the Electron Micrographs.

Nucleotide-free full-length myosin-VI adhered to the carbon-coated grids in two main orientations (Fig. S1) providing a front view with the neck along the long axis of the head domain (straight conformation) and a side view with the head and neck at an angle of ∼53° (bent conformation). To interpret the electron micrographs we compared the class averages with published crystal structures of myosin-VI. We generated a modeled structure comprising a nucleotide-free structure of the catalytic domain (2BKI) (23) and the neck region with two calmodulins and the subsequent three-helix bundle. Because the second calmodulin was modeled on in the published structure (23), which also lacked the subsequent three-helix bundle, we combined it with a second crystal structure, which resolved the lever arm with two calmodulins and the three-helix bundle (3GN4) (26). This alignment was achieved using the FATCAT rigid pairwise alignment method (45). The significance of similarity between the structures had a P value of 1.79E−06. The combination of the aligned structures yielded a modeled structure of myosin-VI (aa1–847) with the catalytic domain from 2BKI and the lever arm from 3GN4 (Fig. S1A).

To interpret the planar projections of myosin-VI in negative stain EM, a set of 799 projections in 2D of the modeled structure in different orientations and a set of 799 projections of a myosin-VI–ADP structure (46) were generated, as described previously (47). The projections were low pass filtered to simulate the lower resolution of the EM images. After adjusting the scales, all projections were cross-correlated with the real EM averages of the straight and bent molecules at low and high calcium and ranked according to their cross-correlation coefficient (CCC; Fig. S1 B–D). Finally the crystal structures in the optimized projection were overlaid onto the EM class averages to interpret structural details (Fig. S1 E–G). For viewing and editing the structures, we used Jmol (jmol.sourceforge.net) and Swiss-PDBViewer (48).

Straight and Bent Conformations of Myosin-VI.

For the EM average in the bent conformation, a very good fit could be achieved with the optimized projection of the modeled crystal structure (Fig. S1E). For the straight conformation, the best fit was obtained by optimizing the projection of the catalytic domain of the model (Fig. S1F). The orientation of the modeled structure optimized for the bent conformation had to be rotated by 89° around the y axis and tilted in the y/z plane (vector r; Fig. S1A, structure on the right). In this projection, the calmodulin-binding lever arm, however, lies underneath the catalytic domain (Fig. S1F, Upper), inconsistent with the EM dataset where the molecules were stretched out along the long axis of the catalytic domain (Fig. S1F, Lower). We also fitted a myosin-VI–ADP structure in the prepowerstroke state (46), again using the catalytic domain to determine the optimized projection (Fig. S1G). The orientation of the lever arm in the EM data, however, was also inconsistent with this prepowerstroke structure. We concluded that the straight conformation was not a result of a different nucleotide state or molecules in a prepower stroke conformation, but probably reflected a different binding mode to the surface, including distortions of the molecule caused by interactions between the protein molecule, stain, and grid. Therefore, we focused the structural analysis on the bent conformation, in which the projections of the calmodulins were also larger in the image plane (compared with the straight molecules) so that calcium-induced conformational changes could be better resolved.

CCCs to Identify Calcium-Sensitive Parts of the Structure.

For the bent conformation, comparison of the CCC values showed that the structural differences between low and high calcium conditions were mostly due to the neck-tail region, specifically the orientation of the second calmodulin (CaM2) and the tail (Fig. S2). The better fit of the catalytic domain to the modeled structure at high calcium was consistent with the absence of the backfolded tail. The statistics on the EM data (Fig. S3) suggested that the calmodulin binding neck region was also the critical factor for the orientation of myosin-VI on the EM grid at low and high calcium. The inverted ratios of straight and bent molecules at low and high calcium were probably due to changes in the surface charge on the calmodulin binding neck regions. Generally at high calcium, calcium-calmodulin at the neck seemed to favor surface attachment in the bent conformation, whereas at low calcium, the second calmodulin in the apo-state appeared to favor the straight conformation (Fig. S3B). At high calcium and high calmodulin, the presence of the second Ca²⁺-calmodulin on the neck seemed to favor the binding in the bent conformation even more (Fig. S3C) compared with the low calmodulin condition, where only short molecules were found because the second calmodulin was either lost or in variable position and the bent conformation was favored somewhat less. Of the two different orientations, related by mirror symmetry with respect to the long axis of the catalytic domain of myosin (called left and right; Fig. S3 B and C), the binding in left orientation was generally favored, except for the short molecules at high calcium and low calmodulin, which indicated a role of the second calmodulin for the binding in left orientation in general.

Movies Based on Electron Micrographs.

To reveal regions of flexibility within the structure of myosin-VI some classes were aligned and classified using specific masks. Previous structural data suggested that, for the most part, the catalytic domain of myosin does not feature much flexibility, whereas other parts are flexible to some extent (e.g., converter region–light chain binding lever arm) or highly flexible (e.g., tail domain). Therefore, a mask was applied covering only the catalytic domain. This mask resulted in a class average that resolved this part of the molecule in great detail, whereas the parts not covered by the mask appeared blurred. The subsequent classification of the blurred parts revealed the extent of flexibility of this region. The class averages with sufficient resolution were sorted and turned into a movie to illustrate the extent of flexibility of the structural parts outside the mask.

Tryptophan Fluorescence.

Dissociation constants for tryptophan-containing peptides (W) and calmodulin (C) were determined by direct fluorometric titrations (λ_ex = 290 nm and λ_em = 323 nm). As described previously (49), the direct fluorometric titrations were fitted to the following equation:

$$Fluorescence={{\displaystyle F}}_{\left(C\right)}\left[C\right]+{{\displaystyle F}}_{\left(W\right)}\left[W\right]+{{\displaystyle F}}_{\left(CW\right)}\left[CW\right],$$

where F is the molar fluorescence intensity. The dissociation constant K_d(W) was obtained from a nonlinear least squares fit to this equation with concentrations calculated by solving the following equation, according to ref. 49:

$${\left[CW\right]}^2-\left(K_{d\left(W\right)}+C_T+W_T\right)\left[CW\right]+C_T{W}_T=0.$$

The subscript T denotes total concentrations. A factor X_W was included in the fitting equation to correct for errors in the peptide concentration, i.e., the actual concentration was W_TX_W, and experiments with X_W > 1.1 or X_W < 0.9 were rejected.

Dissociation constants for peptides (W) without tryptophan were determined using a fluorescence competition assay in which the nonfluorescent peptide was used to displace a tryptophan-containing peptide from its complex with either apo- or calcium-calmodulin. As described previously (49), the optical signal for the displacement assay was fitted to the following equation:

$$Signal=F_{\left(C\right)}\left[C\right]+F_{\left(w\right)}\left[W\right]+F_{\left(S\right)}\left[S\right]+F_{\left(CW\right)}\left[CW\right]+F_{\left(CS\right)}\left[CS\right],$$

where S is the spectroscopically silent peptide.

As previously described (49), the dissociation constant K_d(S) was obtained from a nonlinear least squares fit to the following equation with concentrations calculated by solving

$${\left[C\right]}^3+\left(-C_T+K_{d\left(S\right)}+K_{d\left(W\right)}+W_T+S_T\right){\left[C\right]}^2+\left(-C_T{K}_{d\left(S\right)}-C_T{K}_{d\left(W\right)}+K_{d\left(S\right)}{K}_{d\left(W\right)}+W_T{K}_{d\left(S\right)}+S_T{K}_{d\left(W\right)}\right)\left[C\right]-C_T{K}_{d\left(S\right)}{K}_{d\left(W\right)}=0,$$

with K_d(W) fixed at the value determined from the direct titration.

The titration experiments were performed with the constructs described in Table S1; the amino acid numbers refer to myosin-VI from human. The blue shift and increase in tryptophan fluorescence intensity on calmodulin binding to the peptide P1 at pCa 4 is shown in Fig. S4A. The fluorescence at λ_em 323 nm increases with increasing calmodulin concentrations until saturation is reached at a 1:1 binding stoichiometry between the peptide and calmodulin. Beyond this point, there is only a slow increase in background fluorescence due to further increasing calmodulin concentrations. The complete, calcium-dependent reversibility of complex formation is shown in Fig. S4B. Mutation of the N- or the C-terminal lobe of calmodulin at the critical amino acids (16) leads to a change in the binding stoichiometry from 1:1 to 2:1, with a single peptide now binding two calmodulins, each via a single lobe only.

Biotinylation of Myosin-VI.

Myosin-VI constructs with a C-terminal biotinylation (BRS) site were expressed to bind a biotin-tag covalently to the C terminus. Biotinylation of myosin-VI–BRS, using d-biotin and biotin-ligase (BirA, Avidity), was performed following the manufacturer’s instructions. Following purification via a HisTrap column, the biotinylated myosin-VI was further purified using a monomeric Avidin column (Thermo Scientific), again following the manufacturer’s instructions. The biotinylation of the purified myosin-VI was confirmed using a Western blot.

Motility Assay.

Procedures were adapted from assays described previously (18, 32, 33). In brief, myosin-VI (∼150 μg/mL) was immobilized on a nitrocellulose-coated glass surface [0.1% (vol/vol) nitrocellulose in amylacetate] using a polyclonal myosin-VI antibody against the globular tail domain (18). To prevent unspecific binding, the surface was blocked with 0.5 mg/mL BSA in assay buffer [AB buffer, in mM: 20 Tris⋅HCl (pH 7.5), 20 imidazole, 25 NaCl, 1 EGTA, 5 MgCl₂, and 10 DTT] before rhodamine-phalloidin–labeled rabbit skeletal actin filaments were introduced in AB buffer supplemented with a scavenger system [10 mM DTT and (in mg/mL) 0.01 catalase, 0.05 glucose oxidase, and 1.5 glucose] and 5 mM ATP.

MST.

The method described by Wienken et al. (39) was adapted to probe myosin-VI tails with myosin-VI motor domains and calmodulin. Purified proteins were serially diluted into PBS and mixed 1:1 (vol:vol) with either 750 nM 913-Head or 300 nM 1060-Tail that were labeled with the RED NHS (amine reactive) MST kit (Nanotemper). The mixtures were transferred to glass capillaries (MST premium coated; Nanotemper) and analyzed in a Monolith NT.115 (Nanotemper). Thermophoresis with temperature jump readings from three independent experiments at 20% RED LED power, at 30 s under 40% laser power, were converted to bound/unbound ratios, and the affinity constants were calculated.