Structural implications of Ca2+-dependent actin-bundling function of human EFhd2/Swiprosin-1

EFhd2/Swiprosin-1 is a cytoskeletal Ca2+-binding protein implicated in Ca2+-dependent cell spreading and migration in epithelial cells. EFhd2 domain architecture includes an N-terminal disordered region, a PxxP motif, two EF-hands, a ligand mimic helix and a C-terminal coiled-coil domain. We reported previously that EFhd2 displays F-actin bundling activity in the presence of Ca2+ and this activity depends on the coiled-coil domain and direct interaction of the EFhd2 core region. However, the molecular mechanism for the regulation of F-actin binding and bundling by EFhd2 is unknown. Here, the Ca2+-bound crystal structure of the EFhd2 core region is presented and structures of mutants defective for Ca2+-binding are also described. These structures and biochemical analyses reveal that the F-actin bundling activity of EFhd2 depends on the structural rigidity of F-actin binding sites conferred by binding of the EF-hands to Ca2+. In the absence of Ca2+, the EFhd2 core region exhibits local conformational flexibility around the EF-hand domain and C-terminal linker, which retains F-actin binding activity but loses the ability to bundle F-actin. In addition, we establish that dimerisation of EFhd2 via the C-terminal coiled-coil domain, which is necessary for F-actin bundling, occurs through the parallel coiled-coil interaction.

The PxxP motif is involved in proper intracellular localisation of target proteins through basic motifs (Arg/ Lys), exposed hydrophobic residues and a pair of Pro residues 28 . These three conserved elements of the PxxP motif are important for phosphoinositide binding, penetration of the lipid bilayer and SH3 domain binding, respectively 28 . Interestingly, the PxxP motif of CD EFhd2 is not only required for association with the B-cell membrane, but was also identified as part of the multiple actin-binding sites 20,26 . Although 10 residues (residues 70-79) of the PxxP motif were disordered in the crystal structure, Pro80, Pro82, Phe86 and Phe89 face towards helix 4 of the EF-hands to form hydrophobic interactions, and Glu85, Glu88 and Tyr83 form hydrogen bonds with Lys95, Arg151 and Arg158 of helix 1 and 4 of the EF-hand domains. As a result, the PxxP motif (ABS1) interacts tightly with the EF-hand domains (ABS2) (Fig. 1c). Furthermore, the EF-hand domains (ABS2) not only interact tightly with the PxxP motif (ABS1) via helix 1 and 4, but also associate with the LM-helix (ABS3) through intramolecular interactions that resemble the intermolecular interactions of Ca 2+ -calmodulin (CaM)-peptide complexes (Fig. 1a,d).
Structural implications of EFhd2 in the absence of Ca 2+ . We failed to determine the structure of the apo form of EFhd2 owing to structural instability during protein purification; however, we could determine the crystal structures of CD EFhd2 mutants defective for one Ca 2+ -binding site (E116A for EF1, CD EFhd2 EF1 ; E152A for EF2, CD EFhd2 EF2 ) (see Supplementary Fig. S3). The overall structures of these two mutants in the presence of Ca 2+ are similar to that of Ca 2+ -bound CD EFhd2 ( CD EFhd2 EF1 , root mean square deviation (RMSD) = 0.34 Å for 102 Cα atoms; CD EFhd2 EF2 , RMSD = 0.61 Å for 105 Cα atoms), which implies a single Ca 2+ -loaded EF-hand is sufficient to maintain a stable fold (Fig. 2a). However, the Ca 2+ -binding loop region of EF1 of CD EFhd2 EF1 (Arg106, Gly107 and Arg108) was observed to be disordered (Fig. 2b). In addition, in the structure of CD EFhd2 EF2 , one water molecule occupied the Ca 2+ position of EF2 and forms several hydrogen bonds with Asp141, Asp143, Asp145 and Lys147. Moreover, Asp143 forms a hydrogen bond with Arg151 and the Ca 2+ -binding loop is slightly shifted (~2.2 Å). As a result, the water molecule is trapped in the Ca 2+ -binding site of EF2 (Fig. 2c). Furthermore, comparison of the crystallographic B-factors between Ca 2+ -bound and EF-hand mutant structures showed that the largest changes in B-factor values were for CD EFhd2 EF1 (35.6 Å 2 ), CD EFhd2 (20.4 Å 2 ) and CD EFhd2 EF2 (21.5 Å 2 ). In particular, B-factor values for EF1 and the C-terminal linker region in the structure of CD EFhd2 EF1 were increased significantly (Fig. 2d). On the other hand, only small changes of B-factor values for EF2 in the structure of CD EFhd2 EF2 were observed and are probably because of small structural perturbations in the absence of Ca 2+ and stabilisation by newly formed hydrogen bonds to the trapped water molecule. These results suggest that the Ca 2+ -binding loop of EF1 adopts a more flexible structure than EF2 in the absence of Ca 2+ , resulting in large conformational fluctuations to EF1 and a concomitant increase in the overall B-factor. Next, we performed CSP analysis using the conditions of Ca 2+ -free and bound states to identify conformational changes to CD EFhd2 in the absence Ca 2+ . The Ca 2+ -dependent conformational changes to CD EFhd2 were monitored by measuring resonance perturbations in 2D 1 H-15 N HSQC spectra (see Supplementary Fig. S4a). Consistent with the crystal structures of the CD EFhd2 mutant, significant CSPs were associated with residues in Ca 2+ binding loop region (Phe101, Asp105, Asp109, Phe111, Ile 112 and Glu116) of EF1. Noticeably, CSPs of hydrophobic residues in the Ca 2+ -binding loop of EF1 are likely to be associated with the failure of structure determination caused by the instability of EFhd2 in the absence of Ca 2+ . Significant CSPs for the disordered region (residues 70-80, PxxP motif) of the crystal structure appears to be associated with the conformational changes to EF1 in the absence of Ca 2+ . On the other hand, noticeable CSPs were not detected in the LM-helix region (see Supplementary Fig. S4b,c). Thus, we expect the LM-helix region to maintain its structure in the absence of Ca 2+ .
Collectively, we postulate that Ca 2+ depletion leads to local conformational flexibility of actin-binding sites (EF1, C-terminal linker) and this reduces the F-actin bundling ability of EFhd2 in the absence of Ca 2+ , as observed in previous results 26 . Ensemble refinement of the CD EFhd2, CD EFhd2 EF1 and CD EFhd2 EF2 . Based ∑ h is the sum over all reflections, and ∑ i is the sum over i measurements of reflection h. c R work = Σ hkl ||F o |-|F c ||/ Σ hkl |F o |; R free is the R value calculated for 5% of the data set not included in the refinement.
we performed ensemble refinement for CD EFhd2, CD EFhd2 EF1 and CD EFhd2 EF2 using Phenix.ensemble refinement 29 . Ensemble refinement is a useful tool to highlight functional protein dynamics through X-ray diffraction data 29 . Ensemble refinement of the CD EFhd2, CD EFhd2 EF1 and CD EFhd2 EF2 yielded a large number of models that represent structural dynamics and decreases in the R free value ( CD EFhd2 = 5.2%, CD EFhd2 EF1 = 3.7%, CD EFhd2 EF1 = 3.8%) compared with the single structure. In the model structures, different degrees of mobility in CD EFhd2, CD EFhd2 EF1 and CD EFhd2 EF2 were observed (Fig. 3). As expected, Ca 2+ -bound CD EFhd2 displayed a rigid conformation and CD EFhd2 EF1 showed the largest degrees of mobility, indicating that Ca 2+ depletion of EF1 has a larger impact on conformational dynamics (Fig. 3a,b). In addition, CD EFhd2 EF2 also showed a moderate  degree of mobility (Fig. 3c). The significant increase of the root-mean-square fluctuation (RMSF) in the EF1 of CD EFhd2 EF1 is consistent with the crystallographic B-factor and CSP analysis, which support an increase in local flexibility of the actin binding sites of EFhd2 in the absence of Ca 2+ (Fig. 3d). Interestingly, the C-terminal linker (residues, 176-184) followed by the LM-helix also showed significant RMSF increases in CD EFhd2 EF1 and CD EFhd2 EF2 (Fig. 3d). These results support the postulate that the EFhd2 core domain forms local dynamic conformations (EF1, C-terminal linker) in the absence of Ca 2+ .  (Fig. 4a). The decrease in entropy upon Ca 2+ binding indicates that the flexible conformation of the Ca 2+ binding site in the absence of Ca 2+ changes to a rigid conformation. The Ca 2+ -binding affinity of EF-hand containing proteins is related to protein stability. In other words, high affinity towards Ca 2+ leads to instability in the Ca 2+ -free state 30 . To assess the effect of Ca 2+ in EFhd2 stability, we measured the Ca 2+ -dependent EFhd2 thermostability thorough a heat aggregation test (Fig. 4b). Consistent with a previous study showing that thermostability of EFhd2 was restored by Ca 2+ at a high temperature 31 , the half aggregation temperature for both EF-hand mutants that bind only one Ca 2+ is significantly lower ( CD EFhd2 EF1 : 62.32 ± 0.14 °C, CD EFhd2 EF2 : 57.90 ± 0.60 °C) than the two Ca 2+ -bound EFhd2 ( CD EFhd2: 84.89 ± 0.01 °C) and is consistent with ref. 30 and 31. Although the crystal structures of EF-hand mutants ( CD EFhd2 EF1 and CD EFhd2 EF2 ) are similar to Ca 2+ -bound CD EFhd2, we observed structural flexibility in the Ca 2+ -binding loop of EF1 and the C-terminal linker (Figs 2 and 3). Based on structural and biochemical results, we hypothesise that changes in the conformation and flexibility lead to exposure of hydrophobic residues around the Ca 2+ -binding loop of EF-hands and the C-terminal linker, and this exposure of hydrophobic residues affects protein stability. We previously reported that F-actin bundling activity decreases in the absence of Ca 2+ 26 . We further analysed the contribution of Ca 2+ binding to each EF-hand domain on F-actin binding and bundling activity (Fig. 5). It is interesting to note that wild-type (Ca 2+ -bound and Ca 2+ -unbound) and each EF-hand mutant (only one Ca 2+ -bound) showed similar F-actin binding activity. Surprisingly, however, F-actin bundling activities were quite different and dependent on the Ca 2+ -bound states. Even depletion of one Ca 2+ site in a two Ca 2+ -binding protein showed a dramatic reduction in F-actin bundling activity to a level that is similar to that of previously reported data for a two Ca 2+ -depleted state 26 . We propose that the increased structural flexibility observed in the Ca 2+ -binding loop and C-terminal linker, which encompass actin-binding sites, cause a reduction in F-actin bundling activity, presumably because coordination of the F-actin binding sites for F-actin bundling is disrupted. The rootmean-square fluctuation (RMSF) of ensemble models for Ca 2+ -bound CD EFhd2 (green), CD EFhd2 EF1 (cyan) and CD EFhd2 EF2 (orange). RMSF in the EF1 of CD EFhd2 EF1 is consistent with the crystallographic B-factor. The disordered region of the PxxP motif in the crystal structure of CD EFhd2 is shown by the blue dashed line.

High Ca
Structural comparison between CD EFhd2 and a homologous protein, allograft inflammatory factor-1 (AIF-1). AIF-1 and EFhd1/EFhd2 are highly evolutionarily conserved proteins, because these genes are generated from common ancestral species of the Bilateria 20 . In particular, EFhd2 and AIF-1 exhibit the same cellular function as an actin-binding protein. Although sequence homology between AIF-1 and EFhd2 is limited to the EF-hand domains, AIF-1 exhibits F-actin binding and crosslinking activity similar to that observed for EFhd2 (see Supplementary Fig. S2b). EFhd2 exhibits F-actin bundling activity in a Ca 2+ -dependent manner, whereas AIF-1 does not exhibit Ca 2+ dependency for F-actin binding and bundling activity 26,[32][33][34] .
The structure of AIF-1 has been determined in the presence and absence of Ca 2+ (PDB IDs: 1WY9 and 2D58) 35 . To investigate the molecular basis of the effect of Ca 2+ on F-actin bundling activity, we compared Ca 2+ -bound and Ca 2+ -free structures of CD EFhd2 and AIF-1. The CD EFhd2 structure is similar to the structures of Ca 2+ -bound and the apo form of AIF-1 (1WY9, Ca 2+ -bound form_Z-score = 4.0, RMSD = 2.48 Å for 76 Cα atoms; 2D58, apo-form_Z-score = 5.5, RMSD = 1.98 Å for 86 Cα atoms), even though we failed to solve the Ca 2+ -free structure owing to protein destabilisation during the protein purification process (see Supplementary Fig. S5a). Interestingly, EF1 of AIF-1 is stabilised by hydrogen bonds between Asn60, Asn62 and Asp66 in the absence of Ca 2+ (see Supplementary Fig. S5b). In addition, a water molecule is trapped in the Ca 2+ binding site of EF2 in the apo structure of AIF-1 similar to that observed for CD EFhd2 EF2 (see Supplementary Fig. S5c). Therefore, CD EFhd2 reveals two Ca 2+ -bound EF-hands, whereas Ca 2+ bound to only EF2 of AIF-1, because Ca 2+ -binding residues (Asp/Glu) are not conserved in EF1 of AIF-1 (see Supplementary Fig. S2) 35 . These structural features of CD EFhd2 and AIF-1 suggest that Ca 2+ is essential for the formation of a stable structure of CD EFhd2, whereas AIF-1 is capable of maintaining a stable structure in the absence of Ca 2+ through hydrogen bonds involving several residues of the Ca 2+ -binding loop of EF1 and a water molecule located in the Ca 2+ -binding site of EF2. These structural differences between CD EFhd2 and AIF-1 support the hypothesis that Ca 2+ is essential for the actin-bundling function of EFhd2 by maintaining a stable structure, whereas AIF-1 exhibits F-actin binding and bundling activity regardless of Ca 2+ dependency [32][33][34] .

EFhd2 displays an actin-bundling function with the parallel coiled-coil domain at the C-terminus.
We already reported that the C-terminal coiled-coil domain is essential for the dimerisation of EFhd2 because we observed EFhd2 lost F-actin bundling activity in the absence of the coiled-coil domain 26 . However, it was unclear whether EFhd2 dimerised by parallel or antiparallel interactions. To establish the molecular architecture of EFhd2, we engineered recombinant fragments corresponding to the predicted coiled-coil domain (residues 199-240), including a Cys residue at either the N-terminus (CC1) or C-terminus (CC2) of the coiled-coil domain 36 . We expected that if the coiled-coil domain assembles as a parallel interaction, formation of a disulfide bond should occur owing to the high proximity between Cys residues of each polypeptide and the dimer should be detected on a non-reducing denaturing gel. Therefore, purified recombinant proteins were resolved on sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) gels under reducing or non-reducing conditions. The disulfide cross-linking assay revealed that SDS-PAGE analysis gave only a monomer band under reducing conditions. In contrast, both CC1 and CC2 protein bands migrated as monomers with the dimeric form observed under non-reducing conditions. These results suggest that the coiled-coil domain of EFhd2 assembles into a parallel dimer (Fig. 6).

Discussion
Ca 2+ is an essential modulator of signal transduction processes required for various cellular functions such as contraction, cell differentiation and proliferation 30 . The presence of EF-hand domains in EFhd2 raises the possibility that EF-hand domains upon Ca 2+ binding may affect its cellular function related to actin dynamics regulation. For example, we found that Ca 2+ or ethylene glycoltetraacetic acid (EGTA) had little effect on EFhd2 binding to F-actin; however, the F-actin bundling activity was significantly reduced in the Ca 2+ -free state and these results were visualised by electron microscopy 26 .
In this study, we have tried to determine the structure of EFhd2 to elucidate the Ca 2+ -dependent F-actin bundling mechanism of this protein. In a search for structures similar to CD EFhd2 using the Dali program 37 , we were able to find ~100 Ca 2+ -bound EF-hands with similar structures (> 10 for Z-scores and < 3.0 Å in RMSD), in which most of the EF-hand matches were CaM and troponin C (TnC). In addition, the structure of CD EFhd2 fits well with those of Ca 2+ -CaM-peptide complexes (see Supplementary Fig. S6) [38][39][40][41] . Ca 2+ -CaM-peptide complexes are more compact than the peptide unbound form because of intermolecular interactions between exposed hydrophobic grooves of CaM and hydrophobic residues of the target molecule. In addition, the Ca 2+ affinity of EF-hands increases with intermolecular interactions, leading to structural stabilisation of the Ca 2+ -bound state 30,42 . Many EF-hand containing proteins can change their diverse biochemical responses through changes in conformation and/or protein stability in the presence or absence of Ca 2+ 30,42 . For example, members of the CaM superfamily are capable of modulating numerous intracellular processes in a Ca 2+ -dependent manner by undergoing conformational changes represented by "close" to "open" structures. On the other hand, several EF-hand containing proteins such as sarcoplasmic Ca 2+ -binding proteins (CaBPs), calcium vector protein (CaVP), calerythrin, and stromal interaction molecule-1 (STIM1) remain in an unstable form in the absence of Ca 2+ in vitro [43][44][45][46][47] . In particular, the structure of STIM1 adopts a compact conformation through a hydrophobic interaction between EF-hands and a SAM domain (sterile α motif) similar to Ca 2+ -peptide-CaM complexes and CD EFhd2. In addition, mutational analysis revealed that Ca 2+ depletion or disruption of hydrophobic interactions between EF-hands and the SAM domain leads to destabilisation of the entire EF-SAM complex 43 . Taking into consideration a previous study and structural similarity between Ca 2+ -peptide-CaM complexes, the observations for STIM1 and CD EFhd2 support the hypothesis that high affinity for Ca 2+ and intramolecular interactions of CD EFhd2 are likely to maximise stabilisation of the EFhd2 fold. In support of this hypothesis is the thermostability results of EFhd2, which showed that the protein thermal stability at high temperature was restored by Ca 2+ 31 . This is further emphasised by the observation that CD EFhd2 remained stable in solution, even at high temperatures in the presence of two Ca 2+ ions (Fig. 4). As mentioned earlier, in the case of CaM, the core region comprising two EF-hand domains shows significant conformational change upon Ca 2+ binding, which leads to structural changes in two lobes and interaction with partner proteins. However, in the case of CD EFhd2, the structural evidence in this report (including mutant structures and CSP analysis) indicates that the core structure of the EF-hand domains is retained regardless of Ca 2+ binding, because two hydrophobic clusters in CD EFhd2 are maintained (see Supplementary Figure S7a). Denessiouk et al. classified EF-hand domains in five groups based on differences in the structural changes in the core region (hydrophobic cluster I and II) upon Ca 2+ binding 48 . CD EFhd2 may belong to type I or IV, because these types have an open conformation in the Ca 2+ -bound form. In the apo state, type I EF-hand domains (Parvalbumin, PVALB) maintain an open conformation; however, type IV EF-hand domains (CaM and Troponin C, TnC) exhibit a closed conformation 48 . This structural difference between types I and IV raised the possibility that the CD EFhd2 may belong to type I, because we expect CD EFhd2 to have an open conformation in the apo state based on the mutant structures and CSP analysis. Additionally, we compared the structures of the single Ca 2+ -bound state in types I and IV. Intriguingly, in the case of type IV (TnC), the single Ca 2+ -bound intermediate state resembled the closed conformation of the apo state (Supplementary Figure S7b) [49][50][51] . The structure of the single Ca 2+ -bound state of type I (PVALB) is close to that of the two Ca 2+ -bound state, although the structural difference between the apo and two Ca 2+ -bound states is marginal (Supplementary Figure S7c) [52][53][54] . This again suggests that CD EFhd2 belongs to type I, because the structures of the single Ca 2+ -bound state of the two CD EFhd2 mutants are similar to the structure of the two Ca 2+ -bound state, and the core structures of CD EFhd2 may not differ even when in a complex with interacting proteins.
We failed to solve the structure of EFhd2 in the absence of Ca 2+ because of protein instability; however, structures of EF-hand mutants, CSP analysis and ensemble refinement analysis showed that CD EFhd2 undergoes changes in local structure and dynamics in the absence of Ca 2+ . The crystal structures of the EF-hand mutants are maintained even when one EF-hand loses Ca 2+ binding capacity (Fig. 2a). However, the Ca 2+ binding loop region that loses Ca 2+ binding activity exhibits structural flexibility (Fig. 2b,d). Furthermore, RMSF values support the premise that F-actin binding sites of EFhd2 form locally dynamic conformations (EF1, C-terminal linker) in the absence of Ca 2+ and this dynamic state reduces F-actin bundling activity (Figs 3 and 5). In particular, greater flexibility of the C-terminal linker between the EF-hands and coiled-coil domain probably leads to incorrect coordination of actin binding sites in dimer formation. Based on these results, we suggest that the EFhd2 core domain comprising the multiple actin-binding sites changes to an unstable structure by changes in local conformational flexibility in the absence of Ca 2+ , and these structural dynamics reduce the F-actin bundling function.
Recently, a structural model for the Ca 2+ -dependent F-actin crosslinking mechanism by non-muscle α -actinin-1 was reported 15 . Non-muscle α -actinin-1 is composed of N-terminal CH domains (actin binding sites), repeated rod domains and C-terminal EF-hands (CaM-like domain; CaMD). Non-muscle α -actinin-1 forms an antiparallel dimer via the rod domain composed by 4 spectrin-like repeats 11,13 . NMR structures of the holo and apo form of CaMD of α -actinin-1 reveal that apo CaMD forms a flexible structure owing to the unstructured linker between N-and C-lobes; however, Ca 2+ binding leads to stabilisation of the linker, resulting in structural rearrangement of CaMD. Consequently, rearrangement of CaMD inhibits proper orientation of adjacent F-actin binding sites for F-actin crosslinking 15 . This observation supports the concept that Ca 2+ -dependent local conformational flexibility of EFhd2 plays a critical role in regulation of F-actin bundling activity by induced reorganisation of actin-binding sites.
Ca 2+ is essential for leading edge formation because several Ca 2+ -related actin-binding proteins modulate cell motility and shape by reorganisation of F-actin structures in a Ca 2+ -dependent manner 11,55 . For example, F-actin crosslinking activity of non-muscle α -actinin and villin at the leading edge of cells is drastically inhibited at high Ca 2+ concentrations (micromolar levels) [8][9]11 . However, Ca 2+ is required for F-actin bundling function of EFhd2 in contrast with what is observed for α -actinin and villin. We speculate that various Ca 2+ -related F-actin bundling proteins may be involved in F-actin reorganisation as suitable regulators in specific cell environments. Furthermore, in our earlier studies, EFhd2 was mainly expressed at the leading edge of cells and improved lamellipodia formation and cell migration 26 . Interestingly, Beerman et al. analysed Ca 2+ transients of migrating immune cells through direct measurement of Ca 2+ signalling using light-sheet microscopy. They demonstrated that Ca 2+ fluctuations were enhanced at the leading edge and reduced at the lagging edge of migrating immune cells 56 . In many EF-hand-containing proteins, including calmodulin, calbindin D9k, and vitamin K-dependent protein S, K d s for Ca 2+ are highly dependent on ionic strength. For these proteins, binding affinity for Ca 2+ is lowered by approximately 2.5-100 fold in the presence of 0.15 M NaCl (close to physiological conditions) [57][58][59] . We were able to measure K d s (70-100 nM) for EFhd2 mutants only at low ionic strength (50 mM Tris-HCl, pH 8.5, 20 mM NaCl); we failed to obtain measurements at higher ionic strengths (even at 100 mM NaCl) because of the instability of EFhd2 mutants at higher ionic strengths in the absence of Ca 2+ . Thus, we hypothesise that the affinity of these mutants for Ca 2+ is much lower than 100 nM, and that both EF hands would not be occupied by Ca 2+ at resting Ca 2+ levels in live cells. This result supports the mechanism of cell migration by EFhd2 because Ca 2+ is essential for the F-actin bundling function of EFhd2.
In conclusion, we demonstrate that EFhd2 shows unique structural and biological features as an EF-hand containing F-actin bundling protein. For F-actin bundling activity, structural stabilisation of the EF-hand domains was found to occur in the presence of Ca 2+ . The core region of EFhd2 maintains its structure in the absence of Ca 2+ ; however, changes in local conformational flexibility reduce F-actin bundling activity of EFhd2 by incorrect coordination of actin-binding sites in parallel dimer formation. Finally, EFhd2 acts as a cytoskeleton-associated adaptor protein that contains two functional EF-hand domains with high Ca 2+ -binding affinity, which might be a useful target for further research involved in its biological functions or various pathologies 25,60,61 .

Methods
Cloning and protein purification of full-length EFhd2 and ΔNTD. A human EFhd2 clone encoding full-length (residues 1-240) and Δ NTD (residues 70-240) were amplified using the polymerase chain reaction (PCR) from pOTB7 (RZPD German Resource Centre, Germany). Full-length EFhd2 was cloned into a modified pET28a vector (Novagen) containing an N-terminal 6× His (His 6 )-tobacco etch virus (TEV) tag. Δ NTD Scientific RepoRts | 6:39095 | DOI: 10.1038/srep39095 was cloned into a modified pET28a vector (Novagen) containing a His 6 -Nus-TEV tag. Recombinant DNA were transformed into E. coli strain BL21 (DE3) and the cells were grown in Luria-Bertani (LB) medium containing 50 μ g/mL kanamycin at 37 °C until the absorbance at 600 nm was 0.7. Expression of recombinant proteins was induced by adding isopropyl β -D-1-thiogalactopyranoside (IPTG) to a final concentration of 0.5 mM and cells were grown for a further 5 h at 37 °C. Cells were harvested by centrifugation (4,000 g) for 20 min at 4 °C. The cell pellet was resuspended in buffer containing 50 mM HEPES-NaOH, pH 7.5, 300 mM NaCl and 5 mM imidazole, and the cells disrupted by sonication. After removal of the cell debris by centrifugation at 14,000 g for 50 min and 4 °C, the soluble supernatant was loaded onto an equilibrated gravity-flow column (Bio-Rad, Hercules, CA, USA) packed with Ni-NTA agarose resin (Peptron, Korea). The protein was eluted with a buffer containing 50 mM HEPES-NaOH, pH 7.5, 300 mM NaCl and 300 mM imidazole. After concentrating the eluate, the protein solution was incubated with TEV protease overnight at 4 °C to remove the N-terminal His 6 or His 6 -Nus tag. To exchange the buffer for crystallisation, the final purified protein was passed through a HiLoad 16/60 Superdex 75 gel filtration column (Pharmacia Biotech) pre-equilibrated with 20 mM HEPES-NaOH, pH 7.5, 150 mM NaCl and 1 mM CaC1 2 . Severe degradation was observed after incubation with TEV protease. Therefore, full-length EFhd2 was cloned into a modified pET-21a vector (Novagen) containing N-terminal His 6 tag. The overall purification procedure was the same as described above. However, the removal process of the N-terminal His 6 tag was omitted, because the modified pET-21a vector (Novagen) does not include a protease cleavage site.
EFhd2 core domain ( CD EFhd2, residues 70-184) identification. We initially tried to crystallise the full-length EFhd2 with Δ NTD. However, crystallisation of this construct failed because of severe degradation during the purification process. Thus, we performed limited proteolysis experiments to identify stable domains and Δ NTD was used for this purpose. Treatment with TEV protease overnight at 4 °C gave a stable fragment, as observed by SDS-PAGE (see Supplementary Fig. S8) and blotted onto a polyvinylidene fluoride membrane to perform N-terminal sequencing analysis (Korea Basic Science Institute, Seoul, Korea). The stable core domain was identified to span residues 70-184 and corresponds to a PxxP motif and two EF-hand domains (see Supplementary Fig. S3).
Cloning and protein purification of CD EFhd2. Human CD EFhd2 (residues 70-184) was amplified using PCR from full-length EFhd2 (residues 1-240) and cloned into the modified pET-28a vector (Novagen) containing an N-terminal His 6 -TEV tag. The expressed recombinant protein was purified using the procedure used to purify full-length EFhd2. For seleno-L-methionine (Se-Met) incorporation, a plasmid encoding the CD EFhd2 was transformed into the methionine-auxotrophic E. coli strain B834 (DE3) (Novagen). Colonies were inoculated into LB medium containing 50 μ g/mL kanamycin and incubated at 37 °C with shaking for ~24 h, and then cells were harvested by centrifugation at 4,000 g for 20 min and 4 °C. The cell pellet was resuspended in minimal medium to wash and remove the LB medium, and washed cells were harvested by centrifugation at 4,000 g for 20 min and 4 °C. After washing, the cell pellet was transferred to a fresh 2 L culture of minimal medium (M9 media) supplemented with 25 mg/mL Se-Met, 2% glucose, 0.1 M magnesium sulfate and amino acids, and grown at 37 °C. Protein expression by the cells was induced by the addition of IPTG to a final concentration of 0.5 mM. After 24 h incubation at 37 °C, the cells were harvested by centrifugation at 4,000 g for 20 min and 4 °C. The overall purification procedure of the Se-Met substituted CD EFhd2 was the same as the native CD EFhd2 protein purification procedure. The purified protein was concentrated using an Amicon Ultra-15 30 K (Millipore) and stored in a deep freezer. During purification, the presence of EFhd2 was confirmed by SDS-PAGE.
Cloning and purification of EFhd2 mutants. To investigate the structural properties of Ca 2+ binding, we have mutated one acidic residue (E116A for EF1, CD EFhd2 EF1 ; E152A for EF2, CD EFhd2 EF2 ) of each EF-hand domain of CD EFhd2 to abolish the Ca 2+ binding ability (see Supplementary Fig. S3). CD EFhd2 mutants were accomplished by PCR and site-directed mutagenesis using the CD EFhd2 cDNA. All mutants were cloned into a modified pET28a vector (Novagen) containing an N-terminal His 6 -TEV tag. The overall purification procedure of the CD EFhd2 mutants was the same as that used for purifying native CD EFhd2. To investigate the Ca 2+ -binding affinity or Ca 2+ -dependent actin-binding and -bundling activity, we have mutated one acidic residue of each EF-hand domain of full-length EFhd2 (E116A for EF1, EFhd2 EF1 ; E152A for EF2, EFhd2 EF2 ) (see Supplementary Fig. S3). Point mutations (EFhd2 EF1 , EFhd2 EF2 ) were accomplished by PCR and site-directed mutagenesis using the full-length EFhd2 cDNA. All mutants were cloned into the modified pET21a vector (Novagen) containing an N-terminal His 6 tag. The overall purification procedure of full-length EFhd2 mutants was the same as that used to purify native full-length EFhd2. CD EFhd2 EF1 and CD EFhd2 EF2 were observed in 0.1 M Tris-HCl (pH 8.5) and 32% (w/v) PEG 2000. All CD EFhd2 crystals were cryoprotected by soaking them for 10 min in mother liquor containing an additional 15% (v/v) glycerol before flash freezing in a stream of nitrogen gas at 95 K. Native and MAD data sets were collected on beamline 7A at the Pohang Accelerator Laboratory (Pohang, Korea). Raw data integration and scaling were performed with the HKL2000 62 . Both the native and Se-Met substituted CD EFhd2 were crystallised in the orthorhombic form and space group P2 1 2 1 2 1 and cell dimensions of a = 37.3, b = 50.7, c = 53.4 Å. A native data set of 1.85 Å resolution was collected and the MAD dataset of Se-Met substituted protein crystals were collected to 2.10 Å. The crystal contains one molecule in an asymmetric unit with a calculated Matthews coefficient of 1.99 Å 3 /Da and an estimated solvent content of 38.6% 63 . Four out of the expected six Se sites in the asymmetric unit were found using the program SOLVE 64 using 2.10 Å resolution data yielding phases with a figure of merit of 0.51. Refinement was performed with PHENIX 65 and manual rebuilding was performed using the COOT program 63 . Cycles of group and individual B-factor refinement were performed with PHENIX 65 . In the last step of the refinement, 117 water and two Ca 2+ ions were added. A final crystallographic R-value of 16.7% (R free = 20.2%) was obtained. The N-terminus residues from 70 to 79 of the PxxP motif were poorly defined in the electron density maps owing to disorder in the crystal lattice. Therefore, we could observe the structure of the predicted PxxP motif (residues 80-90) at the N-terminus, two EF-hand domains (residues 91-163) and the connecting short LM α -helix (residues 170-177) region at the C-terminus (Fig. 1). CD EFhd2 EF1 and CD EFhd2 EF2 datasets were collected at beamline 5C at the Pohang Accelerator Laboratory to 1.95 Å and 1.94 Å, respectively. Both the CD EFhd2 EF1 and CD EFhd2 EF2 structures had the space group P2 1 2 1 2 1 and cell dimensions of a = 36.3, b = 51.5, c = 53.6 Å and a = 35.6, b = 52.1, c = 55.3 Å, respectively. Raw data integration and scaling were performed with HKL2000 62 . The Matthews coefficient for CD EFhd2 EF1 and CD EFhd2 EF2 was calculated as 1.84 and 1.88 Å 3 /Da, respectively, which corresponds to a solvent content of 33.0 and 34.7% assuming one molecule in the asymmetric unit 63 . Initial automatic model building was performed with AutoMR. The model was then refined in cyclic rounds of manual model building in COOT with refinement using PHENIX 65,66 . Refinement of CD EFhd2 EF1 and CD EFhd2 EF2 was performed using PHENIX to R work = 18.2% and R free = 20.7%, and R work = 17.9% and R free = 20.7%, respectively. All structures of the CD EFhd2 mutants were solved by molecular replacement using the refined native CD EFhd2 structure and molecular graphics were created using PyMol 67 . The refinement statistics are given in Table 1.

NMR Spectroscopy.
For NMR experiments, the CD EFhd2-expressing cells were grown in M9 medium containing 15 N ammonium chloride and 13 C glucose as the sole nitrogen and carbon sources, respectively. The overall purification procedure followed the approach used to purify the native protein. To remove pre-bound Ca 2+ , proteins were treated with 25-fold excess EGTA and then dialysed extensively against buffer with or without CaCl 2 . During the purification process, 5 mM 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonate was added to the buffer to retain protein stability. The Ca 2+ -dependent structural changes to CD EFhd2 were monitored by resonance perturbations in the two-dimensional (2D) 1 H-15 N HSQC spectra. NMR data were recorded on a Bruker Avance 800 spectrometer at 25 °C. Data were processed with NMRPipe 68 and analysed with SPARKY program (Goddard TD and Kneller DG, SPARKY 3, University of California, San Francisco). The magnitude of the 1 H-15 N chemical shift differences (∆ δ , ppm) were calculated using the equation: ∆ δ = {(δ H 2 ) + 0.2x(δ N 2 )} 1/2 , where δ H and δ N are changes to the proton ( 1 H) and nitrogen ( 15 N) chemical shift perturbation, respectively. CSPs for peaks that disappeared upon addition of Ca 2+ are set to 1 ppm. We considered CSP to be significant if ∆ δ ≥ 0.2 ppm.
Measurement of Ca 2+ -binding affinity using ITC. Since the full-length EFhd2 (residues 1-240) was more stable than CD EFhd2 in the absence of Ca 2+ , full-length EFhd2 mutants (EFhd2 E116A , EFhd2 E152A ) were used to measure Ca 2+ -binding affinities of EFhd2. Protein samples were treated initially with 25-fold excess EGTA and EDTA for > 20 h at 4 °C to remove pre-bound metal ions. We dialysed extensively against buffer (50 mM Tris-HCl (pH 8.5) and 20 mM NaCl) for 48 h at 4 °C and changing the buffer every 12 h. To measure the residual Ca 2+ concentration after the dialysis step, we used quantitative fluorescence measurement using the Ca 2+ -indicator fura-2 (non acetoxymethyl ester (AM) form, Molecular Probes, Eugene, OR) (see Supplementary Fig. S9). For determining the intensity of fura-2 at various Ca 2+ concentrations, we prepared standard solutions refer to the method of Kong et al. 69 . After the dialysis process, EFhd2 mutants (EFhd2 EF1 , EFhd2 EF2 ) (5 μ M) were mixed with 10 μ M fura-2. Fluorescence spectra of standard solutions and EFhd2 mutants were collected using a FlexStation Ш (Molecular Devices) at room temperature (excitation wavelength: 280 nm to 460 nm, emission wavelength: 510 nm, slit: 4 nm). The residual Ca 2+ concentration used for the ITC measurement was around the 1 nM range, indicating that our dialysis process was sufficient to measure Ca 2+ binding affinity using the ITC experiment. The protein sample (70 and 150 μ M for EFhd2 EF1 and EFhd2 EF2 , respectively) was titrated with 30 injections of ligand (10 μ L) in a VP-ITC calorimeter (MicroCal). The ligand solution (0.6 and 1.2 mM Ca 2+ for EFhd2 EF1 and EFhd2 EF2 , respectively) was prepared in the same buffer. All measurements were conducted at 25 °C, and binding isotherms were analysed using Origin software supplied with the calorimeter. Protein stability measured using a heat aggregation assay. To measure the stability of EFhd2 in a Ca 2+ -dependent manner, the half aggregation temperature of native CD EFhd2, CD EFhd2 EF1 , and CD EFhd2 EF2 were determined spectrophotometrically. The protein solution contained 5 mM Tris-HCl (pH 8.0), 1 mM CaCl 2 and 250 μ M of protein in a final volume of 4.0 mL. The temperature was increased at the rate of 4 °C per 90 s. Turbidity was monitored by the absorption at 470 nm and room temperature using an ultraviolet-visible spectrometer (Ultrospec 2000; Pharmacia Biotech).
Ensemble refinement of CD EFhd2, CD EFhd2 EF1 and CD EFhd2 EF2 . To evaluate structural dynamics at the atomic level, we performed ensemble refinement using the Phenix.ensemble refinement 31 for CD EFhd2, CD EFhd2 EF1 and CD EFhd2 EF2 . Harmonic restraints were applied for all amino acids with visible electron density at a level of 1σ in the 2mFo-DFc electron density map using parameters slack = 1.0 and weight = 0.001.

Protein preparation and the crosslinking experiment of cysteine mutants within the EFhd2 coiled-coil domain (residues 199-240).
To determine whether the C-terminal coiled-coil domain formed a dimer by parallel or antiparallel coiled-coil interaction, we designed recombinant fragments of the coiled-coil domain (residues 199-240) with Cys mutations at the N-terminus (CC1) or C-terminus (CC2) of the coiled-coil domain. The sequence for CC1 starts with CysGlyGly at the N-terminus, whereas for CC2 the C-terminus ends with GlyGlyCys. CC1 and CC2 clones were PCR amplified from the cDNA of the coiled-coil domain of EFhd2. CC1 was subcloned into pGEX-4T-1 and the glutathione S-transferase (GST) tag at the N-terminus was removed by thrombin treatment during the purification process. In addition, CC2 was subcloned into a modified pET-21a vector (Novagen) containing an N-terminal His 6 tag. The purification procedure was the same as those used for the other EFhd2 proteins. CC1 and CC2 proteins were analysed by SDS-PAGE under reducing and non-reducing conditions to identify the disulfide bond between Cys residues that mediate dimerisation.
In vitro actin-binding and -bundling assay. F-Actin binding (co-sedimentation) and bundling assays were performed as reported 26 . In brief, non-muscle actin derived from human platelets was purchased from Cytoskeleton Inc. (Denver, CO, USA). Actin was mixed in G-buffer (5 mM Tris-HCl, pH 8.0 and 0.5 mM CaCl 2 ) to produce an actin stock solution and polymerised in actin polymerisation buffer (0.2 mM Tris-HCl, pH 8.0, 100 mM KCl, 2 mM MgCl 2 and 0.5 mM ATP) at room temperature for 1 h and then incubated with EFhd2 or its mutants from 5 min to 1 h at room temperature. Actin filaments with bound proteins were pelleted by centrifugation at 100,000 g for 2 h at room temperature (for the F-actin binding assay) or 15,000 g for 10 min at room temperature (for the F-actin bundling assay). BSA and α -actinin were used as a negative and positive control, respectively. Equal amounts of pellet and supernatant were resolved by SDS-PAGE and proteins were visualised by Coomassie Blue staining. The percentage of actin in the supernatant (S) and pellet (P) was quantified by densitometry using ImageJ 1.44p.