A Non-Canonical Interaction between Calmodulin and Calcineurin Contributes to the Differential Regulation of Plant-Derived Calmodulins on Calcineurin

Abstract

Calmodulin (CaM) serves as an important Ca²⁺ signaling hub that regulates many protein signaling pathways. Recently, it was demonstrated that plant CaM homologs can regulate mammalian targets, often in a manner that opposes the impact of the mammalian CaM. However, the molecular basis of how CaM homologs mutations differentially impact target activation is unclear. To understand these mechanisms, we examined two CaM isoforms found in soybean plants that differentially regulate a mammalian target, calcineurin (CaN). These CaM isoforms, sCaM-1 and sCaM-4 share >90% and ∼78% identity with the mammalian CaM (mCaM), respectively, and activate CaN with comparable or reduced activity relative to mCaM. We used molecular dynamics (MD) simulations and fluorometric assays of CaN-dependent dephosphorylation of MUF-P to probe whether calcium and protein-protein binding interactions are altered by plant CaMs relative to mCaM as a basis for differential CaN regulation. In the presence of CaN, we found that the two sCaMs’ Ca²⁺-binding properties, such as their predicted coordination of Ca²⁺ and experimentally measured EC₅₀ [Ca²⁺] values are comparable to mCaM. Furthermore, the binding of CaM to the CaM binding region (CaMBR) in CaN is comparable among the three CaMs, as evidenced by MD-predicted binding energies and experimentally measured EC₅₀ [CaM] values. However, mCaM and sCaM-1 exhibited binding with a secondary region of CaN’s regulatory domain that is weakened for sCaM-4. We speculate that this secondary interaction affects the turnover rate (k_cat) of CaN based on our modeling of enzyme activity, which is consistent with our experimental data. Together, our data describe how plant-derived CaM variants alter CaN activity through enlisting interactions other than those directly influencing Ca²⁺-binding and canonical CaMBR binding, which may additionally play a role in the differential regulation of other mammalian targets.

Graphical Abstract

1. Introduction

Calmodulin (CaM) is a 16.7 kDa Ca²⁺ sensor ubiquitously expressed in eukaryotic cells [1] and has an invariant sequence in vertebrates [2]. CaM regulates at least 300 targets in cellular processes, including muscle contraction, proliferation, and immune responses [3, 4]. CaM is also essential to plants and their responses to environmental stimuli [5]. In contrast to vertebrates, plants have different CaM isoforms [6, 7]. Two soybean CaM isoforms, sCaM-1 and sCaM-4, share ∼90% and ∼78% sequence identity with mCaM, respectively, and have been extensively studied [8–10]. Interestingly, these two plant CaM variants show differential or even reciprocal regulation of human protein targets [11, 12]. As an example, calcineurin (CaN) is activated by sCaM-1 with similar kinetics to the human CaM, while sCaM-4 has significantly reduced activity [11]. CaN is a ubiquitously expressed serine/threonine phosphatase in all human tissues [13] and CaM enhances its activity in a Ca²⁺-dependent manner [14]. The positive correlation between sequence similarity and enhancement of CaN activity of these soybean CaMs with mCaM renders CaN as a model system to study the effects of CaM sequence modification on target activation.

CaM mutations found in humans have been reported to disturb Ca²⁺-handling in cardiac cells [15]. Given the rich variety of CaM targets, a complete description of the underlying molecular basis remains elusive [16]. Several mechanisms may explain altered functions in CaM variants [17]. CaM’s Ca²⁺ binding thermodynamics undoubtedly plays an important role in its target activation. It has been widely accepted that Ca²⁺ binding promotes CaM’s N/C domains to expose a hydrophobic patch for target binding [18–20]. Although apo-CaM can bind targets [21], the presence of Ca²⁺ facilitates the formation of native complex [22] for many targets, including CaN. In addition to Ca²⁺-binding, CaM’s interaction with CaM binding region (CaMBR) targets is an also important consideration. A single amino acid mutation in the CaMBR region, for instance, can abolish CaM binding to inositol trisphosphate receptors [23]. Lastly, secondary interactions between CaM and target regions beyond the CaMBR can affect target regulation as well. For example, the ryanodine receptor (RYR) isoforms are differentially regulated by CaM, albeit they share the same CaMBR sequence [24], suggesting that isoform-dependent region beyond the CaMBR may contribute to regulation. Another example is the activation of CaN by CaM, in which a distal helix region C-terminal to the CaMBR interacts with CaM to fully remove the autoinhibitory domain of CaN [25]. Experimental and computational studies [25–29] have suggested a structural model of Calmodulin (CaM)-dependent CaN activation (Fig. 1) [30] to rationalize this mechanism. In this structural model, Ca²⁺-saturated CaM adopts two binding interactions with CaN to fully activate the phosphatase: the first interaction includes the binding of the CaMBR (A391-R414) in a canonical binding mode characterized by CaM wrapping around the alpha-helical CaMBR, while the latter entails binding the CaN ”distal helix” (DH, K441-K466) on the CaM solvent-exposed surface. These mechanisms could be utilized by CaM variants to alter or tune a target protein’s response to changes in the intracellular Ca²⁺.

Figure 1: CaM-dependent CaN activation.

(a) Proposed CaN activation model by CaM [30]. Step 1, Ca²⁺-saturated CaM binds to the CaM binding region (CaMBR) and the distal helix (DH) motif of CaN, which displaces the autoinhibitory domain (AID) from its catalytic site. Step 2, Entry of substrate to CaN’s catalytic site enables dephosphorylation of the substrate. (b) Two soybean CaM isoforms sCaM-1 and sCaM-4 share ∼90% and ∼78% sequence identity with mammalian CaM (mCaM), respectively. The unique residues are colored green and red for sCaM-1 and sCaM-4, respectively. A detailed sequence comparison is in Fig. S1. (c) The three molecular interactions examined in this study to determine the basis of the CaM’s differential regulation of CaN activity.

Here, we use CaN/CaM as a model system to explore the molecular basis of CaM target activation with different CaM variants. We investigated the potential structural mechanisms for how two sCaMs differentially regulate human CaN. We hypothesized that soybean CaM variants could modulate one of the following three mechanisms to alter mammalian CaN activity: 1) Ca²⁺ binding properties of CaM, 2) CaM binding to CaMBR of CaN, and/or 3) region of the CaN regulatory domain distal to the CaMBR, the ‘distal helix’ motif. Our simulations contain overall two consecutive stages to sample the interactions of CaM with the CaMBR region and the DH motif of CaN. The CaM/CaMBR interaction patterns are either from the crystal structure (for mCaM) or from Rosetta modeling (for sCaM). After obtaining the CaM/CaMBR poses, we docked the DH motif onto the CaM/CaMBR complex. After performing extensive MD simulations, we found that the two sCaMs maintain comparable Ca²⁺ and CaMBR binding abilities to mCaM. However, the distal helix/CaM interaction patterns are different. This implicates the distal helix interaction as the cause of the different activation capabilities. We further pinpointed a hot-spot region on CaM that contributes most to the distal helix interaction difference between the two sCaMs and mCaM, This region lies in between residues N60-E87. Lastly, using a revised activation kinetic model, we found CaM activates CaN by converting it to an activated form that exhibits altered kinetic properties. Altogether, our computational and experimental results show that DH interactions are central to the differently-enhanced CaN activities by these CaM variants.

2. Materials and methods

2.1. Calcineurin phosphatase assay using MUF-P substrate

The mCaM, sCaM-1, and sCaM-4 were overexpressed in E. coli and purified as described in [31] using phenyl sepharose. CaN was overexpressed from a pETagHisCN plasmid using E. coli BL21 (DE3) CodonPlus RIL cells. Cells were lysed by sonication, and CaN was purified using Ni-NTA and CaM-sepharose. Protein purity (>95%) was determined by Coomassie staining and SDS denaturing gel electrophoresis. Purified CaN fractions were pooled and dialyzed into 10 mM MOPS, 150 mM KNL, 1 mM TCEP at pH7.0. CaN and CaM concentrations were determined by UV-VIS spectroscopy. Experimental sample conditions were 2 mL of 150 nM CaN in an assay buffer comprised of 200 mM MOPS, 150 mM KCl, 3 mM MgCl₂, 2 mM EGTA at pH 7.0. The MUF-P concentrations were set to 0, 100, 200, 400, and 600 μM for substrate titrations of CaN and held at 100 μM for EC₅₀ [Ca²⁺] and EC₅₀ [CaM] experiments. Assay buffer EGTA concentration was calibrated based on a dose-response of CaM tyrosine fluorescence (corresponding to CaM-C domain) against Ca²⁺ [32]. PerkinElmer spectrometer parameters were as follows: excitation 365 nm, emission 445 nm, data sampling interval 0.2 second, temperature = 20 °C. CaM titrations ranged from 5 nM to 3 μM. Ca²⁺ titrations ranged from pCa9 to pCa4. A minimum of 3 min of raw fluorescence data were collected for each titration point and fit to a linear line using KaleidaGraph. The slope values were normalized and plotted against the titrant (CaM or Ca²⁺). Data was fit to Hill Equation using the CURVE_FIT function from scipy. The resulting EC₅₀ values were plotted using the matplotlib. Statistical significance was determined (95% confidence, p< 0.05) via student’s t-test.

2.2. Kinetic Analyses

To investigate the kinetics of CaM-dependent CaN activation, we proposed a model comprising two reaction pathways (Fig. 2). Since Michaelis-Menten kinetic schemes classically do not contain a regulatory mechanism, we needed to add a CaM regulatory pathway to derive the experimentally-determined parameters. The first pathway is independent of CaM and follows a canonical Michaelis-Menten relationship in which the substrate, S (MUF-P), binds Calcineurin (CaN) to form the enzyme/substrate complex, CaN·S; after this, the substrate is catalyzed to form the dephosphorylated product, P. The second pathway entails CaM binding to CaN to form the CaM·CaN complex, after which the complex undergoes a typical Michaelis-Menten mechanism of substrate binding and catalysis. In this model, we first determine K_m and k_cat under CaM-free conditions. This was done by determining K_m and V_max using the Michaelis-Menten relationship

$$\frac{1}{v}=\frac{K_m}{V_{max}}[S]+\frac{1}{V_{max}}$$ (1)

Figure 2: CaN activation scheme.

A proposed kinetic model that describes CaM-independent and CaM-dependent CaN activation. The model accounts for the multiple molecular interactions involved in CaN’s enzymatic activity after being activated by CaM (see Fig. 1c). This model was used to fit the experimental CaN activity data

For the CaM-dependent pathway, we assumed that mCaM and the soybean isoforms have the same K_a (0.038 ± 0.005 μM), based on experimental data collected in this study. The reciprocal form of the revised model (Eq. 2) was fit to the CaM and sCaM data sets using K_m and V_max from the CaM-free fit and K_a, as described above. For this fit, the free parameters α and β were used to rescale the CaM-free parameters K_m and V_max to the CaM-dependent substrate reaction. Here we used

$$\frac{1}{v}=\frac{1}{\frac{V_{max}S\left(\frac{1}{K_m}+\frac{\alpha A}{\beta K_a{K}_m}\right)}{1+\frac{S}{K_m}+\frac{A}{K_a}+\frac{AS}{\beta K_a{K}_m}}}$$ (2)

where A is CaM, K_a is the binding constant of CaM to CaN, α scales V_max, and β scales K_m.

2.3. Rosetta Comparative Modeling of sCaM-1 and sCaM-4 structures

The structural data for sCaM-1 and sCaM-4 in the PDB databank are for isolated N- or C- domain structures [12] (PDB ID: 2RO8/2RO9 for sCaM-1 and 2ROA/2ROB for sCaM-4, respectively). There is one complete sCaM-4 structure in complex with vacuolar calcium ATPase BCA1 peptide deposited with PDB ID 2L1W [8], but the binding mode of the target peptide is significantly different relative to the mammalian CaM/CaMBR structure. Namely, for the sCaM structure, the N- and C-domain binds to the two ends of the peptide, as opposed to the conventional ‘wrap round’ manner in which N-/C- domains collapse around the middle of the target peptide. Therefore, we built complete sCaM-1 and sCaM-4 structures that bind CaMBR in the same pose as the mammalian CaM. The complete structures of sCaM-1 and sCaM-4 were built using the Rosetta comparative modeling approach [33], which has been suggested to have high predictive accuracy when the sequence identity is at least 15%. Briefly, the protocol has four stages: 1) assembling aligned fragments to obtain the overall topology; 2) fusing the fragments with a loop-closure sampling method; 3) resampling the side-chain conformations; and 4) all-atom optimization and energy-based scoring. For each model, the mammalian CaM/CaMBR complex crystal structure (PDB ID 4Q5U) as well as the N-/C- domain structures were used as templates (the weights of each template are 1.0, 0.8, and 0.64, which are the default values in the RosettaCM hybridize mover). The HYBRIDIZE function of Rosetta was used and the specific protocol and flags are given in Sect. S2.1. For each soybean CaM, 10 structures were modeled (Fig. S2) and the 2 highest-scored (lowest energy) structures were very similar, with ∼1 Å RMSD. The highest-ranked structure was subject to 3 × 2 μs all-atom MD simulation refinement.

2.4. All-atom MD simulations to refine the modeled sCaM structures

All-atom MD simulations were performed to refine the Rosetta modeled soybean CaM structures. The ff14SB [34] force field was used for the protein atoms. The system was solvated in a TIP3P [35] waterbox with the distance between protein to waterbox wall set to 14 Å. K⁺ and Cl⁻ ions were added to maintain a 0.15 M salt concentration that is consistent with a typical cytosolic environment, in which potassium is the predominant cation. This KCl concentration is also consistent with the experiments complementing these simulations. The system was subjected to energy minimization, for which all atoms except hydrogens, water, and KCl ions were constrained by the IBELLY functionality. The cutoff value for non-bonded interactions was set to 10 Å. A 2 fs time-step was chosen, as SHAKE [36] constraints were applied to bonds involving hydrogen atoms. Two heating procedures were performed to heat the system from 0 to 300 K using the Amber18 SANDER.MPI engine [37]. In the first heating stage, the ibelly function was used to keep all atoms except the water and KCl ions fixed. The water box was heated to 300 K over a 100 ps interval under the NVT ensemble. For the second heating stage, the entire system was heated from 0 to 300 K over 500 ps under the NPT ensemble, for which the backbone atoms were constrained by a harmonic potential (force constants of 3 kcal mol⁻¹ Å⁻²). Thereafter, an additional 1 ns equilibrium stage was conducted at 300 K under the same constraints, but with a reduced force constant of 1 kcal mol⁻¹ Å⁻². Langevin thermostat was used during the simulation. These equilibrium simulations were followed by triplicate 2 μs production-level MD simulations. The simulation length to refine the Rosetta modeled structures is based on the reports that the timescale of conformational fluctuation in the MD refinement of modeled structure is approximately 1 μs for systems with a mean residue number around 120 [38], which is approximately the length of CaM. The MD refinement simulations for mCaM were performed following the same protocol, except that the starting structure was based on the crystal structure PDB 4Q5U [27].

2.5. Docking distal helix structure to CaMs

2.5.1. Replica exchange molecular dynamics (REMD) simulations of isolated distal helix region

REMD was performed on the distal helix fragment (residues K441-K466 of CaN, amino acid sequence is shown in Fig. S7) to sample representative conformations of the isolated DH. 12 replicas of the system was constructed and the temperature range of the REMD was 270 to 454.72 K, which was calculated based on the protocol proposed in [39]. The temperature range led to an exchange probability of 0.4, which ensures sufficient exchange between replicas [40]. The system was parameterized with ff99SBildn force field [41] coupled with the generalized Born implicit solvent model. The detailed REMD simulation procedure has been described in [30]. The production run of the REMD was 100 ns and clustering analysis was performed on the trajectory from 297 K. Representative structure of the most populated cluster was selected for docking to CaM structures. The selected DH conformation is shown in Fig. S7.

2.5.2. Docking DH to CaM surface and extensive MD refinement

The representative structure of the DH generated by the REMD simulations was docked to the surface of sCaM-1, sCaM-4, and mCaM via the ZDOCK webserver [42]. The CaM structures that were used for the docking are from the MD refinements of the modeled structures. Specifically, clustering analysis was first performed on the MD trajectories and the representative structures of the most populated cluster were selected. For mCaM, one representative structure was chosen and for the two sCaMs, two representative structures were chosen for each case. The procedure of clustering analysis and identifying the two representative structures is detailed in Sect. S2.3. After preparing the DH and CaM structures, the DH was docked to CaM following the docking procedure in [30]. Briefly, four putative binding sites were defined on CaMs that represent the grooves formed by helix bundles, as these regions are more likely to accommodate helix-helix interactions [43]. The selected four putative sites covered most of the CaM’s solvent-accessible surface. The residues in Table S2 from each site on CaM were specified as contact residues during docking. The docking protocol is illustrated in Fig. 3 and extra details are reported in [30].

Figure 3: CaM/DH docking strategy.

Scheme for docking the REMD-sampled distal helix (DH) conformation to CaM’s solvent-exposed surface via the ZDOCK webserver [42]. For each CaM structure, four putative binding sites on CaM that represent interhelical grooves were specified as contact sites for the docking. The putative site residues of CaMs are listed in Table S2. For each site, the highest-scored pose was subject to 3 × 1.2 μs all-atom MD refinement.

The highest-scored CaM/DH interaction pose at each site was selected for further MD refinement with triplicate 1.2 μs runs, using the MD protocol described in our previous study [30]. 72 μs refinement MD simulations were performed in total with 28.8 μs for each sCaM-1 and sCaM-4 and 14.4 μs for mammalian CaM. All molecular dynamics simulations were performed via Amber16 package [44] and are summarized in Table S1.

2.6. Analyses Methods

Clustering analysis, root mean squared deviations (RMSD)/root mean squared fluctuations (RMSF) calculations, hydrogen bonds, and secondary structure analysis were performed via CPPTRAJ [45]. The reference structure used for these analyses was the CaM/CaMBR crystal structure (PDB ID: 4Q5U [27]). The secondary structure for each residue was calculated using CPPTRAJ with the Define Secondary Structure of Proteins (DSSP) algorithm [46].

2.6.1. Projection of MD sampled conformations onto 2D plane using the sketch-map dimensionality reduction method

Sketch-map (https://github.com/cosmo-epfl/sketchmap) is a non-linear dimensionality reduction method that we used to visualize the CaM/DH complex conformation space sampled in the MD simulations [47, 48]. Each frame from the MD trajectories was represented by a point in the high-dimensional space $\overrightarrow{X_D}=\left(D_1,D_2,\mathrm{..}{D}_N\right)$ and D₁, D₂...D_N are the minimum distances between C_α atoms of CaMBR and the distal helix. In the present study, N was set as 32: D₁ to D₁₄ represents the minimum distance of the 14 residues in CaMBR (E394-A407) to the distal helix. Namely, for residue i, we first calculated its distance to each residue in the distal helix and assigned the smallest value to D_i. D₁₅ to D₃₂ are the reverse minimum distances. Namely, the minimum distance of each residue in the distal helix (S438-D455) to CaMBR residues. This definition of high-dimensional points ensures that the relative conformation information between distal helix to CaM/CaMBR is encoded in the coordinates of high-dimensional points, as illustrated in [49]. These high-dimensional points were projected onto a 2D plane as described in [49] to generate the sketch-map representation of the MD sampling.

3. Results and Discussion

3.1. Experimental assays revealed that sCaM-1 and mCaM exhibited stronger activation of CaN than sCaM-4.

It was previously demonstrated that sCaM-1 activates CaN to a greater extent compared to sCaM-4, using activity measurements of CaM, 100 μM MUF-P, and bovine brain CaN [11]. These results were anticipated based on sCaM-1 having higher sequence similarity with mCaM relative to sCaM-4. To investigate the kinetic mechanisms by which the CaM variants differentially activate CaN, we performed pMUF substrate titrations of CaN alone as well as CaN with different CaM isoforms. As shown in Fig. 4, the presence of CaM (regardless of the isoform relative to isolated CaN) significantly increased the activity of CaN. Consistent with our previous results, sCaM-1 enhanced CaN’s activity to a similar extent as mCaM, while sCaM-4 was about 2-fold less active.

Figure 4: Experimental CaN activity.

CaN enzyme kinetics as measured by the fluorescence of the dephosphorylated MUF-P with saturating Ca²⁺ (100 μM) and CaM (1 μM). Substrate titrations were done in absence of CaM as well as CaN with different CaM isoforms.

3.2. The different activation effects between sCaM-4 and mCaM/sCaM-1 are not caused by their Ca²⁺ binding properties and their CaMBR binding properties

We next sought to determine the molecular mechanisms behind these experimental trends via computational modeling, which included homology modeling, protein-protein docking and molecular dynamics simulations. As shown in Fig. 1a, our assumed model for CaM/CaN activation includes: 1) Ca²⁺ binding to isolated CaM; 2) binding of Ca²⁺-saturated CaM to the CaM binding region (CaMBR) of CaN; and 3) the interaction of the distal helix (DH) motif from CaN to CaM. We investigated theses interactions for the mCaM and sCaM isoforms in the activation of CaN.

Because the complete sCaM-1/4 structures complexed with CaMBR of CaN are not yet available, we performed homology modeling followed by extensive all-atom MD refinement to obtain structural models. We show in Sect. S2.2 that the modeled sCaM/CaMBR complex structures exhibited converged RMSD values over the 2 μs MD simulations, low fluctuations of backbone atoms (<2 Å RMSF) and a more compact structure of sCaM-4 that was consistent with experimental observations (Fig. S3). As an additional confirmation of the MD refined structures’ stability, we evaluated the RMSD, RMSF and radius gyration using the final frame of the simulations as a reference structure and found similar trends as before (Fig. S4). Altogether, these analyses suggest that our MD refined structures have converged to a local steady state. Based on the modeled structures, we examined key molecular interactions that may contribute to the sCaM isoforms’ differential activation of mammalian CaN.

3.2.1. Ca²⁺ binding

Ca²⁺ plays a vital role in CaM’s target regulation [50]. In this section we examined if an altered Ca²⁺ sensitivity might explain the differences in activation among mCaM and the sCaM isofroms. CaM has four EF hands, two in each N-/C- domain. MD simulations indicated that the EF-hands in the modeled sCaM structures were stable as evidenced by their convergence to ∼ 2 Å RMSF values (Fig. S3). Not surprisingly, mCaM had the smallest RMSF of ∼1 Å at the EF-hands, while sCaM-1 had the largest RMSF among the three CaMs.

Nonetheless, sCaM-1 maintained the same Ca²⁺ coordination number as the other two CaMs, that is, each structure bound Ca²⁺ with the same number of protein oxygens. Fig. S5(a,b) reports the integrated radial distribution functions (IRDF) of amino acid oxygens and water oxygens around Ca²⁺. Previously we demonstrated that EF-hand Ca²⁺ affinities tend to correlate with oxygen IRDFs [51, 52]. For all CaM isoforms, the IRDF curves reached a plateau at ∼2.3 Å, suggesting that the average distance between Ca²⁺ and the coordinating oxygens are indistinguishable across all cases. Interestingly, as shown in Fig. 5a, the number of amino acid oxygens and water oxygens varied somewhat: sCaM-1 had the lowest amino acid oxygen number around 5.5 for Ca²⁺ at the first EF-hand of N-domain and second EF-hand of the C-domain but was compensated for by the highest water oxygen numbers (around 2). In addition, the Ca²⁺s bound at the C-domain of the two sCaMs had more water oxygens compared with mCaM. Nevertheless, the total Ca²⁺ coordination numbers were comparable, as all were in the range of seven to eight oxygens (Fig. 5a), which is in agreement with the Ca²⁺ coordination number measured from other Ca²⁺-binding proteins [30, 51, 53–55] and in aqueous solution [56]. The one oxygen variation should have negligible impact on Ca²⁺ binding free energy in the EF-hand site. This is supported by a previous study demonstrating that a difference of ±1 oxygen does not appreciably impact Ca²⁺ binding free energy in the EF-hand sites [52]. Since changes in Ca²⁺ affinity can enlist conformational changes outside of the EF-hands [52, 57], we experimentally measured the EC₅₀ [Ca²⁺] of mCaM and sCaM-4 in activating CaN (Fig. 6(b–c)). This was obtained by measuring CaN activity at varying Ca²⁺ concentrations with saturating CaM. The mean values of the EC₅₀ for mCaM and sCaM-4, although not identical, were not significantly different. These data together demonstrate that the three CaM isoforms maintain comparable Ca²⁺ binding properties, which suggests the differential CaN activation by the CaM variants is unlikely due to altered Ca²⁺ binding.

Figure 5: Ca²⁺ binding properties.

(a) The predicted total number of coordinating oxygens (amino acid oxygen from CaM plus water oxygen) around Ca²⁺ as determined from MD simulations. The two Ca²⁺s at the N-domain EF-hands were indexed as ”Ca1” and ”Ca2”, and the Ca²⁺s at the C-domain were indexed as ”Ca3” and ”Ca4”. The horizontal lines within each bar indicate the number of amino acid oxygens. (b) Experimentally determined dependence of CaN activity on Ca²⁺ under saturating CaM conditions. (c) Experimental EC₅₀ [Ca²⁺] for mCaM and sCaM-4. n.s. p > 0.05.

Figure 6: CaMBR binding properties.

(a) Normalized contact probability based on the residue pair. These probabilities are projected with the greatest number of contacts predicted by MD simulations for the CaMBR of each CaM structure on the mCaM/CaMBR complex structure PDB 4Q5U. The residues numbers comprising the helices are labeled. (b) MM-GBSA calculated binding free energy between CaMBR and CaM. The values are reported as the mean ± standard error of mean. The ΔG difference between the three CaMs were not significant as determined by student’s t-test. (c) Experimentally determined dependence of CaN activity on CaM under saturating Ca²⁺ conditions. (d) Experimental EC₅₀ [CaM] for mCaM and sCaM-4.

3.2.2. CaMBR binding

We next examined if the binding between the CaMBR and CaM differed for the three CaM isoforms. We report in Fig. 6a the contacts between CaMBR and CaM. Overall, the CaMBR/CaM interaction pattern shown for mCaM was largely maintained by the two sCaMs, especially the contacts by the CaM helices formed by residues Q8-S17, T43-N53, K77-R90. In the simulations for the CaMs, the CaMBR exhibited small fluctuations with RMSF < 2 Å (Fig. S3b), which suggests that the CaMBR was highly stable when bound to three CaMs. The largest difference was observed in the C-domain: both sCaMs exhibited contacts with the CaMBR via two helices through residues S101-M109 and D118-I125, while mCaM had negligible contacts with CaMBR at these two helices. However, the MMGBSA-calculated binding free energies between the CaMBR and CaM (Fig. 6b) were comparable across the three CaMs and were highly favorable: −128.8 ± 3.3, −127.1 ± 2.0, −125.4 ± 1.9 kcal/mol for mCaM, sCaM-1 and sCaM-4, respectively. To validate these predictions, we measured the EC₅₀ [CaM] for mCaM and sCaM-4 in activating CaN. This was obtained by measuring CaN activity with varying CaM concentrations under saturating Ca²⁺. As shown in Fig. 6c, the CaN activity for mCaM and sCaM-4 were indistinguishable and the calculated EC₅₀ [CaM] for mCaM and sCaM-4 were not statistically different (∼0.038 μM, Fig. 6d). We assume that CaM and CaMBR binding affinity accounts for the majority of the affinity implied by the measured EC₅₀. Thus, the similar EC₅₀ values for mCaM and sCaM-4 suggest that these two CaMs are highly likely to have the same CaMBR binding affinities. This finidng is consistent with reports that the two sCaMs bind to the target with comparable affinities and are also comparable to other mCaM/targets binding affinities [12, 21]. In sum, the two sCaMs did not show significant difference in binding CaMBR when compared with mCaM, given that: 1) most of the contacts shown in CaMBR/mCaM binding are maintained in the two sCaMs; and 2) the MMGBSA-calculated CaMBR/CaM binding free energies are comparable and 3) the mCaM and sCaM-4 have similar EC₅₀ [CaM] values. Therefore, the different CaN activation ability of sCaMs is not likely due to altered CaMBR binding affinity.

3.3. mCaM/sCaM-1 exhibited stronger binding with the distal helix than sCaM-4

We next examined if the CaM/distal helix binding differed among the three CaMs as a potential mechanism for differential CaN activation (Fig. 1c). To analyze the CaM/DH binding models, we first performed 100 ns REMD simulations of the isolated distal helix to sample the representative distal helix conformation as initially described in [30] (see Fig. S7 for the REMD results). The representative distal helix was then docked to the CaM isoforms via protein-protein docking, followed by extensive all-atom MD simulations. To visualize the CaM/distal helix conformational space, the MD trajectories were projected onto a 2D plane via the Sketch-map method [47], using the distal helix to CaMBR distance as the coordinate basis (Fig. 7a).

Figure 7: DH binding properties.

(a) Poses of the distal helix (DH) interacting with CaM. The poses are colored based on their free energies after projection using Sketch-Map dimensionality reduction method [47]. Region with free energies less than −5.5 k_bT are used as a reference for stable DH/CaM binding. The mean and standard deviation of the region areas are obtained via bootstrapping. In sCaM-4, representative structures from the unstable regions (> −5.5 k_bT) are shown with the DH colored red. (b) Overall α-helicity of residues comprising the DH and AID of CaN from MD simulations. A bar depicts a residue that has α-helix as the dominant secondary structure. The dashed lines depict the residues whose secondary structure were determined to be bends or turns. (c) Sequence comparison of residues N60-E87 of the two sCaMs with mCaM. The region contributing to the DH/CaM contact map difference between mCaM/sCaM-1 and sCaM-4 are shown in Fig. S8. Non-conserved residues are underlined. The location of this sequence region is highlighted as cyan in the mCaM crystal structure PDB ID: 4Q5U.

The densities (ρ) of the projected points reflect the distribution of DH poses on CaM’s surface. For easier interpretation, we converted the densities to free energies using −k_bT ln(ρ/ρ_o) where ρ_o is the density corresponding to the outer-most points projected for sCaM-4. As shown in Fig. 7a, the free energy minimum of DH/CaM interaction were about 8.5 kcal/mol for all three cases, which suggested that there were thermodynamically-favorable binding sites for the distal helix for all CaM isoforms. This is consistent with our experimental data demonstrating that all CaM isoforms are capable of activating CaN. Using < −5.5 k_bT as an arbitrary cutoff to define a thermodynamically favorable area, mCaM and sCaM-1 have comparable areas (or poses) for binding the DH, while sCaM-4 sampled only 58% of the mCaM and sCaM-1 areas (Fig. 7a). The means and standard deviations of these areas were computing using the bootstrapping procedure with 500 iterations as outlined in Sect. S2.5. Together, these data suggest that while all DH/CaM variant complexes exhibited configurations with favorable binding thermodynamics, mCaM and sCaM-1 had a larger number of thermodynamically-accessible states relative to sCaM-4, which is expected to confer the former two CaMs with greater thermodynamic stability. Visual inspection indicated that the high energy conformations of DH from sCaM-4 were widely distributed about CaM, as opposed to being localized to the CaM binding surface. The larger interacting area reflected for sCaM-1 relative to sCaM-4 for the distal helix interaction is consistent with sCaM-1’s complete activation of CaN versus partial activation by sCaM-4 (see Fig. 4).

We next performed contact map analyses to identify the key CaM residues that contributed to CaM/DH interactions among the three CaMs. We found that mCaM had the strongest contacts in terms of the number of contacting residues as well as contact persistence (percent of simulation the residues were in contact) (Fig. S8). sCaM-1 had comparable contact persistence with sCaM-4 but with more pairs of contacts. One notable difference was that mCaM residues (T62, I63 and D64) and sCaM-1 resides (D64, F65, P66 and F68) had more contacts with DH than sCaM-4 (Fig. S8). These residues are located close to the linker of CaM that connects N- and C- domains. In this region, sCaM-1 has six different residues relative to mCaM while sCaM-4 has ten (Fig. 7c). The larger sequence difference of sCaM-4 relative to mCaM/sCaM-1 is likely to weaken the distal helix interaction. Hence, because sCaM-4 appears to exhibit weaker CaM/DH interaction than mCaM and sCaM-1, we attribute impaired DH binding to sCaM-4’s reduced CaN activation ability.

Higher helical content of the DH may facilitate AID displacement from the catalytic site via shortening the distance between DH and AID. We calculated this propensity and determined that the α-helical residues in DH were comparable among the three CaM (Fig. 7b). The distal helix in all cases adopted α-helix structure at the C-termini. When interacting with CaM, the number of α-helical residues were comparable for these three CaM, with 16, 12, and 15 residues for mCaM, sCaM-1 and sCaM-4, respectively. Hence, the CaMs don’t appear to impact distal helix upon activating CaN.

3.4. Kinetic fitting suggested that stronger CaM/DH led to larger apparent k_cat of CaN.

To determine how the CaM/CaN interactions might impact CaN’s kinetics, we fit the kinetic data shown in Fig. 4 to the kinetic model proposed in Fig. 2. This model includes the additional CaM/CaN interactions studied via simulations. As shown in Fig. 1a, we propose that multiple interactions such as the CaM/CaMBR interaction, the CaM/DH interaction and activated CaN/- substrate interactions are involved in the dephosphorylation process. There are therefore three assumptions in the proposed model in Fig. 2: 1) CaM binds only to the free CaN enzyme (CaN) but not the enzyme-substrate complex (CaN·S), in accordance with the mechanism of CaN activation by CaM; 2) the three CaM isoforms share the same K_a as indicated by the EC₅₀[CaM] data where mCaM and sCaM-4 showed nearly identical EC₅₀[CaM] (Fig. 6); and 3) once bound by CaM and activated, CaN exhibits modified kinetic properties described by α and β factors which rescale k_cat and K_m, respectively, relative to values for the CaM-free system.

Firstly, fitting yielded an α value much greater than 1 for all CaM cases, indicating that CaM significantly enhanced the catalysis of CaN (see Table 1). All CaM isoforms also exhibited β > 1, indicating a modified K_m. However, it should be noted that interpreting K_m as the substrate affinity is not appropriate here for the following reasons. For instance, suppose the binding of S to CaN can be described by k_on and k_off, the on and off rates of S to CaN. $K_m=\frac{k_{off}+k_{cat}}{k_{on}}$ therefore approximates $K_{d,S}=\frac{k_{off}}{k_{on}}$ only if $k_{cat}\ll k_{on},k_{off}$. In this case, a change in α is independent of a change in β. However, as our kinetic data revealed, both α and β increased, implying a possible correlation between these two parameters. Consequently, interpretation of K_m as the affinity for substrate is nuanced.

Table 1:

Kinetic parameters obtained by fitting model in Fig. 2 to the experimental data in Fig. 4. Student’s t test was used to determine the significance of the difference between mCaM and the soybean isoforms.

	no CaM
Ka (μM)	0.038 ± 0.0048 @
Vmax (RFU/s)	0.44 ± 0.16
Km (mM)	1.73 ± 0.88
	mCaM	sCaM-1	sCaM-4
α	19.43 ± 6.39	39.69 ± 13.25 *	7.95 ± 1.24 *
β	2.61 ± 0.85	4.70 ± 1.23 *	2.20 ± 0.42
α/β	8.47 ± 1.84	8.41 ± 1.60	3.64 ± 0.15 *

^*p < 0.05.

^@obtained from EC₅₀[CaM] data in Fig. 6

To examine the impact of different CaMs on CaN kinetics, we computed α/β as a measure of catalytic efficiency. Both mCaM and sCaM-1 showed nearly identical α/β that is about twice that of sCaM-4. This indicates that upon CaM binding, CaN is converted into a more activated form (i.e. AID-removed CaN), in contrast to its autoinhibited form. More importantly, this activated form has enhanced catalytic capability. In the previous sections, both experimental and computational data demonstrated that the Ca²⁺ and CaMBR binding did not contribute significantly to the different activation effects of CaM isoforms on CaN. Meanwhile, the three isoforms exhibited differences in CaM/DH interaction. It is therefore possible that the DH/CaM binding to mCaM or sCaM-1 pulled the CaN’s AID away from its catalytic site to a greater extent through the interactions with DH, resulting in a CaN form that has higher catalytic efficiency (α/β). Hence the difference in catalytic efficiency may be due to the altered DH binding. Further studies will be needed to delineate other potential mechanisms, including interactions of CaM with other sites on CaN [58].

4. Conclusions

CaM is an important Ca²⁺ sensor that translates Ca²⁺ signaling to various protein-protein interactions. It was a long standing belief that CaM has an invariant amino acid sequence in vertebrates [59] with only one amino acid difference among several insect CaM sequences [60]. However, recent studies have revealed that CaM mutations in human populations that have so far been related to cardiac diseases [61, 62], likely through disturbing Ca²⁺-handling in cardiac cells [15]. The molecular mechanisms underlying these ‘calmodulinopathies’ are still emerging [16, 63]. Understanding how plant-derived CaM isoforms affect the activation of mammalian targets helps to explain the altered functions of CaM variants found in humans. In the present study, we examined the molecular basis of two soybean CaM isoforms that differentially regulate human CaN activity. Two soybean CaMs, sCaM-4 and sCaM-1, share moderate to high sequence identity to human CaM (mCaM), but exhibit differential regulation of human CaN: sCaM-1 was comparable with mCaM and sCaM-4 was significantly reduced. Based on a recent CaM/CaN interaction model [30], we investigated three molecular interactions that contribute to CaM-dependent activation on CaNL 1) Ca²⁺ binding to CaM. 2) CaM binding to the CaM binding region (CaMBR) of CaN and 3) CaM binding to the distal helix (DH) motif of CaN. To determine a structural basis for this effect, we used a combination of docking and MD refinement to the sCaM structures here that we had originally developed for a study of mCaM [30]. This procedure was validated through enzyme kinetics assays using site-directed mutagenesis of CaM residues along the proposed CaM/distal helix interfaces. However, non-equilibrium enhanced sampling via approaches including Gaussian accelerated MD [64] and metadynamics [65] may uncover additional CaM/distal helix poses that were not sampled using the unbiased molecular dynamics (MD) simulations considered here. Using these MD simulations and kinetics measurements, we excluded Ca²⁺ binding and CaMBR binding as factors that differentiate the sCaMs. Instead, we find that the CaM/DH interaction differentiates sCaM-4 from mCaM and sCaM-1. Namely, mCaM and sCaM-1 both exhibited stronger interactions with distal helix than sCaM-4. Kinetic modeling indicates that the primary effect of the two sCaMs was to form a CaN/CaM complex with different turnover (k_cat) rates that may stem from altered CaM/DH interactions. Our study therefore sheds light on how CaM mutations may affect target activation through interactions distinct from those involved in canonical CaMBR and Ca²⁺ binding. Such noncanonical interactions may serve similar roles in other CaM regulated targets such as CaM protein kinase II (CaMKII [66]), CaM kinase kinase (CaMKK [67]), myosin light chain kinase (MLCK [68] and death-associated protein kinase (DAPK) [69]) that utilize autoinhibitory domains.