Modeling the Annexin A1-S100A11 Heterotetramer: A Molecular Dynamics Investigation of Structure and Correlated Motion

Abstract

Annexin A1 (A1) has been shown to form a tetrameric complex (A1t) with S100A11 which is implicated in calcium homeostasis and EGFR pathways. In this work, a full-length model of the A1t was generated for the first time. Multiple molecular dynamics simulations were performed on the complete A1t model for several hundred nanoseconds each to assess the structure and dynamics of A1t. These simulations yielded three structures for the A1 N-terminus (ND) which were identified via principal component analysis. The orientations and interactions of the first 11 A1-ND residues for all three structures were conserved, and their binding modes were strikingly similar to those of the Annexin A2 N-terminus in the Annexin A2-p11 tetramer. In this study, we provided detailed atomistic information for the A1t. Strong interactions were identified within the A1t between the A1-ND and both S100A11 monomers. Residues M3, V4, S5, E6, L8, K9, W12, E15, and E18 of A1 were the strongest interactions between A1 and the S100A11 dimer. The different conformations of the A1t were attributed to the interaction between W12 of the A1-ND with M63 of S100A11 which caused a kink in the A1-ND. Cross-correlation analysis revealed strong correlated motion throughout the A1t. Strong positive correlation was observed between the ND and S100A11 in all simulations regardless of conformation. This work suggests that the stable binding of the first 11 residues of A1-ND to S100A11 is potentially a theme for Annexin-S100 complexes and that the flexibility of the A1-ND allows for multiple conformations of the A1t.

1. Introduction

The annexin family of proteins exhibits calcium-dependent binding with membranes. There are 12 members present in vertebrates (Annexins A1-A11, A13) with A12 being unassigned.¹ Annexins are composed of a highly conserved core domain (CD) and a N-terminal domain (ND) unique to each family member. Five α-helices make up each of the four homologous repeats in the core domain with the exception of Annexin A6, which has eight repeats.^1–4 The core-domain is curved, with its two faces being referred to as the concave and convex faces. The concave face contains the ND which serves as the site for interactions with S100 proteins. Additionally, phosphorylation and glycosylation sites found within the concave face have been shown to regulate interactions with other proteins^2,5,6 The convex face contains Ca²⁺ binding sites and the primary location of annexin-membrane interactions.^2,6 Annexin A1 (A1), contains 346 amino acids (~38 kDa) with a 44 amino acid ND.^1,2 A1 can bind up to eight calcium ions on the convex protein surface.² Prior to the binding of calcium, the ND exists in a buried state within Repeat III of the core domain.^1,2,7 Upon calcium binding at Repeat III, the ND is ejected from the core domain (CD).^1,2,8 Following calcium binding, the released ND can interact with cell membranes, bind with phospholipid bilayers, and perform additional biological functions.^1,2

A1 is most notably involved in inflammatory responses and apoptotic pathways.² For instance, A1 has been linked to Alzheimer’s disease by its ability to reduce the inflammation marker secretion from Amyloid B plaques. In mice treated with human-recombinant A1, the leakage of the blood-brain barrier was reversed which improved amyloid plaque clearance. With a decrease in amyloid plaque, vascular structure was restored showing promise of A1 in the treatment or mediation of neurodegenerative diseases.⁹ The misregulation of A1 has been found to be present in many forms of cancer including breast cancer, gastric cancer, glioblastoma and certain types of carcinoma.⁷ This may be attributed to the involvement of A1 in pathways regulating EGFR. A1 is also related to the inherent resistance of recurrent ER positive breast cancer to tamoxifen therapy.¹⁰

Annexin family proteins and S100 proteins form heterotetramers via interactions between the NDs of annexins and the hydrophobic pockets of S100 proteins upon Ca²⁺ binding.³ The S100 family is comprised of 10-12 kDa proteins found exclusively in vertebrates. S100 proteins consist of two unique EF-hand Ca²⁺ binding motifs connected via a hydrophobic linker.^11,12 The C-terminal Ca²⁺ binding loop is made of 12 amino acids and has a higher Ca²⁺ affinity compared to the 12 amino acid N-terminal Ca²⁺ binding loop.¹¹ S100s are implicated in regulatory functions, Ca²⁺ homeostasis, and cellular motion. Additionally, S100s have been linked to various inflammatory and neurological conditions/diseases.¹² S100A11 is found cellularly as an antiparallel, non-covalent homodimer. Dimerization of S100 proteins occurs between helices I and IV of two S100 monomers. Upon calcium binding, the S100 dimers create hydrophobic cleft sites for ligand interactions. Like A1, misregulation of S100A11 is a feature of various cancers. For instance, its upregulation is correlated to worse prognoses due to its involvement in tumor cell proliferation and apoptosis.¹³ In the study conducted by Wang, H et al., cell proliferation and tumorigenesis of glioma were considerably inhibited when S100A11 was downregulated.¹³

Like other A1 functions, the A1-S100A11 heterotetramer (A1t) is proposed to regulate calcium-dependent biological processes. A1t has been implicated in the development of keratinocytes and their cornified envelopes.¹⁴ A1t has also been shown to bind and inhibit phospholipase A2 which is pivotal for keratinocyte growth.¹⁵ Furthermore, the cleavage of A1 prevents its association to S100A11 resulting in the unregulated growth of model squamous carcinomas.¹⁶ In addition, A1t has been observed in early endosomes, and the binding of A1 to S100A11 helps regulate the degradation of EGFR via lysosomal degradation.¹⁷ Malfunctions in the EGFR pathway can result in tumor formation, suggesting that A1t is involved in tumorigenesis.

Previously, molecular dynamics (MD) simulations have been performed on monomeric A1. Lewis et al. investigated the conformational changes of A1 during the ejection of the ND from the CD.¹⁸ Simpkins et al. performed MD simulations to study S27 mutations in the ND of A1 and their implications on structural organization and membrane aggregation.¹⁹ In this work, we attempt to study the structure and dynamics of the complete A1t. For the first time, the full structure of the human A1t was constructed and subjected to multiple Molecular Dynamic simulations using AMBER18. The results presented here are intended to provide detailed structural and atomistic information to better understand the interactions within the A1t and guide future experimental studies.

2. Computational Methods

2.1. Modeling the Human A1-S100A11(A1t) Tetramer

To build the structure of the human A1t, the complete A1 structure previously generated by our group and the PDB structures 2LUC (PDB ID: 2LUC) and 1QLS (PDB ID: 1QLS) were used.^18,20,21 The complete A1 structure was previously constructed using the PDB structures 1MCX (PDB ID: 1MCX) and 1HM6 (PDB ID: 1HM6).^8,22 1MCX is a calcium bound X-ray structure for the sus scrofa A1 monomer without the N-terminal. 1HM6 is a calcium-free X-ray structure for the full-length sus scrofa A1 monomer. As described by Lewis et al., the construction of the complete A1 structure used the core domain of 1MCX and the ND from 1HM6; then the ND was fused to the core domain.¹⁸ 2LUC is an NMR structure of a S100A11 dimer in the presence of calcium. 1QLS is a crystal structure of S100A11 in complex with the first 11 residues of the A1-ND in the presence of calcium ions. Each structure was compared with its respective human wild-type sequence via EMBOSS Needle.²³ Differences were identified and modified to match the human wild-type sequence using Modeller.²⁴

To build A1t, two copies of the 1QLS structures were aligned to the 2LUC S100A11 dimer. The NDs of two A1 monomers were then aligned to the A1-ND residues already present in the two 1QLS structures. All protein residues of 1QLS were then removed while leaving the calcium ions from 1QLS in the structure. Calcium ions were added to A1 by first aligning two 1AIN (PDB ID: 1AIN) structures to each A1 monomer and then removing 1AIN protein residues. 1AIN is a calcium bound crystal structure of human annexin A1.²⁵ The resulting structure was a full length, calcium bound model of human A1t.

2.2. Molecular Dynamics Simulation

The topology and initial coordinate files for molecular dynamics (MD) simulation were generated using the tleap module of AmberTools 18.^26,27 The FF14SB force field was used along with the calcium ion parameters of Bartolotti et al.^28–30 Prior to simulation, proteins were solvated explicitly using TIP3P waters in an octahedral box with 16 Å between the protein and box boundary.²⁸ Sufficient chloride and sodium ions were added to neutralize any non-zero charges and bring the salt concentration to 0.15 M.

All MD simulations and calculations were performed on a 4-RTX 2080 Ti workstation housed at East Carolina University. Production run was performed using the pmemd.CUDA module within AMBER 18 while all other steps were carried out utilizing the pmemd.MPI module.²⁷ For all simulations, periodic boundary condition was utilized along with a 10Å non-bonded cutoff to limit the van der Wall non-bonded interactions. A 2 fs timestep was achieved through the constraint of bonds involving hydrogen atoms via the SHAKE algorithm.³¹ In this work, three MD simulations were performed on A1t model. Prior to each MD simulation, the system was subjected to two steps of energy minimization. For the first step of minimization, the steepest descent method was used for 2500, 4000, and 5000 steps for each simulation respectively. For the second step of minimization, the conjugate gradient method was used for 7500 steps in all three simulations.

Following minimization, the systems were warmed in two steps. The first step scaled the temperature from 0-100 K over 5 ps, and in the second step, temperature was scaled from 100-300 K over 50 ps. Finally, isothermal-isobaric (NPT) ensemble simulations were performed for 720, 600, and 560 nanoseconds.

2.3. Simulation Analysis and Figure Generation

Simulation trajectories were analyzed using the cpptraj and ptraj modules of AMBER18.^32,33 Interaction energies were calculated using the MMGBSA.py script.³⁴ All visualization and data manipulation were performed using Python, Pymol, and R.^27,35

3. Results and Discussion

3.1. A1t Structural Conformation

Three simulations were performed on the A1t model for 720, 600, and 560 nanoseconds. These simulations differed only by the number of initial energy minimization steps as described in section 2.2 Molecular Dynamics Simulation. The stabilities of these simulations were determined by the conservation of energy of the simulation system and the maintenance of constant average temperature. One of the simulated structures of A1t was shown in Figure 1. For all three simulations, the distance between two monomers of A1 within A1t and the root mean square coordinate deviation (RMSD) of α-carbons were calculated via the ptraj module of AMBER18 (see Figures 2 and S1). As shown in Figure 2, the three simulations converged. The distances between two A1 monomers in A1t tetramer stabilized at ~75 Å, which is similar to the experimental Cryo-EM results reported by Lambert et al. In their study, Lambert et al. investigated the organization of annexin and S100 proteins in the event of membrane aggregation/junctions.³⁶ Their Cryo-EM studies found that the A2-S100A10 tetramer induced junctions between two membranes. The Cryo-EM density analysis of the junction between two membranes indicated that the distance between two A2 monomers in A2-S100A10 tetramer is about 60Å.³⁶ Our A1t adopted a conformation with ~75Å between A1 monomers. Due to the relative lengths of A1-ND (44 residues) and A2-ND (32 residues), it is reasonable for A1t to adopt a conformation which is slightly larger than that of the A2-S100A10 tetramer reported in the literature.

Figure 1.

The cartoon representation of the simulated A1t. Two S100A11 monomers (S-1 and S-2) are in blue and red. The N-terminal domains (ND) of A1 are in purple, and the four structural repeats of A1 core domain (I-IV) are in yellow, orange, green, and cyan. The bound calcium ions for the A1t are shown in silver. A1-1 and S-1 calcium ions are labeled in silver while calcium ions bound to A1-2 and S-2 are labeled in black.

Figure 2.

Distance between the centers of mass of two A1 subunits.

3.1.1. PCA Analysis of A1 N-terminal domain (ND) within the A1t

Orientations of the A1-ND in A1t were explored using principal component analysis (PCA). The trajectories were aligned based on the S100A11 dimer, and PCA was performed for the first 23 residues of both A1-NDs of all three simulations. The data collected for the six NDs were combined into a single PCA plot where the trajectory frames were projected onto the first two principal components. A plot of the projections along with a kernel density estimate (KDE) overlay is shown in Figure 3A. From the KDE overlay, three regions with high densities of structures were identified. The most representative structures were extracted from the three regions (see Figure 3B–3D).

Figure 3.

A) Scatter plot of projections of the trajectories for the N-terminal domain onto the first two principal components with a kernel density estimate overlay. B-D) Three representative structures. Color scale represents the strength of each individual residue’s total interaction with the S100A11 dimer. E) Alignment for all six A1-ND in complex with the S100A11 dimer for representative structures B (yellow), C (teal), and D (purple). F) Close-up view of the difference in side-chain interaction between A1 W12 and S100A11 M63; linear conformation (yellow and purple) and kinked conformation (teal). G) Alignment of the A1-ND of Structure C (Yellow), 1QLS (purple), and 4HRE (teal).

The three representative structures were colored based on their total interaction energies with the S100A11 core (see Figure 3B–3D, Table 1). Structures C and D adopted a similar conformation which had a kink at W12 (see Figure 3C, 3D). Structures C and D differed only by the angle of the kink which resulted in slightly different interactions. In Structure B, the ND was a continuous helix with no kink (see Figure 3B).

Table 1:

Total interaction energies (in kcal/mol) of the first 12 residues of A1-ND with the S100A11 complex.

Residue	Structure B	Structure C	Structure D
A2	−10.21	−13.64	−20.18
M3	−9.51	−12.80	−16.73
V4	−16.40	−20.08	−20.60
S5	−10.28	−13.25	−16.00
E6	−1.44	−0.46	−1.04
F7	−16.94	−14.63	−10.25
L8	−14.80	−16.42	−15.38
K9	−9.11	−9.56	−11.76
Q10	−3.58	−4.33	−3.07
A11	−3.83	−5.08	−4.66
W12	−32.56	−28.56	−22.15

Inspection of the side chain interactions between A1-ND and S100A11 for both kinked structures revealed a W12-M63 interaction between A1-ND and S100A11(see Figure 3F). The angled alpha helix placed W12 of A1 near the M63 sulfur of S100A11. This interaction was not possible for the linear conformation as shown in Figure 3F.

3.1.2. A1t Similarities to Crystal Structures

Figure 3E highlights the similarities of the first 11 residues of the simulated structures. For all 6 simulated NDs, the conformation of the first 11 residues was conserved as a helix. After residue 11, the structures diverged depending on the presence of the W12 kink. Comparison to the 1QLS crystal structure (Figure 3G in purple) revealed that the binding mode of the first 11 residues of our simulated A1-NDs was consistent with that of the A1-ND present in the 1QLS crystal structure. As previously described, 1QLS is the crystal structure of S100A11 in complex with the first 11 residues of the A1-ND in the presence of calcium. In the 1QLS structure, only the first 11 residues of the A1-ND were crystallized; this implicated that those remaining residues were more flexible and less stable in structure which was reflected in our simulations. Furthermore, the interaction between A2-ND and p11 within the A2-p11 tetramer (PDB: 4HRE) was also very similar to the interaction between A1 and S100A11 in our simulated A1t tetramer (Figure 3G). The first 10 residues of A2 from the x-ray structure (PDB Code: 4HRE, Figure 3G in teal) of the A2-p11 tetramer adopted a helical structure. Starting at residue 11, the A2-ND adopted a loop conformation until Repeat I of the A2 core domain. This was strikingly consistent with our A1t structure.³⁷ Additional ongoing studies in our lab have observed the loss of A1-ND helical structure for simulations in the multi-μs timescale for A1 monomer upon calcium binding. This leads us to believe that eventually the A1-ND of our A1t may also lose some helical structure following its stable binding segment (first 11 residues) and become a disordered loop similar to the 4HRE structure.

Based on these observations, it appears that the first ~11 residues of Annexin NDs play a key binding role in forming Annexin-S100 tetramers. We suggest that the binding mode of the first 11 residues of Annexin NDs will be consistent for all the Annexin-S100 tetramer combinations and a common feature among these complexes.

3.2. Secondary Structure Analysis

Secondary structure analysis was performed to examine the overall change in A1 and S100A11 structures upon tetramer formation. A1 calcium binding sites are found at calcium binding loops connecting the helixes of its conserved core.

Secondary structure analysis was performed using cpptraj. The fractions of alpha helix content for the three simulations were overlapped for A1-1/S-1 Figure 4A and A1-2/S-2 Figure 4B. Additionally, the fractions of alpha helix content for crystal structures 1MCX and 2LUC were overlapped with A1-1/A1-2 and S-1/S-2, respectively, to track the overall structural change upon tetramer formation. It should be noted that the N-terminus is not found in the 1MCX structure, thus the plot in black begins at residue 41. Furthermore, human A1 contains 6 calcium ions as opposed to the porcine annexin (1MCX), which contains 8 calcium ions. Consequently, calcium labels and numbers are based on the 6 binding sites of human annexin which are shared between the structures.

Figure 4.

Fraction of Alpha helix content of the three simulations (Simulation 1:green, Simulation 2:blue, Simulation 3:red) for A1t divided into its components A) A1-1/S-1 and B) A1-2/S-2. Black represents the helix content from the 1MCX/2LUC PDB structures. Content is divided based on repeat, and calcium locations are shown with the dashed red lines.

The simulated and crystal structures were mostly consistent in their secondary structures with no drastic conformational changes occurring within the A1 core domain upon A1 and S100A11 association. Analysis showed both A1-1 and A1-2 maintained all five alpha helices for each of the four Repeats (I-IV) in all three simulations. Additionally, the A1-1 and A1-2 calcium binding loops were preserved in the A1t.

The most significant differences between the simulated tetramer and the crystal structure were observed in the S100A11 dimer and the A1-NDs. A loss of helical content occurred at the Calcium 7 binding loop (EF1 calcium binding loop in S100A11). This expansion in the Calcium 7 binding loop was present for all three simulations. Additionally, a decrease in alpha helix content was observed at the site of S100A11 and A1 binding (S100A11 residues 48-55 and A1 residues 1-11). The loss of alpha helical structure within the A1-NDs was responsible for creating kinks in the A1-NDs. The different locations of the kinks contributed to the different ND conformations seen in Figure 3. In all three simulations, one A1-ND formed a kink while the other A1-ND retained its alpha helix. Additionally, variable levels of alpha helix structure were observed from residues 48-55 within the S100A11 regions.

3.3. Strong Interactions

3.3.1. Interactions between Representative N-terminal domain Structures and S100A11

MMGBSA calculations were performed to determine the strong interactions between 1) each representative structure from Figure 3B–D and the S100A11 dimer and 2) each subunit of A1t. All calculations were performed using the MMGBSA.py script of AMBER18 with a modified salt concentration of 0.15M.³⁸ For the energy calculations of the A1t subunits, 200 frames were sampled over the final 20 ns of each simulation. This ensured that each simulation had reached a stable conformation prior to energy calculations. The total energy contributions between each residue of the three representative structures with their respective S100A11 dimers were reported in Table 1. Consistent interaction energies were observed between the A1-ND and the S100A11 core for all three representative structures. Specifically, the interactions involving the first 12 residues were conserved. The following residues of the A1-ND were identified to have the strongest binding interactions with S100A11: W12, F7, V4, L8, S5, A2, and M3.

The residue-residue interaction pairs between S100A11 (S-1/S-2) and the A1-ND were reported in Table S1. For all three structures, the A1-ND had strong interactions with both S100A11 subunits. The strongest interactions involved either members of the first 12 A1-ND residues or acidic A1 residues (E-15 and E-18). Bolded in Table S1, A1-W12 interactions were similar between all three structures. However, Structures 2 and 3 had the additional strong interaction between W12 of A1 and M63 of S-2 which was not present for Structure B.

3.3.2. Interactions within the A1t

The 10 strongest residue-residue interactions between each tetramer component were summarized in Table S2. Strong interactions were reported for A1 monomers and both S100A11 subunits. Both S-1 and S-2 residues E11, I14, and E15 had multiple strong interactions with A1-1 and A1-2 ND residues respectively. The strongest interactions of the A1-S100A11 dimer pairs (A1-1:S-1 and A1-2:S-2) were highly consistent and involved Annexin ND residues, specifically, the first 11 residues (M3, V4, S5, E6, L8, K9). A1-S5 has been shown to be critical for tetramer formation, as its phosphorylation inhibits A1-S100A11 association.³⁹ S5 of both A1-1 and A1-2 had strong interactions with E11 of S-1 and S-2 respectively. This interaction was present for all three structures demonstrating its importance within the A1t.

For A1-2:S-1 and A1-1:S-2, S100A11 residues K97 and M94 were important interaction sites. However, interacting residues within A1-1 and A1-2 differed for these regions. A1-2:S-1 involved residues 12-20 of the ND and Repeat II/IV residues (R213/R292) of A1-2. A1-1:S-2 relied on residues F7, W12, and Y39 of the ND and Repeat I residues (Q79, Q86). Additionally, A1-1:S-2 included the W12-M63 interaction which was reflected in the kinked conformation of A1-1. This structural feature facilitated interactions between Repeats I of A1-1 and S-2 which accounted for the asymmetric interactions and motion of the A1t (see section 3.5 Correlated Motions).

In Figure 5A, positive correlation was observed between A1-1 Repeat I and the S100A11 core while A1-2 displayed negative correlation between the same regions. This appeared to stem from the orientation of A1-1 with respect to the S100A11 dimer. As shown in Figure 5B, Repeat I of A1-1 was more closely associated to S-1 and S-2 than that of A1-2. Consequently, A1-1 Repeat I moved as a unit with the A1-ND-S100A11 complex hence the positive correlation. This demonstrates the importance of the A1-ND for the global motion of the A1t complex.

Figure 5.

A) Cross-correlation heatmap of alpha carbons for A1t complex. Dark purple represents anticorrelated motion while yellow is representative of positively correlated motion. B) Graphical depiction of the first normal mode.

3.4. Calcium Binding

It is well-known that calcium binding is required for the formation of all Annexin-S100 heterotetramers except for the A2/S100A10 heterotetramer.⁴⁰ The calcium coordination was evaluated for the three simulations using cpptraj. Coordination sites were defined as oxygen atoms within 3.5 Å of calcium atoms. To screen for stable coordination sites, only oxygen atoms coordinated for at least 10% of the simulation were reported in Table 2. Simulated coordination sites for A1-1 and A1-2 were compared to the calcium coordination in the 1MCX crystal structure. 1QLS was used to compare calcium coordination sites for S100A11 residues. All 3 simulations were relatively consistent in their results. The coordination data for one simulation is reported below while the remaining results are in Supplemental Materials (see Table 2, S3, S4).

Table 2:

Calcium coordination comparison for simulated A1t and 1MCX and 1QLS. The fraction columns represent the fraction of the simulation during which the coordination was present.

				A1-1/S-1		A1-2/S-2
Calcium	Type	Repeat	Crystal Comparison	Trajectory	Fraction	Trajectory	Fraction
1	II	I	-	G59@O	0.783	G59@O	0.998
			V60@O	V60@O	0.784	V60@O	0.998
			E62@E1	E62@OE1	0.964	E62@OE1	0.999
			E62@E2	E62@OE2	0.922	E62@OE2	1
			-	D334@OD1	0.195	-	-
			-	D334@OD2	0.414	-	-
2	III	I	K97@O	K97@O	0.166	K97@O	0.11
			L100@O	L100@O	0.139	-	-
			-	E105@OE1	0.838	E105@OE1	0.91
			E105@OE2	E105@OE2	0.85	E105@OE2	0.81
			-	E106@OE1	0.67	E106@OE1	0.754
			-	E106@OE2	0.692	E106@OE2	0.763
3	II	II	-	A126@O	0.63	-	-
			M127@O	-	-	-	-
			G129@O	G129@O	0.226	-	-
			G131@O	G131@O	0.4	-	-
			-	D133@OD1	0.877	-	-
			-	D133@OD2	0.502	-	-
			-	E134@OE1	0.376	-	-
			-	E134@OE2	0.326	E134@OE2	0.766
			-	-	-	E136@OE1	0.829
			D171@OD1	D171@OD1	0.879	D171@OD1	0.995
			D171@OD2	D171@OD2	0.935	D171@OD2	0.977
4	II	III	-	G210@O	0.114	G210@O	0.516
			-	E211@OE1	0.883	E211@OE1	0.632
			-	E211@OE2	0.891	E211@OE2	0.75
			R213@O	-	-	-	-
			G215@O	-	-	-	-
			E255@OE1	E255@OE2	0.997	E255@OE1	0.986
			E255@OE2	E255@OE1	0.985	E255@OE2	0.991
5	II	IV	M286@O	-	-	M286@O	0.975
			G288@O	-	-	G288@O	0.761
			G290@O	-	-	G290@O	0.762
			-	D329@OD1	0.627	-	-
			-	D329@OD2	0.578	-	-
			E330@OE1	E330@OE1	0.776	E330@OE1	0.996
			E330@OE2	E330@OE2	0.888	E330@OE2	0.992
6	III	IV	L328@O	L328@O	1	L328@O	0.716
			T311@O	T331@O	0.991	T331@O	0.711
			E336@OE1	E336@OE1	0.23	E336@OE1	0.925
			-	E336@OE2	0.804	E336@OE2	0.371
7	III	S100A11	D28@OD2	D28@OD2	0.999	D28@OD2	1
			N30@O	D28@OD1	0.999	D28@OD1	0.997
			K33@O	T33@O	0.998	T33@O	0.999
			E38@OE1	E38@OE1	0.886	E38@OE1	0.944
			E38@OE2	E38@OE2	0.883	E38@OE2	0.95
			-	Q74@OE1	0.109	-	-
			-	-	-	T105@O	0.51
8	III	S100A11	D68@OD1	D68@OD1	0.721	D68@OD1	0.683
			-	D68@OD2	0.283	D68@OD2	0.322
			D70@OD2	N70@OD1	0.979	N70@OD1	0.638
			-	D72@OD1	0.372	D72@OD1	0.526
			D72@OD2	D72@OD2	0.976	D72@OD2	0.996
			Q74@O	Q74@O	1	Q74@O	1
			-	D76@OD1	0.373	D76@OD1	0.505
			-	D76@OD2	0.628	D76@OD2	0.467
			E79@OE1	E79@OE1	1	E79@OE1	1
			E79@OE2	E79@OE2	1	E79@OE2	0.994

The simulated calcium coordination sites were consistent with those of the crystal structures indicating that the simulations performed in this study were reasonable. By comparison, we observed that the simulated calcium coordinations for A1-1 and A1-2 were consistent. Similarly, the simulated calcium coordinations for S-1 and S-2 were also consistent. Particularly, we noticed that the X-ray coordinations for S100A11 calcium ions 7 and 8 were present for at least 60% of the simulation.

Overall, our results show good agreement with available experimental crystal structures. Experimental conditions may influence calcium coordination numbers for structures determined through X-ray crystallography. Our simulated results were not expected to be identical to X-ray data as MD simulations investigate the dynamics of each coordination. In this work, we observed the dynamic interactions between calcium and the A1t and provided detailed atomistic information pertaining to the calcium coordination within the A1t. The calcium binding information we reported may be valuable for future mutation studies investigating the impact of calcium binding on the biological functions of the A1t.

3.5. Correlated Motions

Cross-correlation analysis has been an important tool for assessing structural domain dynamic correlation during MD simulations. In this work, cross correlation analysis was employed to explore the correlated motions within A1t. Cross correlation analysis between the α-carbons of all residues was performed via the cpptraj module of AMBER.

The correlation heatmaps generated for all three simulations were consistent in their overall trends. One plot is shown in Figure 5A while the other plots are found in Supplementary Information (Figure S2, S3). Overall, the heatmaps demonstrated correlated motion throughout A1t. A prominent feature shared by all three correlation maps was the strong positive correlation between the S100A11 dimer and the NDs of both A1-1 and A1-2. Shown in Figure 5A, the S100A11 dimer was negatively correlated to Repeat II in both A1-1 and A1-2. The asymmetric nature of A1t was highlighted when comparing the correlations of A1 Repeats I and IIIwith the S100A11 dimer. For A1-1, Repeat I was positively correlated and Repeat III was negatively correlated with the S100A11 dimer. Conversely, Repeats I and III of A1-2 exhibited negative and positive correlation respectively.

Within each A1 monomer, Repeats I and IV, II and III were positively correlated. However, Repeat III was negatively correlated with Repeat I. Between the two A1 monomers, strong negative correlations were observed between A1-1 Repeats II and III together and Repeat III of A1-2, and between A1-ND and Repeat I together and Repeat I of A1-2. This long-distance phenomenon was visualized through the depiction of first normal modes (Figure 5B). The normal modes depicted a rocking motion of the tetramer with the A1-1 Repeats II and III moving in the opposite directions of A1-2 Repeats II and III and A1-1 Repeat I moving in opposite directions of A1-2 Repeat I. The A1 region in close association with the S100A11 core, the ND of A1, was positively correlated to both S-1 and S-2, and they moved together.

The remaining heatmaps (Figure S2, S3) also reflected the asymmetric correlation within A1t albeit through different A1 repeats. This was a consequence of the multiple orientations of A1 monomers allowed by the different A1-ND conformations. However, the defining trend seen in all our A1t conformations was the strong positive correlation between the A1-ND and the S100A11 dimer. Based on the similarity to the A2t (see Section 3.1: A1t Structure and Conformation) and the consistency among the three simulated structures of the A1t, we believe this stable, correlated complex between Annexin-NDs and the S100 proteins will be a theme among the other Annexin-S100 tetramers.

4. Conclusions

In this study, multiple molecular dynamics simulations were performed on a complete model of the A1-S100A11 heterotetramer (A1t) for the first time. Three simulations resulted in stable structures of A1t with ~75Å distances between two A1 monomers. A1t structures adopted a range of conformations afforded by the flexibility of the A1-ND. When comparing these structures, the orientations of the first 11 residues of the NDs, with reference to the S100A11-dimer, were conserved. Furthermore, the S100A11 interaction with the first 11 residues of the simulated A1-ND were consistent with both the crystal structure of A1-ND and S100A11 in 1QLS and the crystal structure of A2 and S100A10 in 4HRE. PCA was used to identify 3 representative structures of the A1-ND which adopted two different conformations. These conformations differed by the presence of a kink in alpha helix structure at W12 of A1 which resulted in a new interaction between W12 and M63 of S100A11. MMGBSA results show strong, consistent interaction energies for all three representative structures for the first 11 residues of the A1-ND (A2, M3, V4, S5, F7, L8) with the S100A11 core. Additionally, strong interactions were identified between the A1-NDs and both S100A11 subunits present in the A1t including the A1-S5 and S100A11-E11 interaction. A1-S5 has been demonstrated in the literature to be important for the association of A1 and S100A11. Calcium binding of the A1t was evaluated and compared to available crystal structures showing strong agreement and stable coordination especially within the S100A11 dimer. The detailed coordination data reported in this work may be important for future mutation studies on the A1t. Furthermore, strong, correlated motion was observed throughout the A1t. Despite having different orientations of the A1 monomers, all three simulations displayed strong positive correlation between A1-NDs and the S100A11 core. These results not only demonstrate the importance of the A1-ND for the structure and dynamics of the A1t, but also provide insight into Annexin-S100 complexes. Based on the conserved interactions of the first 11 residues of A1 between our different structures and their consistency with available crystal structures for the A2t, we believe that the stable complex between Annexin ND and S100 dimer core will be a common feature of other Annexin-S100 complexes.