S100A4 is a calcium-binding protein that regulates cytoskeletal dynamics, motility, and epithelial–mesenchymal transition (EMT). Upon binding to calcium in its EF-hand domains, S100A4 undergoes a conformational shift that exposes its hydrophobic pockets, which are required for interacting with downstream effectors that drive migration and proliferation (9). Dysregulated S100A4 activity has been implicated across multiple malignancies, with elevated expression commonly observed in oral squamous cell carcinoma (OSCC), a metastatic form of head-and-neck cancer. Although small-molecule inhibitors targeting S100A4’s hydrophobic interaction surfaces have been explored, no therapeutic strategy has been developed to block its calcium-dependent activation, representing an unmet mechanistic target. Here, we designed and screened for three engineered small-molecule inhibitors that bind to the EF-hand of the calcium-binding pocket in S100A4. Using a protein-based drug screening platform, we evaluated each compound and its ability to bind calcium-dependent S100A4 in silico. One candidate observed effective binding to the EF-hand region and demonstrated the potential for interference with calcium binding, and was identified as a suitable molecule for further in vitro testing. These findings establish the feasibility of designing EF-hand–targeted inhibition of S100A4 in silico for studying an unexplored therapeutic strategy aimed at mitigating S100A4-driven progression in oral and head-and-neck cancers.

• Drug screening

• Head and neck cancer

• Cell motility

• Oral squamous cell carcinoma

• S100A4

Behzadian Y, Guan D, Iqbal T, Kim Y, Tseng T, et al. (2026) Targeting S100A4-Driven Calcium Signaling in Head and Neck Cancer: Identifica tion of Novel Calcium-Site Inhibitors through High-throughput Drug Screening. JSM Head Neck Cancer Cases Rev 6(1): 1012.

Oral squamous cell carcinoma (OSCC) is a malignancy that arises from the squamous epithelium of the oral cavity and remains a significant clinical challenge. Although advances have been made in the understanding of its molecular drivers, OSCC continues to exhibit high recurrence and aggressive metastatic behaviour [1]. More than 90% of deaths associated with OSCC result from metastatic spread rather than the primary tumor [2]. These observations have gained the interest of studying the molecular pathways that enable OSCC cells to migrate and invade surrounding tissues, with recent evidence highlighting epithelial–mesenchymal transition (EMT) as a central contributor to these phenotypes. Given these findings, S100A4 has emerged as a key EMT-associated protein that contributes to cytoskeletal reorganization, cell motility, and metastasis. S100A4 is a small hydrophobic calcium-binding protein belonging to the S100 family [3]. It can form homodimers or heterodimers with S100A1, although the functional requirement of dimerization remains unresolved [4]. Each monomer contains two EF hand motifs that bind calcium independently. Although the C-terminal EF-hand exhibits higher affinity, the N-terminal EF-hand binds calcium first [4]. Calcium binding induces a conformational transition that exposes a hydrophobic pocket required for engaging partners such as actin, nonmuscle myosin II isoforms, tropomyosin, p53, and p57 [5,6]. These interactions underlie the established association between elevated S100A4 levels and metastatic progression in multiple cancers. 

Although S100A4 has been widely studied as a regulator of metastasis, its calcium-binding interface has not been examined as a potential therapeutic target. The EF-hand motifs coordinate calcium and initiate the structural changes that allow S100A4 to interact with downstream effectors. The reliance of S100A4 on this calcium-dependent activation step suggests that disrupting calcium binding may suppress its pro-metastatic functions. Numerous studies have elucidated the structure of the EF motif responsible for calcium coordination, raising the possibility of designing small molecules that interfere with this process. The ability to prevent the calcium-induced conformational activation of S100A4 presents a potential strategy for reducing its biological activity. Unlike approaches that target the hydrophobic interaction surface directly, inhibiting calcium binding may block S100A4 activation upstream of hydrophobic pocket exposure. This concept provides a foundation for targeting the EF-hand region as a therapeutic site that remains untested.

Given the dependence of S100A4 on calcium-induced activation, we sought to determine whether the EF hand calcium-binding site could serve as a viable point of therapeutic intervention. To explore this possibility, we carried out a structure-guided virtual screening of small molecules designed to interact with residues that coordinate calcium within the EF-hand motifs. By integrating structural information with ligand-binding predictions, we identified compounds with the potential to interfere with calcium coordination and thereby limit S100A4 activation. Establishing whether the EF-hand region can be targeted in this way provides an important foundation for developing upstream inhibitors of S100A4 function and introduces a mechanistically distinct strategy for reducing its contribution to metastatic progression in OSCC.

Visualizing S100A4 in PyMOL and Identifying Ca²? Binding Residues

The Ca²?-bound crystal structure of S100A4 (PDB ID: 3CGA) was downloaded from the Protein Data Bank (PDB) and opened in PyMOL [7]. The structure was examined in cartoon representation to locate the canonical (C-terminal) and pseudo (N-terminal) EF-hand motifs.

Each EF-hand motif was inspected manually by tracing the helix–loop–helix regions characteristic of Ca²?-binding domains. Residues directly coordinating Ca²? ions were identified by selecting atoms within 3.5 Å of each Ca²? ion using PyMOL’s distance-selection tools. These residues were recorded and later used to define the docking pocket for inhibitor screening.

Ray-traced renderings were generated using PyMOL’s ray command to visualize the Ca²?-binding sites and overall structural arrangement of the EF-hand motifs.

Screening Potential Inhibitors Using Mcule

Structure-based virtual screening was conducted using Mcule’s 1-Click Docking platform [8]. The Ca²?- coordinating EF-hand residues identified in PyMOL were used to define the docking center. With the EF-hand motifs 

set as the binding center, Mcule’s default docking engine (AutoDock Vina) generated docking poses for thousands of compounds from the Mcule compound library. Docking scores for each compound (ΔG) were recorded, and the top-scoring compounds were selected for further evaluation. For each selected molecule, we also recorded the Mcule provided molecular weight, LogP, hydrogen- bonding profile, and Lipinski Rule of Five statistics. In addition, Mcule’s built-in toxicity checker was used to predict potential toxicity liabilities.

Analyzing Protein–Ligand Interactions Using PLIP

The top-scoring compounds were submitted to the Protein–Ligand Interaction Profiler (PLIP) to identify specific interactions formed within the EF-hand binding site. PLIP-generated interaction maps highlighting hydrogen bonds, hydrophobic contacts, water bridges, and polar interactions between the ligand and S100A4 [9-11]. Interaction profiles were compared directly to the Ca²?-coordinating residues to determine whether the compounds were positioned to interfere with calcium binding. PLIP output files were imported back into PyMOL to visualize ligand orientation and confirm spatial overlap with the EF-hand binding loops.

Identifying Ca²?-Binding Residues in EF-Hand Motif of S100A4

A necessary first step in evaluating whether the EF-hand region of S100A4 could be targeted by small molecules was to identify the residues responsible for coordinating Ca²?. Since any inhibitor must interact with this specific structural environment to disrupt Ca²? binding, we examined the Ca²?-bound S100A4 structure in detail.

Using PyMOL, the helix–loop–helix architecture surrounding each bound Ca²? ion was inspected, allowing identification of the coordinating residues. In the pseudo EF-hand, Ca²? coordination involved residues such as Asp25, Glu23, Glu33, Ser20, and Lys28, whereas the canonical EF-hand included Asp63, Glu74, Asn65, Glu69, and Asp67. These coordination geometries are displayed in Figure 1.

Figure 1 Structural visualization of Ca²?-binding residues within the EF hand motifs of S100A4. Calcium ions are shown in green, while coordinating residues are displayed as sticks coloured by element (carbon in grey, oxygen in red, nitrogen in blue). Polar interactions are indicated with yellow dashed lines, with distances reported in ångströms. Panel A presents the Ca²?-bound S100A4 homodimer (PDB ID: 3CGA) in cartoon representation, with the N-terminal pseudo EF-hand coloured in cyan and the C-terminal canonical EF-hand in pink. Panel B illustrates the coordination environment of the C-terminal canonical EF-hand, and Panel C shows the corresponding Ca²?-binding environment in the N-terminal pseudo EF-hand. Together, these structural views highlight the residues responsible for Ca²? coordination and define the EF-hand binding sites examined throughout this study.

Together, these observations confirm that both EF-hand motifs contain well-defined, spatially constrained Ca²?- binding pockets. Because Ca²? coordination is essential for the conformational activation of S100A4, these residues form a structurally and chemically meaningful target for inhibitor design.

Virtual Screening of Potential Inhibitors for Calcium Binding

Having established the residues required for Ca²? coordination, we next evaluated whether small molecules could occupy these EF-hand pockets. Structure-based virtual screening using Mcule’s 1-Click Docking workflow produced a wide range of potential molecules. Three ligands were selected for being the highest-scoring candidates. These compounds were MCULE-5204697437, MCULE-3195257311-0-1, and MCULE-9710401326-0-1,

which we will now refer to as Molecule 1, Molecule 2, and Molecule 3, respectively. As shown in Table 1, Molecule 1 and Molecule 2 docked near the N atom of Lys28 within the canonical EF-hand, whereas Molecule 3 docked near the OD2 atom of Asp63 in the pseudo EF-hand.

Molecule 2 and Molecule 3 showed the strongest predicted affinities, each with a docking score of –7.0, while Molecule 1 displayed a weaker score of –5.4. Toxicity predictions further distinguished the candidates: Molecule 1 was classified as toxic, whereas Molecule 2 and Molecule 3 were both predicted to be non-toxic (Table 2). Additional Mcule generated parameters revealed clear physicochemical differences among the three candidates. Molecule 1 satisfied all Lipinski criteria, with a LogP of 4.58, 7 hydrogen-bond acceptors, and 1 donor, but was flagged as toxic. Molecule 2 had the highest hydrophobicity, with a LogP of 6.19, 7 acceptors, 1 donor, and one Lipinski violation, though it was predicted to be non-toxic. Molecule 3 exhibited a more polar profile, with 8 acceptors, 3 donors, a LogP of 3.36, and no violations, and was also classified as non-toxic.

Table 1: Mcule results for small molecule binding to S100A4 (PDP Accession Code: 3CGA). Bond-line representations are used to show the ligand structure of each molecule. The chemical formula and Mass (Da) were obtained using MCULE’s property calculator.

Table 2: Mcule derived physicochemical and toxicity properties of candidate inhibitors for EF-hand binding in S100A4. Docking scores, Lipinski parameters, LogP values, hydrogen-bond donor/acceptor counts, and toxicity predictions were obtained using the Mcule 1-Click Docking platform, Mcule Property Calculator, and Mcule Toxicity Tracker for each molecule (PDB: 3CGA).

Together, these screening results demonstrate that multiple chemically distinct small molecules can stably occupy the EF-hand Ca²?-binding pockets of S100A4, supporting the feasibility of targeting this site to disrupt calcium-dependent activation of the protein.

Protein–Ligand Interaction Analysis using PLIP

Performing a protein ligand interaction analysis was necessary to determine whether the candidate inhibitors formed interactions with the residues known to coordinate Ca²?, and to assess whether their binding orientations could realistically interfere with calcium-induced activation of S100A4.

Using the Protein–Ligand Interaction Profiler (PLIP), we generated interaction profiles for each compound. PLIP identified hydrogen bonds, hydrophobic contacts, and polar interactions formed between the ligands and residues surrounding the EF-hand loops. Figure 2 shows the binding poses of Molecule 1, 2, and 3, aligned within the EF-hand region of S100A4. In the first two models, the ligands bind within the C-terminal canonical EF-hand and position themselves near the key Ca²?-coordinating.

Figure 2 Interaction profiles of screened small-molecule candidates with the calcium-binding regions of S100A4. The S100A4 protein (PDB: 3CGA) is shown in light green surface representation in the right panels, with each ligand displayed in pink to illustrate overall pocket occupancy. In the left panels, the same ligands are shown in stick representation, where ligand carbon atoms appear in cyan, and S100A4 atoms follow standard colouring (carbon in grey, oxygen in red, nitrogen in blue, sulfur in yellow). Hydrogen bonds are indicated with blue dashed lines, and hydrophobic interactions with grey dashed lines. Panel A shows the binding interactions between Molecule 1 and the C-terminal canonical EF-hand of S100A4. Panel B displays the interaction profile of Molecule 2 within the canonical EF-hand Ca²?-binding region. Lastly, panel C illustrates the binding interactions between Molecule 3 and the N-terminal pseudo EF-hand, illustrating engagement of residues associated with the typical Ca²?-binding environment.

residues Lys28, indicating that both compounds occupy the correct structural region required for potential competition with Ca²?. In contrast, the third molecule binds within the Ca²?-binding environment of the N-terminal pseudo EF-hand, interacting with residues that overlap with the known Ca²?-coordinating network, including Asp63, suggesting that this compound may also interfere with calcium coordination through an alternative EF-hand pocket.

The interaction profiles generated through PLIP reveal clear differences in how the three selected molecules engage the EF-hand binding sites of S100A4. As shown in Table 3, Molecule 1 forms several hydrogen bonds with residues in the canonical EF-hand, including Asn87, Lys28, and Asn65, but overall displays a relatively limited number of stabilizing contacts, consistent with its weaker predicted binding behaviour. Molecule 2 establishes a broader and more diverse interaction network, combining multiple hydrophobic contacts with EF-hand residues such as Phe27, Asp67, Glu69, and Lys28, along with hydrogen bonds to Lys28 and Asp71, indicating a more extensive.

Table 3: Molecule 1, 2, and 3 docking interactions with calcium binding regions of S100A4. PLIP was used to record key docking interactions between each inhibitor and the calcium-binding regions of S100A4.

engagement of the canonical binding pocket. In contrast, Molecule 3 binds within the pseudo EF-hand and positions itself closest to key Ca²?-coordinating residues, forming hydrogen bonds with Asp63, Ser60, Arg66, and Asn68, a pattern that closely overlaps with the native Ca²?-binding environment.

These interaction patterns demonstrate that all three candidates engage residues essential for Ca²? coordination within the EF-hand motifs, indicating that the calcium- binding interface of S100A4 can indeed accommodate small-molecule binders capable of competing with Ca²? for activation of the protein.

EF-Hand Motifs Are Valid and Druggable Targets on S100A4

This study investigated whether the Ca²?-binding EF-hand motifs of S100A4 can be targeted using small molecules. Through structural visualization, virtual screening, and interaction profiling, all three candidate molecules successfully localized within the EF-hand pockets, forming interactions that overlapped with residues required for Ca²? coordination. This confirms that the EF-hand region is structurally accessible and chemically capable of accommodating ligand binding.

These findings validate our central hypothesis: interfering with Ca²? coordination at the EF-hand site may prevent S100A4 activation and limit its downstream pro-metastatic signaling.

Differences in Binding Strength Reflect Distinct Inhibitor Profiles

Although each molecule occupied either the canonical or pseudo EF-hand environment, they displayed differing levels of engagement and stability. Molecule 1 showed the weakest stabilizing contacts and also carried predicted toxicity, limiting its suitability for further development despite modest interactions with Ca²?-coordinating residues. Molecule 2 adopted a stronger and better- oriented binding pose and demonstrated favourable toxicity predictions, although part of its interaction profile relied on weaker hydrophobic contacts that may affect stability within the pocket. Molecule 3 exhibited the most extensive and well-positioned interaction network, binding in close proximity to several key EF-hand residues necessary for Ca²? coordination.

Together, these profiles illustrate how ligand chemistry, such as hydrophobic surface area, polarity, and capacity for hydrogen bonding, influences inhibitors’ fit within the EF-hand loops.

Chemical Features Associated With Strong EF-Hand Engagement

Across the three compounds, several structural trends emerged as influential for successful EF-hand targeting. Molecule 3 demonstrated the most effective balance of polar and hydrophobic contacts, allowing it to sit close to central Ca²?-coordinating residues while still engaging surrounding side chains. This pattern suggests that a carefully tuned polarity is important, not just for forming specific interactions with acidic and polar residues in the EF-hand loop, but also for anchoring the ligand in a defined orientation rather than relying solely on nonspecific surface contacts. Molecule 2, by contrast, illustrated how increasing hydrophobic character can enhance occupation of the EF-hand pocket and may support membrane permeability, but at the cost of reduced solubility and a greater risk of off-target binding to other hydrophobic regions in the cell. Molecule 1, with its smaller size and higher polarity, was able to reach the binding site but lacked sufficient surface area and interaction diversity to establish a stable, high-quality binding network. Collectively, these observations suggest that effective EF-hand inhibitors will likely require an intermediate profile: enough polarity to engage key coordinating residues  through  hydrogen  bonds  and  electrostatic interactions, coupled with sufficient hydrophobic surface area to promote stable pocket occupancy and intracellular bioavailability, without tipping into poor solubility or nonspecific binding.

Implications for Targeting S100A4 to Limit Metastatic Progression

S100A4 contributes to EMT, cytoskeletal regulation, and metastatic behaviour in OSCC through a Ca²?-dependent conformational activation. Because the EF-hand motifs initiate this activation, small molecules that prevent Ca²? binding may block S100A4 at its earliest regulatory step, upstream of interactions with downstream partners. The ability of multiple distinct scaffolds to engage this region in silico supports the feasibility of this mechanism-based therapeutic approach. These results also suggest several avenues for refinement, including increasing interaction surface area, repositioning aromatic groups, or adjusting hydrogen-bonding motifs to improve selectivity and stability within the Ca²?-binding loops.

Limitations

Although virtual screening provides a rapid and accessible approach for evaluating small-molecule interactions, several limitations arise when Mcule docking is used as the primary tool for selecting inhibitory candidates. Mcule predicts ligand placement within a binding pocket to find favorable poses. However, these predictions do not fully capture the dynamic environment experienced by a ligand in a living system. Proteins such as S100A4 undergo dynamic changes in signalling, particularly in response to Ca²? binding. Mcule is not able to model these conformational changes with high accuracy, and therefore, the predicted poses may not reflect the true binding events occurring in vitro or in vivo.

A virtual screen that positions a molecule directly within the EF-hand does not guarantee that the ligand will successfully compete with calcium under physiological conditions. As a result, molecules that appear to bind effectively in silico may fail to displace Ca²? or may not engage with the protein when the cellular environment is considered. Even predicted interactions that appear structurally stable on the computer may not accurately be reflected in biological systems. The gap between computational predictions and biological interactions remains a source of uncertainty, and results generated through Mcule must therefore be interpreted as initial screens rather than validated outcomes.

Overall, while Mcule provides a useful first step in identifying small molecules capable of interacting with the EF-hand of S100A4, its predictions require experimental validation. Structural, biochemical, and cellular assays are necessary to determine whether the compounds truly bind in the presence of Ca²?, whether they maintain stability within the EF-hand pocket, and whether they display effective inhibition in a physiological context.

Future Directions

Future work should expand on the virtual screening to determine whether the candidate molecules produce pharmacological effects on S100A4 function in biological systems. In vitro experiments will be essential for validating the predicted inhibition of Ca²? binding. A calcium-detection assay could be used to quantify changes in intracellular or protein-associated Ca²? levels following treatment, allowing direct assessment of whether the compounds disrupt S100A4’s EF-hand–mediated calcium coordination. Such assays would confirm whether the interactions observed in silico correspond to functional outcomes in a cellular or biochemical context.

Additional structural refinement of the small-molecule candidates could also be pursued. Electrostatic surface potential analysis of S100A4 would provide insight into regions of negative or positive charge that could be exploited to improve ligand complementarity. This information could guide adjustments to hydrogen-bonding patterns, hydrophobic contacts, and overall molecular orientation to increase stability within the EF-hand pocket and strengthen affinity for the target site. These future directions allow for the development of inhibitors that are able to more effectively inhibit the activity of S100A4.

Together, these findings show that the Ca²?-binding EF- hand motifs of S100A4 constitute a viable and previously untapped therapeutic target. By integrating structural inspection with computational screening and interaction profiling, this study demonstrates that small molecules can reliably access the EF-hand pocket and engage residues essential for calcium coordination. Although the three candidate scaffolds differ in chemical behaviour, their shared ability to occupy the Ca²?-binding environment provides the first proof-of-concept that S100A4 can be inhibited at the level of its activation mechanism rather than at its downstream interaction surfaces. These results establish the foundational rationale for developing inhibitors that modulate S100A4 by preventing its calcium- dependent conformational shift. In this way, the work offers a conceptual framework that supports continued refinement and experimental validation, and underscores the broader therapeutic potential of targeting EF-hand– mediated activation in metastasis-associated proteins.

We would like to thank Dr. Murray Junop for his guidance throughout this project and for providing valuable insight into the structural biology of S100A4. We also thank Victoria Marie Marner for her assistance and helpful feedback during the development of this report.

1. Imbesi Bellantoni M, Picciolo G, Pirrotta I, Irrera N, Vaccaro M, Vaccaro F, et al. Oral Cavity Squamous Cell Carcinoma: An Update of the Pharmacological Treatment. Biomed. 2023; 11: 1112.
2. Natarajan J, Hunter K, Mutalik VS, Radhakrishnan R. Overexpression of S100A4 as a biomarker of metastasis and recurrence in oral squamous cell carcinoma. J Appl Oral Sci. 2014; 22: 426-433.
3. Barraclough R. Calcium-binding protein S100A4 in health and disease. Biochim Biophys Acta. 1998; 1448: 190-199.
4. Kim EJ, Helfman DM. Characterization of the metastasis-associated protein, S100A4. Roles of calcium binding and dimerization in cellular localization and interaction with myosin. J Biol Chem. 2003; 278: 30063-30073.
5. Boye K, Maelandsmo GM. S100A4 and metastasis: a small actor playing many roles. Am J Pathol. 2010; 176: 528-535.
6. Fei F, Qu J, Zhang M, Li Y, Zhang S. S100A4 in cancer progression and metastasis: A systematic review. Oncotarget. 2017; 8: 73219-73239.
7. The PyMOL Molecular Graphics System. Version 3.0 Schrodinger. LLC.
8. Kiss R, Sandor M, Szalai FA. A public web service for Drug Discovery. J Cheminform. 2012; 4: 17.
9. Schake P, Bolz SN, Linnemann K, Schroeder M. PLIP 2025: introducing protein-protein interactions to the protein-ligand interaction profiler. Nucleic Acids Res. 2025; 53: W463-W465.
10. D’Ambrosi N, Milani M, Apolloni S. S100A4 in the Physiology and Pathology of the Central and Peripheral Nervous System. Cells. 2021; 10: 798.
11. Pathuri P, Vogeley L, Luecke H. Crystal structure of metastasis- associated protein S100A4 in the active calcium-bound form. J Mol Biol. 2008; 383: 62-77.