Peptide immobilization was done via a copper-catalyzed cycloaddition reaction, by coupling the ethynyl group of the C-terminal propargylglycine residue to azide-modified SPR gold chips (Xantec AZD-50 L) (Fig. S1). These chips feature a 50 nm thick, low-coverage azide-derivatized carboxymethyldextran hydrogel layer that provides the mobility and flexibility required to ensure high accessibility for the conjugated peptide combined with low nonspecific binding. Although copper-catalyzed click couplings are widely used in organic synthesis and bioconjugations, they are not standard immobilization methods for SPR and, therefore, need optimization. To this end, we conducted pH scouting experiments to determine the optimal acidity conditions for maximizing peptide immobilization efficacy (Supporting Information). Our findings revealed that optimal conditions for peptide immobilization are achieved at pH values near the isoelectric point (pI) of each peptide. This aligns with previous reports12 and can be explained by electrostatic considerations. Whereas a higher pH will promote electrostatic repulsion between the negatively charged peptides and the polyanionic carboxylated dextran 3D matrix, lowering the pH below the pKa of the surface carboxylic acids will suppress binding due to a decrease in surface charge density. Acetate buffer was chosen to conduct the immobilization assays to avoid copper sulfate precipitation in the sensitive microfluidic system of the SPR instrument. For binary mixtures (a1/rc3 and a2/rc3), a 1:1 ratio of the respective peptides was prepared in acetate buffer at pH 5. After pH scouting experiments were conducted, it was found that a pH that lies in the middle of the optimal pH range for the individual peptides provided the best results in terms of immobilization efficiency (Figure S2). Using optimized conditions, the peptides were immobilized individually or in binary mixtures, and the immobilization ratio was estimated from the change in response units (RUs) as detailed in Table 1.
Our results showed that immobilization of a2 resulted in the highest surface coverage, approximately 10 times higher than a1 and nearly twice that of rc3 (Supporting Information). For the dual epitope combinations, the highest immobilization efficacy was achieved by the a1/rc3, with an 8-fold increase compared to the single immobilization of a1. On the other hand, the a2/rc3 mixture resulted in a diminished immobilization performance compared with a2. The differences in immobilization efficiency between single- and double-peptide immobilizations may be influenced by the different isoelectric points, molecular weights, and intermolecular interactions between the mixed peptides. This last aspect was examined from molecular dynamics simulations in binary mixtures (vide infra).
Binding affinity of single peptide sensors to RBD variants
The single-peptide sensors were tested for their binding affinity to the receptor-binding domain (RBD) of three β-coronavirus proteins, namely SARS-CoV-2 Alpha, SARS-CoV-2 Delta, and SARS-CoV-1, using multi-cycle kinetic experiments. In these tests, multiple analyte injections are performed over the same sensor surface, and after each injection, the surface is regenerated to remove any remaining analyte or complex, allowing for a fresh interaction in subsequent cycles. This approach is useful when the binding kinetics are complex. Our multi-cycle kinetic experiments used five sequential RBD concentrations ranging from 3.1 to 100 nM. Figure 2 shows the SPR sensorgrams for the interaction between the β-type corona RBD variants and the single peptide sensors obtained from a1 and a2.
Representative SPR sensorgrams for the binding of three β-type coronavirus RBD variants (SARS-CoV-2 Alpha, SARS-CoV-2 Delta, and SARS-CoV-1) to the immobilized peptides a1 and a2. Fitted curves are inserted as dashed lines.
Sensorgrams were recorded and were globally fitted to a 1:1 Langmuir model12. Fit quality was assessed primarily by the Chi² values reported by BIAevaluation and complemented by calculating the root mean squared error (RMSE) in response units (RU) and the coefficient of determination (R²) from the exported traces (Table S6). The 1:1 Langmuir kinetic model was chosen and kept consistent for all sensograms because it provides the standard framework for describing the peptide–protein interactions across all variants and sensor surfaces. We note, however, that the 1:1 model alone has inherent limitations, as it does not explicitly account for mass-transport effects at higher ligand densities, heterogeneity in peptide presentation, or possible rebinding events during dissociation. These limitations were not apparent for the Alpha variant, where the fitted curves closely overlapped with the experimental data. By contrast, the Delta-variant and SARS-CoV-1 sensograms adheared less well to the model suggesting a more heterogeneous interaction with the peptide surface. These deviations were most pronounced on the a2 sensor, which is possibly linked to the higher immobilization densities of these sensors (Table 1) that in turn could impact ligand conformation and access. Despite these modest discrepancies, the overall χ² values remained low, and the extracted kinetic parameters (Table S7) and KD ranking (Table 2) were robust and in agreement with the molecular dynamics analysis.
The kinetic data and individual dissociation constants for each run are specified in the Supporting Information. A comparative analysis of the binding interactions between the a1- or a2-based sensors and various RBD variants revealed concentration-dependent analytical responses and KD values in the nanomolar range (16–89 nM) but with significant differences between the sensors. The highest affinity was observed using the a1 sensor, which showed a binding preference in the order SARS-CoV-2 Delta > SARS-CoV-2 Alpha > SARS-CoV-1. The tighter binding of the Delta variant aligns with other reports and reflects the selection pressure towards mutants displaying a higher affinity for ACE2 and, in turn, infectivity. For a2, the relative binding affinities between the SARS-CoV-1 and SARS-CoV-2 variants were reversed with SARS-CoV-1 now showing the strongest binding with a KD of 21 nM. Among the SARS-CoV-2 variants, the Delta RBD displayed the highest affinity also in this case. Data corresponding to rc3 is provided as Supporting Information, as only weak interactions were detected between the sensors containing this peptide and all RBDs.
Binding affinity of single peptide sensors to the full Spike protein of SARS-CoV-2
Additional SPR experiments were conducted to test the capacity of the immobilized peptides (a1, a2, and rc3) to recognize the full spike protein of SARS-CoV-2 (Fig. 3; Table 2, Table S8,). Our findings revealed that a1 binds to the full spike protein with detectable SPR signals even at low nanomolar concentrations, yielding a KD of 1.2 nM and a RUmax of 605, which indicates a very high affinity and significant binding capacity. The measured dissociation constant is significantly lower than the estimated values for other sensors targeting the SARS-CoV-2 spike protein and comparable to those of available antibodies for this protein. Additionally, the interaction with the full spike protein is stronger than the interaction with the RBDs, possibly reflecting multivalent interactions due to the trimeric nature of the protein (Table 2). Turning to a2, this epitope showed a KD of 19 nM with the full spike protein of SARS-CoV-2 and an RUmax of 123, reflecting weaker binding than a1, although again stronger than that observed between a2 and SARS-CoV-2 Alpha RBD. Again, the rc3 peptide did not show measurable signals attributable to effective interactions with the spike protein, as expected from its negligible interaction with the RBD (data provided as Supporting Information). The selectivity of a1 and a2 for SARS-CoV-2 was evaluated by testing the binding of the immobilized epitopes to the spike protein of HCoV-NL63, an a-type coronaviral variant that primarily targets the upper respiratory tract (URT) and induces milder symptoms in comparison to beta coronaviruses such as fever, cough, runny nose, and difficulty breathing13,14,15,16,17. SPR experiments confirmed the absence of measurable signals for the interaction of a1 and a2 with this protein, underscoring the selectivity of the tested epitopes for β-type coronaviruses.
Representative SPR sensorgrams for the binding of the full Spike proteins of SARS-CoV-2 and HCoV-NL63 on sensors containing the immobilized ACE2 peptides a1 and a2. Fitted curves are inserted as dashed lines.
Binding affinity of dual epitope sensors SARS-CoV-1 RBD
Noticing the different binding preferences of the single epitope receptors, we investigated dual epitope systems to explore whether epitope combinations could also influence the variant selectivity (Fig. 4; Table 3, Table S9). Binary mixtures a1/rc3 and a2/rc3 were immobilized on SPR gold chips using equimolar peptide mixtures in acetate buffer (pH 5) to combine the small, random-coil conformation epitope rc3 with the longer epitopes from the primary ACE2 binding region. SPR experiments revealed that the dual sensors achieved remarkable improvements in binding affinity for SARS-CoV-1 RBD (Fig. 4; Table 3). Notably, the dual sensor a1/rc3 showed a KD of 6 nM, marking a 9-fold enhancement compared to the single a1 sensor, while the sensor a2/rc3 with a KD= 21 nM was not affected by the presence of rc3.
SPR sensorgrams for the binding of three β-type coronaviral variants (SARS-CoV-2 Alpha, SARS-CoV-2 Delta and SARS-CoV-1) RBDs with mixed epitopes a1 and a2 with rc3. Fitted curves are inserted as dotted lines.
For SARS-CoV-2 Alpha RBD, a1/rc3 and a2/rc3 yielded KD values of 17 nM and 22 nM, respectively, representing 1.7-fold and 4-fold improvements over their single counterparts (KD= 29 nM for a1 and KD = 89 nM for a2). Interestingly, while single sensors based on a1 exhibited the highest affinity for SARS-CoV-2 Delta and the weakest for SARS-CoV-1 (KD = 16 nM for Delta and 55 nM for SARS-CoV-1), the dual sensors reversed this trend.
These results underscore the cooperative advantage provided by dual sensors in improving affinity, particularly for challenging targets. The a1/rc3 dual epitopes consistently demonstrated superior binding affinity across tested variants, while a2/rc3 showed dramatic improvements relative to single a2. The reversal in binding preference between variants highlights how epitope combination influences binding behavior, potentially offering a strategy to tune sensor selectivity. Collectively, these findings support the potential of multivalent interactions in optimizing receptor-based detection and targeting strategies.
Molecular dynamics simulations
Molecular Dynamics (MD) simulations were conducted to gain molecular-level insight into our SPR findings regarding the binding affinity of single and dual epitope sensors to β-type coronaviral RBD variants. First, we examined the binding properties of the three peptide epitopes (a1, a2, and rc3) to the RBDs of SARS-CoV-2 Alpha, SARS-CoV-2 Delta, and SARS-CoV-1 to evaluate whether the isolated peptides, modified at the C-termini with the linker moiety that connects the amino acid sequence to the hydrogel layer in the SPR sensor (Figure S3), maintain their conformation and recognition capacity for the binding domains. An a-type variant (HCoV-NL63) was also modeled as a negative control to evaluate the peptides’ selectivity toward β-coronavirus.
Trajectory analysis revealed that the peptides explore distinct conformations and binding regions on the RBDs. Only a1 maintained binding modes and conformations similar to those of the epitope in ACE2. In contrast, a2 and rc3 showed greater mobility and bound to diverse regions on the RBD, including areas not available for peptide interaction in the full-length spike protein context (Figure S4). Figure 5A provides a pictorial summary of the simulation results, showing the most frequently observed association regions for each of the studied systems. Our results indicate that a1 binds to the recognition region in all β-type RBDs, whereas a2 recognizes this region only in the complex with SARS-CoV-1 RBD. Peptide RMSD calculations in protein-aligned MD trajectories align with these observations, showing that a1 has very low mobility in its complexes with SARS-CoV-2 Alpha and SARS-CoV-2 Delta RBDs (Fig. 5B). This suggests the formation of stable interactions with these receptors, which is consistent with the lower affinity constants measured for these systems by SPR (Table 2). For a2, the lowest peptide mobility is observed in the complex with SARS-CoV-1 RBD, which also aligns with our experimental results. For rc3, large mobilities are observed in all b-type RBDs, evidencing loose contacts with the proteins. High mobilities were also observed in all three complexes with hCoV-NL63, which is consistent with the minimal binding affinity of the peptides for this receptor detected by SPR (Fig. 3). Additionally, we evaluated the occupancy of the peptide-RBD complexes to assess the permanence of intermolecular contacts in each system (Fig. 5C). Occupancy is defined as the fraction of MD frames in which the peptide interacts with the protein through at least one intermolecular contact at a distance equal to or less than 5 Å. Complexes with a1 and a2 have occupancies greater than 80% across all three β-types RBDs, while rc3 shows significantly lower complex occupancies (< 40%). This aligns with the minimal SPR response observed experimentally for complexes formed by this latter peptide (Supporting Information).
A Pictorial representation of the preferred interaction regions for the peptide-RBD complexes formed by a1, a2, and rc3 with the RBD domains of SARS-CoV-2 Alpha, SARS-CoV-2 Delta, and SARS-CoV-1 and hCoV-NL63. Images were obtained from 500 ns MD trajectories for single-peptide complexes with the RBDs, considering the most frequent regions of interaction for each peptide. The structures of the complexes for each RBD are superimposed to illustrate the distinct binding modes exhibited by each peptide. B RMSD distribution data for peptides a1, a2, and rc3 in single-peptide complexes with the RBDs of SARS-CoV-2 Alpha, SARS-CoV-2 Delta, and SARS-CoV-1 and hCoV-NL63. C Occupancy of the peptide-protein complexes formed by a1, a2, and rc3 and the RBDs tested in this study. D Binding free energy estimates (kcal/mol) for the peptide-RBD complexes addressed in this study. MM/GBSA calculations were carried out on 1000 MD frames retrieved from 500 ns MD simulations.
To estimate the strength of the peptide-RBD interactions, we used Molecular Mechanics/Generalized Born Surface Area (MM/GBSA) calculations. MM/GBSA is a well-established computational technique to estimate the binding free energy of a supramolecular complex18,19,20,21,22. The free energy difference between the bound and unbound states is calculated using an end-point strategy from the contributions of the molecular mechanics energy (comprising both intra- and intermolecular components) and the solvation free energy (which includes both polar and non-polar terms). While the entropy of the process can be calculated, it’s often omitted under the assumption that systems of similar size and structure, like our single-peptide complexes with various RBDs, should have comparable entropic contributions. While not as rigorous as more computationally demanding methods like alchemical free energy calculations, MM/GBSA offers a powerful balance of computational efficiency and predictive accuracy, making it a valuable tool for understanding and predicting the interactions within peptide-protein systems. Figure 5D summarizes the results of MM/GBSA binding free energy estimates for the series of peptide-RBD complexes under study. For a1, the MM/GBSA results show a higher affinity for the SARS-CoV-2 Alpha and Delta variants, which align with our SPR experiments. For a2, the results are also consistent with the experimental data, showing a preference for association with the SARS-CoV-1 RBD. Additionally, the peptides show negligible interaction energies with the RBD of hCoV-NL63, accounting for their selectivity to b-type variants. The qualitative agreement between our theoretical and experimental results suggests that our computational models, which are based on interactions between free peptides and receptors, reflect the interactions that would occur in the context of the SPR sensor. This is likely due to the high conformational flexibility of the polymer chains that link the peptide to the chip’s metallic surface, which allows the immobilized peptide to interact with the receptor in a manner very similar to its unbound state.
For a1, the only peptide that binds the RBDs in the region targeted by ACE2, we compared the intermolecular contact maps obtained from MD simulations with the structural models taken as reference in this study (6M17, 2AJF, and 7TEW). Heatmaps showing the frequency of a1-RBD contacts throughout the simulated MD trajectories are displayed in Fig. 6A. The residues that are part of the interaction site in the structural models (considering a distance cutoff of 5 Å) are highlighted under each heatmap and displayed in colored surfaces in Fig. 6B. The contact maps show that a1 interacts with a conserved region of the RBD domains that is also involved in ACE2 recognition. For the Alpha variant, this region includes L455, A475, Y489, and Y495, which are relevant for the receptor’s binding to the virus. These findings highlight the potential of a1 to mimic the intermolecular recognition properties of the natural receptor ACE2 toward b-type coronaviruses, which is a relevant outcome to support the design of biomimetic sensors based on this epitope.
A Interaction heatmaps for complexes formed by a1 and the RBD domains of SARS-CoV-2 Alpha, SARS-CoV-2 Delta, and SARS-CoV-1. Highlighted residues correspond to the recognition region for the natural receptor in the crystallographic structures 6M17, 2AJF and 7TEW, considering a distance cutoff of 5 Å. B Snapshots of the MD trajectories in which the recognition regions of the natural receptor are displayed as colored surfaces (blue: SARS-CoV-2 Alpha, green: SARS-CoV-2 Delta, red: SARS-CoV-1) and a1 is represented as colored ribbons.
For our dual peptide mixtures, we first conducted MD simulations to examine potential peptide-peptide interactions that might occur on the surface of SPR receptors. We reasoned that rc3 could not alter the recognition of the RBDs in dual mixtures due to its negligible individual binding capacity. Therefore, the observed variations in RBD affinity for the dual-peptide sensors might stem from a cooperative effect driven by the interaction between the peptides. While this is a tentative approach, our experimental setup allows us to assume that such interactions are indeed possible, given the length and flexibility of the polymer chains attached to the peptides and the high immobilization efficiency achieved for the dual-peptide sensors (Table 1). We investigated the potential interaction between peptides using MD simulations of binary a1/rc3 and a2/rc3 systems, comparing the behavior of the mixture to that of the individual peptides. Our results indicate that rc3 interacts with both a1 and a2, forming transient contacts that stabilize the α-helix secondary structures in the longer peptides (Fig. 7). This stabilizing effect is more pronounced in the a1/rc3 system, particularly in the region spanning amino acids 16–30. This conformational restriction could affect the interaction with RBDs by either reducing the entropic cost of achieving the helical structure needed for an effective interaction with the recognition region or by favoring stabilizing/destabilizing contacts with the surface of the target protein.
Analysis of secondary structures for a1 and a2 as individual peptides and in binary mixtures with rc3. Data was obtained from 500 ns MD simulations.
To further attempt to understand the effect of the peptide mixture, we performed additional MD simulations on RBD complexes with the a1/rc3 and a2/rc3 peptide mixtures. The results show a marked reduction in peptide mobility compared to that observed in the corresponding single-peptide complexes, mediated by inter-peptide interactions on the RBD recognition region (Figure S5). The distribution of peptide RMSD in the dual complexes confirms this observation, suggesting a favorable effect on the stabilization of peptide-RBD contacts compared to complexes with isolated peptides (Fig. 8A). On the other hand, MM/GBSA binding free energies only suggest a moderate stabilization in the complex formed by the a1/rc3 mixture and the SARS-CoV-1 RBD (Fig. 8B). For the other complexes, no significant variations in binding free energy were observed compared to single-peptide complexes. In this regard, it is worth noting that the MM/GBSA method only captures the enthalpic components of the peptide-RBD association. A more exhaustive approximation of the global recognition process would require addressing the entropic contribution. That said, the lower calculated interaction energy for the complex between a1/rc3 and the SARS-CoV-1 RBD could account for a strengthening of the enthalpic terms responsible for RBD recognition. This could be further enhanced by a decrease in the entropic cost necessary for the peptides to adopt the required conformation and secondary structure for an effective interaction with the RBD’s recognition region.
A RMSD distributions for a1/rc3 and a2/rc3 peptides throughout 500 ns MD trajectories in dual-peptide complexes with the RBD of SARS-CoV-2 Alpha, SARS-CoV-2 Delta, and SARS-CoV-1. B MM/GBSA binding free energy estimates (kcal/mol) for the interaction between the a1/rc3 and a2/rc3 mixtures with the RBD of SARS-CoV-2 Alpha, SARS-CoV-2 Delta, and SARS-CoV-1 calculated from 500 ns MD simulations.