Exploring the energy landscape of a molecular engineered analog of a tumor-homing peptide

Abstract

Recently a new non-coded amino acid was designed as a replacement for Arg, to protect the tumor-homing pentapeptide CREKA (Cys-Arg-Glu-Lys-Ala) from proteases. This constrained Arg analog, denoted c₅Arg, was engineered to also promote the stability of the CREKA bioactive conformation. The conformational profile of the CREKA analog obtained by replacing Arg by c₅Arg has been extensively investigated in this work. Two molecular dynamics simulations-based strategies have been employed: a modified simulated annealing and replica exchange. Results obtained using both techniques show that the conformational features of the new analog fulfill the purpose of its design. The new CREKA analog not only preserves the main structural attributes found for the bioactive conformation of the parent peptide but also shows lower flexibility. Moreover, the conformational profile of the mutated peptide narrows towards the most stable structures previously observed for the parent CREKA peptide.

Introduction

The systematic use of nanostructures for clinical applications is rapidly becoming a central milestone in modern nanomedicine. Among the possible nano-approaches, treatment systems based on nanoparticles able to target tumor cells rapidly took the lead.¹ Even though the first generation of nanoparticles relied on vessel “leakiness” for preferential accumulation in tumors—which limited the efficiency of such particles because of their low penetration capability into the targeted tumors—they were trapped by the sustained high pressure of the interstitial tumor fluid.²

An alternative approach is targeting the nanoparticles to specific molecular receptors in blood vessels: tumors express many molecules that are not significantly populated in normal tissues and the receptors are available for direct binding of species in the blood stream.^3–5 Different nanosystems have been proposed, such as the super-paramagnetic iron oxide (SPIO) nanoparticles coated with dextran, by Ruoslahti and co-workers. These nanoparticles are able to home to tumors while amplifying their homing activity through a mechanism that resembles formation of blood clots by platelets.⁶ The bioactive part of these nanoparticles is the coating peptide, which can recognize selectively clotted plasma proteins.⁶ This short linear peptide (CREKA, Cys-Arg-Glu-Lys-Ala) was recently discovered by in vivo screening of phage-display peptide libraries^3,7 for tumor-homing peptides in tumor-bearing MMTV-PyMT transgenic breast cancer mice.⁸ Synthesized CREKA labeled with the fluorescent dye 5(6)-carboxyfluorescein (FMA) was detectable in human tumors from minutes to hours after intravenous injection, whereas it was essentially undetectable in normal tissues.⁹

For the last three years we have devoted our efforts to devise protection strategies of CREKA against proteolytic cleavage with concomitant stabilization of the bioactive folded conformation.^10–12 Following determination of the bioactive conformation, specific replacements were designed using non-coded amino acids. This strategy is expected to grant proteolytic protection and, simultaneously, allow control of the conformational preferences. Exploration of CREKA conformational preferences was performed under conditions in which the homing peptide was experimentally shown to bind to the clotted plasma proteins: as a free peptide, attached to a nanoparticle, and inserted into a viral capsid protein (in a phage display library).¹⁰ Independently of the simulation conditions, the most favored organizations often featured turn motifs, which generally involved Cys and either Glu or Lys. Among this variety of turn organizations, a common pattern was observed in all simulated cases: a β-turn arrangement that involves the formation of a hydrogen bond between Cys and Lys residues. This organization was identified as the bioactive motif, which enabled the formation of salt bridges between the side-chains of all ionized amino acids, since the curvature site made them protrude outward. Once the bioactive structural profile was delineated, non-coded amino acids were used in order to enhance the protection against proteolytic cleavage and the homing activity.¹³

A key question relates to the final choice of the non-coded amino acid in the substituted position which should satisfy the bioactive conformational requirements. Examination of the conformation adopted by Arg in the bioactive organization shows that it is located at position i + 1 of a β-turn.¹⁰ If one searches the conformational preferences of small peptides and proteins, this position is usually occupied by Pro.^14–17 Hence, initially, a new amino acid with a Pro scaffold and Arg side chain, named (^γPro)Arg, was modeled.¹² Subsequent work¹² showed that inclusion of this residue significantly reduced the conformational flexibility of the peptide, enhancing its tendency to accommodate partially folded conformations centered at the Arg site. In addition, inspection of the lowest energy conformation indicated that the ionized side chains faced the same side of the molecule, a feature that seems to have a role in the peptide bioactivity. However, insertion of this Arg surrogate presented drawbacks: despite the similarities described above, the turn conformations adopted when (^γPro)Arg replaced Arg did not match the most favored conformations previously described,¹⁰ since it shifted the conformational preferences of the CREKA analog towards γ-turn formation. Therefore, the search for appropriate amino acid backbones continued.

Within that context, a specific family of non-coded amino acids, the 1-aminocycloalkane-1-carboxylic acids^18,19 (known in the abbreviated form as Ac_nc, with n referring to the ring size), have been recently proposed as a structural tool for conformational control and enhancement of naturally occurring structural motifs.^19–23 This series of residues exhibits a restricted conformational space characterized by a high propensity to adopt φ, ψ backbone angles typical of the 3₁₀-/α-helix (with some distortion in the case of Ac₃c).^23–25 As we previously mentioned, in the majority of the most favored CREKA conformations Arg is at position i + 1 of a β-turn motif.¹⁰ In such arrangements, the backbone dihedral angles adopted by the residue are in the regions corresponding to the aforementioned helical structures. Hence, in order to take advantage of Ac_nc conformational preferences, a new amino acid was designed by attaching the Arg functionalized side-chain to the cycloalkane moiety in Ac₅c, the resulting compound being denoted c₅Arg (see Scheme 1).²⁶ c₅Arg is a substituted Ac₅c-like derivative of nor-arginine, where nor refers to a chain-length reduction of one carbon atom with respect to the Arg side-chain.²⁶ A complete conformational characterization showed that the new residue preserved the conformational preferences corresponding to those expected for its Ac₅c skeleton,²⁵ which makes it suitable for tests as a CREKA homing peptide-enhancing modification.

Scheme 1.

This work reports the conformational preferences of a new CREKA analog, hereafter denoted Cc₅REKA, obtained by substituting Arg by the cis enantiomer of c₅Arg, where cis refers to the relative position of the guanidinium group in the cycloalkyl segment. Specifically, results obtained in previous quantum mechanical calculations on both the cis and trans isomers of c₅Arg suggest that, a priori, the former is the best to fulfill the structural requirements of the peptide.²⁶

The energy landscape of the new peptide analog has been explored by techniques based on molecular dynamics (MD) simulations, i.e., modified simulated annealing MD (SA-MD)²⁷ and replica exchange MD (REMD).²⁸ The conformational features identified in Cc₅REKA are compared with those previously reported of CREKA¹⁰ and the advantages of using c₅Arg as Arg replacement are inferred from such a comparison.

Methods

Conformational search and force-field calculations

The conformational preferences of Cc₅REKA were explored using both modified SA-MD and REMD sampling strategies. The first methodology, which was previously used to study the conformational preferences of the parent peptide,^10,11 is based on the minimization of conformations generated at the initial and intermediate states of several SA-MD cycles. In the SA-MD technique the high starting temperature is gradually reduced during the simulation, allowing the system to surmount energy barriers. In spite of this, in practice, it is known that such a sampling technique does not always lead the system to the most stable region at the end of the simulation. However, recent studies showed that this limitation can be overcome, since very low energy structures can be reached by minimizing structures generated during the SA-MD cycles.²⁷ Accordingly, such modification of this sampling technique was found to be robust enough to obtain conformations close in energy to the global minimum but located in different regions of the potential energy hypersurface, when the extracted conformations from each SA-MD cycle are minimized.^10–12

The same conformational profile was studied using REMD.²⁸ This MD based technique speeds convergence relative to brute force conventional MD.²⁷ The method is based on the generation of a number of copies (“replicas”) of the system that span from the temperature of interest (e.g., physiological temperature) to heated states, which facilitates overcoming the free energy barriers. Periodic swaps of neighboring replicas, which are performed while preserving an overall Boltzmann-weighted ensemble at each temperature (Monte Carlo based criterion of swap acceptance), enable conformations to heat up and cool down. Nonetheless, recent work has indicated that for short peptides the optimum convergence of the sampling can be obstructed by a combination of two factors:²⁹ the intrinsically high flexibility of these systems and the limited thermal agitation associated with the reference temperature. In this work we show that under normal conditions flexible systems may not reach some conformations of medium energy rank, affecting significantly the final bioactive ensemble of conformations.

Computational details

Energy calculation.

The conformational energy for all the simulated systems was calculated using the Amber force field.³⁰ All bonding and non-bonding parameters were extracted from Amber libraries³⁰ except for those describing the non-coded residue which were previously optimized and tested.²⁶

Molecular models.

The simulated system consisted of the Cc₅REKA peptide attached to a surface through the sulfhydryl group of the Cys residue. The N-and C-termini of the peptide backbone were capped with acetyl (Ac, MeCO−) and methylamide (−NHMe) groups, respectively. The surface was formed by 100 spherical particles distributed in a 10 × 10 square (45.90 × 45.90 Ų), with van der Waals parameters R = 2.35 Å and ε = 0.90 kcal mol⁻¹ and no electric charge. This system, which is identical to that considered for CREKA in our previous work,¹⁰ mimics the experimental conditions,^3,7 i.e. the peptide linked to the surface of a nanoparticle. Therefore, the results obtained in this work for the CREKA analog have been compared with those reported for natural CREKA attached to an identical surface (system III in ref. 5), unless otherwise indicated. The Cc₅REKA analog attached to the surface was placed in the center of a cubic simulation box (a = 45.90 Å) filled with 2663 explicit water molecules, which were represented using the TIP3 model.³¹ Two chloride ions and one sodium ion were added to the simulation box to reach electric neutrality (net charges were considered for Arg, Lys, and Glu at neutral pH).

Simulated annealing

Prior to the production cycles with the modified SA-MD, the system was equilibrated. 0.5 ns of NVT-MD at 500 K were used to homogeneously distribute the solvent and ions in the box. Next, thermal equilibration for 0.5 ns in the constant-NVT ensemble at 298 K, followed by density relaxation for 0.5 ns in the constant-NPT ensemble at 298 K were performed. The last snapshot of the NPT-MD was used as the starting point for the conformational search process. This initial structure was quickly heated to 900 K at a rate of 50 K ps⁻¹ to force the molecule to jump to a different region of the conformational space. Along 10 ns, the 900 K structure was slowly cooled to 500 K at a rate of 1 K ps⁻¹. A total of 500 structures were selected and subsequently minimized during the first cycle of modified SA-MD. The resulting minimum energy conformations were stored in a rank-ordered library of low energy structures. The lowest energy structure generated in a modified SA-MD cycle was used as starting conformation of the next cycle.

Atom pair distance cut-offs were applied at 14.0 Å to compute the van der Waals and electrostatic interactions. In order to avoid discontinuities in the potential energy function, non-bonding energy terms were forced to slowly converge to zero, by applying a smoothing factor from a distance of 12.0 Å. Both temperature and pressure were controlled by the weak coupling method, the Berendsen thermobarostat,³² using a time constant for heat bath coupling and a pressure relaxation time of 1 ps. Bond lengths were constrained using the SHAKE algorithm³³ with a numerical integration step of 2 fs. All MD simulations were performed using the NAMD program.³⁴

Replica exchange molecular dynamics

Simulations were performed using the AMBER 10 program.³⁵ In order to use the low complexity of the system to ensure a reasonable replica swaps ratio of acceptance (approaching a Partial Replica Exchange Molecular Dynamics model), 8 replicas were exponentially distributed in the temperature range from 283.8 K to 418.7 K.³⁶ Exchanges were attempted every 40 ps between all neighboring replicas with an average acceptance rate of 16%, above the minimal acceptance rate for the complexity of the studied peptide.³⁶ The REMD trajectories resulted in a cumulative simulation time of 41 ns. Between replica exchanges, the system was evolved using NVT Langevin MD³⁷ with a damping coefficient of γ = 2.5 ps⁻¹ and an integration step of 2 fs. The replicas were previously equilibrated by a set of short runs (isothermal and isobaric equilibration), and completed with a final NVT run of 0.5 ns to ensure that each replica reached the target temperature. In all cases the surface particles were fixed at the initial positions and only peptide, water molecules, and ions were allowed to move. Atom pair distance cut-offs were applied at 14.0 Å to compute the van der Waals and electrostatic interactions. Bond lengths involving hydrogen atoms were constrained using the SHAKE algorithm³³ and explicit water molecules of the TIP3 model³¹ were used in every replica except for evaluating the replicas energies before the swapping process. In order to increase the global efficiency of the technique, a hybrid solvent model was used to evaluate each replica total energy.³⁸ Specifically, 118 explicit water molecules were taken into account as a first hydration shell while the rest of solvent contribution was computed using Generalized Born Model implemented by Hawkins and coworkers,³⁹ using the late parametrization of Tsui and Case.⁴⁰ The REMD data were collected from the last 10 ns of simulation at the targeted temperature of 300 K to perform clustering analysis.

Conformation classification and clustering analysis

In order to construct a list of unique minimum energy conformations, each set of structures provided by either modified SA-MD or REMD were compared among them. The list was organized by rank ordering all the unique minimum energy conformations found following an increasing order of energy. In the case of SA-MD, previously listed conformations that appear at a new cycle were discarded. The criterion to identify unique minimum energy conformations was already developed in previous work, and it is based on defining virtual dihedral angles combined with the computation of the interaction pattern, i.e. identification of salt bridges, hydrogen bonds and dipole–dipole interactions.¹⁰

Five virtual dihedral angles were defined considering the α-carbon atoms of the five residues, the methyl carbon atom of the acetyl and N-methylamine capping groups, and one acetyl hydrogen atom. The existence of different interactions is accepted on the basis of the following geometric criteria: (a) salt bridges: distance between the centers of the interacting groups shorter than 4.50 Å; (b) hydrogen bonds: H⋯O distance (d_H⋯O) shorter than 2.50 Å and angle higher than 120.0°; and (c) dipole–dipole: distance between dipoles shorter than 3.00 Å and the interaction has not been counted as a hydrogen bond. Two structures were considered different if they differ in at least one of their virtual dihedral angles by more than 60° or in at least one of the interactions counted. All the structures categorized as different were subsequently clustered according to a criterion based on the presence of the intramolecular interactions mentioned above.

Results and discussion

Clustering analysis

Following the previously developed analysis strategy, we classified the microstructures generated by both SA-MD and REMD following a hierarchical clustering.¹⁰ In this process, each structure is compared with the rest of the structures obtained using the same strategy and included in a cluster that presents the same main-chain conformation and the same polar interaction types. This analysis provides important information about the population of conformations generated for the studied system and, additionally, it serves as a simple and efficient criterion of convergence for SA-MD (i.e. when no new conformations are obtained in the annealing cycles the search is concluded). In Fig. 1 the convergence of the exploration clearly shows that after 9 cycles the number of new conformations does not increase. Moreover, the number of different structures produced for the peptide analog, 1125, is close to the number of different structures described for the wild type peptide, 1306, after an equivalent number of SA-MD cycles.¹⁰ However, the significant reduction of the number of different minima, 14%, seems to indicate that the replacement has succeeded, at least in terms of reduction of the conformational flexibility. On the other hand, much faster convergence was reached by REMD considering the same criterion (i.e. the number of newly explored conformations), a plateau being obtained after only 8 ns of cumulative simulation time, and the maximum number of explored conformations was found to be only 550. Below, we further show that this relatively small number of different structures may be enough to represent the most relevant conformational features of the bioactive organization for the Cc₅REKA peptide.

Fig. 1.

Variation of the number of unique minimum energy conformations found for Cc₅REKA against the number of conformations generated using the SA-MD (red line, solid circles) and REMD (blue line, solid squares) exploration methods.

Although the main goal of the current study was assessing the suitability of c₅Arg as a homing enhancer, comparison between two different conformational search strategies for short peptides is necessary. Despite the general perception that REMD is the most suitable strategy to explore the conformational space of any given molecule, recent works have pointed out the flaws of the method when dealing with highly flexible peptides.⁴¹ These limitations were attributed to the inherent efficiency of REMD for locating deep basins. Thus, when the energy profile that separates the lowest energy structure and other low energy arrangements is not sufficiently rough, the exploration can be confined to those structures of lower conformation energy. In this study, the rapid convergence is apparently reached by biasing the conformational ensemble towards those structures that presented lower potential energy.

Fig. 2, which depicts the energy distribution of all the generated structures, shows a Gaussian and a bimodal function for the SA-MD and REMD methods, respectively. These profiles indicate that for REMD the lower energy basins are overpopulated while zones that might represent intermediate potential energy ranges are not explored. Despite the apparent independent evolution of each trajectory, these results show that replicas close to the target temperature may present some degree of structural correlation. Thus, in cases of very flexible peptides longer cumulative time should be used to complete a proper exploration of the conformation space (i.e., higher computational effort). Taking into account that only the last 10 ns of simulation were collected to obtain statistical information about the conformational distribution at the given temperature, the computational resources used by REMD for obtaining 40 ns of cumulative simulated time were significantly larger than those committed for reaching the search convergence using SA-MD. It is worth noting that in our case this methodology has allowed us to obtain an energy distribution closer to a canonical ensemble than the one reached by REMD. The main difference between the two methods is the introduction of the temperature as a variable in the conformational search process, which imposes a heavier thermal hindrance for reaching those conformations that are kinetically less favored. As we show later, the first peak of the bimodal distribution derived from the REMD results corresponds to those conformations that are similar to the wild type bioactive motif, while the second peak corresponds to organizations that do not resemble those of lower energy. The central void in the population of medium energy structures may denote kinetic traps that preclude the formation of transient structures at 300 K. These structures, which are needed to reach the conformations that belong to the lowest energy rank, are only described by the SA-MD method.

Fig. 2.

Distribution of energies for the unique minimum energy conformations found for Cc₅REKA using SA-MD (red line) and REMD (blue line). The profile previously obtained for CREKA (ref. 10) using SA-MD (green line) is displayed for comparison. Energies are relative to the corresponding global minimum.

It was previously demonstrated that the bioactive profile of CREKA mainly corresponds to the lowest energy conformations.¹² Because both the brute force-based simulations (SA-MD) and the statistically rigorous exploration (REMD) techniques drove toward the most energetically relevant structures, they should be equally efficient in assessing the conformational preferences of the Cc₅REKA analog. Our REMD results still might suggest the presence of some degree of deficient exploration. These apparent restrictions though can be surpassed by considering a higher number of replicas, which would allow overcoming the REMD limitations when it is used to search the energy landscape of small flexible peptides. Nevertheless, it should be emphasized that this option increases remarkably the computer resources devoted to the conformational search.⁴¹

Conformational features

A detailed analysis of the conformational exploration performed by both SA-MD and REMD shows interesting features and differences that are inherent to these techniques. Fig. 3, which shows the Ramachandran plots for the five residues of Cc₅REKA, suggests that SA-MD covers wider extent of the main chain conformational space than REMD. However, the former method explores conformations that are accessible, without representing the effect of the temperature agitation (i.e. the kinetics accessibility of these conformations at a determined temperature). Thus, most of the conformations obtained by SA-MD might not be reachable at the studied reference temperature. As we have previously indicated, this is not a surprising feature since this technique was designed to speed up the search of the most relevant conformational minima at a specific temperature. Despite this difference, the lower energy structures derived from the two techniques are almost coincident (Fig. 3). In other words, at 300 K the most relevant conformations for the studied pentapeptide analog are well represented with relatively short trajectories and with a limited amount of replicas. It is worth noting that at such a temperature some conformations that are separated by 4 or 5 kcal mol⁻¹ from the lowest energy structure, which were detected via SA-MD, are not found by REMD. These conformations that had a meaningful contribution in the conformational profile of the wild type segment^10–12 require higher thermal agitation to be reached. This question though does not invalidate our bioactive motif of both wild type and the peptide analog, since the physiological conditions are represented by temperatures that feature almost 10 degrees above the studied conditions.

Fig. 3.

Ramachandran plot of the main chain dihedral angles (φ, ψ) for the five residues of Cc₅REKA considering the unique minimum energy conformations obtained using the SA-MD and REMD methods. Grey and dark-brown dots represent all the conformations derived from SA-MD and REMD, respectively. The 5 conformations of lower energy obtained by SA-MD and REMD are indicated by blue and black dots, respectively.

Regarding the conformational space accessible by c₅Arg (position 2 in Cc₅REKA), Fig. 3 shows that the non-coded amino acid has achieved its primary goal: it restricts the conformational freedom of the peptide and, simultaneously, favors specific conformation profiles. If comparing the accessible conformations of both CREKA^10–12 and Cc₅REKA peptides, it is noticeable that the latter peptide explores smaller number of different main chain conformations than the former one, with roughly the same amount of generated minima. In other words, the number of backbone positions that are accessible to the peptide when c₅Arg is present is significantly reduced, i.e., the global flexibility of the homing peptide analog is lower than the parent wild type segment.

At the same time, such flexibility reduction is selective because the cyclic nature of the Arg surrogate favors the adoption of the conformation that the coded Arg presents in the bioactive conformation.¹⁰ Independently of the exploration method, the main-chain conformations found for the second residue of the Cc₅REKA analog mainly clustered at the two α regions of the Ramachandran plot, which correspond to the more favored conformations found for a single c₅Arg residue in aqueous solution.²⁶ In both cases the structures of lower energy (blue and black dots in Fig. 3) present conformations for Cc₅REKA’s 2nd position that are located at the α_L region, which corresponds to that adopted by Arg in the parent CREKA peptide. On the other hand, the lower efficiency exhibited by the SA-MD method for locating the more stable conformations widens the range of explored conformations for c₅Arg, with a significant number of Cc₅REKA arrangements with relative energies higher than 5 kcal mol⁻¹ versus the tight conformation distribution observed using REMD. However, in the case of the SA-MD exploration, those Cc₅REKA structures that were within 2–3 kcal mol⁻¹ from the absolute minimum again showed conformations at the α_L region, which was the main objective of introducing c₅Arg as Arg surrogate.

Despite the initial success, in order to enhance the homing activity, the biased conformation of c₅Arg in the peptide analog should also favor the formation of turn conformations beyond the residue position, in order to adopt the same conformation described as bioactive. As we mentioned before, we had already tried enhancing the stability of the β-turn featured by the bioactive conformation of CREKA replacing the Arg by another surrogate of low conformational flexibility, (^γPro)Arg.¹² In fact, the incorporation of this Pro derivative into CREKA also reduced the conformational flexibility, especially at the position 2. However, it caused the disruption of the desired β-turn conformation, apparently because of the drastic structural changes that the Pro skeleton introduced in the dynamics of the adjacent residues. In the case of c₅Arg, despite fixing the conformational preferences of position 2, the higher flexibility of the Ac₅c skeleton²⁵ facilitates the adjacent Glu adopting the proper conformation for the continuation of the turn (Fig. 3). Thus, in the minimum energy conformations found for the Cc₅REKA analog, Glu usually falls in the same region that was featured for the parent peptide, making possible the formation of the β-turn.^10–12

Structural comparison between the lowest energy conformations

Examination of the backbone conformations for the most favored arrangements of Cc₅REKA reveals several features that are closely related with the exploration efficiency and the effect of c₅Arg on the conformational properties of CREKA. Fig. 4 plots the Root Mean Square Difference (RMSD) correlation between the 10 lowest energy structures obtained in each landscape exploration (Cc₅REKA generated by SA-MD and REMD, named SAn and REn, respectively, where n refers to the rank position and number 01 the lowest energy structure) and their structural correlation with the previously determined bioactive conformational pool of the wild type homing peptide¹⁰ (named WTn following an analogous nomenclature). As can be seen, the most favored conformations generated by REMD are significantly similar between themselves. The largest main-chain difference between the structures is 1.83 Å, which corresponds to a pair of almost iso-energetic conformations, RE04 and RE05. Again, this feature indicates that the conformational space explored by REMD is essentially restricted to peptide organizations with low energy, which is consistent with the results showed in the previous sections.

Fig. 4.

Backbone structural pair-correlation among the 10 unique conformations of lower energy obtained using SA-MD for CREKA (WTn, ref. 10) and Cc₅REKA (SAn), and using REMD for Cc₅REKA (REn). The correlation was computed as the Root Mean Square Deviation (RMSD) between two conformations. Each color represents a specific RMSD interval of values in which the deviation value of two intersecting structures is situated: green color corresponds to a RMSD lower than 1.3 Å, white color defines a RMSD comprised between 1.3 Å and 2.0 Å, yellow color refers to the interval between 2.0 Å and 3.0 Å and, finally, orange higher than 3.0 Å.

On the other hand, SA-MD appears a more efficient tool when exploring the conformation of small peptides since high structural diversity is observed among those structures of lower energy. This feature becomes especially remarkable for the SA02 conformation, which features a γ-turn that does not correspond to the β-turn motif frequently found in Cc₅REKA. However, differences among the rest of the conformations of Cc₅REKA obtained using both methodologies are relatively small, essentially reflecting small variations in the β-turn motif with respect to the most favored organization. Hence, the use of c₅Arg has achieved another primary goal: its incorporation in CREKA instead of Arg drastically restricts the potential organizations to those that characterized the bioactive structure for the parent pentapeptide (i.e. the β-turn).

Thus, Fig. 4 reveals that the majority of backbone conformations are very similar to that described for the bioactive organization of CREKA, with the exception of the aforementioned structure SA02. These results indicate that the most accessible arrangements for the Cc₅REKA analog correspond to those describe as bioactive for the CREKA peptide.

Table 1 lists the main electrostatic interactions detected in the four arrangements of lower energy for both the parent peptide and Cc₅REKA. It is worth noting that WT01, WT03 and WT04, despite their evident differences, share important features that make them part of the bioactive profile of CREKA. The three conformations are folded into a β-turn and all expose the charged side chain to the solvent on the same side of the peptide chain.^10–12 When comparing their interaction patterns, both present remarkable similarities in the central segment, –REK–. Specifically, in WT01 (Fig. 5a) and WT04 the backbones of Lys (N–H) and Cys (C=O) interact through a hydrogen bond, closing a type II β-turn. On the other hand, in WT03 the Lys backbone (N–H) forms a hydrogen bond with the terminal acetyl blocker (Ac), defining an α-turn that may easily transform into a β-turn if the competing Ala C=O forces tightening the turn (Ac and Ala are mutually blocked by both amide groups forming two interlocked hydrogen bonds). The topological outcome of this complex arrangement resembles WT01 and places the ionized side-chains in the proper position for the formation of salt bridges. In summary, with the exception of WT02, the structures mainly differ in the conformation adopted by both terminal groups.

Table 1.

Comparison among the interaction patterns^a of the four unique conformations of lower energy obtained for CREKA^b and Cc₅REKA using SA-MD and REMD^c

	Cc5REKA (SA-MD)		Cc5REKA (REMD)		CREKA (SA-MD)
SA01	HB Lys(N-H)⋯(O=C)Cys1	RE01	HB Glu(N-H)⋯(O=C)Ace1	WT01	HB Lys(N-H)⋯(O=C)Cys1
	HB Ala(N-H)⋯(O=C)Glu2		HB Lys(N-H)⋯(OOC−-,side chain)Glu		SB Arg⋯Glu
	HB c5Arg(C=O)⋯(+H3N, side chain)Lys		SB Lys⋯Glu		SB Lys⋯Glu
	SB Lys⋯Glu
SA02	HB c5Arg(N-H)⋯(O=C)Ace2	RE02	HB Glu(N-H)⋯(O=C)Ace1	WT02	HB Ala(N-H)⋯(O=C)Arg1
	HB Glu(N-H)⋯(O=C)Cys2		HB Lys(N-H)⋯(O=C)Ace3		HB NMe(N-H)⋯(O=C)Glu1
	HB c5Arg(N-H2 guanidinium)⋯(O=C)Glu		HB c5Arg(N-H2 guanidinium)⋯(O=C)Ala		SB Arg⋯Glu
	HB Ala(N-H)⋯(OOC−–,side chain)Glu		SB Lys – Glu
	HB NMe(N-H)⋯(OOC−–, side chain)Glu
	SB Lys⋯Glu
SA03	HB Lys(N-H)⋯(O=C)Cys1	RE03	HB Glu(N-H)⋯(O=C)Ace1	WT03	HB Lys(N-H)⋯(O=C)Ace3
	HB c5Arg(N-H)⋯(O=C)Lys2		HB Lys(N-H)⋯(O=C)Ace3		HB Ala(N-H)⋯(O=C)Ace4
	SB Lys⋯Glu		SB Lys⋯Glu		HB NMe(N-H)⋯(O=C)Cys4
					SB Arg⋯Glu
SA04	HB Lys(N-H)⋯(O=C)Cys1	RE04	HB Ala(N-H)⋯(O=C)Cys3	WT04	HB Lys(N-H)⋯(O=C)Cys1
	HB c5Arg(N-H2 guanidinium)⋯(O=C)Cys		SB Lys⋯Glu		HB Ala(N-H)⋯(O=C)cys3
	HB c5Arg(N-H2 guanidinium)⋯(O=C)Lys				HB NMe(N-H)⋯(O=C)Cys4
	SB Lys⋯Glu				SB Arg⋯Glu
					SB Lys⋯Glu

^aInteraction pattern: hydrogen bonds and salt bridges are labeled as HB and SB, respectively.

^bFour minima generated according to ref. 10.

^cClassification of tight turns as a function of the distance (in number of residues) between the hydrogen bonding donor and acceptor:

¹β-turn (i, i + 3),

²γ-turn(i, i + 2),

³α-turn (i, i + 4) and

⁴π-turn, (i, i + 5).

Fig. 5.

Detailed representation of selected low energy conformations obtained for CREKA and Cc₅REKA. The main chain is represented by solid cartoons, with standard protein conformational colors (blue and white for turn motifs). In the side chains heavy atoms are represented with solid scaled balls and their bonds with solid sticks following the CPK colors convention. Hydrogen atoms of the side groups are omitted for clarity. The depicted structures are: (a) WT01 (extracted from ref. 10); (b) RE01; (c) RE04; (d) SA02; and (e) SA03 (the meaning of the codes is described in the text).

Analysis of the four lower energy conformations obtained for Cc₅REKA using REMD (Table 1) demonstrates that they all have identifiable counterparts in the structures characterized for the parent peptide, with the exception of structure RE01 (Fig. 5b). The latter conformation features a single β-turn at the C terminus, which involves the N–H of Glu and the C=O of the acetyl blocking group that was not observed in any CREKA conformation. This organization is partially due to the interaction between the guanidinium group of c₅Arg and the C=O of Ala, which precludes the formation of other hydrogen bonding patterns more similar to those observed in CREKA. R02 and R03 define arrangements intermediate between the α- and β-turns, leading to turns at the C termini with geometries slightly wider than that typically expected for a conventional β-turn. However, a simple rearrangement of the N terminal group would easily facilitate the formation of the proper turn since in those two conformations the hydrogen bonding acceptor is the C=O of the terminal Ac, whereas in the bioactive conformation of CREKA it is the C=O of Cys. Finally, R04 features an α-turn that presents a similar topology to that observed in the parent peptide (Fig. 5c). Thus, c₅Arg favors the formation of a turn that makes possible the correct spatial location of the Glu and Lys ionized side chains allowing their interaction (Fig. 6a).

Fig. 6.

Superimposition of selected structures of Cc₅REKA with the lowest energy conformation of CREKA (denoted WT01 in the text; ref. 10). In all cases main chains are represented by solid cartoons, green and cyan colors depicting the WT01 and each particular Cc₅REKA structure, respectively. In the side chains heavy atoms are represented with solid scaled balls and their bonds with solid sticks following the CPK colors convention. Hydrogen atoms of the side groups were removed for clarity. The superimposed structures are: (a) RE04; (b) SA01; and (c) SA03 (the meaning of the codes is described in the text).

Results obtained with SA-MD show a better match with the expected arrangements for Cc₅REKA. This search strategy allows locating low energy structures that adopt the β-turn featured by WT01 and WT04. Thus, the N-termini organization adopted by SA01, SA03 and SA04 (Fig. 6b and c, respectively) correspond to that found for WT01, even though some differences are observed at the C terminus (Fig. 5e). Finally, SA02 presents a double γ-turn conformation (Fig. 5d), which was also predicted for other Ac₅c-containing peptides.^21,23

Conclusions

The conformational preferences of a homing peptide synthetic analog have been widely explored by different techniques based on MD simulations. Independent of the strategy, our results have indicated that the recently engineered residue, c₅Arg,²⁶ does accomplish its design aim, which is to reduce the conformational freedom of the CREKA peptide,¹⁰ biasing the accessible conformations towards those that feature the bioactive one. The lower energy conformations of Cc₅REKA involve the formation of β-turn motifs centered around the second and third residues of the synthetic analog, similar to those observed in the parent peptide.¹⁰ At the same time, this conformational bias aids the formation of an interaction network between the side-chains of the central residues, which was described as a necessary feature for the bioactivity of CREKA (i.e., the orientation of the ionized central side chains towards the same side of the turn allows their interaction with the tumor vessel receptors). In this context, it is worth noting that even in those conformations that were not described by the parent peptide, such as RE01 and SA02, the final topological distribution of the side chains guarantees the formation of the required interaction pattern.

On the other hand, differences between the two methodologies demonstrate that the temperature is an important variable in the Cc₅REKA conformational preferences. Even though the energy differences among the ten absolute lowest energy conformations are less than 5 kcal mol⁻¹ some of them are not detected at the REMD reference temperature. Thus, the lower thermal agitation featured under normal conditions precludes the complete coverage of the bioactive conformational profile. In contrast, the SA-MD, which is a technique that explores the most relevant regions of the energy landscape without help from higher temperatures, allows obtaining conformations of Cc₅REKA that were previously reported as relevant for CREKA’s bioactivity. Both the lack of thermal distribution in the generated ensembles and the redundant production of conformations starting from those that are already located in low energy regions facilitated the identification of quasi-degenerated energy arrangements that were not detected at the REMD targeted temperature.

In summary, results derived from the exploration of the conformational bioactive profile of Cc₅REKA demonstrate that inclusion of the engineered residue c₅Arg as Arg replacement does achieve the objectives of its design: not only could the new surrogate protect the homing peptide from the protease activity but also increase the stability of the bioactive conformation previously determined for the wild type peptide CREKA. These results confirm our strategy^{12,13,19–23,26} that is aimed to re-design natural amino acids to improve properties of biotechnological interest in natural peptides.