‘À La Carte’ Cyclic Hexapeptides: Fine Tuning Conformational Diversity while Preserving the Peptide Scaffold.

Cyclic peptides have recently emerged as promising modulators of challenging protein-protein interactions. Here we report on the design, synthesis and conformational behavior of a small library composed of C2 symmetric cyclic hexapeptides of type c(Xaa-DPro-Yaa)2, where Xaa and Yaa are chosen from alanine, isoleucine, serine, glutamic acid, arginine and tryptophan due to the favorable properties of the side chains of these residues to recognize complex protein surfaces. We used a combination of nuclear magnetic resonance and molecular dynamic simulations to perform an extensive conformational analysis of a representative set of cyclic hexapeptides. Our results indicated that both the chemical nature and the chirality of the variable Xaa and Yaa positions play an important role in the cis/trans configuration of the Xaa-D-Pro bonds and in the conformational preferences of this family of peptides. This structural tuning can be exploited in design strategies seeking to optimize the binding efficiency and selectivity of cyclic hexapeptides towards protein surfaces.


Introduction
Macrocycles have found broad application in chemical biology, biotechnology and pharmaceutical research. [1] Over the last decade, particularly intense efforts have been channelled into the development of macrocycles and their use in the field of molecular recognition. [2] Nowadays, most synthetic compounds reported to interact with proteins are small molecules that bind to well-defined cavities on the protein structure. [3] More challenging, however, is the design of molecules with the ability to bind to the solvent-exposed protein surface and specifically recognize the large interfacial areas involved in protein-protein interactions, with the purpose of interfering in the functional activity of the target protein. [4] The growing appreciation that these complex protein-protein interfaces appear to be 'undruggable' by small ligands has coincided with renewed interest in macrocycles, and cyclic peptides in particular, as potential protein binders, based on the premise that these cyclic molecules may be more suited to mimic natural biorecognition processes. [5] Cyclic peptides have a number of advantages over their linear counterparts. [6] For example, they show improved resistance to proteolytic degradation by exopeptidases due to the lack of both amino and carboxyl termini. Moreover, protein recognition is a conformation-dependent process, and converting linear sequences into their cyclized form is a common approach to modulate the performance of peptide sequences. A classic and elegant example is the chemical development of Cilengitide, a cyclic pentapeptide (c(Arg-Gly-Asp-D-Phe-Val)) that is selective for αv integrins and that was under investigation for the treatment of glioblastoma. [7] Research on a variety of linear and cyclic penta-and hexapeptides containing the tripeptide sequence RGD (arginine, glycine and aspartic acid) showed that cycles inducing distinct conformations of RGD sequences were able to recognize various integrin receptors. [8] These pioneering studies showed that the preorganization of the crucial side chains by cyclization significantly affects the functional activity of the amino acid sequences. Both the binding affinity and the selectivity of a recognition motif introduced into a cyclic peptide may increase when its bioactive conformation is included in the restrained conformational space of the cyclic molecule. In this case, cyclization reduces the entropic penalty of binding, thereby improving the binding interaction. However, cyclization often results in a significant decrease or even a total loss of activity. [9] This occurs when the bioactive conformation of the recognition motif is excluded from the conformational space of the constrained peptide. It is also possible that, although accessible in the restricted conformational ensemble of the cycle, the bioactive conformation of the peptide is scarcely populated. Cyclic peptides still have considerable backbone mobility, but usually display rugged conformational landscapes that include a number of energy local minima that may not be uniformly sampled. Reaching some conformational states may involve the rare crossing of significant energy barriers.
Therefore, the cyclic molecule would thermally fluctuate among a subset of states at a given temperature. These crucial considerations should not be overlooked in screening projects based on "ligand-oriented design" strategies. The chances of finding suitable protein ligands will depend on the availability of a robust synthetic approach to generate a variety of diverse molecules, combined with detailed knowledge of the conformational behavior of the molecules generated.
Motivated by these observations, we sought to synthesize a small library of C2 symmetric cyclic hexapeptides of the type c(Xaa-D-Pro-Yaa)2 (Scheme 1 and Table S1). We hypothesized that this family of constrained cyclic peptides may have potential applications in the recognition of different types of protein surfaces. In fact, this library has recently demonstrated to be useful as a source of ligands able to recognize two completely unrelated proteins, such as the human vascular endothelial growth factor (VEGF) [10] and the bacterial OmpA. One OmpA inhibitor selected from this library was shown to prevent bacterial infection in an in vivo model. [11] Scheme 1.

Results and Discussion
Design of the library of cyclic hexapeptides Here we aimed to synthesize a family of cyclic peptides with a preorganized structural scaffold that could optimally present a set of side chains selected for their favorable properties for protein surface recognition. We focused on C2 symmetric cyclic hexapeptides due to their straightforward synthesis and because they adopt conformations featured by two fused β-turn arrangements. [13] Initially, we synthesized a set of peptides of the type c(Xaa-D-Pro-Yaa)2 (Scheme 1, peptides 1-15) including a D-amino acid to enhance the β-turn propensity. [14] Xaa and Yaa could be Trp, Arg, Glu, Ser, Ala and Ile. These residues were rationally selected from all proteinogenic amino acids on the basis of the physicochemical, synthetic and statistical parameters considered in our previous work. [15] Tryptophan was chosen because it is over-represented at protein interfaces and is found on many hot spots. Like lysine, arginine (pI = 10.76) displays a positive charge at physiological pH. However, significant differences in the hydration energy and hydrogen bonding potential of these cationic amino acids make arginine more favorable to establish inter-residue contacts. Alanine is a highly abundant residue in natural systems. Since it has a small lateral side chain, it can act as a spacer for various interaction sites on peptides. Glutamic acid (pI = 3.22) contributes to negative charges. Serine is a polar non-charged amino acid that outperforms threonine in solidphase peptide synthesis (SPPS). Finally, isoleucine is an aliphatic β-branched amino acid with a significantly bulky side chain. It has a higher presence in active sites than valine and leucine. Proline was selected as it is the only genetically encoded N-alkyl amino acid and it often participates in crucial hydrophobic-hydrophobic interactions. In addition, D-proline is known to act as a turn-inducing element to promote peptide cyclization. [16] The initial library was then expanded to include new peptides (peptides 16-31) (Scheme 1 and Table S1) with the aim of studying the influence of the chemical nature and/or chirality of the Xaa and Yaa positions on the conformational behavior of this family of cyclic peptides.

Synthesis of the C2 symmetric cyclic hexapeptides
Cyclic hexapeptides were synthesized following a standard solid-phase Fmoc/tBu peptide synthesis protocol using 2chlorotrityl chloride (CTC) resin (Scheme 2). To compensate for the lower reactivity of the secondary amine groups of the prolyl residues, HATU was chosen as coupling reagent. All linear peptides were then synthesized as H-Xaa-Pro-Yaa-Xaa-Pro-Yaa-OH, with its respective residue chirality. To prove that no epimerization was taking place in the C-terminus of these peptides during the synthesis, all the corresponding C-terminus epimers for each peptide of the initial library were synthesized (Scheme S1). Peptide cyclization was performed in solution using PyBOP/DIEA as coupling reagent. The observation of three individual spin systems in the NMR spectra of all four peptides confirmed the C2 symmetric nature of these molecules ( Figure S1). This observation is consistent with these peptides populating either a unique conformation or different states coexisting in fast equilibrium in the NMR timescale. A complete resonance assignment is provided in the Supporting Information (Table S2). The resemblance of structure-sensitive NMR parameters such as secondary chemical shifts, 3 JNHHα values, temperature coefficients of amide protons (Table 1) (Table 1). [17] The  The NH amide protons of the Xaa and Yaa residues within this set of peptides exhibited a well-differentiated trend.
Irrespective of the chemical nature of the amino acid, the NH of Xaa residues showed higher-field chemical shifts (Table S2) and smaller temperature coefficient values (-ΔδNH/ΔT) than those of the Yaa residues ( Table 1).
The reduced temperature dependence of the Xaa amide protons, ranging from 1.0 ppb K -1 in peptide 11 (Xaa = Ser) to 4.5 ppb K -1 in peptide 3 (Xaa = Arg), indicates their involvement in intramolecular hydrogen bonds. In contrast, the NH protons of the Yaa residues showed large temperature coefficients (7.7 -8.5 ppb K -1 ) characteristic of solvent-exposed backbone amides. The 3 JNHHα values observed for these peptides were in the range 7.7 -9.6 Hz, consistent with a β-like conformation ( Table 1)     . [20] The spatial proximity between the Trp and Pro side chains also pointed to the occurrence of CH-π interactions ( Figure S5b).
In  (Table S3). Comparing the NMR spectra of peptide 17 (LDD chirality) and 7 (LDL chirality), a general observation emerges, namely that the former peptide displays higher conformational flexibility.
As an example, the chemical shift difference ∆δCβγ (D-Pro) of peptide 17 (∆δCβγ = 5.8 ppm) was slightly larger than that of signals coexisting with those of the two major species was also observed but they could not be assigned due to their lower intensity and to severe spectral overlap. A recent study [23] reported that a peptide with the enantiomeric chirality of  The cis Xaa-Pro segment of type VI β-turns has been proposed as a potential binding motif for various peptidyl prolyl cis/trans isomerases (PPIases), which are emerging as important biomedical targets related with diseases such as cancer, Alzheimer's disease, and asthma. [24] Given that peptide 18 (DDL chirality) seemed to adopt, at least to some extent, a  (Table S4 and Figure S13).
The NMR-derived populations of conformers at 298 K are shown in Table 2 previous report that proposed that aromatic side chains preceding D-Pro residues tend to favor the cis conformation of amide bonds. [25]   and by D-Ala (peptide 28) ( Table S5).
Removal of the aromatic side chain decreased the cis/cis content in both cyclic hexapeptides (

Conclusions
Here we show how the chemical nature and/or stereochemistry of the amino acids in cyclic hexapeptide scaffolds can modulate the conformational preferences of the backbone. By changing these two properties, we were able to obtain various b-turn-like conformations that may be useful in the field of molecular recognition. Moreover, we reveal that the chemical nature and the stereochemistry of the amino acids present in this kind of cyclic peptide also affects the orientation of side chains, affording distinct types of stabilizing contacts between side chains, such as CH-p or cation-p interactions. REMD simulations were able to overcome the multiple energy barriers that are needed to be surmounted to sample the diverse conformational states of cyclic hexapeptides in solution. This in silico approach also allowed us to analyze the Trp side chain c1 angles of the studied peptides to predict possible interactions between side chains. Our results contribute to the further development of constrained cyclic peptides for molecular recognition processes. Moreover, they pave the way to create a skeletal conformational diversity, which is of great interest in the recognition of challenging protein surfaces.