Macrocyclisation of small peptides enabled by oxetane incorporation

Head-to-tail peptide macrocyclisations are significantly improved, as measured by isolated yields, reaction rates and product distribution, by substitution of one of the backbone amide C 
<svg xmlns="http://www.w3.org/2000/svg" version="1.0" width="16.000000pt" height="16.000000pt" viewBox="0 0 16.000000 16.000000" preserveAspectRatio="xMidYMid meet"><metadata>
Created by potrace 1.16, written by Peter Selinger 2001-2019
</metadata><g transform="translate(1.000000,15.000000) scale(0.005147,-0.005147)" fill="currentColor" stroke="none"><path d="M0 1440 l0 -80 1360 0 1360 0 0 80 0 80 -1360 0 -1360 0 0 -80z M0 960 l0 -80 1360 0 1360 0 0 80 0 80 -1360 0 -1360 0 0 -80z"/></g></svg>
 O bonds with an oxetane ring.


Introduction
Currently more than 40 cyclic peptide drugs are in clinical use with the vast majority derived from natural products (e.g. cyclosporine, vancomycin). 1 Compared to their linear counterparts, cyclic peptides benet from enhanced cell permeability, increased target affinity, and resistance to proteolytic degradation. 2 Moreover, they are capable of acting as inhibitors against some of the most challenging targets, including protein-protein interactions (PPIs). 3 One major obstacle to the discovery and development of new cyclic peptide drugs relates to the difficulties associated with their synthesis. 4 The head-to-tail cyclisation of short peptides containing seven or fewer amino acids is especially challenging. Common problems encountered during cyclisation are C-terminal epimerisation, cyclooligomerisation and the appearance of side products arising from polymerisation (Scheme 1). Consequently, there is a pressing need to discover new macrocyclisation strategies that can provide easy access to a variety of small cyclic peptide scaffolds suitable for use in drug discovery programmes. 5 The incorporation of turn-inducing elements such as a pseudoproline, an N-alkylated or a D-amino acid into a linear peptide oen enhances peptide macrocyclisations. 4c,6 Recently we have shown that substitution of a backbone amide C]O with an oxetane ring induced turns in tripeptides, 7 suggesting that oxetane introduction into the backbone of a linear peptide might offer a general method to improve macrocyclisation efficiency by promoting conformational preorganisation of the linear peptide such that the C-and N-termini are brought into close proximity (Scheme 1).
Importantly, the use of oxetanes as bioisosteres in smallmolecule drug discovery is well established, 8 with emerging applications in peptide science. 9,10 For example, the graing of oxetanes onto the cysteine side chains of proteins and antibodies can signicantly improve their stability and activity. 9 Additionally, the serum half-life and in vivo analgesic properties of linear peptides such as Leu-enkephalin can be improved by introduction of an oxetane. 10 In this Edge Article, we show that many peptide macrocyclisations can be greatly improved by oxetane incorporation, and that the resulting modied peptides are biologically active, supporting the hypothesis that oxetane modied cyclic peptides (OMCPs) represent a powerful new class of peptidomimetic for drug discovery.

Synthesis of oxetane modied cyclisation substrates
To study the impact of oxetane modication on peptide macrocyclisation efficiency, a range of linear peptide precursors were synthesised. Many of the substrates were made in solution using Bn ester/Z-protection of the C-and N-termini to allow the synthesis of salt-free precursors by use of a nal hydrogenolysis step. This enabled accurate comparisons in product yields to be made across different substrates without having to correct for the impact of salts on the cyclisations. The synthesis of oxetane modied peptides was also achieved using the more practical approach of solid-phase peptide synthesis (SPPS). In most cases, control peptides without the oxetane modication were made for comparison purposes (see ESI †).
Two distinct methods were used to make the linear peptides. In the rst, oxetane introduction was achieved by: (i) conjugate addition of the N-terminus of the growing peptide to 3-(nitromethylene)oxetane; (ii) nitro group reduction and in situ coupling of the resulting amine to the next protected amino acid, preactivated as its succinyl ester. 7,11 This method proved very convenient for the solution-phase synthesis of oxetane modied glycine residues as illustrated by the preparation of pentapeptide 1 (Scheme 2). Alternatively, preformed Fmocprotected dipeptide building blocks containing the oxetane modication were incorporated into the growing peptide chain as shown for the synthesis of pentapeptide 2 (Scheme 3). This second method is suitable for the introduction of oxetane modied derivatives of other L-amino acids, 10 is fully compatible with standard Fmoc/ t Bu SPPS, 11a and avoids the need to subject the growing peptide chain to strongly reductive conditions. The assembly of resin-bound tetrapeptide 17 by SPPS provides an illustration of the utility of this method (Scheme 4), with additional examples in the ESI. †

Macrocyclisation of oxetane modied pentapeptides
To study the impact of oxetane introduction on the efficiency of macrocyclisation, ring closure of the pentapeptide LAGPY in the presence and absence of oxetane modied glycine was studied under a variety of conditions. As shown in Table 1 (entries 6-8), this substrate was chosen because cyclisation of the corresponding unmodied pentapeptide is very low yielding (13-23%) even under high dilution. 12 Upon introduction of the oxetane modication, a 3-fold yield improvement was observed. The best yield was obtained using PyBOP (Table 1, entry 3), although DEPBT and HATU were also efficient with similar levels of improvement seen. Later studies exploring the wider scope of this methodology led us to favour DEPBT as the reagent of choice in most cases. Of practical importance, the reaction can be conducted on a larger scale (0.5 mmol) under less dilute conditions (0.005 M) without loss of yield (Table 1, entry 1).
Insights into the impact of oxetane introduction on the efficiency of this macrocyclisation were revealed by LC-MS analysis (Fig. 1). In the cyclisation of the unmodied peptide, there are signicant quantities of unreacted linear pentapeptide 3 le even aer 48 h. Additionally, linear and cyclic decapeptides arising from substrate dimerisation were evident alongside a second cyclic pentapeptide arising from C-terminal epimerisation (Fig. 1, bottom). In contrast, for the oxetane containing precursor, essentially complete consumption of starting material is seen on this timescale. The reaction is also much cleaner with the corresponding by-products seen in only trace quantities (Fig. 1, top). Thus, signicant improvements in this difficult macrocyclisation are realised through introduction of the oxetane motif.
To gain a deeper understanding of reaction rates and the products formed, time-course studies of an inherently difficult to cyclise tetrapeptide WLGG (6) and corresponding oxetane modied WLGOxG (7) (where GOx ¼ oxetane modied glycine) were undertaken. For this investigation, substrates containing a tryptophan residue were used to allow quantitative monitoring by UV spectroscopy. Both substrates were subjected to the DEPBT method and conversions monitored over 74 h. From these data, it is clear that the initial rate of formation of oxetane containing cyclic peptide 9 is considerably faster than 8, even though both linear precursors 6 and 7 are consumed at similar rates (Fig. 2). The isolated yields for these macrocyclisations are given in Fig. 3 and again demonstrate the benets of the oxetane motif. Appreciable quantities of the unwanted dimer and    cyclodimer were produced in the cyclisation of 8, explaining the lower conversion and isolated yield (see ESI †). In contrast, for the oxetane-modied peptide 7 clean conversion to the cyclic product 9 was observed. Taken together, these studies establish that head-to-tail ring closures to form small cyclic peptides proceed more quickly, give higher yields and produce less side products when one of the backbone carbonyl groups is replaced by an oxetane ring.

Oxetane modied cyclic peptides: scope and limitations
To test the generality of these initial ndings, the head-to-tail cyclisation of a variety of oxetane modied substrates was studied and the isolated yields compared with those for the unmodied systems (Fig. 3). Oxetane introduction clearly leads to appreciable improvements across a range of ring sizes as illustrated for tetrapeptides 9-12, pentapeptides 4 and 13 and hexapeptide 14. OMCPs were obtained in yields ranging from 49-65% whereas all yields for the control systems were substantially lower (0-33%). In a striking illustration of the power of this new method, cyclisation to tetrapeptide 10 was achieved in 54% yield in the presence of the oxetane modication, however cyclodimerisation to octapeptide was the predominant reaction pathway (20%) for the unmodied system. The successful formation of 12 containing an alanine modied residue conrms that this methodology is not simply limited to glycine substitution. Generally, DEPBT was favoured as the activating agent as it gave efficient cyclisations and minimised formation of difficult to separate diastereomers (e.g. 13: DEPBT: 51%, d.r. >95 : 5; PyBOP: 64%, d.r. 3 : 1) arising from epimerisation at the C-terminus prior to cyclisation. In some instances, impurities derived from the DEPBT reagent proved hard to remove, in which case PyBOP was used as an alternative.
Next, the synthesis of OMCPs by SPPS techniques was examined. In this way, it is possible to circumvent the need to purify any of the intermediates, simplifying and accelerating the process and potentially enabling automation. This was realised through the synthesis of cyclic tetrapeptide 18 in an impressive 39% yield from commercial H-Trp(Boc)-2-ClTrt (15) (Scheme 4). First, the linear peptide 17 was prepared from 15, 16 and Fmoc-Leu-OH using Fmoc/ t Bu SPPS, 11a cleaved from the resin using TFE then cyclised to 18 using DEPBT conditions and puried by column chromatography. Macrocycles 19-21 based on different ring sizes were readily made through further generalisation of this SPPS approach (Fig. 4 and ESI †).
Macrocyclisations not involving head-to-tail ring closure have also been examined. Specically, the formation of disulphide containing macrocycles through oxidative cyclisation of cysteine side chains has been explored (Scheme 5). Treatment of 2, whose synthesis was described in Scheme 3, with iodine provided the 17-membered macrocycle by way of trityl deprotection and disulphide bond formation. Further reaction with TFA in the presence of the cation scavenger triisopropylsilane (TIS) facilitated removal of all the side chain protecting groups providing 22 aer reverse-phase HPLC purication. Importantly, the four-membered oxetane ring survives the strongly  acid conditions required to globally deprotect the side-chains including the arginine Pbf group. Smaller 11-and 14membered macrocycles 23 and 24 were also conveniently produced in good yields using this methodology.

Comparison with other amino acid modications
To understand how signicant these improvements are in the broader context of peptide macrocyclisation, a series of alternative modications were made to the central G residue of LAGAY and the efficiencies of the ring closures compared. N-Methyl glycine, 2-methylalanine (Aib), ethylenediamine, dimethylethylenediamine and b-alanine were all introduced in place of the glycine (Scheme 6). These modications could improve cyclisation efficiency by: (i) increasing the conformational exibility of the peptide backbone by deletion of one of the amide bonds; (ii) enlarging the size of the macrocycle by introduction of an additional methylene group; (iii) bringing the reacting ends closer together by favouring the cis-amide conformation; or (iv) introducing a potentially benecial Thorpe-Ingold effect. In fact, only the oxetane modication led to marked improvement in isolated yield of the derived cyclic peptides, suggesting that oxetane introduction is particularly benecial.
Macrocyclisation to pentapeptide 13 by formation of all four possible amide bonds was examined to see whether the location of the oxetane relative to the amide bond being formed is important (Fig. 5). The yields were compared with those obtained making pentapeptide 25 by the same disconnections. This approach removes inherent differences in cyclisation efficiency associated with amide bond formation between different amino acid residues. For all four pairs of cyclisation studied, the oxetane modied system outperformed the C]O system leading to higher product yields. However, larger improvements are seen when the modication is more centrally located along the precursor backbone.

Structural impact of oxetane modication
To understand why oxetane incorporation improves the macrocyclisations, we undertook solution-state NMR studies on the linear cyclisation precursor LAGAY and its oxetane modied variant. In order to accurately mimic the active ester intermediate without the effect of zwitterion formation, which may alter the solution conformation, we chose to study the corresponding methyl esters: LAGAY-OMe and LAGOxAY-OMe. 2D 1 H-1 H TOCSY and NOESY spectra were acquired as detailed in the ESI † in order to assign the 1 H signals in both peptides and probe changes in conformation, especially near the site of modication. As shown in Fig. 6a and b, oxetane modied LAGOxAY-OMe yielded a larger number of nuclear Overhauser enhancements (NOEs) than unmodied LAGAY-OMe. The observed NOEs were used to construct an NOE-connectivity map (Fig. 6c) which summarises sequential (i.e. residue i to residue i + 1) and medium-range NOEs that are commonly observed in a-helices,  (d aN , d NN , d bN ), medium-range NOEs are only observed in peptide LAGOxAY-OMe containing the modication. Specically, d NN (i, i + 2) and d aN (i, i + 2) NOEs are clearly observed in the oxetane modied peptide between residues Gly3 and Tyr5 indicating their close proximity, data consistent with the formation of a turn. These NMR studies provide the rst direct experimental evidence that the oxetane modication is indeed turn-inducing, and brings the peptide termini closer in space to facilitate efficient cyclisation (Scheme 6).
To gain insights into how oxetane modication impacts the structure of the derived cyclic peptides, molecular dynamics (MD) computer simulations with distance restraints derived from NMR experiments were carried out on cLAGAY (25) and cLAGOxAY (13) in DMSO. 2D 1 H-1 H NOESY spectra were used to compile a set of nuclear Overhauser enhancement (NOE)derived experimental distance restraints that were incorporated into MD simulations using the CHARMM forceeld 16 with modications for the oxetane ring. 7 All simulations were performed using the Gromacs 5.1.4 simulation package. 17 Each cyclic peptide was simulated for 100 ns in DMSO at 500 K and from this trajectory ve distinct conformations of each peptide were selected for a further 100 ns at 300 K. For these small, relatively rigid peptides, 500 ns total simulation time was sufficient for the simulations to converge, given that no new structural information was obtained aer $320 ns.
Cluster analysis was performed on 12 500 structures extracted from the trajectories at 40 ps intervals. Five and four distinct structures were found for peptides 25 and 13 respectively, with the rst three clusters representing $99.9% of the entire trajectory for both peptides. The trajectories were dominated by conformations belonging to a single cluster that represented 93.7% and 99.1% of the total population of 25 and 13, respectively (see ESI †). Ten representative structures from the dominant cluster of each peptide are overlaid in Fig. 7a. The cluster analysis indicates that the backbones of both structures have little conformational exibility (as expected for small cyclic peptides), but that the OMCP is slightly more rigid. H-bond analysis revealed that the unmodied cyclic peptide 25 did not form any signicant intra-peptide H-bonds whereas OMCP   13 was characterised by a H-bond with 86.3% occupancy between the NH group of Leu1 and the O atom of Ala4 (see Fig. 7b). The backbones of 25 and 13 in their dominant congurations are shown overlaid in Fig. 7b. Oxetane introduction only alters the backbone conformation of residues in proximity to the modication itself. In this case, an inversion of the amide bond between Ala4 and Tyr5 arises on oxetane introduction, which allows the structure to be stabilised by an intra-peptide H-bond across the macrocycle. Ramachandran plots, showing the distribution of the F and 4 backbone dihedral angles of each residue and the entire peptide (averaged over the entire simulation trajectory), are presented in Fig. 7c. The Ramachandran plots show that 13 explores less of the F/4 space than 25, which supports our observation above that 13 is more rigid. The Ramachandran plots also highlight the effect of the oxetane introduction on the structure of the residues in proximity to the modication, with Gly3, Ala4 and Tyr5 shiing to different regions of F/4 space following the introduction of the oxetane, which is in agreement with the visual representation of the backbone in Fig. 7b.

Metalloprotease inhibition studies
Studies to explore the impact of oxetane modication on the bioactivity of the derived cyclic peptides were undertaken. cCNGRC (26) is known to target Aminopeptidase N (APN), a transmembrane zinc-dependent metalloprotease involved in a variety of processes, including blood pressure regulation, cell migration, viral uptake, cell survival, and angiogenesis. 18,19 Both oxetane modied derivative 22 and cCNGRC (26) were examined for their inhibitory activity towards porcine APN using a spectrophotometric assay (Table 2). 19 Although both compounds are rather weak inhibitors, the fact that similar IC 50 values were observed for 22 and 26 suggests that the oxetane motif is an excellent bioisostere of the amide bond it replaces.

Conclusions
A new way to make small macrocyclic peptides has been discovered and evaluated. The approach relies on the introduction of an oxetane into the backbone of the cyclisation precursor. Two strategies for the assembly of these molecules were developed involving either chain elongation of the peptide by reaction with 3-(nitromethylene)oxetane, or alternatively through coupling preformed Fmoc-protected dipeptide building blocks containing this modication. The rst is convenient for the insertion of glycine modications, whereas the second is more general, being suitable for introduction of oxetane modied derivatives of other L-amino acids, 10 and for integration with conventional SPPS methods. 11a Oxetane incorporation led to improvements in all the headto-tail cyclisations studied as assessed by isolated yields, reaction rates and analysis of product distributions. The method works across a range of ring sizes and enabled the synthesis of a range of cyclic tetra-, penta-, hexa-and heptapeptides. Major improvements in yields are seen using glycine modied residues, with levels of epimerisation and dimerisation signicantly reduced. The extension of this method to other modied L-amino acids was demonstrated. It can also be applied to sidechain to side-chain macrocyclisations, specically oxidative ring closure of cysteine residues by disulphide bond formation. As a tool to improve head-to-tail macrocyclisations, our evidence suggests that oxetane incorporation is superior to a number of other common backbone modications. Importantly, OMCPs are compatible with the harsh acidic conditions needed to deprotect amino acid side chains.
A key question we sought to address in this study was why does oxetane introduction improve macrocyclisations? For maximum benet, we have shown that the oxetane is best located centrally along the peptide backbone, an observation consistent with the hypothesis that this modication is turninducing. 7 Moreover, using solution-state NMR spectroscopy, we have obtained the rst direct experimental evidence that oxetane introduction induces a turn in the peptide backbone, through the observation of d NN (i, i + 2) and d aN (i, i + 2) NOEs. The formation of a turn presumably helps bring the termini closer together enabling the macrocyclisation.
To be useful in drug discovery programmes as well as being easy to make, OMCPs must also have useful properties. Intuitively, one might expect oxetane introduction to only alter the backbone conguration of the cyclic peptide close to the site of modication. Indeed, this was borne out in detailed NOErestrained MD simulations conducted on 13. The full retention of bioactivity of 22 relative to cCNGRC suggests that the replacement of a C]O amide bond by an oxetane ring has little impact on binding, making it a viable bioisosteric replacement. Next steps will involve a fuller assessment of the properties of OMCPs and their application in drug discovery programmes.

Conflicts of interest
There are no conicts to declare.