Synthesis of Membrane-Permeable Macrocyclic Peptides via Imidazopyridinium Grafting

Abstract

Macrocyclic peptides (MPs) are a class of compounds that have been shown to be particularly well suited for engaging difficult protein targets. However, their utility is limited by their generally poor cell permeability and bioavailability. Here, we report an efficient solid-phase synthesis of novel MPs by trapping a reversible intramolecular imine linkage with a 2-formyl- or 2-keto-pyridine to create an imidazopyridinium (IP⁺)-linked ring. This chemistry is useful for the creation of macrocycles of different sizes and geometries, including head-to-side and side-to-side chain configurations. Many of the IP⁺-linked MPs exhibit far better passive membrane permeability than expected for “beyond Rule of 5” molecules, in some cases exceeding that of much lower molecular weight, traditional drug molecules. We demonstrate that this chemistry is suitable for the creation of libraries of IP⁺-linked MPs and show that these libraries can be mined for protein ligands.

Graphical Abstract

◼. INTRODUCTION

Macrocyclic peptides (MPs) and related molecules have attracted considerable interest as probe molecules and drug leads, particularly for addressing difficult protein targets.¹ Macrocyclization of peptides imparts many favorable properties, including increased resistance to proteases, decreased conformational flexibility and, in certain special cases, increased cell permeability by promoting intramolecular hydrogen bonds that mask otherwise highly hydrated amide N−H moieties.² Moreover, powerful methods exist for the synthesis and screening of huge libraries of genetically encoded MPs, such as phage display³ and mRNA display.⁴ Split and pool synthesis of DNA-encoded libraries (DELs) of macrocycles is another route to access large numbers of these molecules.⁵ For such applications, there continues to be a need for new macrocyclization reactions that proceed with high efficiency under gentle reaction conditions, allow the introduction of novel functional groups, and suffer from little or no competitive intermolecular coupling.^6,7 Here, we report a new method for the macrocyclization of peptides that meets all of these criteria and confers on the products favorable pharmacological properties.

One common approach to the synthesis of MPs and other macrocyclic compounds is to trap an intramolecular imine linkage with an internal or external nucleophile.^7,8 Depending on the imine trapping strategy, the products can range from simple amines to interesting heterocycles. Based on the work of Aron⁹ and others,¹⁰ we hypothesized that an alternative imine trapping strategy would be to expose the imine to a 2- formyl- or 2-keto-pyridine to create a stable, imidazopyridinium (IP⁺) unit (Scheme 1), a heterocycle that, to our knowledge, has not been incorporated into MPs previously. Given the broad availability of substituted pyridines, this would constitute a simple method to introduce additional diversity into MP libraries over and above that of the amino acids. Moreover, as explained below, we hoped that the IP⁺ moiety would improve certain pharmacokinetic properties of MPs, particularly their passive membrane permeability.

Scheme 1.

Proposed Solid-Phase Synthesis of MPs Linked by an IP⁺ Ring^a

^a(A) Established chemistry for the synthesis of the IP⁺ moiety. (B) Proposed application of IP⁺ chemistry to the synthesis of MPs.

In this study we demonstrate that cyclization via IP⁺ formation is indeed a remarkably efficient and flexible method for the solid-phase synthesis of MPs. We also show that the reaction lends itself to the creation of bead-displayed libraries of IP⁺-linked macrocycles and report the identification of a novel streptavidin (SA) ligand from such a library. Finally, we find that the IP⁺ ring imbues these MPs with remarkably good passive membrane permeability, such that even molecules far beyond Lipinski’s Rule of 5 (Ro5)¹¹ permeate membranes at rates comparable to low molecular weight, drug-like compounds.

◼. RESULTS

Solid-Phase Synthesis of IP⁺-Linked Macrocycles.

The IP⁺ unit has been utilized in the synthesis of dyes¹² and compounds of pharmaceutical interest^13–15 However, the existing methods for synthesis of this heterocycle employ one equivalent each of formaldehyde, a primary amine, and a pyridine-2-carboxaldehyde (P2CA) (Scheme 1A), or, alternatively, two equivalents of P2CA and one equivalent of amine. These protocols cannot be utilized to create macrocycles using the approach shown in Scheme 1B.

To explore the formation of macrocycles using this chemistry, we synthesized the model starting material shown in Figure 1A on 160 μm TentaGel beads with a Rink-amide (RAM) linker. Oxo represents the glyoxamide formed by NaIO₄-mediated oxidation of an N-terminal serine.¹⁶ The Mmt protecting group was removed using a 1:3 mixture of hexafluoroisopropanol/DCM, then the beads were exposed to 3 equiv of P2CA in a 1:1 mixture of acetic acid (AcOH)/ trifluoroethanol (TFE) for 10 h at 30 °C. (While the reaction also could be carried out using 5% AcOH/TFE for 10 h, 50% AcOH/TFE was found to reduce the “stickiness” of the beads to plastic microtiter plates and thus make the protocol more convenient.) The products were released from the beads using trifluoroacetic acid (TFA) and analyzed by liquid chromatography/mass spectrometry (LC−MS). As shown in Figure 1, macrocycle MP1 was formed in nearly quantitative yield. NMR analysis of the crude material (Supporting Information Figure S1) confirmed the high purity of the material as well as its structure. Indeed, when MP1 was prepared on a large (50 μmol) scale, nearly pure compound was obtained in 72% yield by simply precipitating the material released from the bead with cold ether. No purification was necessary.

Figure 1.

Solid-phase synthesis of an IP⁺-linked MP. (A) Reaction was carried out on a 50 μmol scale based on loading. Mmt = Monomethoxytrityl. (a) 25% HFIP/DCM, r.t., 2 × 30 min. (b) AcOH/TFE (1:1, v/v), P2CA (3.0 equiv), rt., 10 h. (c) TFA, r.t., 40 min. (B) Crude analytical HPLC trace of the released cyclic peptide MP1. (C) High-resolution mass spectrum of the crude material released from beads (MP1).

Given this promising result, we examined the scope of the reaction with respect to the P2CA unit. The results are shown in Figure 2. A wide variety of substituted P2CAs, carrying methyl, methoxy, chloro, bromo, piperidyl, and phenyl groups, provide the IP⁺-linked macrocycle in excellent yield and purity (>85%; Py1−4, Py6−7, Py9−11, and Py14−15). P2CAs bearing electron-withdrawing groups such as −F, −CN, or a carboxylic acid were less efficient substrates (Py8, Py12, and Py13). We were also pleased to see that quinoline carboxaldehydes are also excellent substrates (Py17, Py18, and Py20). Somewhat to our surprise, macrocyclization proceeded smoothly even with several 2-ketopyridines (Py21−24), further extending the number of substituents that can be introduced into the macrocycle using this strategy. We conclude that the scope of the reaction in the pyridine moiety is broad, though electron-withdrawing substituents result in somewhat lower yields.

Figure 2.

Various aldehydes containing P2A motif for macrocyclization. Aldehydes and ketones containing P2CA motif for cyclization. Nap [l-Ala(2-naphthyl)−OH] was used as UV indicator. All the reactions were carried out on a 2 μmol scale, and the purity of crude product was determined by LC−MS. For the primary data, see Supporting Information Figures S5−S29.

Given the excellent result obtained with Py14 and with an eye toward facile labeling of IP⁺ MPs for future protein-binding studies, we synthesized Py25 (Figure 2), with a piperazine group bearing a free secondary amine, and asked if this compound was a suitable substrate. Gratifyingly, macrocyclization proceeded efficiently (Supporting Information Figure S30). The amine was readily modified with common labels such as biotin or carboxyfluorescein (Figures 2 and Supporting Information S31 and S32).

We next probed the scope of the reaction with respect to ring size. A series of linear precursors were synthesized in which 0−9 alanine residues separated the reactive Oxo and lysine (Lys) units (Figure 3A). Beads displaying these molecules were then subjected to the reaction conditions described above using 2-formyl-quinoline to close the macrocycle. After release from the beads, the products were analyzed by LC−MS (see Supporting Information Figures S33−S39). Compound MP2, in which no spacer alanine was present, was produced in only about 5% LC yield. A single alanine spacer resulted in about half of the starting material being converted to the desired product. The linear peptides containing 2−5 alanine units provided a high yield of clean product. The yield decreased again using peptides containing 6−9 alanine residues. We conclude that this chemistry is optimally suited to provide MPs containing 15−24 atoms from canonical amino acids and can also yield useful amounts of MPs with 12 and 27−30 atoms in the ring.

Figure 3.

Side-to-head cyclization on TG Beads. (A) Each linear peptide precursor was prepared on a 2 μmol scale on TG beads. Red balls represent the former glyoxamide unit after cyclization (B). All the reactions were carried out on a 10 μmol scale based on loading. The purity of the crude product was determined by HPLC and is shown in the figure. See Supporting Information for details about isolated yields. Ahx: 6-aminohexanoic acid; Ipa: Isonipecotic acid.

We then turned to the synthesis of various MPs in this size range that contain a variety of residues. As shown in Figure 3B, various protected amino acids were compatible with this reaction using the standard conditions with various P2CAs (for polar amino acids, the protecting groups were removed after cyclization). Cyclization proceeded efficiently (see Supporting Information Figures S40−S45) with substrates containing Tyr (MP12), Trp (MP13), Arg (MP13), Lys (MP14), Cys (MP15), His (MP16), and Glu (MP17). Methionine residues are oxidized quantitatively to the sulfoxide (see compound MP14) by NaIO_4, as expected.¹⁷ The structures of these macrocycles were confirmed by LCMS and NMR spectroscopy (see Supporting Information Figures S40−S45).

We also revisited the synthesis of mixed residue MPs with ring sizes greater than 24 atoms. It is possible that the fall off in yield for substrates containing more than five consecutive alanine residues (Figure 3A) is due to a propensity to form helices that hinder cyclization. Therefore, we constructed linear peptides in which more flexible units (Ahx and Ipa) were placed between the reactive amine and Oxo residues. As seen in Figure 3B, this indeed resulted in efficient macrocyclization for compounds (MP18 and MP19) with 37 and 39 atoms, respectively, in the ring (see Supporting Information Figures S46 and S47). Based on these data, we speculate that the IP⁺ chemistry will be useful for the efficient synthesis of macrocycles spanning a much wider range of ring sizes than might have been suggested from the data shown in Figure 3A, so long as there is sufficient flexibility in the linear precursor to allow formation of the intramolecular imine intermediate.

We also examined the application of IP⁺ chemistry to the synthesis of MPs in which the linkage is between an oxo-modified side chain and an amine-containing side chain, such as a Lys or Dap residue (see Supporting Information Figures S48−S51 for the synthesis of these peptide precursors). As shown in Figure 4, side chain-to-side chain-connected MPs with different ring sizes were obtained in excellent purity using a variety of P2CAs.

Figure 4.

MPs synthesized using side chain-to-side chain linkages. All the reactions were carried out at 10 μmol scale based on bead loading. The purity of the crude product determined by HPLC is given. Dap: L-2,3-diaminopropionic acid; βAla: β-Alanine; Nle: l-norleucine. Red balls represent the glyoxamide-derived residues after cyclization.

Stapled Peptides via IP⁺ Formation.

Motivated by the broad success of the macrocyclization reactions presented above, including the side chain-to-side chain connections, we examined the utility of the IP⁺-forming reaction for “peptide stapling” (Figure 5). This is a technique in which appropriately positioned side chain residues are linked covalently to stabilize an α-helical conformation of the peptide.¹⁸ Stapled peptides are of broad interest as inhibitors of protein−protein interactions. For example, a stapled peptide inhibitor of p53-Mdm2/MdmX binding has recently entered clinical trials for the treatment of certain cancers with wild-type p53.

Figure 5.

Creation of stapled peptides via IP⁺chemistry. All the reactions were carried out at 10 μmol scale based on bead loading. The purity of the crude product determined by HPLC is given. The spectra were obtained using 100 μM of the stapled peptide dissolved in 30% TFE/PBS (pH = 7.4).

As a model, we focused on the 12 amino acid peptide ITFEDLLDYYGP-NH₂, which is a ligand for the HIV capsid protein that disrupts capsid assembly in vitro but not in cultured cells due to its cell impermeability. Zhang et al. used olefin metathesis chemistry to create a stapled version of this peptide (NYAD-1: IPFXDLLXYYGP, where X = a residue with an alkene-containing side chain) with much higher helical content than the native peptide and improved cell penetration properties. NYAD-1 is a modestly potent inhibitor of HIV replication in cultured cells.¹⁹

We constructed the five linear substrates shown schematically in Figure 4B, in which the amino- and Oxo-containing residues had an i−i + 4 or i−i + 7 spacing. The beads displaying these compounds were treated with the indicated P2CA under the standard conditions. After cleavage, the products were analyzed by LC−MS. In each case, good to excellent conversion to the desired stapled peptides was observed (see Supporting Information Figures S52−S56). The degree of helicity was determined by circular dichroism spectroscopy. As shown in Figure 5, the linear peptide was a random coil. All of the stapled MPs displayed an enhanced degree of helical content. The best of these (≈74% helical) was MP24, in which the i and i+4 residues had been stapled with P2CA. It is interesting that MP24, 25, and 26, which contain the same ring but were formed using a different P2CA, display markedly different levels of helicity. The two i−i + 7-stapled peptides (MP27 and MP28), linked using P2CA and Py17, also displayed measurably different levels of helicity. These data suggest that the use of different P2CA units in the creation of IP⁺-stapled MPs will provide a novel tool to finetune the helical content of stapled peptides.

Parallel Solid-Phase Synthesis of IP⁺-Linked MP Libraries.

Given the high level of efficiency and broad scope of the IP⁺-forming macrocyclization reaction, we examined if this chemistry is suitable for the synthesis of MP libraries. Specifically, it was of interest to determine if the entire process could be miniaturized so as to be carried out on a small amount of TentaGel resin in the wells of a microtiter filter plate, which is the most convenient format for the creation of combinatorial libraries,²⁰ including DELs.²¹ To evaluate this issue, 20 different peptides, including eight tetrapeptides, four pentapeptides, and four hexapeptides, were synthesized on 2 mg of 160 μm TentaGel RAM beads in individual wells of a 96-well microtiter filter plate open to the air. Eighteen of the 20 peptides had Mmt-protected Lys as the C-terminal residue, included a UV-active Nap, tyrosine (Tyr), or tryptophan (Trp) residue to facilitate subsequent LC−MS analysis, and terminated in a serine residue (subsequently oxidized to a glyoxamide). L15 and L16 contained the protected Lys as the third residue. L17 has a glyoxamide installed on a side chain. After completion of the peptide chain, the Lys and serine residues were deprotected and the peptide was oxidized with NaIO₄ to create the Oxo unit. Finally, three equivalents of a P2CA were added to each well to create the IP⁺-linked macrocycle. This protocol was carried out on five identical plates, but in each case, a different P2CA unit was added at the end. After incubation and washing, the material was released from the beads and completely deprotected using TFA. The products were analyzed by LC−MS.

Gratifyingly, the overwhelming majority of the products were produced in good to excellent purity (Figure 6). 83 of the IP⁺-containing MPs were >85% pure and 13 were 75−85% pure. Only 4 of the 100 MPs were produced in <75% purity (see Supporting Information Figures S58−S77 for the primary data). These data strongly support the idea that the IP⁺ grafting chemistry is suitable for the creation of libraries of MPs.

Figure 6.

Solid-phase synthesis of 100 IP⁺-linked MPs in individual wells of a microtiter plate. The numbers indicate % conversion to product.

With this result in hand, we turned to the synthesis of a model screening library. The tripeptide sequence HP(Q/Y) has been shown to be a modest affinity ligand for SA. To test the feasibility of making a small library of IP⁺-linked MPs and screening it against a target protein, we constructed a library of 480 compounds on 10 μm TentaGel beads by parallel solid-phase synthesis in microtiter filter plates and tested each set of beads for their ability to capture fluorescently labeled SA. The general structure of the library is shown in Figure 7. A C-terminal Lys residue and an N-terminal Oxo unit flanked four variable positions, in which one of 14 different amino acids was employed (Figure 7A). To bias the library toward peptides that resemble HP(Q/Y), amino acids aa1, aa2, aa3, and aa11 were employed in the first position, amino acids aa1−aa3 and aa8−aa10 in the second position, amino acids aa12−aa14 in the third position, and amino acids aa4−aa7 in the fourth position. The linear precursors were then treated with one of five different P2CAs (Figure 7A) to complete the synthesis.

Figure 7.

Synthesis and screening of an IP⁺MP library. (A) General structure of the library and the building blocks used in its construction. The position(s) at which each amino acid building block was employed in construction of the library is indicated at the bottom of the Figure (A1, A2, etc.). (B) Top: schematic of the screening protocol used to identify ligands for fluorescently labeled SA. Bottom: increase in fluorescence polarization as a result of titration of fluorescein-labeled MP29 (structure shown on the right without the fluorescein) with unlabeled SA. See the text for details of the library construction and screening.

After thorough washing and equilibration in an aqueous buffer, the beads were blocked with starting block to discourage nonspecific protein binding, then incubated for 1 h with Alexaflour 647-labeled SA (A647-SA, 150 nM) at room temperature. After thorough washing, the amount of fluorescence retained by the beads in each well was measured using a fluorescent plate reader. This protocol was conducted in triplicate. The five bead-displayed MPs that retained the highest levels of A647-SA in each run were synthesized as C-terminal fluorescein conjugates, and their solution affinities for unlabeled SA were determined by titration, followed by an increase in fluorescence polarization (Figure 7B). These data showed IP⁺-linked MP29 to be the best SA ligand with a K_D of approximately 7.0 μM (Figure 7B). These data demonstrate that the IP⁺ chemistry is suitable for the construction of useful screening libraries of novel MPs.

Evaluation of the Passive Membrane Permeability of IP⁺-Linked MPs.

As mentioned above, a major limitation of most MPs is that they display poor passive membrane permeability, limiting their utility for engaging intracellular targets. The IP⁺ motif is an interesting functional unit with respect to potentially influencing membrane permeability. It is relatively hydrophobic, especially in the case where quinoline or other more highly substituted P2CAs are employed to create the macrocycle, yet it carries a permanent positive charge (i.e., not due to a protonation event). It seems reasonable to hypothesize that the positive charge might concentrate the compound on the cell surface through electrostatic interactions and that the hydrophobic character of the heterocycle might stimulate movement across the membrane, resulting in improved passive membrane permeability.

To evaluate this idea, we prepared 20 different IP⁺-linked MPs of various sizes and compositions and measured their rates of membrane transit using PAMPA (parallel artificial membrane permeability assay). Propranolol, a highly permeable low molecular weight (259 Da) drug, and the much less permeable charged small molecule Ranitidine were employed as controls. In this assay, compounds displaying a −log Pe below 6.0 are considered to be highly permeable, while those with a −log Pe between 6.0 and 7.0 are considered to be moderately permeable.

As shown in Figure 8A, a remarkable 45% of the MPs tested displayed a −logP_e value below 6.0 (green dots), despite the fact that the masses of these compounds are all well above 500 Da (Figure 8B). One of these macrocycles (MP40), with a molecular weight of 846 Da, actually traversed the membrane more rapidly than propranolol. The particular P2CA used to create the macrocycle clearly had a significant effect on permeability. For example, four MPs with the same peptide component, but formed using different P2CAs (second row of Figure 8B), displayed significantly different permeabilities, ranging from −logP_e = 4.6−6.0.

Figure 8.

Passive membrane permeability of IP⁺ MPs. (A) Rates of membrane passage (expressed in units of −logP_e) for the compounds indicated as measured using PAMPA. (B) Structures of the most permeable MPs (see Supporting Information Figure S28 for a complete listing of the compounds analyzed). (C) Structures of MP36 analogues lacking the IP⁺ unit. Their (poorer) passive membrane permeability is shown in A (yellow dots).

We also tested the permeability of two very large compounds, MP18 and MP19 (see Figure 3B for structures), which have molecular weights of 1071 and 1234 Da, respectively. Remarkably, even these >1000 Da compounds displayed moderate membrane permeability in the PAMPA (−logP_e of 6.4 and 6.3, respectively). Taken together, these data suggest that IP⁺-containing MPs display much better membrane permeability than typical peptide macrocycles.

To directly compare the permeability of an IP⁺ macrocycle with non-IP⁺ analogues, we chose MP36 as a model. This compound displays a −logP_e value (5.94) close to the median of the 20 MPs analyzed. We then synthesized various analogues of MP36 (Figure 8C). MP47 and MP48 were created using intramolecular amination or acylation chemistry to close the ring. This is the equivalent of opening the pyridinium ring and deleting the positive charge. MP49 and MP50 were created by closing the ring through thioether bond formation and thus lack any trace of the IP⁺ moiety. The passive permeabilities of these four molecules were assessed using PAMPA. As shown in Figure 8A (yellow dots), each of these molecules displayed a passive permeability between 10- and 100-fold poorer than the IP⁺-containing MP36. This comparative assessment shows that the IP⁺ unit has a major positive effect on the passive membrane permeability of the MPs.

Entry of IP⁺ MPs into Living Cells.

While the PAMPA method is used routinely to assess passive membrane permeability, it employs an artificial membrane. To obtain a preliminary assessment of the ability of IP + MPs to enter living cells, we employed the chloroalkane penetration assay (CAPA).²² In this protocol, cells expressing HaloTag protein (HTP) are incubated with a chloroalkane-tagged (Ct) molecule of interest (MOI) for a defined period of time. Excess Ct−MOI is washed away, and the cells are then treated with a Ct-coupled fluorescent dye. Finally, excess dye is washed away, and the level of fluorescence retained by the cells is assessed by fluorescence cell sorting. This assay is repeated at several different MOI−CA concentrations. The more permeable the Ct-MOI conjugate, the less intense the cellular fluorescence will be since the HTP active site is blocked from reacting with the Ct-dye conjugate.

The chloroalkane-bearing IP⁺ macrocycles (MPs 51−54) shown in Figure 9 were synthesized, purified, and subjected to the CAPA using HEK293 cells stably expressing HTP. A Lipinski-compliant tryptophan-Ct conjugate was used as a control for comparative purposes. These MPs are analogues of MP35, MP36, MP37, and MP40 (Figure 8) but were constructed using Py25 (Figure 2B) to allow attachment of the Ct tag to the secondary amine of the piperidine moiety. As shown in Figure 9, the IP⁺ MP-Ct conjugates display CP₅₀ values between 0.9 μM and 2.7 μM, whereas the Trp-Ct Lipinski-compliant molecule displayed a CP₅₀ value of 0.11 μM (see Supporting Information Figures S88−S93 for the raw data). Thus, the MP−Ct conjugates are approximately 8- to 25-fold less cell permeable than the Trp-Ct control. We consider this to be a promising result given the much higher molecular masses of the MPs. A more comprehensive analysis of the cell permeability of these MPs is underway and will be reported in due course.

Figure 9.

Determination of the cell permeability of IP⁺ MPs. HEK293 cells stably expressing HTP were incubated with the indicated compound for 5 h at 37 °C. After washing, the cells were lysed, and the extract was treated with a biotin-Ct conjugate to alkylate HTP not already occupied by the MP-Ct conjugate. The extracts were then subjected to sodium dodecyl sulfate−polyacrylamide gel electrophoresis and western blotting using fluorescently labeled SA. The amount of fluorescent protein (normalized to total HTP) under the conditions employed are shown on the left. The structures of the compounds employed in this experiment are shown on the right.

Stability of IP + MPs under Physiological Conditions.

IP⁺ macrocycles have the potential to act as electrophiles. To determine if typical cellular nucleophiles are reactive with IP⁺ MPs, MP12, MP14, and MP16 were incubated with glutathione (5 mM) or 2-mercaptoethanol (5 mM) at 37 °C for 24 h. LC−MS analysis of the solutions showed that the IP⁺ MPs were stable under these conditions; no new peaks were observed. A similar experiment was performed using the amine nucleophiles hydrazine and piperidine with the same result. We conclude that the IP⁺ MPs are stable under physiological conditions with respect to reaction with nucleophiles.

The stability of IP⁺ MPs to acid, basic, oxidizing, and reducing conditions was evaluated by incubating the macrocycles with 5 mM NaIO₄, sodium dithionite (Na₂S₂O₄), or triisopropyl silane (TIPS) at 37 °C for 24 h. Again, no products were observed by LC−MS. Finally, the macrocycles were incubated in buffers with pH values ranging from 1 to 11. LC−MS analysis showed the macrocycles to be stable under all of these conditions.

The primary data for all of these experiments is shown in Supporting Information Figures S94−S96.

◼. DISCUSSION

MPs have been shown to be capable of engaging difficult-to-drug proteins with shallow binding pockets, but the relatively poor membrane permeability and bioavailability of most MPs have limited their utility in addressing intracellular targets. Therefore, there is a high degree of interest in the discovery of novel MPs with improved pharmacokinetic properties. Several investigators have developed variants of cell-penetrating peptides to solve this problem.²³ These are cationic oligomers that are thought to bind to the plasma membrane and trigger the formation of an endocytotic vesicle.²⁴ This is very different from achieving passive permeability in that the MP must eventually escape from the endosome. While there has been significant progress in this vein,²⁵ there is lingering concern about the toxicity of employing this mechanism of action to gain entry of MPs into cells.

It is well established that one of the principal barriers to the passive permeation of peptides across membranes is the presence of multiple polar, highly hydrated N−H amide bonds.^2,26 Indeed, the unusually high bioavailability of certain peptide-based natural products, such as cyclosporine, can be ascribed to a high degree of N-methylation as well as the ability to adopt conformations that “hide” the remaining N−H units from the solvent through intramolecular hydrogen bonding.² Thus, many strategies for improving the passive membrane permeability of MPs have focused on mimicking these characteristics. For example, some stapled peptides achieve this goal by stabilizing a helical structure with multiple intramolecular hydrogen bonds. Hydrophobic staples also generally improve permeability.²⁷ An emerging strategy, nicely represented by a recent report from Suga and colleagues,²⁸ is to incorporate into the macrocycle non-canonical amino acid units (in this case, one with two pyridine rings) that are positioned to enter into intramolecular hydrogen bonds with N−H units in the peptide chain.

An alternative approach, represented by the present study, is to include in the ring units that improve the passive permeability of the MPs irrespective of the conformation of the peptide. We hypothesized that the permanent positive charge on the IP⁺ heterocycle would attract IP⁺-containing MPs to a membrane and that the hydrophobic nature of the ring system would facilitate movement across it. The PAMPA data shown in Figure 8 strongly support this notion. 45% of the IP⁺-containing MPs analyzed displayed excellent passive permeability (−logP_e < 6.0), despite the fact that their molecular weights ranged from 650 to 850 Da. Even very large macrocycles with molecular weights of >1000 Da (MP18 and MP19) displayed moderate passive permeability (−logP_e = 6.4 and 6.3, respectively). This is clearly due to an effect of the IP⁺ unit, as demonstrated by a comparison of the membrane permeabilities of comparable peptides cyclized using different linkage strategies (Figure 8, yellow dots). Preliminary CAPA data also show that these IP⁺ MPs readily access the cytosol of living cells (Figure 9).

Another interesting point is that the substitution on the IP⁺ unit also influences permeability substantially (compare MP38-MP41 in the second row of Figure 8B). While we cannot rule out the possibility that this influences the conformation of the macrocycle, it seems more likely that this is a direct effect of the chemical and physical properties of the IP⁺ moiety. For example, the hydrophobic, electron-donating piperidine ring in MP40 appears to promote permeability to a greater degree than fused aromatic rings (MP38, MP39, and MP41). It will be interesting in the future to thoroughly assess substitution effects on the permeability of otherwise identical IP⁺ MPs.

It is useful to note that all of these molecules contain a C-terminal primary amide unit formed by cleavage of the RAM linker. This is detrimental to cell permeability. Thus, it is likely that even more permeable MPs can be developed by employing different linkage chemistry for the solid-phase synthesis. Likewise, incorporation of N-methylated amino acids would likely lead to improved permeability as well.

Combined with the broad utility of IP⁺ formation for the efficient closure of macrocyclic rings of different sizes and compositions, these data suggest that this class of molecules is ideally suited for use in combinatorial library synthesis and screening, with a reasonable expectation that the derived ligands against intracellular targets will show activity in living cells and animals.