Structure–Activity Relationships for the N-Me- Versus N-H-Amide Modification to Macrocyclic ent-Verticilide Antiarrhythmics

Abstract

The synthesis of all N-Me and N-H analogues of ent-verticilide is described, enabling a structure–activity relationship study based on cardiac ryanodine receptor (RyR2) calcium ion channel inhibition. The use of permeabilized cardiomyocytes allowed us to correlate the degree of N-methylation with activity without concern for changes in passive membrane permeability that these modifications can cause. A key hypothesis was that the minimal pharmacophore may be repeated in this cyclic oligomeric octadepsipeptide (a 24-membered macrocycle), opening the possibility that target engagement will not necessarily be lost with a single N-Me → N-H modification. The effect in the corresponding 18-membered ring oligomer (ent-verticilide B1) was also investigated. We report here that a high degree of N-methyl amide content is critical for activity in the ent-verticilide series but not entirely so for the ent-verticilide B1 series.

The conformation and stereoelectronic characteristics of tertiary amides are extremely different compared to secondary amides with as little as a substitution of methyl (N-Me) for hydrogen (N-H). These differences extend to their influence on small molecule and peptide conformation and, particularly, with their direct engagement with a biological target and metabolic stability.¹ Exploration of the effect of interconverting N-H and N-Me substructures is therefore a critical variable during early stage hit-to-lead studies.² Nature has also leveraged this variable during post-translational modifications to increase natural product diversity.³ This is especially true in the peptide and depsipeptide natural product space, where the extremes of all-N-H and all-N-Me amide are well-represented, and most every degree in between when multiple amides are present.

We recently reported the discovery that 1, the enantiomer of verticilide (a naturally occurring cyclic oligomeric depsipeptide),⁴ is a potent antiarrhythmic that inhibits the intracellular cardiac calcium release channel, RyR2.⁵ Pathological RyR2-mediated calcium leak is associated with several atrial and ventricular arrhythmias.⁶ent-Verticilide is a 24-membered macrocycle bearing four N-methyl amides (Figure 1). The extent to which this high degree of methylation is required for activity is not yet known. Our recent finding that the 18-membered macrocyclic oligomer (ent-verticilide B1) exhibits a similar effect to that of ent-verticilide provided a second platform with which to examine the importance of each N-methyl amide.⁷ In this study, we show that N-methylation is critical for activity in both hit compounds but that some residual activity can be observed in certain cases. Furthermore, assays in cardiomyocytes with permeabilized membranes remove the possibility that such changes affect activity by changing the membrane permeability.

Figure 1.

ent-Verticilide template (n = 4) and ent-verticilide B1 (n = 3) for N-Me amide scan evaluation in the calcium sparks inhibition assay.

We hypothesized that the N-Me amides play a role in ent-verticilide activity, since each contributes to the conformation and rigidity of the 24-membered ring. Unlike other cases where N-methyl → N-H replacements will either modify receptor engagement directly or through a remote effect, the oligomeric nature of ent-verticilide was attractive as a template that would likely retain some form of the pharmacophore despite a single N-Me → N-H replacement. This hypothesis was validated in a ring-size oligomer study wherein the 18-membered oligomer exhibited only a modest loss of activity despite the dramatic change in ring size.⁷ Decreasing ring size further, or extending to larger ring sizes, led to inactive compounds. A concise, complete examination of the importance of N-methyl amides to ent-verticilide was expected to provide further insight into its pharmacophore and a second opportunity to determine the activity as sequential modifications are made to each depsipeptide. The first series of analogues are listed in Figure 2, ordered by decreasing N-Me amide content for ent-verticilide for a total of six compounds (1–6). While only one isomer is possible for [NMe]₄ (1), [NMe]₃ (2), [NMe]₁ (3), and [NMe]₀ (6), adjacent and alternating isomers are possible for [NMe]₂ (4, 5). Collectively, we expected that these analogues would reveal whether a substructure of ent-verticilide constitutes a minimal pharmacophore requiring an N-Me. For example, ent-verticilide may contain as many as four pharmacophore copies if a single didepsipeptide (or smaller) oligomer is responsible for RyR2 binding, while activity for [NMe]₀ would suggest that the α-oxy amide and α-amino ester substructure is primarily responsible for target engagement.

Figure 2.

Analogue design: decreasing amide N-methylation series (1 → 6), including two [NMe]₂ regioisomers.

Our original route to verticilide oligomers involved preparation of 1 by Mitsunobu-based oligomerization/macrocyclization.⁸ The permethylation step produced ent-verticilide in 78% yield.^9,10 This synthetic route afforded ent-verticilide concisely in six steps (7 → 6 → 1, Scheme 1). This method of synthesis offers several advantages, with the most obvious being brevity. Instead of employing a traditional approach involving chain elongation through condensative couplings, the elongation steps and cyclization are accomplished in one step without protecting groups. This allows for a drastically reduced step count and rapid access to analogues.⁸

Scheme 1. Preparation of ent-Verticilide (1) and N-H Analogues (6, 4) Using a Macrocyclooligomerization (MCO) Approach.

Aside from permethylation product 1, the alternating dimethylated analogue (4) could be formed in 6% yield when limiting the potential for permethylation by decreasing the reaction time.¹¹ In this experiment, both ent-verticilide (1) and 4 were formed and could readily be separated and isolated via preparatory high-performance liquid chromatography (HPLC). Analogues 2, 3, and 5 were not readily prepared by partial methylation, so we turned to a traditional approach to their synthesis that used conventional amide bond formation as outlined in Scheme 2. Sunazuka has reported several pathways to nat-verticilide using fragment couplings in both solution and solid phases,¹² but our enantioselective synthesis of the α-oxyacid⁷ provided the most effective means to access the N-methyl amide series for ent-verticilide described here. As outlined in Scheme 2, hexanal was converted into α-hydroxy ester 8 in 47% yield as previously described.^7,13 Coupling of alcohol 8 with either N-Me or N-H alanine provided didepsipeptides 9 & 10 in consistently high yield. In this manner, both N-H and N-Me didepsipeptide analogues could be accessed at a 15 g scale. This subsequent pathway was bifurcated, with half of the didepsipeptide material subjected to hydrogenolysis to produce acids 11 & 12 and the other half treated with trifluoroacetic acid to produce amines 13 & 14. This pair of intermediates was coupled using bromotripyrrolidinophosphonium hexafluorophosphate (PyBroP) to form the tetradepsipeptide (15–17) with yields ranging from 86 to 94%. In similar fashion to the didepsipeptides, the tetradepsipeptides were bifurcated with half subjected to hydrogenolysis to produce acids 18–20 and the other half treated with trifluoroacetic acid to produce amines 21–23. Subsequent coupling using PyBroP afforded octadepsipeptides 24–26 with yields ranging from 69 to 89%. Global deprotection was completed first using hydrogenolysis, followed by Boc-deprotection to afford the deprotected octadepsipeptide. Cyclization utilizing benzotriazol-1-yloxytripyrrolidinophosphonium hexafluorophosphate (PyBOP) afforded macrocycles 2, 3, and 5 with yields ranging from 11 to 23% after isolation via preparatory HPLC.

Scheme 2. Preparation of Analogues 2, 3, and 5 Using a Stepwise Approach.

The effects of the analogues on cardiomyocytes were studied using a Casq2 gene knockout mouse, a validated model of severe human catecholaminergic polymorphic ventricular tachycardia, an arrhythmia induced by pathologically increased RyR2 activity. Ventricular cardiomyocytes were isolated, permeabilized with saponin, and incubated with a control (dimethyl sulfoxide (DMSO)) or 25 μM of a selected compound for 15 min. RyR2 activity was measured in the form of calcium sparks, which are Ca²⁺ release events generated by spontaneous openings of intracellular RyR2 Ca²⁺ release channels. Normalized spark frequency for each analogue, including the hit compound ent-verticilide (at two concentrations: 3 and 25 μM), are summarized in Figure 3. In this series, each N-H for N-Me substitution made resulted in complete loss of activity. The ablation of activity when serially converting each of the four N-Me to N-H amides was not unexpected, but the lack of a clear and identifiable trend led us to recapitulate this in a different oligomeric series (Figure 4).

Figure 3.

Normalized activity of 24-membered 2–6 in the calcium sparks assay using permeabilized cardiomyocytes. Data presented as median with full range (error bars) and the 25th–75th percentiles bounded by the box. The hashed line provides a reference for DMSO (vehicle) conditions. (*) Bonferroni-adjusted p-value < 0.0001 vs DMSO.

Figure 4.

Analogous analogue design for ent-verticilide B1: decreasing amide N-methylation series.

We recently reported that ent-verticilide B1 (27), the 18-membered oligomer of ent-verticilide, inhibited calcium sparks similar to ent-verticilide.⁷ This provided a second series (Figure 4) with which to investigate the degree of N-methylation and its effect on activity. In the 18-membered ring series, all the N-Me/N-H analogues were accessed in a single preparation via 7 and the use of methylation conditions to generate the two partially N-methylated analogues 28 and 29 (Scheme 3) in corresponding yields of 22% and 34%.

Scheme 3. Synthesis of ent-Verticilide B1 (27) and N-Methyl/N-H Analogues.

These 18-membered ring analogues were evaluated in the calcium sparks assay as well (Figure 5). Interestingly, the monodemethylated amide analogue (28) of ent-verticilide B1 exhibited potent activity in the reduction of calcium sparks and, consequently, other properties of RyR2-mediated calcium leak (see Supporting Information). The didemethylated analogue 29, however, retained partial activity but lost some potency. The all-N-H amide analogue 30 was similarly potent as 29 but less potent than 28. These results show a clear dependence on N-methylation on activity with the 18-membered oligomer but certainly an unexpected trend in the context of the behaviors exhibited by the 24-membered analogues.

Figure 5.

Normalized activity of 18-membered 28–30 in the calcium sparks assay using permeabilized cardiomyocytes. Data presented as median with full range (error bars) and the 25th–75th percentiles bounded by the box. The hashed line provides a reference for DMSO (vehicle) conditions. (*) Bonferroni-adjusted p-value < 0.0001 vs DMSO.

The degree of N-methylation of nitrogen atoms is a common and vital chemical modification made in nature to regulate biological function, including N-methylation of histones and DNA methylation.¹⁴ Furthermore, there are countless examples of diverse, naturally occurring N-methylated linear and cyclic peptides. We anticipated that this modification might alter the biological activity of ent-verticilide depsipeptides, whether through conformational changes or disruption of intramolecular hydrogen bonding capabilities that modify binding.¹⁵ This tactic has been used successfully in studies of small peptides to probe early-stage structure–activity relationship (SAR).¹⁶ In a study of hemiasterlin, potency was doubled by N-methylation of a single amide without increasing its cytotoxicity (Figure 6).¹⁷ In this same study, the potency and cytotoxicity of SPA110 were each halved upon N-methylation. Although N-Me for N-H exchanges are expected to impact membrane permeability, the use of permeabilized cardiomyocytes in our work enabled a more direct interpretation of active versus inactive derivatives.

Figure 6.

Effects of N-Me → N-H modifications of hemiasterlin and HTI-286.¹⁷

Evaluation of analogues 2–6 in the calcium sparks assay revealed a strong dependence on amide per-N-methylation, leading to inactive analogues. This decrease in activity with increasing N-H amide content may be due to direct modification of the pharmacophore, or if there is a minimal pharmacophore within the cyclic oligomer, perhaps conformational changes substantially modulate target engagement. Furthermore, a new hydrogen-bond donor is added in these series with each N-Me → N-H exchange. It is well-known that N-H cyclic peptides tend to form tight, intramolecular hydrogen-bonding networks,^18,19 and this could significantly change the orientation of the aliphatic side chains. We have seen evidence of this type of internal hydrogen bonding in solid-state structures for both the 24- and 18-membered rings (6 & 30, Figure 7). In the 24-membered ring (6), two of four the N-H amides are buried inside the macrocycle, forming intramolecular hydrogen bonds with ester carbonyls. This intramolecular bonding in the solid state orients the four lipophilic pentyl side chains on the same face of the macrocycle. We hypothesize that this arrangement could (1) provide a nonpolar face for ent-verticilide to interact with the lipophilic membrane, (2) enable binding to RyR2 through either an all-nonpolar face or an all-polar face, or (3) provide polar and nonpolar faces to different molecules of RyR2 to promote the formation of the multimeric protein assembly. This same behavior is seen with the 18-membered N-H cyclodepsipeptide (30),^8c in which two of the three N-H amides are buried in intramolecular hydrogen bonds (Figure 7). While the N-methylated analogues have yet to succumb to crystallization, future efforts will be directed toward acquiring additional structural data for comparisons and insight to this activity difference. Overall, our data indicate that N-methylation is very important for activity.

Figure 7.

X-ray crystal structures of all-N-H amide analogues of ent-verticilide and ent-verticilide B1. Views of 30 (left column) and 6 (right column) through the macrocycle (first row) and from the side (second row) for each. Intramolecular hydrogen bond(s) denoted by dashed red line.

The degree of N-Me amide content in the oligomeric cyclic depsipeptide ent-verticilide was found to directly affect its activity as an inhibitor of calcium release. The unsymmetrical nature of these ent-verticilide analogues was addressed by either methylation of an N-H depsipeptide precursor or through a stepwise synthesis of linear precursors with the proper N-H/N-Me patterns. The degree of N-methylation was also investigated for the 18-membered ring size (ent-verticilide B1). This study uncovered an additional analogue with significant sparks inhibition. That this new hit resides within the 18-membered oligomer series and not in the 24-membered series was unexpected. These results suggest that a significant portion of the ent-verticilide hit constitutes the minimal pharmacophore but that its conformation may be more critical than the oligomeric structure suggests.

Glossary

Abbreviations

SAR

structure activity relationship

RyR2

cardiac ryanodine receptor

CASQ2

cardiac calsequestrin

PyBrop

bromo-tris-pyrrolidino-phosphonium hexafluorophosphate

PyBOP

benzotriazol-1-yloxytripyrrolidinophosphonium hexafluorophosphate

DIAD

diisopropyl azodicarboxylate

EDCl

1-(3-dimethylaminopropyl)-3-ethylcarbodiimide hydrochloride

DMAP

4-dimethylaminopyridine

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmedchemlett.2c00377.

• Complete experimental details; NMR data (PDF)

• X-ray crystallographic information (CIF)

• X-ray crystallographic information (CIF)

Author Contributions

^∥ Equal contribution.

The authors declare no competing financial interest.