Madelaine P.
Thorpe
a,
Abigail N.
Smith
a,
Daniel J.
Blackwell
b,
Corey R.
Hopkins
c,
Bjorn C.
Knollmann
b,
Wendell S.
Akers
d and
Jeffrey N.
Johnston
*a
aDepartment of Chemistry, Vanderbilt Institute of Chemical Biology, Vanderbilt University, Nashville, TN 37235-1822, USA. E-mail: jeffrey.n.johnston@vanderbilt.edu
bDepartment of Medicine, Vanderbilt Center for Arrhythmia Research and Therapeutics (VanCART), Vanderbilt University Medical Center, Medical Research Bldg IV, Room 1265, 2215B Garland Ave, Nashville, TN 37232-0575, USA
cDepartment of Pharmaceutical Sciences, College of Pharmacy, University of Nebraska Medical Center, Omaha, NE 68198, USA
dPharmaceutical Sciences Research Center, College of Pharmacy, Lipscomb University, Nashville, TN 37204, USA
First published on 15th August 2024
The number of peptide-like scaffolds found in late-stage drug development is increasing, but a critical unanswered question in the field is whether substituents (side chains) or the backbone drive passive permeability. The backbone is scrutinized in this study. Five series of macrocyclic peptidic compounds were prepared, and their passive permeability was determined (PAMPA, Caco-2), to delineate structure–permeability relationships. Each series was based on the cell-permeable antiarrhythmic compound ent-verticilide, a cyclic oligomeric depsipeptide (COD) containing repeating ester/N-Me amide didepsipeptide monomers. One key finding is that native lipophilic ester functionality can impart a favorable level of permeability, but ester content alone is not the final determinant – the analog with highest Papp was discovered by a single ester-to-N-H amide replacement. Furthermore, the relative composition of esters and N-Me amides in a series had more nuanced permeability behavior. Overall, a systematic approach to structure–permeability correlations suggests that a combinatorial-like investigation of functionality in peptidic or peptide-like compounds could better identify leads with optimal passive permeability, perhaps prior to modification of side chains.
The elucidation of structure–permeability relationships for peptide and peptide-like compounds remains a critical need, and cyclic oligomeric depsipeptides8 provide a powerful platform for filling this gap. The cyclic depsipeptide ent-verticilide is the first potent and selective inhibitor of the intracellular calcium channel RyR2 (Fig. 1).9–11 Its activity has been recapitulated in vivo with animal models of disease,12 and a lower oligomer exhibits similar inhibition.13 More recent findings indicate a favorable pharmacokinetic–pharmacodynamic (PK–PD) profile.14 Our interest in defining the pharmacophore through in vitro assays15 led to a broader study to examine molecular features of ent-verticilide that enhance or diminish its permeability. Moreover, the study of ‘Beyond Rule of 5’ (bRo5) molecules, especially macrocyclic compounds, demands a better understanding of their physicochemical properties.16 The recent report of Sando and coworkers17 prompts us to report a concomitant investigation comparing the passive permeability18 of a cyclic depsipeptide, building on similar reports by others19–22 interested in the physicochemical characteristics (e.g. solubility, permeability) of compounds in this size regime.23,24 Use of a cyclic oligomeric depsipeptide template for this study provided an opportunity to systematically correlate single, double, triple, and quadruple point-replacements of esters in the parent tetradepsipeptide structure with both N-H and N-Me amides.15
Fig. 1 The ent-verticilide cyclooligomeric depsipeptide structure, its general features, and this study's approach to deconvolute the effect of backbone modifications on membrane permeability. |
ent-Verticilide and a complete set of 24 analogs over five series (Series 1–5) promised the opportunity to disentangle direct effects of these modifications with potential property-driving changes in conformation, polarity, and hydrogen bonding. An analysis of marketed drugs scrutinized the structural features of oral and injectable drugs, noting that structural differences were more often within the molecular backbones than the side chains of those scaffolds.25 The combination of ester and amide functional groups in the ent-verticilide backbone provides a unique opportunity to further probe the hypothesis that backbone modifications can substantially affect passive permeability independent of side chain modification. Furthermore, the oligomeric nature of ent-verticilide provides an additional angle to query to what extent these effects might be additive. We report herein the first outline of common bioisostere interconversions, and correlation of these structural changes with passive permeability. The macrocyclic nature of these depsipeptides and peptides, and their relationship to a cell-permeable antiarrhythmic, broadens the relevance of these findings, increasing the possibility that similar correlations will be identified for use in bRo5 drug development.26
N-Me | N-H | logP | P app (10−6) | logPapp | Diffusion (%) | Recovery (%) | HBD | HBA (Est/Am) | |
---|---|---|---|---|---|---|---|---|---|
a Definitions: N-Me: N-Me amide count in structure; N-H: N-H amide count in structure; logP: determined from standard curve to measure [D] and [A], using equation detailed in SI 2.4b; Papp: determined from a ratio-based method, given as mean ± standard deviation; diffusion: [acceptor]/[acceptor + donor]; recovery: acceptor + donor/total analyte; HBD: potential hydrogen bond donor; HBA: potential hydrogen bond acceptor, Est: ester, Am: amide. Italic font highlights compounds of higher permeability: bold italic ≥ 30% diffusion, non-bold italic = 20–30% diffusion. b General details describing methods and analysis, including standard deviation, are in the ESI. c For reference, cyclosporin A is a cell-permeable immunosuppressant with 4 N-H and 7 N-Me amides. The N-H amides combine with one hydroxyl for a total of 5 HBDs. Also for CsA, 11 amides and one hydroxyl provide a total of 12 HBAs. | |||||||||
CsAc | 7 | 4 | −8.96 | 8.28 | −5.01 | 28 | 86 | 5 | 0/11 |
ent-verticilide | 4 | 0 | −8.99 | 6.92 ± 0.66 | −5.21 ± 0.09 | 25.6 ± 2.8 | 94 | 0 | 4/4 |
1.1 | 3 | 1 | −9.18 | 5.41 ± 0.56 | −5.27 ± 0.04 | 19.4 ± 1.6 | 91 | 1 | 4/4 |
1.2 (adj) | 2 | 2 | −9.14 | 4.87 ± 0.35 | −5.31 ± 0.03 | 19.0 ± 2.0 | 88 | 2 | 4/4 |
1.3 (alt) | 2 | 2 | −9.82 | 1.15 ± 0.06 | −5.94 ± 0.02 | 5.1 ± 0.4 | 91 | 2 | 4/4 |
1.4 | 1 | 3 | −9.72 | 1.14 ± 0.23 | −5.95 ± 0.09 | 5.7 ± 1.3 | 91 | 3 | 4/4 |
1.5 | 0 | 4 | −9.36 | 3.51 ± 0.54 | −5.46 ± 0.07 | 12.1 ± 2.5 | 87 | 4 | 4/4 |
2.1 | 4 | 1 | −8.82 | 12.49 ± 2.49 | −4.91 ± 0.09 | 36.9 ± 4.6 | 88 | 1 | 3/5 |
2.2 (adj) | 4 | 2 | −8.94 | 9.04 ± 0.71 | −5.05 ± 0.03 | 30.4 ± 2.3 | 89 | 2 | 2/6 |
2.3 (alt) | 4 | 2 | −9.03 | 5.93 ± 0.51 | −5.23 ± 0.04 | 25.0 ± 1.6 | 90 | 2 | 2/6 |
2.4 | 4 | 3 | −9.17 | 4.76 ± 0.75 | −5.33 ± 0.07 | 19.1 ± 2.3 | 89 | 3 | 1/7 |
2.5 | 4 | 4 | −9.14 | 4.87 ± 0.48 | −5.31 ± 0.04 | 17.5 ± 2.3 | 86 | 4 | 0/8 |
3.1 | 5 | 0 | −9.09 | 6.14 ± 0.80 | −5.22 ± 0.05 | 22.4 ± 2.6 | 84 | 0 | 3/5 |
3.2 (adj) | 6 | 0 | −9.18 | 4.28 ± 0.40 | −5.37 ± 0.04 | 18.6 ± 2.4 | 94 | 0 | 2/6 |
3.3 (alt) | 6 | 0 | −9.21 | 3.32 ± 0.24 | −5.48 ± 0.03 | 17.2 ± 0.9 | 91 | 0 | 2/6 |
3.4 | 7 | 0 | −9.32 | 3.81 ± 0.32 | −5.42 ± 0.04 | 12.9 ± 0.9 | 93 | 0 | 1/7 |
3.5 | 8 | 0 | −9.43 | 2.83 ± 0.40 | −5.55 ± 0.06 | 11.0 ± 2.2 | 90 | 0 | 0/8 |
4.1 (=3.5, 5.1) | 8 | 0 | −9.43 | 2.83 ± 0.40 | −5.55 ± 0.06 | 11.0 ± 2.2 | 90 | 0 | 0/8 |
4.2 | 7 | 1 | −9.15 | 4.78 ± 0.41 | −5.32 ± 0.04 | 19.9 ± 1.7 | 89 | 1 | 0/8 |
4.3 (adj) | 6 | 2 | −9.20 | 4.05 ± 0.24 | −5.39 ± 0.03 | 18.0 ± 1.5 | 91 | 2 | 0/8 |
4.4 (alt) | 6 | 2 | −9.42 | 2.13 ± 0.18 | −5.68 ± 0.04 | 11.3 ± 0.7 | 92 | 2 | 0/8 |
4.5 | 5 | 3 | −9.52 | 1.50 ± 0.15 | −5.83 ± 0.04 | 9.0 ± 1.0 | 88 | 3 | 0/8 |
4.6 | 4 | 4 | −9.93 | 0.34 ± 0.07 | −6.48 ± 0.08 | <5 | 91 | 4 | 0/8 |
5.1 (=3.5, 4.1) | 8 | 0 | −9.43 | 2.83 ± 0.40 | −5.55 ± 0.06 | 11.0 ± 2.2 | 90 | 0 | 0/8 |
5.2 | 7 | 1 | −9.10 | 5.73 ± 0.38 | −5.24 ± 0.03 | 23.8 ± 1.5 | 91 | 1 | 0/8 |
5.3 (adj) | 6 | 2 | −9.14 | 3.81 ± 0.32 | −5.42 ± 0.04 | 20.9 ± 2.1 | 90 | 2 | 0/8 |
5.4 (alt) | 6 | 2 | −9.26 | 4.01 ± 0.45 | −5.40 ± 0.05 | 15.9 ± 2.4 | 90 | 2 | 0/8 |
5.5 | 5 | 3 | −9.42 | 2.73 ± 0.35 | −5.57 ± 0.06 | 10.0 ± 2.1 | 88 | 3 | 0/8 |
5.6 (=2.5) | 4 | 4 | −9.14 | 4.87 ± 0.48 | −5.31 ± 0.04 | 17.5 ± 2.3 | 86 | 4 | 0/8 |
Series 2 explores the hypothesis that the native ester functionality is beneficial to permeability, but replacement of one or more esters with an N-H amide could provide a HBD that might improve permeability by balancing the number of HBAs. A scenario where a fraction of the macrocycle's HBAs are internally bonded by HBDs is envisioned to provide an alternative to intermolecular water solvation. Five analogs were prepared that sequentially replaced each ester with an N-H amide, including the double replacement [NH]2 that offers adjacent (2.2) and alternating (2.3) sequences for the didepsipeptides. A single replacement (2.1) increased diffusion substantially from ent-verticilide (26 → 37%). This increase held with the adjacent, double replacement as well (2.2), increasing diffusion to 30%. The alternating double replacement (2.3) exhibited 25% diffusion, and subsequent replacement of esters with N-H amides lowered diffusion further (2.4: 19%, 2.5: 18%).
Series 3 is analogous to Series 2, but questions whether replacement of ester with N-Me amide (instead of N-H amide as in Series 2) will retain or diminish the permeability. The N-Me amide provides increased polarity but HBA capability similar to an ester. The increased conformational mobility of tertiary amide, relative to ester, might lead to nuanced effects on the compound's ability to diffuse through the membrane. Based on these expectations, the diffusion profile for this series was perhaps narrower than expected (22 → 11%) but followed a nearly linear decrease in diffusion with each replacement of ester for N-Me amide. This includes a 3% decrease from [NH]1 (3.1) to adjacent-[NH]2 (3.2), and a further 2% decrease for alternating-[NH]2 (3.3). Since the replacement of an ester for N-Me amide does not alter the HBA or HBD qualities, it is tempting to use this trend to propose a 2–4% decrease in diffusion for ester → N-Me amide replacement. Conversely, the N-Me → ester replacement may provide a 2–4% increase in diffusion, similar to the recent findings of Sando.17
Peptide Series 4 design includes the replacement of esters (of parent ent-verticilide) with N-Me amides, while investigating the demethylation of alanine nitrogen to probe the effect of an increase in polarity and HBD/HBA level. Compared to the parent all-N-Me peptide 4.1 (11% diffusion), analog 4.2 provided a positive spike in permeability, at 20% diffusion. A second replacement of N-Me for N-H sustained this spike (4.3, 18% diffusion), suggesting that internal hydrogen bonding may both attenuate the increase in hydrophilicity and encourage a more lipophilic conformation. Further N-Me to N-H conversion provided regular decreases in diffusion (4.4–4.6), with a 3 N-H:5 N-Me ratio providing 9% diffusion, and a 4:4 ratio delivering diffusion below the limit of detection.
Series 5 explores substitution of the esters with combinations of N-Me and N-H amides, beginning from the all-N-Me analog 3.5 (relabeled 5.1). Analog 3.5/4.1/5.1 was among those with the poorest passive diffusion (see also 1.3, 1.4). A single N-H amide among seven N-Me amides provided analog 5.2 with diffusion restored to 24%, nearly equivalent to ent-verticilide. Two N-H amides among 6 N-Me amides was slightly less membrane-permeable in the adjacent isomer (5.3). The alternating positional isomer (5.4) exhibited diminished permeability (16%). The lowest permeability in this Series was observed with 5.5 (11% diffusion) at a 5:3 N-Me/N-H ratio, whereas replacing a fourth N-Me amide with N-H resulted in an analog with 18% diffusion. Interestingly, this N-Me/N-H amide analog 5.6 is the same as 2.5, which was the least permeable analog in Series 2, although its Series 2 analogs collectively exhibited good permeability.
• The number of hydrogen bond donors (HBDs) in a macrocycle does not always negatively affect permeability. Compounds with the maximal number (4 HBDs) of N-H amides ranged from 12% (1.5) to 18% diffusion (2.5, 5.6), compared to the least permeable compounds 1.3 (2 HBDs) and 1.4 (3 HBDs).
• Within a series containing no HBDs, but varying relative ester/amide content, permeability generally increased as ester content increased from 11% for all-N-Me amide (3.5) to 22% for 3 esters/5 N-Me amides (3.1). This trend mirrors that identified in Sando's recent study.17
• The ester content alone does not drive permeability. Compound ent-verticilide (4 esters, 26%) was at the high-permeability range, while 1.4 (4 esters, 6%) exhibited the lowest permeability of 24 analogs.
• Considering the compounds composed only of HBAs, higher permeability correlated with a balance between ester and N-Me amide (ent-vert, 26%; 3.1, 22%), whereas those with a ratio favoring N-Me amides over esters exhibited lower permeability (2 ester:6 N-Me amide: 3.2, 19%; 3.3, 17%; 1 ester:7 N-Me amide: 3.4, 13%; only N-Me amide: 3.5, 11%).
• Considering the compounds that offered a symmetry element, resulting in the possibility of adjacent or alternating positioning of the unique didepsipeptides (4 pairs: 1.2/1.3, 2.2/2.3, 3.2/3.3, 5.3/5.4), all adjacent isomers exhibited higher permeability (>19%), while all alternating isomers exhibited lower permeability (<16%) with one exception (2.3, 23%).
• All high N-H-content compounds (4 HBDs) were in the lower half of permeability measured (<18%). However, the absence of N-H-content did not at all correlate with a permeability trend: ent-vert, 3.1, and 3.2 were >18% diffusion, while 3.3, 3.4, and 3.5 were <17% diffusion.
ent-Verticilide's eight residues are simplified by its oligomeric nature [α-hydroxy-heptamide-D-Ala], providing key differences in size and symmetry, both of which offer additional analogs and data for passive membrane permeability. The Sando study focused on a cyclic hexapeptide (Tyr-Leu-DLeu-Leu-Leu-DPro) model selected for its known low membrane permeability, finding that replacement of N-H amide at any position except DLeu3 enhanced permeability. Double substitutions or higher were not explored. Paradoxically, these permeability-increasing substitutions involved both solvent-exposed and unexposed N-H amides. Sando also showed that the amide-to-ester substitution has little conformational effect, while an N-Me amide substitution changed the conformation significantly but, in this case, did not change the number of solvent-exposed N-H amides.
The symmetry of ent-verticilide provides possible compensatory interactions via distal functional groups. One finding is consistent: the ester functional group can provide enhanced membrane permeability despite the presence of other N-H or N-Me amides. However, the template studied here addresses the potential synergism between residues across the macrocycle, including the possible compensatory role a residue may play to drive permeability retention. Compensatory behavior may result in the chameleon characteristic often invoked for some peptides (i.e. cyclosporin A)35 where a conformational change is believed to interconvert exposed and unexposed functionality as needed for transition from aqueous to membrane to aqueous environments to enter a cell.21,36 This may contribute to the finding here that introduction of a single N-H amide, at the expense of an ester, led to the analog with highest passive permeability (2.1). The second-highest permeability was found in an analog where two N-H amides (adjacent didepsipeptides) replaced two esters (2.2), and its non-adjacent isomer was essentially equipermeable to ent-verticilide. Importantly, the solid-state structure for N-H ent-verticilide (1.5) reveals a conformation that presents an all-pentyl lipophilic face on one side of the ring plane, and an all-methyl face on the opposite side.13 Within the 24-membered ring backbone of 1.5, two of the four N-H amides form intramolecular hydrogen bonds with ester carbonyls, while the remaining two are solvent-exposed (Fig. 4).
Fig. 4 Comparison of Series 1 analogs with a map of hypothesized hydrogen bonds based on the solid-state structure of 1.5. |
In these cases, the beneficial effect of amide N-methylation can be scaffold-dependent. In others, the effect of methylation may be more nuanced. For example, analog 1.5 exhibits two intramolecular N-H⋯OC intramolecular H-bonds, and two solvent-exposed amide N-H bonds.13 While it is true that N-methylation of each amide increases permeability, the change is rather abrupt from low permeability (1.3/1.4/1.5) to moderate permeability (ent-verticilide/1.1/1.2). Analog 1.3 is especially peculiar, since it has two intramolecular H-bonds that can form analogous to 1.5, leaving two solvent-exposed N-Me amides. This network should lead to good permeability, but it does not (5% diffusion). Similarly, analog 1.2 can engage in an intramolecular hydrogen bond with one N-H amide, while the other would be solvent-exposed. Permeability of this analog was similar to 1.1 whose ‘exposed’ amide is an N-Me (19% vs. 20% diffusion).
Some inter-Series observations are important. In all Series, substantially better diffusion was exhibited by the adjacent positional isomers. For example, a comparison of adjacent analog 1.2 to the alternating isomer (1.3) in Series 1 revealed a 19% vs. 5% diffusion. This is a rather dramatic example where the “total polar content” of the backbone alone does not drive permeability. Instead, its arrangement within the structure can either preserve or diminish its diffusion properties. This trend was observed in all other cases as well, with 30% vs. 25%, 19% vs. 17%, 18% vs. 11%, and 21% vs. 16% for Series 2–5, respectively.
Series 3 suggests that ester and N-Me amide functionality may be somewhat interchangeable with an increase in diffusion associated with the ester. This behavior mirrors that reported recently.17 However, this correlation may only hold when the overall HBA/HBD content is unchanged, as it was in Series 3.
Series 2 provided both the highest level of increased permeability for an analog (2.1), and the broadest range of passive diffusion (18–37%) via decreasing ester content by converting the depsipeptide to a mixed N-Me/N-H peptide. Unlike the trend noted in Series 3, this trend is opposite to recent findings that increasing ester content beginning from a peptide always improves passive permeability. Such an effect may be more nuanced, with some combination of backbone ester/N-Me/N-H amide constitution driving optimal permeability. The remaining Series provided somewhat similar overall ranges of diffusion: 10, 13, ∼13, and 14 for Series 1, 3–5 respectively. Such a consistent range might lead one to speculate that the backbone composition has an ultimately limited effect, but the unmeasurable permeability for 4.6 clearly demonstrates the dominance of backbone over aliphatic side chains.
Fig. 5 Assay of select compounds by Caco-2 permeability assay (left) and PAMPA/Caco-2 permeability correlation plot (right). aGeneral details describing methods and analysis are in the ESI.† Donor concentration (Caco-2) = 10 μM. ND = not detectable. ER = efflux ratio (AB/BA). Right panel: the number of ester functional groups for each compound is listed inside the box. Colors highlight compounds in different Series (green = Series 1, blue = Series 2, red = Series 3). Only compounds with both PAMPA and Caco-2 measurements are included. Dotted line indicates [PAMPA/Caco-2] = 1. PAMPA data is summarized in Table 1. |
Depsipeptide 1.5 is the all-N–H analog of ent-verticilide that exhibited attenuated diffusion (12%) as expected relative to ent-verticilide (26%). Its evaluation in the Caco-2 assay revealed a Papp (A–B) = 0.186, consistent with attenuated diffusion, and a 0.482 efflux ratio. Series 2 was chosen for complete evaluation since this series displayed a near-linear trend and expanded passive permeability in PAMPA (37% → 18% diffusion) for analogs 2.1 → 2.5. These analogs sequentially replaced the esters in verticilide with N-H amides, with the expected decrease in PAMPA permeability. In the Caco-2 assay, Papp (A–B) values were 0.978, 0.349, ND,37 2.28, 2.89 for 2.1–2.5, respectively (Fig. 5).
Evidence for transport-mediated efflux was observed for 2.2 with an efflux ratio of 6.09. Analog 3.5 was also evaluated in Caco-2 as a low-permeability peptide analog with all N-Me amides since methylation of amides is known to improve permeability. Analog 3.5 exhibited moderate permeability in Caco-2 with Papp (A–B) = 1.99 and an efflux ratio of 0.928. Overall, both the Caco-2 and PAMPA data generally indicate that these select analogs were generally low to moderate permeability compounds. However, Papp estimates in the PAMPA assay were consistently higher for each compound than observed in the Caco-2 assay. From an absolute permeability perspective two distinct groupings were noted. Group 1 (ent-verticilide, 1.5, 2.1) whose permeability values by PAMPA were higher relative to Group 2 (2.4, 2.5, 3.5). This permeability relationship was inverted in the Caco-2 assay with compounds in Group 2 exhibiting higher permeability values compared to compounds in Group 1. However, a similar rank order of permeability was observed within Group 1 (2.1 > ent-verticilide > 1.5) and Group 2 (2.5 > 2.4 > 3.5) for both the PAMPA and Caco-2 assay. Compound 2.2 or 2.3 could not be evaluated in the absolute or relative permeability comparisons due to a high efflux ratio and undetectable compound levels, respectively, in the Caco-2 assay. Overall, differences in permeability between the two assays most likely reflect the differences inherent among them, specifically passive vs. transport-mediated behavior.
Using the octadepsipeptide ent-verticilide lead, we prepared 24 analogs, within a relatively narrow molecular weight range, and measured the passive permeability of each using PAMPA. Numerous trends emerged that highlight the ability to substantively modify – either enhance or decrease – passive permeability without side chain modification. In one case, Papp increased 80% over ent-verticilide by the non-intuitive replacement of one backbone ester for an N-H amide. The overall trends can be used as a critical informer study to prioritize specific backbone modifications in peptide drug development. For example, for cases when ester-to-amide substitution reduces inhibitory activity in cells but not in vitro, the ability to disentangle passive permeability effects could speed further development. Overall, this study provides a rigorous starting point for understanding the relative contributions of polar backbones with lipophilic side chains in a manner that could be leveraged in optimization studies with non-oligomeric lead compounds. The molecular basis for these behaviors is the subject of ongoing studies and will be reported in due course.
Footnote |
† Electronic supplementary information (ESI) available: Complete experimental details (PDF); NMR data (PDF). See DOI: https://doi.org/10.1039/d4sc02758b |
This journal is © The Royal Society of Chemistry 2024 |