Stereochemical Effects on the Antimicrobial Properties of Tetrasubstituted 2,5-Diketopiperazines

Abstract

Antimicrobial drug resistance is a looming health crisis facing us in the modern era, and new drugs are urgently needed to combat this growing problem. Synthetic mimics of antimicrobial peptides have recently emerged as a promising class of compounds for the treatment of persistent microbial infections. In the current study, we investigate five cyclic N-alkylated amphiphilic 2,5-diketopiperazines against 15 different strains of bacteria and fungi, including drug-resistant clinical isolates. Several of the 2,5-diketopiperazines displayed activities similar or superior to antibiotics currently in clinical use, with activities coupled to both the cationic and hydrophobic substituents. All possible stereoisomers of the lead peptide were prepared, and the effects of stereochemistry and amphiphilicity were investigated via 1D and 2D NMR spectroscopy, solution dynamics, and membrane interaction modeling. Clear differences in solution structures and membrane interaction potentials explain the differences seen in the bioactivity and physicochemical properties of each stereoisomer.

Antimicrobial drug resistance (AMR) is a global health crisis, and antibiotic resistant bacteria are becoming commonplace.¹ It has been estimated that 2.4 million people from high OECD areas will die over the coming 30 years due to infection by AMR bacteria, with an accompanying added annual healthcare cost of $3.5 billion USD.² New drugs are urgently required to combat AMR, and research and development in this field needs to uncover new classes of drugs with novel mechanisms of action.³ One such class of compounds is cationic antimicrobial peptides (cAMPs), which have an inherent potential for high specificity, high potency, low toxicity and, perhaps most importantly, a low incidence of AMR.⁴⁻⁶ Innate cAMPs are produced by animals and plants as a first line of defense against invading pathogens and can be highly potent broad-spectrum antimicrobials.^7,8 The predominant mode of action of cAMPs is at the plasma membrane, and this is thought to be the determining factor behind the low rates of AMR.^4,6 It was originally thought that cAMPs interacted with the plasma membrane in an achiral manner by merely disrupting the plasma membrane and causing cell lysis.⁹⁻¹¹ This theory is supported by many studies illustrating that enantiomeric pairs of AMPs often display similar bioactivities, with differences in efficacy attributed to the increased stability of AMPs containing unnatural d-amino acids.^9,10,12 However, some recent studies dispute this theory and posit more complex, and often multiple, mechanisms of AMP interactions with plasma membranes.¹²⁻¹⁵ Bacterial membranes are comprised of a complex mixture of components that often have set chiralities,^14,16 and it has recently emerged that cAMP–membrane interactions can be modulated via stereochemistry.^13,14,17

Despite the promising activity of cAMPs, developing them as drug candidates is a major challenge due to their inherent instability to enzymatic degradation, which results in short pharmacokinetic half-lives and low oral bioavailability.^5,18−20 The incorporation of unnatural amino acids has been successfully employed to increase the potency and prolong the half-lives of cAMPs, and recent research has expanded to other more stable backbone scaffolds.¹⁹⁻²³ 2,5-Diketopiperazines (DKPs) present an ideal scaffold from which to build novel antimicrobial compounds, as they are chemically accessible, highly stable, and amenable to extensive synthetic derivatization at several positions.²⁴⁻²⁷ 2,5-DKPs are cyclic dipeptides that exist in a planar, chair, pseudoboat, or twist form, with a small energy gap between the conformers (1.3–1.7 kcal/mol) that varies depending on the ring substituents.^25,28

The nature of DKP substitution and the stereochemistry around the C_α position can drastically vary the 3D structures and the bioactivities of these 2,5-DKPs.^29,30 Our previous study investigating N_α-alkylated-2,5-diketopiperazines as antimicrobial candidates, which focused primarily on diastereomeric mixtures, showed that the evaluated DKPs exhibited similar or superior antimicrobial activities when compared to those of structurally related linear peptides.²³ These compounds were designed around an antimicrobial pharmacophore, which dictated that the minimum requirement for potent antimicrobial activity was two regions of hydrophobicity and two regions of cationic charge.^23,31 For comparison, an enantiopure sample was also prepared and shown to display improved activity against all tested bacteria when compared to the diastereomeric mixture.²³ This observation was not further explained but suggests that stereochemistry can modulate the bioactivity of antimicrobial DKPs.

The present study aimed to investigate the role of stereochemistry on the bioactivity of cAMPs using a lead antimicrobial 2,5-DKP as a model. Three diastereomers (Figure 1) of an arginine-derived DKP (1) were synthesized using enantioselective methods. These stereoisomers were subsequently evaluated against a panel of human Gram-positive and Gram-negative bacterial pathogens (including several drug-resistant clinical isolates), yeasts, and a filamentous fungus. The compounds were also assessed for off-target toxicity toward mammalian cells. The results add fundamental new insights into the important role of stereochemistry in DKP bioactivity and for cAMPs in general. A combination of NMR and in silico molecular dynamics (including membrane interactions) were employed to understand the mechanism of interaction at a molecular level.

Figure 1.

Stereoisomers of cyclo(N-Bip-Arg-N-Bip-Arg) (l,l)-(1), (d,d)-(2), and (l,d)-(3), along with cyclo(N-Bip-Orn-N-Bip-Orn) (l,l)-(4) and cyclo(N-Bn-Arg-N-Bn-Arg) (l,l)-(5).

DKPs 1–3 (Figure 1) were synthesized as the three enantiopure stereoisomers using our recently published methodologies (Scheme 1).²⁷ DKP 4 and DKP 5 were included as controls to differentiate the physicochemical effects of the charged group and the hydrophobic contributions (Table 1).²⁷

Scheme 1. Synthesis of DKPs 1–3.

Reagents and conditions are as follows: (a) HATU, 6-Cl-HOBt, DIPEA, and DMF at rt for 5 h; (b) (i) TFA/DCM (1:1, v/v) at rt for 1 h and (ii) 0.1 M AcOH in s-BuOH and NMM at reflux for 16 h; (c) (i) H₂, Pd/C, and AcOH/DCM (1:1, v/v) at rt for 16 h and (ii) N,N′-di-Boc-1H-pyrazole-1-carboxamidine, DIPEA, and MeCN/H₂O (9:1, v/v) at rt for 16 h; and (d) (i) 4-phenylbenzyl bromide and KHMDS (1.0 M in THF) at −40 °C for 16 h, (ii) TFA/DCM (1:1, v/v) at rt for 1–2 h, and (iii) 0.1 M HCl.

Table 1. Physicochemical Properties of DKPs 1–5.

compound	MW (g/mol)	αD	Rfa	CSV (Å3)b	cLogD8.1
1	664.8	–104.9	19.4	145.6	–0.73
2	664.8	+104.9	23.4	145.6	–0.73
3	664.8	0	5.8	145.6	–0.73
4	560.7	–67.9	8.5	145.6	0.39
5	492.6	–15.0	0.6	87.9	–4.56

^aHPLC retention factor.

^bConnolly solvent-excluded volume for the hydrophobic side chains only.

Diastereomeric mixtures of DKPs have previously exhibited moderate to high antibacterial activity against a range of bacteria.²³ DKPs 1–4 displayed potent activities toward the bacterial and fungal strains investigated (Table 2). Of the 12 bacteria in this study, such as Klebsiella pneumoniae and methicillin-resistant Staphylococcus aureus (MRSA), five are listed as either critical or high on the WHO priority list of bacteria.³²

Table 2. Minimum Inhibitory Concentration (MIC) of DKPs 1–5 against a Range of Human Pathogens, Including Drug-Resistant Isolates.

MIC (μM)
	Compound					Antimicrobial controls
Cell type	1	2	3	4	5	ammox.a	polymyx.b	ampho.c
Bacteria
S. aureus	4	4	4–8	4	>64	1	64	–
MRSA	4	4	4–8	4–8	>64	>64	32–64	–
E. faecalis	2–4	4	16	8	>64	8	8	–
VRE	8	4–8	8–16	8	>64	64	>64	–
S. pneumoniae	8	8–16	16	8	>64	0.5–1	8	–
S. pyogenesd	8	8	16	8–16	64	0.25	32–64	–
E. coli	8	8	64	16	>64	16	0.25	–
E. colie	8	16–32	>64	16	>64	–	4–8	–
K. pneumoniaef	32–64	64	>64	8	>64	–	16–32	–
A. baumannii	64	64	>64	64	>64	–	1	–
A. baumanniig	16–32	32	>64	32	>64	–	0.25	–
P. aeruginosa	8	8	64		>64	>64	–	–
Fungi
C. albicans	4	2	8	64	>64	–	–	1
C. utilis	4	4	2	32	>64	–	–	2
A. fumigatus	8	8	64	32	>64	–	–	4
Control
NHDFh	63	99	175	30	>512	>512	>512	–

^aAmoxicillin.

^bPolymyxin B.

^cAmphotericin B.

^dMacrolide- and tetracycline-resistant S. pyogenes HKU16.

^eMultidrug-resistant (MDR) E. coli MS8345.

^fPandrug-resistant K. pneumoniae MS6671.

^gMDR A. baumannii AB5075.

^hLD₅₀ against normal human dermal fibroblasts (NHDF). A dash (−) indicates that the compound was not tested against the indicated organism or strain. See SI Table S1 for the complete strain list.

Clear differences in the physicochemical properties of the peptides could be experimentally observed (Table 1), while the computed theoretical lipophilicity was the same for DKPs 1–3. DKP 3 was significantly less hydrophobic than DKPs 1 and 2 (as judged by RP-HPLC), and these differences in polarity were reflected in the bioactivities of the peptides. All three diastereomers exhibited potent activity toward drug-sensitive S. aureus (MIC 4–8 μM) and MRSA (MIC 4–8 μM). DKPs 1–3 also exhibited potent activity toward the other Gram-positive bacteria, including Enterococcus faecalis, vancomycin-resistant E. faecium (VRE), Streptococcuspneumoniae, and macrolide- and tetracycline-resistant Streptococcus pyogenes. A slight reduction in antimicrobial activity was observed against the Gram-negative bacteria, but the compounds were still active. This observation is in agreement with previous studies on linear analogs.²² DKP 1 was the most potent across all Gram-positive strains (MIC 2–8 μM), with DKPs 2 and 3 being only marginally less active (Table 2). The cis-enantiomers 1 and 2 exhibited significantly higher activities against Gram-negative bacterial strains Escherichia coli (MDR isolate), Pseudomonas aeruginosa, K. pneumoniae, and Acinetobacterbaumannii in comparison to the trans-diastereomer DKP 3. DKPs 1 and 2 exhibited MIC values ranging from 8 to 64 μM, while DKP 3 was largely inactive at 64 μM, only inhibiting E. coli at the highest concentration tested. DKP 5 was inactive in all assays, presumably due to insufficient hydrophobic contributions.

The active DKPs were confirmed to act through a bactericidal mechanism, as determined by a minimum bactericidal concentration (MBC) assay (SI Table 13). In most instances, the MBC was observed to be equivalent or within one twofold dilution of the respective MIC. Most notable was the activity of the DKPs toward the six drug-resistant bacterial strains examined. While amoxicillin presented a MIC toward S. aureus (1 μM) comparable to those of DKPs 1–3, it was completely inactive (MIC > 64 μM) toward MRSA and VRE. In contrast, DKPs 1–3 maintained their potent activity toward the resistant strains. DKP 1 also exhibited an activity against polymyxin-resistant E. coli (8 μM) similar to those of the current antibacterial treatment polymyxin B (4–8 μM). Against the pandrug-resistant K. pneumoniae, the free-amine-bearing DKP 4 exhibited the most potent activity of all the evaluated compounds, with a MIC value of 8 μM.

DKPs 1–3 displayed potent activity toward the evaluated strains of yeast and filamentous fungus (Table 2). All three diastereomers displayed low micromolar activities toward both Candida albicans and Candida utilis yeasts, with MIC values ranging from 2 to 8 μM. These MICs were comparable to that of the related linear lead peptide LTX-109, which was previously shown to display superior activity to both amorolfine and terbinafine.³³ Toward the filamentous fungus Aspergillus fumigatus, DKPs 1 and 2 inhibited growth at 8 μM, while the trans-DKP 3 was significantly less active, only inhibiting growth at 64 μM. Notably, the activities of DKPs 1 and 2 toward the investigated clinical isolates of both C. utilis and A. fumigatus are comparable to that observed for the clinical gold-standard amphotericin B.³⁴

While native cAMPs are often highly selective toward pathogens,⁷ the clinical relevance of synthetic cAMPs has been hampered by toxic effects such as hemolytic activity and a narrow therapeutic index.⁵ Aside from DKP 5 (LD₅₀ > 512 μM), all DKPs displayed cytotoxicity within the test range toward NHDF cells (Table 2 and Figure S1), with values ranging from 30 to 175 μM. Among DKPs 1–3, the strong influence of the stereochemical configuration was observed, which was akin to the antimicrobial activity. This was again most pronounced for DKP 3 (LD₅₀ = 175 μM); however, enantiomeric DKPs 1 and 2 also showed varying toxicities, whereby the l-configuration (1) was generally favorable for microbial selectivity. One may conclude that the trans-diastereomer 3 is the most selective DKP with its reduced toxicity and Gram-negative activity but high potency toward Gram-positive bacteria and yeasts. This is in line with previous studies of AMPs with mixed stereochemistry^5,15,35 and illustrates how stereochemistry can toggle the therapeutic window. The high toxicity of some of the compounds is a drawback; free-amine derivative 4 is the most toxic of the five DKPs (LD₅₀ = 30 μM) but remains a lead for future development due to its potent activity toward K. pneumoniae.

Previous studies of cAMPs reported that diastereomers with mixed stereochemistry formed less defined amphipathic solution structures relative to the enantiopure l- and d-forms, as evidenced by reduced RP-HPLC retention times and loss of bioactivity.^12,36 A similar trend was observed in this study. The marked difference between the retention factors (R_f) of DKPs 1 (19.4) and 2 (23.4) and that of DKP 3 (5.8) implies the stereoisomers adopt different solution conformations (Table 1). 2D NMR studies were therefore performed to investigate the DKP structure and rationalize the differences in the observed bioactivities (Figure 2). Previous studies have correlated the decreased bioactivities and retention factors of cAMPs to a “shielding” effect exerted by the guanidine group on the hydrophobic element of the molecule.¹² When the hydrophobic moiety is in close proximity to the charged group, the hydrophobic element is “concealed” from the aqueous environment, thus increasing the aqueous solubility of the molecule.

Figure 2.

¹H NMR spectra (in MeOH-d₄) of DKP stereoisomers 1–3 (upper) showing the weak upfield shift in the H-4 and H-5 protons of DKP 3, which is characteristic of shielding effects from the hydrophobic groups. ROESY spectra of (l,l)-DKP 1 (lower left) and (l,d)-DKP 3 (lower right) indicating the close proximity between the arginine side chain (H-3 to H-5) and the biphenyl groups in DKP 3 through the presence of ROE cross peaks that are absent in DKP 1.

The same effect was seen in our study through an upfield shift in the resonances of the 1D ¹H NMR spectrum of DKP 3. The H-4 and H-5 signals were shifted upfield by 0.3 and 0.2 ppm, respectively, relative to corresponding cis enantiomers (Figure 2) due to the aromatic biphenyl group exerting a shielding effect on the arginine side chains. The close proximity between the arginine side chain (H-3 to H-5) and the biphenyl groups was further corroborated by 2D ROESY, where rotating Overhauser effects were observed (Figure 2). This observation further validates the observed differences in both the physicochemical properties and the antimicrobial activity. Conversely, H-2 was shifted downfield by 0.2 ppm for DKP 3. The coupling constants for H-2 in the cis enantiomers 1 and 2 and the trans-DKP 3 were also different. cis-DKPs 1 and 2 exhibited coupling constants of 9.3 and 4.3 Hz, respectively, whereas DKP 3 exhibited coupling constants of 5.5 and 2.9 Hz. (Figure 2). This observed difference for H-2 is consistent with the cis versus trans relationship at the C_α position.^37,38

To further understand the importance of stereochemistry for the DKP bioactivity, in silico molecular dynamics and membrane insertion studies were performed. Solution dynamic modeling predicted that both DKPs 1 and 2 would have largely amphipathic structures in solution. Both charged groups were predicted to be proximal to one another, extending away in the axial position at the C_α position, and the hydrophobic units were predicted to extend out in the opposite direction away from the DKP ring (Figure 3). This predicted conformation is in agreement with the NMR data and is also supported by previously postulated structures of bioactive tetrasubsituted DKPs.²⁶

Figure 3.

Predicted solution-phase conformations of cis-DKPs (a) 1 and (b) 2. The preferred amphiphilic structure on the left of each pair is highlighted.

Both DKPs 1 and 2 were predicted to form a minor secondary conformation. In both situations, the shift involved positioning a biphenyl group toward the arginine side chains, forming a tripod-shaped structure. These predicted conformations were observed for approximately 15 ns for both enantiomers, and the relative potential energy change for the (l,l)- and (d,d)-enantiomers was estimated to be +10.6 and +5.1 kJ/mol, respectively. The (l,d)-trans-DKP 3 has two distinct conformations, both of which are nonamphiphilic structures (Figure 4). The energy minimization studies predicted one conformation in which the arginine side chains sit trans to each other across the DKP ring with one of the side chains axial and one equatorial to the DKP ring, while the bulky groups remain unchanged at the nitrogen position. This conformation resulted in a “tripod” like structure. The second conformation occurred when one of the biphenyl groups “flipped” to the opposite side of the DKP core to lie proximal to the previously isolated arginine side chain (Figure 4). The calculated change in potential energy for the two conformations was 10.3 kJ/mol, indicating that the initial conformer is preferred. Despite the apparent asymmetry within DKP 3, the NMR data suggest that the two hydrophilic units and the two hydrophobic units are still in identical environments due to the lack of additional signals.

Figure 4.

Predicted solution-phase conformations of trans-(l,d)-DKP 3. The tripod like structure on the left is highlighted, while the structure on the right best matches the observed NMR spectra.

The predicted hydrogen bonding interactions between the DKPs and a model membrane were monitored to evaluate the interactions between the guanidine moiety and the phosphate groups on the membrane surface. The arginine side chains have a high capacity to form hydrogen bonds and are able to form stable bidentate hydrogen bonds with the phosphate groups of the plasma membrane.³⁹ The simulation of DKP 1 and 2 revealed an initial phase of membrane interaction during which hydrogen bonds formed between the positively charged guanidine moieties and the negatively charged phosphate group of the phospholipid head of the membrane. Both salt bridge and hydrogen bonding interactions were observed (Figure 5). Interestingly, DKP 3 was considerably slower in forming these interactions with the surface of the membrane and as a result spent much of the simulation in the solvent space (SI Figure 2).

Figure 5.

Initial interactions of (l,l)-DKP 1 with the membrane surface. Hydrogen bonding interactions are depicted with yellow dashed lines, salt bridges are depicted with pink dashed lines, and cation−π interactions are depicted with green dashed lines.

The interactions in the simulations of DKPs 1 and 2 continued for roughly 50 ns for DKP 1 compared to 10–20 ns for DKP 2. After these initial interactions, the arginine side chains were buried into the membrane, with the biphenyl groups remaining outside the membrane. A rotation of the entire compound (1 and 2) at 60–70 ns then allowed for a single biphenyl group to come into close proximity with the membrane, leading to a cation−π stack between a positively charged free amine on the membrane and the biphenyl group. This rotation of the DKP led to the phosphate being positioned between the arginine side chains. Evidently, this interaction caused the DKP to insert farther into the head layer of the membrane, with only one biphenyl group remaining outside. The frequency of hydrogen bonding interactions (between 4–6 interactions) is maintained during this period (SI Figure 2). At this stage, a shift was observed in the DKP conformation to a chair- or tripod-like structure due the rotation of an arginine side chain to the biphenyl side of the DKP. This change brought the entirety of the molecule into the hydrophilic head of the membrane. Shortly after, the biphenyl groups shifted into the hydrophobic space, while the arginine side chains were retained in the hydrophilic space. DKP 3 showed two distinct conformations during the simulations. The initial conformation was the tripod conformation with no salt bridges or hydrogen bonding interactions between the guanidine moiety and the phosphate group of the membrane. At 40 ns, DKP 3 shifted to the second conformation (Figure 4), which led more readily to binding with the membrane at 50 ns. A large number of relatively minor and rapid of interactions were subsequently observed until the compound stabilized at 75 ns; here, the compound was partially inserted into the membrane, with one biphenyl moiety still residing in the hydrophilic space. (SI Figure 3). To support the theoretical membrane interaction study, DKP 1, 3, and 5 were also assessed in an E. coli membrane integrity assay, which illustrated that the potent antimicrobial compounds are membrane disrupters (SI Figure S4).

In the current study we have prepared five cyclic N_α-alkylated amphiphilic 2,5-diketopiperazines and reported their antimicrobial activities against 15 different strains of bacteria and fungi, including drug-resistant clinical isolates. Improved or comparable antimicrobial activities with respect to those of clinically used antibiotics were observed against several of the resistant bacteria, including MRSA, VRE, and MDR-resistant E. coli. The role of stereochemistry on the bioactivity was studied, and the biological effects were explained by differences in solution structures as supported by NMR and molecular dynamics. A clear dependence on stable amphiphilic solution structures dictates the bioactivity of this class of small antimicrobials. Our observation provides fundamental insights into the structural requirements for optimal membrane interactions for these compounds but may also be extrapolated to rationalize the activities of larger native cAMPs. These effects impact both the antibacterial activity and the off-target cytotoxicity, enabling the design of peptides with a wider therapeutic index. Our study further illustrates how small mimics of cAMPs can be prepared in the cyclic DKP format while their activity is maintained in the clinically relevant range.

Commercially available starting materials were obtained from Sigma-Aldrich, Merck, or AK scientific and were used as received. Reactions performed at low temperatures were cooled with a Julabo FT402 immersion cooler. Reactions were monitored by thin-layer chromatography on silica gel plates (Merck) using UV light as a visualizing agent and an ethanolic solution of vanillin, potassium permanganate or ninhydrin, and heat as the developing agents. The retention factors and purities of all target compounds were assigned using achiral reverse-phase HPLC (Dionex P680 system using a Phenomenex Gemini C18–Si column (150 mm × 4.6 mm, 5 μm)). Compounds were eluted with an isocratic gradient of 26:74 A/B over 90 min at 1 mL/min (for retention factor analysis) or a gradient from 100:0 A/B to 0/100 A:B over 10 min at 1 mL/min (for purity); solvent A was water (+0.1% v/v trifluoroacetic acid), and solvent B was acetonitrile (+0.1% v/v trifluoroacetic acid). NMR spectra were recorded at room temperature in CDCl₃, MeOH-d₄, or DMSO-d₆ solutions on either Bruker DRX400 spectrometers operating at 400 MHz for ¹H nuclei and 100 MHz for ¹³C nuclei or a Bruker DRX-500 spectrometer operating at 500 MHz for ¹H nuclei and 125 MHz for ¹³C nuclei. Chemical shifts are reported in ppm and were measured relative to the solvent in which the sample was analyzed (CDCl₃ δH 7.26 ppm or δH 0.00 ppm (TMS) and δC 77.16 ppm; MeOH-d₄ δH 3.31 ppm and δC 49.00 ppm; DMSO-d₆ δH 2.50 ppm and δC 39.52 ppm). High-resolution mass spectra (HRMS) were obtained using a micrOTOF-Q II mass spectrometer.

The compounds were prepared according to Grant et al.²⁷ The DKPs were used at >95% purity. NMR spectra and the full spectroscopic characterization of all synthesized compounds are reported in the SI. MIC and MBC assays against bacterial strains (Tables 2 and S1) were performed in accordance with the recommended protocol from the Clinical and Laboratory Standards Institute (CLSI) (SI) in 96-well polypropylene microtiter plates at ∼5 × 10⁵ CFU/mL.⁴⁰ The susceptibility of the yeasts C. albicans SC5314 (type strain) and C. utilis SVB-Y1 was assessed by microdilution in accordance with the CLSI-recommended protocols at final inoculums from 0.5 × 10³ to 2.5 × 10³ CFU/mL.⁴¹ Antimicrobial susceptibility of Aspergillus fumigatus SVB-F136 was assessed by microdilution in accordance with the CLSI recommended protocols with final inoculums 0.4 × 10⁴ to 5 × 10⁴ CFU/mL.⁴² All assays were performed independently on three occasions, with the reported MIC defined as the lowest concentration for which agreement was observed for all three biological replicates. Compounds 1–5 were assessed for mammalian cytotoxicity alongside the control antibiotics amoxicillin and polymyxin B and the vehicle control DMSO against adherent NHDF cells (ATCC PCS-201=010) employing standard MTT methodology (SI). Cytotoxicity data are reported as the percentage of viable cells relative to the untreated controls from at least three independent biological replicates, and each compound concentration was tested in technical duplicate.

Molecular models of all compounds were created and visualized using Maestro ver. 11.8. DKP structures were optimized using MacroModel, and energies were calculated using OPLS3e (Schrödinger Release 2020-1: MacroModel, Schrödinger, LLC, New York, NY), in preparation for molecular dynamics simulations. The simulations were carried out using Desmond protocols.⁴³ The system was built by placing DKPs in an orthorhombic boundary (15 Å × 15 Å × 15 Å), and water molecules were described using the TIP3P potential.⁴⁴ The system energy was relaxed using the multistep protocol of Desmond.⁴³ Each simulation time was 50 ns, including a recording interval of 50 ps that generated 1000 frames. Energies were noted with an interval of 1.2 ps. Pressure and temperature were stabilized using a Martyna–Tobias–Klein barostat⁴⁵ and Nosé–Hoover chain thermostat,⁴⁶ respectively. Long-range electrostatic interactions were treated using the Ewald mesh summation with a 9.0 Å cutoff. Each simulation was conducted three times to give a total summation time of 150 ns per DKP. Distance plots of the DKPs over the simulation period were generated using the Simulation Interaction Diagram tool of Desmond.

A series of molecular dynamic simulations were then carried out to predict the interaction of each enantiomer with a phospholipid bilayer. Two 100 ns simulations were conducted using Desmond protocols.⁴³ The system was built in an orthorhombic boundary with a volume size of 40 Å × 40 Å × 120 Å. An optimized POPE membrane was inserted in addition to SPC water molecules and NaCl at a 0.15 M concentration.⁴⁷ The membrane was optimized using the Desmond default membrane relaxation protocol and used an NPγT ensemble class.⁴³ The DKP was manually placed outside the membrane at least 3 Å from the membrane surface. Each simulation was 100 ns long, including a recording interval of 100 ps that generated 1000 frames. Energies were noted with an interval of 1.2 ps. Simulations were carried out twice, giving a total of 200 ns per DKP. Hydrogen bonding interactions were calculated with Maestro.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmedchemlett.1c00683.

• Synthesis details, ¹H and ¹³C NMR spectra for all compounds, additional experimental details, and biological and membrane interaction data (PDF)

Author Contributions

Conceptualization: D.R., P.C., M.A.B., and J.S. Methodology: D.R., P.C., M.A.B., T.M.G., S.A.F., and J.S. Investigation: T.M.G., A.L.K., H.J.A., A.C., S.M., and S.A.F. Data curation: D.R. and T.M.G. Preparation of the original draft: T.M.G., D.R., P.C., M.A.B., and J.S. Review and editing: D.R., P.C., M.A.B., A.C., and J.S. Visualization: H.J.A. and T.M.G. Supervision: D.R., J.S., P.C., G.M.C., and M.A.B. Project administration: D.R., P.C., and M.A.B. Funding acquisition: P.C. and M.A.B. All authors have read and agreed to the published version of the manuscript.

The authors are grateful for support from the New Zealand Ministry of Business, Innovation, and Employment via Smart Idea Grant CAWX1805 (T.G., D.R., J.S., P.C., and M.A.B.) and Endeavor Grant UOAX2010 (M.A.B. and A.C.).

The authors declare no competing financial interest.