Amphipathic Barbiturates as Mimics of Antimicrobial Peptides and the Marine Natural Products Eusynstyelamides with Activity against Multi-resistant Clinical Isolates

We report a series of synthetic cationic amphipathic barbiturates inspired by the pharmacophore model of small antimicrobial peptides (AMPs) and the marine antimicrobials eusynstyelamides. These N,N′-dialkylated-5,5-disubstituted barbiturates consist of an achiral barbiturate scaffold with two cationic groups and two lipophilic side chains. Minimum inhibitory concentrations of 2−8 μg/mL were achieved against 30 multi-resistant clinical isolates of Gram-positive and Gram-negative bacteria, including isolates with extended spectrum β-lactamase−carbapenemase production. The guanidine barbiturate 7e (3,5-di-Br) demonstrated promising in vivo antibiotic efficacy in mice infected with clinical isolates of Escherichia coli and Klebsiella pneumoniae using a neutropenic peritonitis model. Mode of action studies showed a strong membrane disrupting effect and was supported by nuclear magnetic resonance and molecular dynamics simulations. The results express how the pharmacophore model of small AMPs and the structure of the marine eusynstyelamides can be used to design highly potent lead peptidomimetics against multi-resistant bacteria. ■ INTRODUCTION There is a desperate need for developing new antimicrobial agents to meet the worldwide emergence and spread of resistant bacteria. Resistant bacteria are currently causing deaths of 33,000 European patients annually, and the worst scenarios estimate 10 million deaths by 2050 per year if no measures are effectuated. WHO announced in their Global action plan on antimicrobial resistance that access to and appropriate use of existing and new antimicrobial drugs are absolutely mandatory to maintain the ability to treat serious infections. Increasing antimicrobial resistance has also dramatic consequences for common medical interventions in cancer treatment, caesarean sections, and organ transplantations. Large pharmaceutical companies show nevertheless little interest in antimicrobial drug development, mainly due to economic reasons. Academia and smaller research institutions are now conceivably the most important contributors for discovery and synthesis of new lead compounds for antimicrobial drug development. The eusynstyelamides are in this setting a fascinating class of antimicrobials isolated from the marine Arctic bryozoan Tegella cf. spitzbergensis and the Australian ascidian Eusynstyela latericius. The eusynstyelamides display moderate antimicrobial activity, and a method for the synthesis of (±)-eusynstyelamide A is reported. An intriguing structural feature of the eusynstyelamides is that they consist of two cationic groups Received: April 23, 2021 Published: July 27, 2021 Article pubs.acs.org/jmc © 2021 The Authors. Published by American Chemical Society 11395 https://doi.org/10.1021/acs.jmedchem.1c00734 J. Med. Chem. 2021, 64, 11395−11417 D ow nl oa de d vi a 88 .9 0. 80 .6 5 on A ug us t 1 8, 2 02 1 at 0 9: 08 :1 2 (U T C ). Se e ht tp s: //p ub s. ac s. or g/ sh ar in gg ui de lin es f or o pt io ns o n ho w to le gi tim at el y sh ar e pu bl is he d ar tic le s.


■ INTRODUCTION
There is a desperate need for developing new antimicrobial agents to meet the worldwide emergence and spread of resistant bacteria. 1 Resistant bacteria are currently causing deaths of 33,000 European patients annually, and the worst scenarios estimate 10 million deaths by 2050 per year if no measures are effectuated. 2,3 WHO announced in their Global action plan on antimicrobial resistance that access to and appropriate use of existing and new antimicrobial drugs are absolutely mandatory to maintain the ability to treat serious infections. 4 Increasing antimicrobial resistance has also dramatic consequences for common medical interventions in cancer treatment, caesarean sections, and organ transplantations. Large pharmaceutical companies show nevertheless little interest in antimicrobial drug development, mainly due to economic reasons. Academia and smaller research institutions are now conceivably the most important contributors for discovery and synthesis of new lead compounds for antimicrobial drug development.
The eusynstyelamides are in this setting a fascinating class of antimicrobials isolated from the marine Arctic bryozoan Tegella cf. spitzbergensis and the Australian ascidian Eusynstyela latericius. 5,6 The eusynstyelamides display moderate antimicrobial activity, and a method for the synthesis of (±)-eusynstyelamide A is reported. 5,7 An intriguing structural feature of the eusynstyelamides is that they consist of two cationic groups (amine or guanidine) and two lipophilic groups attached to a five-membered dihydroxybutyrolactam ring ( Figure 1). This amphipathic structural arrangement of cationic and lipophilic groups satisfies the pharmacophore model of small antimicrobial peptides (AMPs) that we and others have studied extensively by design of peptidomimetics of AMPs [also named synthetic mimics of AMPs (SMAMPs)]. 8−11 AMPs play a crucial part of innate immunity in virtually all species and constitute the first line of defense against infections by bacteria, virus, fungi, and parasites. 12−14 Natural AMPs are however rather large cationic peptides (+2 to +9) consisting of 12−50 amino acid residues where 20−50% are lipophilic residues. They have an amphipathic characteristic that is essential for their membrane disruptive mode of action against bacteria. 12,15 The limitation of AMPs as drugs is related to their pharmacokinetic properties, such as low proteolytic stability, low oral bioavailability, and potential immunogenicity. 16 The design of SMAMPs can offer a solution to these limitations.  In the present study, we report a peptidomimetic amphipathic scaffold inspired by the marine antimicrobials eusynstyelamides and fulfilling the pharmacophore model of small AMPs (Figure 1). A barbiturate ring was used as a structurally simplified mimic of the more complex dihydroxybutyrolactam ring of the eusynstyelamides, providing a scaffold without stereogenic centers. Different lipophilic and cationic groups could then be introduced on the barbiturate scaffold and provide a variety of amphipathic barbiturates ( Figure 2). Selection of lipophilic side chains was based on our previous work with SMAMPs. 17,18 The present amphipathic barbiturates were then investigated for their antimicrobial activity against bacterial reference strains and multi-resistant clinical isolates, and toxicity against human cell lines. One selected compound was investigated in vivo using a peritonitis model in mice to determine the efficacy against Gram-negative clinical isolates. The mode of action was studied in vitro using two luciferase-based membrane assays. To gain further insights into the membrane interaction of the amphipathic barbiturates, conformational analysis by nuclear magnetic resonance (NMR) in a membrane mimicking environment and molecular dynamics (MD) simulations of the interaction progression of compounds with an inner Escherichia coli cell membrane were performed.
■ RESULTS AND DISCUSSION Synthesis. Reported methods for the synthesis of substituted barbiturates include the condensation of alkylated malonate esters with urea, 19−21 cyclization with N-alkylated urea and diethyl malonate or malonic acid, 22,23 Knoevenagel condensation of barbituric acid and aldehydes or ketones, 20,24,25 and alkylation of barbituric acid. 26 We first focused on a divergent synthetic strategy to gain quick access to tetrasubstituted, amphipathic barbiturates by cyclization of N,N′-dialkylated ureas and disubstituted diethyl malonates. Unfortunately, no suitable reaction conditions for the cyclization of a number of malonate derivatives with N,N′dialkylated urea with a short C 2 linker to the cationic groups were found (see the Supporting Information; Section 1 for details). Depending on the reaction conditions, the dialkylated urea proved to be either unreactive, decomposed, or led to undesired side products. As this strategy did not deliver the desired results, we turned our attention to a different approach.
The condensation of dialkylated malonate esters with urea followed by N-alkylation became a successful strategy for the synthesis of amphipathic barbiturates (Scheme 1). Symmetrically disubstituted malonates 2a−g were obtained from diethyl malonate 1 by dialkylation with the appropriate arylmethyl halides and were subsequently cyclized with urea by treatment with NaH in dimethylformamide (DMF) to provide the 5,5disubstituted barbiturates 3a−h in yields of 70−92%. Dry conditions were imperative to the yield. Cyclization of malonate 2f (3,5-di-CF 3 ) gave low yields (27%) due to decarboxylation under the reaction conditions. The 5,5disubstituted barbiturates 3a−h were alkylated with an excess of 1,4-dibromobutane under basic conditions (K 2 CO 3 in DMF) to afford N,N′-dialkylated barbiturates 4a−h in 40− 96% yield. These were converted to the corresponding azides 5a−h with NaN 3 (2−3 equiv) in DMF (68−100% yield). Reduction of the azides to amines with NaBH 4 and a catalytic amount of propane-1,3-dithiol, 27 and subsequent Bocprotection, provided Boc-protected diamines after purification by flash chromatography. Boc-protection was important to increase the yield and ease the purification.
Structure−Activity Relationship Study against Reference Strains and Human Erythrocytes. Two series of amphipathic barbiturates were prepared, in which series 6 consisted of barbiturates with two cationic amino groups and series 7 encompassed barbiturates with two cationic guanidine groups ( Figure 2). Note that an abbreviation for the lipophilic side chain substituents is included in parentheses to aid the discussion. The barbiturates were initially screened for antimicrobial activity against antibiotic susceptible Grampositive and Gram-negative reference strains (Table 1). Hemolytic activity was tested against human red blood cells (RBCs) as a measurement of toxicity.
Amine Barbiturates of Series 6 against Reference Strains. For the amine barbiturates in series 6, the minimum inhibitory concentration (MIC) values ranged from 0.25 to 64 μg/mL against the Gram-positive strains Staphylococcus aureus and Corynebacterium glutamicum and MIC values from 2 to 128 μg/mL against the Gram-negative bacteria E. coli and Pseudomonas aeruginosa (Table 1). Higher antimicrobial activity was thereby in general observed against Gram-positive bacteria than against Gram-negative bacteria, although the differences were marginal for the most potent amine barbiturates of series 6. Considering a membrane-disruptive mode of action (see below), the outer cell wall of Gramnegative bacteria may provide additional protection and thereby result in higher MIC values compared to Grampositive bacteria. For comparison, the four different eusynstyelamides isolated from Tegella cf. spitzbergensis display MIC values of 6.25−12.5 μg/mL against the Gram-positive bacteria S. aureus and C. glutamicum and 12.5−25 μg/mL against the Gram-negative bacteria E. coli and P. aeruginosa. 5 The most potent amine barbiturate was 6h (3,5-di-tBu), which had two super-bulky lipophilic 3,5-di-tBu-benzylic side chains and displayed MIC values in the very low range of 0.25−4 μg/mL against all Gram-positive and Gram-negative reference strains. The side chain Clog P of 6h (3,5-di-tBu) (Clog P: 6.29) was the highest calculated for all the lipophilic side chains included in the study (Table 1). Derivative 6h (3,5di-tBu) showed, however, unacceptable high hemolytic toxicity (EC 50 : <5 μg/mL).
The two barbiturates 6e (3,5-di-Br) and 6g (4-tBu) were the second most potent derivatives displaying MIC values of 1−8 μg/mL against the bacterial reference strains and were both less hemolytic (6e EC 50 : 79 μg/mL and 6g EC 50 : 145 μg/mL). These had smaller lipophilic side chains and implied a correlation between side chain size or calculated side chain Clog P and antimicrobial activity.
The 3,5-di-substituted derivative 6f (3,5-di-CF 3 ) was less potent and displayed MIC values of 16 μg/mL against all strains except for the very susceptible strain C. glutamicum (MIC: 4 μg/mL). The C. glutamicum strain is a valuable strain for identifying antimicrobial agents in screenings since it is so susceptible but is otherwise not of any medical importance. Its high susceptibility resulted in that none of the barbiturates from series 6 (nor series 7) displayed MIC values above 4 μg/ mL against C. glutamicum.
It is noteworthy that the calculated Clog P of 6e (3,5-di-Br) was lower than the calculated Clog P of the less potent 6f (3,5di-CF 3 ), showing that not only the lipophilic effects of the side chains affected the antimicrobial potency but possibly also the size and electronic effects. With respect to electronic effects, a difference in electron distribution was observed both in 13 C NMR and when calculating the electron density of the bromine and trifluoromethyl substituents of 6e (3,5-di-Br) and 6f (3,5di-CF 3 ). The electron distribution in the side chains of 6e (3,5di-Br) and 6f (3,5-di-CF 3 ) was different hosting an overall more negative partial charge on the CF 3 groups compared to the bromine substituents (results not shown). This may affect the electron distribution of the aromatic side chains and possibly affect the lipophilic side chains in their interaction with the bacterial membrane and especially related to localization in the water−lipid interface region of the membrane. This may also explain why 6f (3,5-di-CF 3 ) displayed much lower hemolytic activity (EC 50 : 177 μg/mL) than 6e (3,5-di-Br) (EC 50 : 79 μg/mL). The 3,4-disubstituted derivative 6d (3-Cl, 4-Br) displayed high antimicrobial activity against the Gram-positive reference strains (MIC: 1−4 μg/mL) but was clearly less potent than the previous derivatives against the Gram-negative reference strains (MIC: 16−32 μg/mL). Derivative 6d (3-Cl, 4-Br) also showed very low hemolytic activity (EC 50 : 172 μg/mL).
A surprisingly low antimicrobial activity was observed for the least lipophilic derivative 6a (4-CF 3 ), which only had acceptable antimicrobial activity against C. glutamicum but very low potency against the remaining reference strains (MIC: 64−128 μg/mL). Derivative 6a (4-CF 3 ) was also all together non-hemolytic within the concentration range tested (EC 50 : >398 μg/mL).
Guanidine Barbiturates of Series 7 against the Reference Strains. Guanylation of the amine barbiturates in series 6 resulted in a striking increase in the antimicrobial activity of the resulting guanidine barbiturates in series 7 ( Table 1). The highly potent guanylated barbiturates of series 7 displayed a narrow range in the MIC values of <0.13−2 μg/ mL against the Gram-positive strains S. aureus and C. glutamicum and MIC 1−8 μg/mL against the Gram-negative bacteria E. coli and P. aeruginosa. One exception lacking increased potency against P. aeruginosa was 7a (4-CF 3 ) (MIC: 64 μg/mL), which was the smallest guanidine derivative (in volume) and least lipophilic derivative.
The general increase in the hemolytic activity following guanylation can be a result of the larger guanidine group forming more intricate electrostatic and hydrogen-bonding interactions than a primary amine group and thereby interact with both anionic and zwitterionic phospholipids (PLs). As we and others have reported, there is little consistency, and both increase and reduction of RBC toxicity is observed when amine groups are interchanged by guanidine groups. 17,30−34 Antimicrobial Activity against 30 Multi-resistant Clinical Isolates. The amine and guanidine barbiturates were screened against a panel of 30 multi-resistant clinical isolates of Gram-positive and Gram-negative bacteria ( Table  2). These isolates represented different resistance mechanisms, in which the Gram-positive isolates were methicillin-resistant S. aureus (MRSA) and vancomycin-resistant Enterococcus faecium (VRE), and the Gram-negative isolates included multi-resistant E. coli, P. aeruginosa, Klebsiella pneumoniae, and Acinetobacter baumannii with extended spectrum β-lactamase−carbapenemase (ESBL−CARBA) production. Three strains were also resistant to the last resort antibiotic colistin. Cytotoxicity was also determined against human hepatocyte carcinoma cells (HepG2) and human lung fibroblast cells (MRC-5).
Antimicrobial activity against the multi-resistant clinical isolates was high with MIC values as low as 2−4 μg/mL for the most potent barbiturates, thereby following the same tendencies as against the antibiotic susceptible reference strains. As opposed to RBC toxicity, the guanidine barbiturates of series 7 were less cytotoxic against human HepG2 and MRC-5 cells compared to the amine barbiturates of series 6 ( Table 2). The interplay between the two different cationic groups and the various lipophilic side chains thereby influenced the antimicrobial potency, hemolytic toxicity, and human cell cytotoxicity differently.
For the amine barbiturates of series 6, highest antimicrobial potencies (MIC: 2−16 μg/mL) were achieved against the Gram-positive multi-resistant clinical isolates of S. aureus and E. faecium and the Gram-negative isolates of E. coli. The overall most potent amine barbiturate of series 6 was 6h (3,5-di-tBu), closely followed by 6e (3,5-di-Br). These amine derivatives showed high potency also against the clinical challenging isolates of P. aeruginosa, K. pneumonia, and A. baumannii. The high cytotoxicity against human HepG2 and MRC-5 cells (IC 50 : 1−17 μg/mL) displayed by the active amine barbiturates of series 6 was unsatisfactory.
All the investigated amphipathic barbiturates displayed antimicrobial activity against the three colistin-resistant clinical isolates K. pneumoniae K47-25, K. pneumoniae 50531633, and A. baumannii K63-58 in the same range as against the colistinsusceptible clinical isolates. The mechanism of resistance of these clinical isolates is thought to involve altered lipopolysaccharide (LPS) outer cell wall composition and charge, changes that affect the mechanism of action of the last-resort cationic antibiotic colistin (pers. commun. prof Ø. Samuelsen).
The altered LPS structure seemed not to have any major impact on the binding and activity of the most potent amphipathic barbiturates.
In Vivo Efficacy of 7e (3,5-di-Br) in a Murine Neutropenic Peritonitis Model. The overall most potent guanidine barbiturate 7e (3,5-di-Br) was investigated in vivo using an established murine peritonitis model at Statens Serum Institut (SSI, Denmark). 35 Our aim was to determine the efficacy of 7e (3,5-di-Br) in mice infected with clinical isolates of E. coli (EC106-09) and K. pneumoniae (KP3010). Initially, the MIC of 7e (3,5-di-Br) was determined to be 4 μg/mL against both strains, which was in coherence with our previous  subtilis. Light emission normalized to an untreated water control (negative control) is plotted as relative light units (RLUs) over time (seconds) with untreated luminescence set to 100 RLU. After addition of the bacterial cell suspension (with 1 mM D-luciferin for the membrane integrity assay) to the analytes in each well, the light emission was measured each second for 150 s. Each line represents the kinetics of 150 subsequent data points of the analyte concentration. Each analysis was repeated at least three times independently. The figure shows a representative data set. screening results. A maximal tolerated dose (MTD) was determined prior to evaluation of in vivo efficacy. In brief, the MTD was determined by intraperitoneal (i.p.) injection of escalating doses of derivative 7e (3,5-di-Br). Derivative 7e (3,5-di-Br) was well tolerated up to 2.8 mg/kg after i.p. injection with no or mild clinical signs of discomfort. At 3.6 mg/kg, moderate signs of discomfort were observed, but the mice recovered within a few hours. The MTD was determined to be 7 mg/kg. In our vehicle controls, a log colony-forming unit per mL (CFU/mL) of 6.4 was determined for E. coli, indicating a 0.8 log CFU increase at the end of the experiment. A log CFU/mL of 5.7 was determined for K. pneumoniae corresponding to an approximately 0.6 CFU/mL increase at the end of the experiment. In contrast, treatment with 7e (3,5-di-Br) caused a 1.7-log (98%) reduction of the bacterial loads of E. coli already at a concentration of 1.4 mg/mL ( Figure 3A). Treatment with 1.4 mg/kg of 7e (3,5-di-Br) against K. pneumoniae resulted in a 1 log CFU/mL reduction (90%) compared to treatment with vehicle ( Figure 3B). A repeated injection after 3 h with 7e (3,5-di-Br) resulted in a 1.6 log CFU/mL (97%) reduction of the bacterial load. Despite limitations regarding the MTD, our results demonstrated that 7e (3,5-di-Br) could significantly reduce the number of viable bacterial cells in this in vivo model. We can conclude that the complex environment of the peritoneal cavity and the peritoneal fluid did not lead to a rapid inactivation of 7e (3,5-di-Br). However, at this point, we can only speculate about the time range 7e (3,5-di-Br) is present in sufficient concentrations for effective bacterial killing. Pharmacokinetic studies as well as different routes of administration have to be undertaken in order to fully reveal the potential of this type of compound in vivo.
Mode of Action Studies. The amphipathic amine barbiturate 6e (3,5-di-Br) and guanidine barbiturate 7e (3,5di-Br) were compared in a mode of action study using two luciferase-based biosensor assays in Bacillus subtilis 168 and E. coli HB101 (Figures 4 and 5). 36,37 The two different biosensor systems evaluate the effects on bacterial viability and membrane integrity, respectively, which are closely linked functionalities in bacterial cells (see the Supporting Information; Section S9 for detailed information regarding the assays). The bacteriolytic agent chlorhexidine (CHX), known for its membrane-disruptive properties, was analyzed for comparison. 38 The overall results demonstrated a strong and immediate membrane disrupting activity for both compounds. A more rapid membranolytic effect was observed against the Grampositive B. subtilis compared to Gram-negative E. coli. We also observed the differences in the rate of membrane lysis related to the test concentrations, in which concentrations higher than the MIC value led to a more rapid lysis, that is, a concentration-dependent killing effect.
The observed effects in the viability assay corresponded well with the respective MICs [6e (3,5-di-Br): 6.3 μg/mL and 7e (3,5-di-Br): 3.1 μg/mL against both B. subtilis and E. coli biosensor strains], in spite of an initial 1000-fold higher concentration of bacteria in the inoculum compared to the MIC assay. The decrease in light emission was rapid, dosedependent, and similar to the CHX control, suggesting a membrane-related mode of action against both strains ( Figures  4A and 5A). In order to confirm that the rapid decrease in bacterial viability was due to membrane damage, the membrane integrity assay was performed. Also, in this assay, a dose-dependent effect was observed against both strains ( Figures 4B and 5B). The effects (rapid peak emission due to the influx of D-luciferin into the cells) were for the most part coinciding with the respective MIC values, indicating that membrane damage was indeed a major effect. The well-by-well measurements allowed for catching the actual light peaks, apart from measurements with 7e (3,5-di-Br) in B. subtilis, which seemed to act substantially faster than 6e (3,5-di-Br) and CHX and therefore only showed a decrease in light emission from a level substantially higher than the control ( Figure 4B).
When comparing the results obtained from the viability assay ( Figure 4A) and the membrane integrity assay ( Figure  4B) in B. subtilis for compounds 6e (3,5-di-Br), 7e (3,5-di-Br), and CHX, the patterns appeared somewhat similar, indicating a rapid membranolytic activity for all compounds. However, in the membrane integrity assay in B. subtilis, we were not able to determine a peak in light emission for any concentration above MIC for 7e (3,5-di-Br) ( Figure 4B). Light emission declined immediately, indicating that peak emission already had occurred before the first measurement, that is, within 2 s after analyte addition. At MIC (3.1 μg/mL), a small peak in light emission was observed after approximately 5 s, but the emission did neither decrease nor increase substantially within the measurement window. Altogether, the effect of 7e (3,5-di-Br) on B. subtilis shown in the viability assay seemed to be immediate ( Figure 4A) and corresponded to the membranolytic effect shown in the membrane integrity assay ( Figure 4B).
In E. coli, the observed overall picture was somewhat different. A rise or peak of light emission in the membrane integrity assay for 6e (3,5-di-Br) coincided with an immediate decrease of light emission in the viability assay (similar to the results in B. subtilis) ( Figure 5). However, an emission peak was not reached for the lowest (1−4× MIC) concentrations of 7e (3,5-di-Br) within the 150 s test window in the membrane integrity assay ( Figure 5B). On the other hand, the concentration-dependent reduction in viability observed with the guanidine barbiturate 7e (3,5-di-Br) resembled the results of the guanidine-containing CHX ( Figure 5A), but the decrease in viability was substantially slower than for similar concentrations in B. subtilis ( Figure 4B). In general, the membrane integrity effects of all tested compounds seemed to occur at a slightly slower rate in the Gram-negative E. coli compared to the Gram-positive B. subtilis. It is tempting to speculate that especially for 7e (3,5-di-Br), the outer membrane of E. coli acted as a barrier, causing a delayed action in the membrane integrity assay. This would however not explain the presence of light production at a time point where the viability assay emits almost no light at all and accordingly indicates complete metabolic shutdown. This effect, even though less pronounced, was also observable for 6e (3,5-di-Br) and the CHX control. Although ATP is necessary for replenishment of the fatty aldehyde pool, this might indicate that reduction equivalents were the limiting factor for light emission of the viability sensor assay and that ATP under these conditions was not a limiting factor after treatment with 6e (3,5-di-Br), and especially, 7e (3,5-di-Br) until after the measurement window ended. Alternatively, there were different subpopulations of bacterial cells present, with different susceptibility to the analytes, resulting in an average light emission, which does not represent any of the subpopulations.
While the main mode of action against B. subtilis for both 6e (3,5-di-Br) and 7e (3,5-di-Br) seemed to be disruption of membrane integrity, our results did not exclude the possibility that especially 7e (3,5-di-Br) might have additional targets than the bacterial cytoplasmic membrane. Further work is needed to elucidate if 7e (3,5-di-Br) possibly targets other components of the cell and if there is a dual mode of action.
Conformational Analysis and Membrane Interaction Simulations. To gain insights into the interactions of the amphipathic barbiturates with a PL membrane surface, we determined the most stable conformations of the barbiturates, followed by a membrane interaction simulation. Density functional theory (DFT)-based geometry optimizations of amine 6e (3,5-di-Br) and guanidine 7e (3,5-di-Br) gave similar distortions and energy differences and indicated three lowenergy conformations mainly differing in the orientation of the benzylic side chains ( Figure 6). In the up (7e up ), down (7e down ), and up−down (7e up−down ) conformations, the benzylic side chains were either directed upward in a Wshape, downward, or having one side chain pointing up and the other pointing down. The 7e up conformation was lowest in energy, whereas 7e up−down and 7e down were 4.9 and 9.8 kcal/ mol higher in energy, respectively (see Supporting Information Section S10 for more details of the conformational analysis). An X-ray structure of 7b (2-Nal) supported the low energy conformation suggested by DFT calculations (see Supporting Information Section S11 for details).
The ROESY spectra acquired in water and micelle [sodium dodecyl sulfate (SDS)] solutions of the guanidine barbiturate 7e (3,5-di-Br) were used to qualitatively assess the conformation experimentally (see Supporting Information Section 12 for details of the NMR conformational analysis). The structural NMR data in water (Figure 7, left side) supported the orientations of the benzylic side chains described by the DFT calculations. It was evident from the ROESY detectable correlations between H7 and H10−H12 (5−10% of the reference volume) that the benzylic side chains and the barbiturate ring adopted the W-shape (similar to the 7e up conformation in Figure 6). There were no dramatic conformational changes in SDS, but there was a shift of populations that made the guanidine side chains spend more time closer to the 3,5-dibromophenyl rings (Figure 7, right side). This was reflected in the volumes of the H7/H11,12 cross-peaks that increased from ∼10 to ∼40% of the reference volume.
The course of the membrane insertion was tracked by following the location of the sp 3 carbon opposite from the carbonyl carbon (C 5 ), as noted by the z-coordinate position in the simulation box ( Figure 8B). The lipid bilayer surface (black lines) is shown as the average position of the phosphorous atoms of the PL headgroups (z-coordinate, −20 and 20 Å). The blue line shows the time evolution for location of the C 5 carbon of 7e (3,5-di-Br). The MD simulations for compounds 6e (3,5-di-Br), 6g (4-tBu), 7e (3,5-di-Br), and 7g (4-tBu) revealed a rapid membrane insertion between 7 and 35 ns, which was as expected due to the electrostatic interaction between the negatively charged membrane surface and the positively charged compounds.
The starting conformation of 7e (3,5-di-Br) in the MD simulations was up. In the shown simulation parallel in Figure  8C, tracking of the two angles c 1 (blue) and c 2 (orange), representing the two benzylic side chains, revealed that 7e (3,5-di-Br) remained in the up conformation throughout this simulation. This is shown by the blue and orange lines both oscillating around 80°, as opposed to if one of the lines was also oscillating around 140°, indicating an up−down conformation ( Figure 8C). As shown in the Supporting Information, however, the conformations of all modeled compounds varied between the up and up−down conformations in at least one of the three parallels, and the changes from up to the up−down conformation occurred sometime between 60 and 255 ns (Table S3 and Figures S6−S10). In most parallels of the MD simulation, the compounds remained incorporated in the membrane throughout the duration of the simulation. Except for 6a (4-CF 3 ) as described below, if a molecule left the membrane, it was only for a few nanoseconds before it returned to the membrane environment, as can be seen from the time evolution of the C5 z-coordinate for the other modeled compounds.
A simplified side view of the MD simulation system is presented in Figure 8D, which shows the interaction of 7e (3,5-di-Br) with an E. coli inner membrane model. This includes a water pad over and under the PL bilayer, a PL bilayer in the middle, the phosphorous atoms of the lipid headgroups, and the location and time evolution of 7e (3,5-di-Br) when interacting with the model membrane.
A reference set of simulations were also run with 6a (4-CF 3 ) to investigate the selectivity of the membrane model. As described above, compound 6a (4-CF 3 ) was much less potent against E. coli (MIC: 128 μg/mL) compared to the other modeled compounds. The simulations also showed that 6a (4-CF 3 ) had less affinity to remain in the model membrane environment compared to the other compounds (Table S3 and Figure S6). In all the three parallels of MD simulations of 6a (4-CF 3 ), it entered and left the membrane environment several times. This contrasted with the behavior seen in the simulations of 6e (3,5-di-Br), 6g (4-tBu), 7e (3,5-di-Br), and 7g (4-tBu) where once incorporated, the compounds remained in the membrane environment. The conformation of 6a (4- CF 3 ) varied between up and up−down, but there was an increase in events where 6a (4-CF 3 ) returned from up−down to the up conformation ( Figure S6). This behavior was not observed for other compounds in the MD simulations where only the shift from up to up−down was observed. As can be seen from Figure S6, 6a (4-CF 3 ) also traveled out from the top of the simulation box and appeared at the bottom side of the simulation box and did this several times during the 260 ns simulation (Table S3). The periodic boundary conditions in the MD simulations allowed the free flow of molecules in and out of the simulation box. The behavior of 6a (4-CF 3 ) compared to the other modeled compounds suggested that 6a (4-CF 3 ) did not find favorable interactions in the membrane environment, and this may in part explain its low antimicrobial potency against E. coli.

■ CONCLUSIONS
In order to succeed transforming AMPs with non-optimal pharmacokinetic properties into clinical useful antimicrobials, an innovative strategy is to develop SMAMPs with imperative functional side chains embodied on a peptidomimetic scaffold. We have in the present study developed a novel peptidomimetic scaffold that fulfills the pharmacophore model of small AMPs and that was inspired by the marine antimicrobials eusynstyelamides. Compared to the structure of the eusynstyelamides, this novel series of cationic amphipathic barbiturates is achiral and easy to modify synthetically with respect to variation in cationic and lipophilic groups for optimization studies. The relative ease of synthesis has important implications for reducing future production costs and enabling large-scale production, which is an argument often raised against several classes of AMPs. We achieved improved antimicrobial activity compared with the eusynstyelamides, and several of the barbiturates displayed high antimicrobial activity against a panel of 30 multi-resistant clinical isolates of Gram-positive and Gram-negative bacteria. This included high activity against Gram-negative ESBL− CARBA isolates and strains resistant to the last resort antibiotic colistin. A pilot in vivo study using a murine neutropenic peritonitis model demonstrated that the overall most potent lead peptidomimetic 7e (3,5-di-Br) significantly reduced the number of viable bacterial cells of clinical isolates of E. coli and K. pneumoniae. Although further structural optimizations are required to improve the MTD in mice, as well as pharmacokinetic studies including exploration of different routes of administration, demonstration of in vivo efficacy gives hope to the drug potential of this class of SMAMPs for treatment of serious infections.

■ EXPERIMENTAL SECTION
Chemicals and Equipment. All reagents and solvents were purchased from commercial sources and used as supplied with the exception of the starting material 1-(bromomethyl)-4-fluoronaphthalene, which was synthesized from the 4-fluoro-1-naphthoic acid according to the literature procedures. 40 Anhydrous DMF was prepared by storage over 4 Å molecular sieves. The reactions were monitored by thin-layer chromatography (TLC) with Merck precoated silica gel plates (60 F 254 ). Visualization was accomplished with either UV light or by immersion in potassium permanganate or phosphomolybdic acid (PMA), followed by light heating with a heating gun. Purifications using normal phase flash chromatography were either done by normal column chromatography using Normal Sil 60, 40−63 mm silica gel, or by automated normal phase flash chromatography (heptane/EtOAc) with the sample preloaded on a Samplet cartridge belonging to a Biotage SP-1. Purification of reactions by RP C 18 column chromatography (water with 0.1% TFA/ acetonitrile with 0.1% TFA) was also executed on an automated purification module with the sample preloaded on a Samplet cartridge. All samples used for biological testing were determined to be of >95% purity. The analyses were carried out on a Waters ACQUITY UPC 2 system equipped with a Torus DEA 130 Å, 1.7 μm, 2.1 mm × 50 mm column coupled to a Waters ACQUITY PDA detector spanning from wavelengths 190−650 nm. The derivatives were eluted with a mobile phase consisting of supercritical CO 2 and MeOH containing 0.1% NH 3 and a linear gradient of 2−40% MeOH over 2 or 4 min, followed by isocratic 0.5 min of 40% MeOH. The flow rate was 1.5 mL/min. NMR spectra were obtained on a 400 MHz Bruker Avance III HD equipped with a 5 mm SmartProbe BB/ 1 H (BB = 19 F, 31 P− 15 N). Data are represented as follows: chemical shift, multiplicity (s = singlet, d = doublet, t = triplet, q = quartet, p = pentet, h = heptet, and m = multiplet), coupling constant (J, Hz), and integration. Chemical shifts (δ) are reported in ppm relative to the residual solvent peak (CDCl 3 : δ H 7.26 and δ C 77.16; CD 3 OD: δ H 3.31 and δ C 49.00). Positive and negative ion electrospray ionization mass spectrometry (ESI-MS) was conducted on a Thermo electron LTQ Orbitrap XL spectrometer.
Synthesis. Dialkylated Malonate Ester (2a−g). General Procedure. To a stirred solution of diethyl malonate in DMF (≈100 mg/mL) and base was added arylmethyl halide (∼2 equiv). The reaction was continuously stirred at room temperature overnight. The reaction mixture was diluted with EtOAc (30 mL) and washed with water (25 mL), aqueous 5% LiCl solution (3 × 25 mL), and brine (25 mL). The organic phase was dried over Na 2 SO 4 , filtered, and concentrated. The crude product was dissolved in CH 2 Cl 2 (20 mL) and adsorbed on Celite. The product was purified on a silica column using 1−5% EtOAc in pentane as the mobile phase.
Nuclear Magnetic Resonance. All spectra were acquired on a Bruker Avance III HD spectrometer operating at 600 MHz for protons and equipped with an inverse TCI probe with cryogenic enhancement for 1 H, 2 H, and 13 C. NMR samples were prepared by dissolving 1 mg of 7e in 500 μL of H 2 O/D 2 O 9:1 in a 5 mm NMR tube. SDS was subsequently added to this sample in a 20:1 M ratio, resulting in a clear solution. Experiments were acquired using TopSpin 3.2, with gradient selection, adiabatic pulses, and excitation sculpting where applicable.
Molecules 6a (4-CF 3 ), 6e (3,5-di-Br), 6g (4-tBu), 7e (3,5-di-Br), and 7g (4-tBu) were built in PyMol. 54 Each of the compounds was given a starting structure where both phenyl groups are oriented in the up conformation. A simple minimization was performed in the builder tool of PyMol to clean the structures. Each molecule was assigned atom types, parameters, and charges with the CGenff online program. 55,56 Three parallels of all-atom MD simulations were performed for all systems with the molecular modeling software NAMD and the CHARMM36 force field. 57,58 A 10,000 step conjugate gradient and line search minimization was performed to ensure a stable starting structure for the MD simulations. Each membrane system parallel was run for 260 ns, and each water system parallel was run for 100 ns. All simulations were run at 310.15 K with a 2 fs time step and periodic boundary conditions. Particle Mesh Ewald was used for calculating the electrostatic interactions. 59 For non-bonded interactions, the scaled 1−4 principle was used for exclusion and 1.0 was used for scaling coefficient. A smoothing function was applied to the non-bonded forces with a cutoff of 12.0 Å and a switching distance of 10.0 Å. A pair list for the calculation of non-bonded interactions was updated every 20 steps, called one cycle, and the maximum distance for inclusion in the pair list for a pair of atoms was set to 16.0 Å. The pair list was regenerated twice every cycle. Bond lengths for hydrogen atoms were constrained with the SHAKE algorithm. 60 Both full electrostatic forces and the non-bonded forces were evaluated at every time step. The NPT ensemble was used for all simulations. Pressure control for the simulations was performed with Nose-Hoover Langevin piston with a target pressure of 1 atm. 61,62 A flexible simulation cell was used for the membrane system. Langevin dynamics were used for temperature control. Trajectory files were written every 1000 steps and energies were recorded every 125 steps.
Analysis of the MD trajectories was performed with the VMD GUI and VMD scripts. Figures were made with VMD and PyMol, and all graphs were generated with pandas, seaborn, and Matplotlib. 63−65 X-ray Crystallography. A rod-like specimen of 7b (2-Nal) was used for X-ray crystallographic analysis. The X-ray intensity data were measured with the Cu source (λ = 1.54178 Å) of an in-house Bruker D8 Venture system. Frames were integrated using the Bruker SAINT software package, and the structure was solved and refined using the Bruker SHELXTL software package. The structure factors of 7b have been deposited with the Cambridge Crystallographic Data Centre with deposition number 2026641. The integration of the data using a monoclinic unit cell yielded a total of 21874 reflections to a maximum θ angle of 66.75°(0.84 Å resolution), of which 6692 were independent (average redundancy 3.269, completeness = 99.6%, R int = 3.12%, and R sig = 2.71%) and 5846 (87.36%) were greater than 2σ(F 2 ). The final cell constants were 17.6014(15), 15.4212 (12), and 16.0233(17) Å with β = 106.833(4)°. The final anisotropic refinement converged with an R 1 /wR 2 of 6.8/21% with a GoF of 1.04. The structure of the asymmetric unit of 7b (2-Nal) with thermal ellipsoids is shown in Figure S2.
(PDF) 1H and 13C NMR spectra and SFC analysis data of the synthesized compounds; detailed description of the biosensor assays; comprehensive discussion of the conformational analysis; and molecular formula strings and MIC and toxicity data for compounds 6a−h and 7a−h (CSV)