Redesigning Arenicin-1, an Antimicrobial Peptide from the Marine Polychaeta Arenicola marina, by Strand Rearrangement or Branching, Substitution of Specific Residues, and Backbone Linearization or Cyclization

Arenicin-1, a β-sheet antimicrobial peptide isolated from the marine polychaeta Arenicola marina coelomocytes, has a potent, broad-spectrum microbicidal activity and also shows significant toxicity towards mammalian cells. Several variants were rationally designed to elucidate the role of structural features such as cyclization, a certain symmetry of the residue arrangement, or the presence of specific residues in the sequence, in its membranolytic activity and the consequent effect on microbicidal efficacy and toxicity. The effect of variations on the structure was probed using molecular dynamics simulations, which indicated a significant stability of the β-hairpin scaffold and showed that modifying residue symmetry and β-strand arrangement affected both the twist and the kink present in the native structure. In vitro assays against a panel of Gram-negative and Gram-positive bacteria, including drug-resistant clinical isolates, showed that inversion of the residue arrangement improved the activity against Gram-negative strains but decreased it towards Gram-positive ones. Variants with increased symmetry were somewhat less active, whereas both backbone-cyclized and linear versions of the peptides, as well as variants with R→K and W→F replacement, showed antimicrobial activity comparable with that of the native peptide. All these variants permeabilized both the outer and the inner membranes of Escherichia coli, suggesting that a membranolytic mechanism of action was maintained. Our results indicate that the arenicin scaffold can support a considerable degree of variation while maintaining useful biological properties and can thus serve as a template for the elaboration of novel anti-infective agents.


Obtaining a quantitative measure of sequence symmetry
To obtain a quantitative measure of symmetry, we inverted each sequence and calculated Nidentthe number of identical residues after aligning the original and inverted sequence using Needleman-Wunsch algorithm [48]. The quantitative measure of sequence symmetry S was defined as the ratio of identical residues for the original and inverted sequence to the total sequence length N:

Preparing initial conformations
Two types of initial conformations were used to simulate the folding and dynamics of the peptides: an extended linear conformation, and a β-sheet structure. Linear conformations for all peptides were created manually in PyMol [49]. An experimental NMR structure (PDB: 2JSB) [4] was used to simulate the original arenicin-1 (AR) and its cyclic variant (ARcycl). Initial β-sheet structures for ARin-s and ARs-C were obtained by replica-exchange Monte Carlo simulation in the Quark program [50]. For peptides ARs-N-B and ARs-C-B, having non-standard amino acids, β-sheet structures were constructed manually using PyMol. All peptides except AR were first simulated with both linear and β-sheet structures lacking a disulfide bond to ensure unobstructed sampling of the conformational space. In the case of ARs-C, the disulfide bond was introduced for the next computational experiment after the cysteine residues came in close contact (<4 Å). The peptide models were placed in a simulation box which was filled with SPC water molecules. The system was neutralized by the addition of sodium and chlorine ions to a final concentration of approximately 0.15 M.

Simulation parameters
The resulting MD trajectories and the corresponding simulation parameters are summarized in Table S1.

Calculation of the kink and twist angles
The twist was defined as the angle between two vectors a and b, where a pointed from the backbone C atom of the second amino acid (W 2 in AR) to the backbone N atom of the last amino acid (W 21 in AR), and b pointed from the backbone C atom of the last amino acid before the turn (V 10 in AR) to the backbone N atom of the first amino acid after the turn (V 13 in AR). To analyze the kink angle, we chose the residues in the center of each β-strand, i.e. the 6th and 17th amino acids in the case of arenicin-1. The kink of the first strand was defined as the angle between a vector c1, pointing from the backbone C atom of the center residue to the C atom of the first residue in a strand (W 2 in AR), and a vector d1, pointing from the backbone C atom of the center residue to the C atom of the last residue in a strand (V 10 in AR). The same calculation was made for the second strand, and the overall kink was estimated as the mean of the two angles.

Monitoring peptide folding
The process of the folding was monitored by determining the root-mean-square deviation (RMSD) from the initial linear structure (a) and the formation of secondary structure elements (b). Data for the rapid folding of the peptide ARs-C is presented in Figure S1a and S1b, respectively. Figure S1. (a) Folding of the peptide ARs-C, monitored by a distance (RMSD) from initial linear conformation; (b) formation of secondary structure elements during the folding of the peptide ARs-C, monitored by DSSP program.

Analysis of peptide dynamics and stability
The dynamics of the peptide backbone were assessed by calculating root-mean-square fluctuations (RMSF) the C α atoms. The characteristic for β-hairpin peptides graphs with labile N-, C-termini and turn region are given in Figure S2a. The stacking of N-terminal tryptophan residues, which contributes to higher rigidity of the peptide ARs-C-B, is shown in Figure S2b.

S3. Why use geometric mean of MIC as an assessment of overall antimicrobial activity?
There are different measures and approaches to assess so called "general tendency" of obtained data, in other words to represent a dataset by a single value. The most common measures for this are the mean, which can be arithmetical, geometric, harmonic, etc., the median and the mode [55]; other methods, e.g. weighted sums, can also be involved. Which one to use in any particular case depends on the problem that needs to be solved, and on the question(s) that need to be answered. So, why use the geometric mean of minimal inhibitory concentrations (MIC) to compare the antimicrobial activities of substances?
A first step is to formulate the problem. Given two substances, A and B, with activities AA and AB, respectively, we need a way to quantitatively compare these activities, so that in the end we can state "the activity of A is higher than that of B by a factor of f ". This can be represented as: If we test the efficacy of substances A and B against only one bacterial strain, we just compare the commonly used measure of antimicrobial activity, the MIC value. The lower the MIC, the higher is the activity, so potency is inversely proportional, and if the MIC of substance A [MIC(A)] is 2 μM and MIC of substance B [MIC(B)] is 8 μM, then the antimicrobial activity of A is higher than the activity of B by a factor of 4, and Equation s1 becomes: How should one combine individual f values when the activity of substances A and B are assayed against several different bacteria? Should we sum them, or intermultiply, or, use some more complex treatment?
Let's consider an exemplary situation. Substances A and B were tested against two bacteria. Substance A was 4 times more effective than substance B against bacteria 1 [e.g. MIC(A)1 = 2 μM, MIC(B)1 = 8 μM, so that f1 = 4], but substance B was 4 times more effective than A against bacteria 2 [e.g. MIC(A)2 = 8 μM, MIC(B)2 = 2 μM, so f2 = ¼ ]. Putting aside any additional considerations about which bacterium is more desirable to eliminate, we can say that overall, the activities of A and B are equal. Mathematically speaking, we anticipate that overall AA = AB so that according to Formula s1, f = 1. The easiest way to obtain a value of 1 by operating with f1 = 4 and f2 = ¼ is to intermultiply them. So a first model for calculating the overall f (ftotal) in case of several tested bacteria is as the product of multiplication of f-s for individual bacteria: where i is the index of an individual bacterial strain, fi is factor f, calculated for bacterial strain i according to Formula s2, and n is the total number of tested strains. We can test Equation s3 in another exemplary situation. Substances A and B were tested against 3 bacteria, and in each case A was found to be 4 times more active than B: f1 = f2 = f3 = 4. Common sense leads one anticipates that the overall activity of A is higher than the overall activity of B by a factor of 4. However, according to the Equation s3 ftotal = f1 • f2 • f3 = 4 3 = 64. Moreover, if we test A and B against a 4th bacterium with the same result of f4 = 4, ftotal will become 4 4 , and in case of n bacteria tested with equal results it will be ftotal = 4 n . This does not make sense so our model clearly requires optimization. For instance, we could calculate some average value for f and so remove the incremental dependency on the number of bacterial species tested (n). It will additionally allow us to compare activities of substances tested on a different number of different bacterial species, at least to some extent.
An obvious method to remove the n power from 4 n in the current simplified example is to calculate its n-th root: (4 n ) 1/n = 4, which was the anticipated answer. In general, the f values can be different for each bacterial strain (e.g. f1 ≠ f2 ≠ … ≠ fn), and the upgraded equation for ftotal will look like: By definition, the geometric mean of a value (in this case G-MIC) is calculated as: where i indicates an individual bacterial strain, so MICi is MIC of the substance against bacterial strain i, and n is the total number of tested strains. Thus, combining s1, s4 and s5, a final equation for comparing antimicrobial activities (AA and AB) of two substances A and B is: This retains the advantage of using a mean which takes into account that a different number of strains may be used with different substances, so the geometric mean is a reasonable choice to assess the general tendency in a set of MIC values, obtained by investigating the activity of substances against different bacteria, if these activities are to be compared.