Multistep optimization of a cell-penetrating peptide towards its antimicrobial activity

Multidrug resistant (MDR) bacteria have adapted to most clinical antibiotics and are a growing threat to human health. One promising type of candidates for the everlasting demand of new antibiotic compounds constitute antimicrobial peptides (AMPs). These peptides act against different types of microbes by permeabilizing pathogen cell membranes, whereas being harmless to mammalian cells. Contrarily, another class of membrane-active peptides, namely cell-penetrating peptides (CPPs), is known to translocate in eukaryotic cells without substantially affecting the cell membrane. Since CPPs and AMPs share several physicochemical characteristics, we hypothesized if we can rationally direct the activity of a CPP towards antimicrobial activity. Herein, we describe the screening of a synthetic library, based on the CPP sC18, including structure-based design to identify the active residues within a CPP sequence and to discover novel AMPs with high activity. Peptides with increased hydrophobicity were tested against various bacterial strains, and hits were further optimized leading to four generations of peptides, with the last also comprising D ow naded rom http://pndpress.com /bchem j/article-oi/10.1042/BC J20200698/5/bcj-2020-0698.pdf by gest on 07 Jauary 2021 Bchem al Jornal. This is an Acepted M ancript. ou re encuraged to se he Vrsion of R eord tat, w en puished, w ill relace his vesion. he m st up-tote-version is avilable at https://drg/10.1042/BC J200698


Introduction
Since the discovery of arsphenamine (also known as Salvarsan) by Paul Ehrlich in 1907 1 , and later on of penicillin by Alexander Fleming in 1928 2 , the number of antibiotics as well as their use for treatments in different areas, such as agriculture 3 , human and veterinary medicine 4 , has steadily increased. As a result of this over usage resistances in bacteria evolved and have more and more become a real global threat, demanding for the development of novel antibiotics with improved activity spectra to counter the development of multidrug resistant strains 5,6 . During the past years, great efforts have been made in this direction leading, for instance, to the discovery of the macrolide azithromycin 7 , or the cephalosporine ceftriaxone 8 . Both antibiotics are effective against many bacteria used as clinical standards but meanwhile also face the problem that bacteria find a way to adapt to them 9 .
A novel potential class of antibiotics is seen in antimicrobial peptides (AMPs), a structurally highly diverse group of peptides that kill bacteria mainly by membrane permeabilization 10 . This mechanism is supported by their structural properties including a usually high content of positively charged, as well as hydrophobic amino acid residues 11 . These two types of amino acids are often orientated along two faces within different kinds of secondary structures, like alpha helices 12 or beta sheets 13 , promoting a more basic, charged face and a nonpolar face. The positively charged side of the peptides perfectly interacts with negatively charged components at the outer surface of bacterial membranes, while hydrophobic interactions of the other face enhance insertion of the peptide into the hydrophobic core of the lipid bilayer parasites 20 . Examples include Magainin isolated from African clawed frog 21 or cationic antimicrobial peptide CAP18 obtained from rabbits 22 . Until today, the set of natural AMPs is complemented by synthetically derived peptide sequences that share the advantage of chemical modification 23 . Interestingly, many antimicrobial peptides exhibit anti-cancer activity, and indeed, such anti-cancer peptides (ACPs) may provide a promising, new approach in chemotherapy 24, 25 . ACPs are usually cationic and therefore, attracted by the more negatively charged cell membrane of cancerous cells 26,27 .
Within this work, we aimed to modify the sequence of the cell-penetrating peptide sC18 28 , and to change its activity to act as an AMP. sC18 is a C-terminal fragment of the cationic antimicrobial peptide CAP18, which, for itself, exhibits only very moderate antimicrobial activity. In recent work, we developed sC18 as a very effective transporter for various cargos, e.g. metal organic complexes 29 , cytostatic drugs 30 , or peptides 31 . From our previous studies, we have also recognized that sC18 forms an amphipathic alpha-helical structure when in presence of a lipid environment 30 . Since CPPs and AMPs act differently to certain organisms (e.g. mammalian cells, or bacteria), but often share common physicochemical characteristics 32 , we wondered if we could switch the activity of sC18 to become an AMP. Moreover, the development of AMPs based on CPP structures, from which it is already known that they are non-toxic, might be beneficial in terms of safe application.
Previously, it was reported that an increase in the amount of hydrophobic amino acids led not only to increased antibacterial activity 33 , but also to an enhanced overall selectivity between bacteria and mammalian host cells 34 . Further options to improve such properties were seen in the incorporation of unnatural amino acids that further increase stability against enzymatic hydrolysis 23,35 . Thus, within this work, we run several cycles of evolution on the structure of sC18, including rational amino acid displacement with hydrophobic and/or unnatural amino acids, to yield more stable and more selective and still highly active new antimicrobial peptides.

Peptide synthesis
Peptides were synthesized using orthogonal Fmoc/tBu strategy on Rinkamide resin (loading 0.48 mmol/g, 0.015 mmol scale). Amino acid coupling was realized using an automated peptide synthesizer from MultiSynTech and following double coupling steps with 8 equivalents (eq.) Fmoc-aa-OH, Oxyma pure ® and DIC. Fmoc-protecting group was removed with piperidine (20 % in DMF). 5,6-carboxyfluorescein (CF)labeled peptides were synthesized according to Ref. 31 . Peptides were labeled at the N-terminus with CF. Finally, peptides were removed from the resin using TFA/TIS/H2O (95:2.5:2.5 v/v/v) for 3 h and precipitated in ice-cold diethyl ether.
Peptides were purified using preparative RP-HPLC, and fractions were analyzed by analytical RP-HPLC ESI-MS. Final purity of all compounds was >95 % (see also

Circular dichroism (CD) spectroscopy
Peptides were dissolved in 10 mM phosphate buffer (pH 7) or 10 mM phosphate buffer (pH 7) with the addition of trifluoroethanol (TFE) (1:1) to yield a concentration of 20 µM. Spectra were accorded from 180 to 260 nm in a 1 mm thick quartz cuvette.
Conversion of θ measured in degree to the characteristic θ for the peptide was realized by using the following equation: Bacterial cultures of an optical density of more than 0.7 at 600 nm were used. In a 96-well plate 180 μL of MH-medium, 10 μL of bacteria suspension and 10 μL of peptide solution were mixed. The resulting cultures were screened at several different peptide concentrations (as triplicates). As negative control pure water and as positive control 35% ethanol in medium were used. All samples were then incubated at 37 °C for 6 h. Afterwards, 10 μL of a 1 mg/mL solution of iodnitrotetrazolium chloride in pure DMSO was added to each well, and samples were further incubated for 30 min at 37 °C. Finally, the absorption of formazan at 560 nm in each well was measured using a plate reader. The MIC was calculated as the average of three independent experiments performed in triplicate. For P. aeruginosa six singular experiments were performed.
To test the stability in culture supernatant, bacterial cultures of an optical density of more than 0.7 at 600 nm were used. The bacterial culture was centrifuged, and the supernatant gathered for further experiments. 200 µL peptide solutions of 4 x MIC (regarding the used bacterium) were prepared in culture supernatant and incubated overnight. As control, a solution of medium only was used. The next day, peptide solutions were diluted 1:3 in fresh medium and an antibacterial assay was performed as already described. As control, peptide samples pretreated with growth medium only were used.

Hemolysis assay
Human red blood cells (hRBCs zenbio, USA) ) ( were washed in PBS buffer and  software.

Scanning electron microscopy (SEM)
Bacteria were cultured as previously described. gold. Samples were analyzed using a FEI Quantum 250 FEG scanning electron microscopy.

Results and discussion
Stepwise AMP design by sequence-based refinement of sC18 Inspired by the work of Süssmuth et al. 36 , we systematically replaced specific amino acids in the sequence of sC18 with hydrophobic ones. Aim was to increase the general hydrophobicity of the peptide, which was already shown to be a promising way of enhancing antimicrobial activity 33,36 . As it is assumed that membrane active peptides use a mechanism of coiling into a double-faced alpha helix to interact with the cell membranes 37 , we first carefully inspected the helical wheel projection of sC18. From this we suspected three positions to have an exceptional effect. For instance, it can be appreciated that substitution of position Arg10 and Lys16 might be favorable, since this enlarges the hydrophobic face of the amphipathic helix ( Figure   1, Figure S1). On the other side, substituting position Glu15 eliminates the negative charge delivered by glutamate that would otherwise inhibit the interaction of the positively charged amino acid residues with negatively charged compounds at the To prove our hypothesis, we performed an isoleucine scan of sC18 to find out how substitution with a single hydrophobic amino acid would influence the AMP activity of sC18, resulting in the first generation of novel possible AMPs 1a-1o (Table 1). All peptides were synthesized by Fmoc-based solid phase peptide synthesis strategy.

TABLE 1
The new peptides 1a-1o were tested in a first screen by incubating different grampositive (B. subtilis, M. luteus) and gram-negative (P. fluorescens) bacteria for 6 h with two different peptide concentrations ( Figure 1). As expected from the helical wheel projections, the results showed increased antimicrobial activity especially for exchanges at position Arg10 (1j), Glu15 (1n) and Lys16 (1o). As a consequence, we concluded that these positions represent important nodes within the sC18 sequence.

FIGURE 1
Therefore, we further tested the impact of these positions on AMP activity analyzing double and triple isoleucine mutants 2a-2d (Table 1) Table 2).
To get more insights into the membrane-activity of the herein tested peptides, the different bacterial strains were chosen according to their differences in cell-wall composition and organization. Here, we aimed at improving peptide potency by incorporating phenylalanine as an even more hydrophobic amino acid. In addition to the mutants bearing single exchanges at positions Arg10, Glu15 and Lys16 (3a-3c), we directly generated the double and triple mutated variants (3d-3g) ( Table 1 and Figures S7, S8). All peptides were analyzed concerning their antibacterial activity as described before for the Ile mutants ( Figure 2B and Figure  Having observed optimized activity when the negative charge at Glu15 was substituted, we removed this amino acid out of the sC18 sequence, yielding peptide 4a that served as a control for peptides of this 4 th generation. Additionally, we included 4b as an analogue to 3g, bearing phenylalanine substitutions for Arg10 and Lys15. In a further attempt to optimize the parent sequence, we increased the hydrophobicity even more by introducing fluorinated amino acids. As fluorination was reported to lead to minimal steric alterations 41 , we used the mono-, di-and pentafluorinated phenylalanine variants X 1 -X 3 (Table 1 and Figure S8, S9) and replaced the same positions as in 4b, yielding peptides 4c-4e. Again, all novel peptide variants were obtained in high purities and directly evaluated concerning their biological activity as described before. Our results showed that the antimicrobial activity of 4a, in comparison to sC18, was slightly increased against C. glutamicum  Table 2 all determined MIC 50   values are listed.   TABLE 2 Within a last and preliminary step, we have analyzed the activity of the most potent peptides 3g, 4b and 4e against pathogenic Pseudomonas aeruginosa ( Figure S6).
Estimated MIC 50 values were in the range of 9 µM for 3g, 19 µM for 4b and 11 µM for 4e, thus, considerably higher than those measured for, e.g. P. fluorescence.
However, considering that P. aeruginosa belongs to the group of ESKAPE pathogens that show increased resistance against various commonly used drugs and cause health-care associated infections, our findings might contribute to the development of more potent new drugs against this species.
In summary, our step-by-step rational design provided novel peptides with improved AMP activity demonstrating that a switch to antibacterial activity is possible. However, although the MIC values are high for most of the analogues, only peptides 3g and that out of generation 4 are likely to be good candidates as antimicrobials.

Peptide stability
One major concern for the future development of AMPs is their limited stability to proteolysis in physiological conditions 43 . Therefore, we tested the stability of the fluorinated peptide 4e, which turned out to be the most promising AMP in our initial as well as another major degradation product for 4a ( Figure 3A). In addition, all peptides were rapidly cleaved after Arg5, and 4a also quickly after Arg10, producing the peptide fragments LRKFR and NKIKK, respectively, that were the major products already after 10 minutes ( Figure S11). Contrarily, 4b and 4e were mainly cleaved after Lys8, Lys12 and Lys14, leading to the two major products FFNK and FXNK after more than 1 h incubation ( Figure S12-S13). Interestingly, the degradation process seemed to be a bit slower for 4e in comparison to 4b, possibly as a result of the incorporation of the fluorinated amino acids 35 .
We next synthesized several of these fragments and tested if they retained antimicrobial activity against B. subtilis and P. fluorescens ( Figure 3B). However, none of these fragments showed any significant antimicrobial activity when tested in a concentration of up to 50 µM. From this we concluded that the complete peptide sequence is needed for antimicrobial activity and that the mechanism of action must be very fast to avoid proteolytic inactivation.
Furthermore, to test the stability under more physiological conditions, we exposed the peptides 4a, 4b and 4e to bacterial cell culture-conditioned supernatant and fresh medium as control. After 18h incubation in the respective media, we performed an antimicrobial activity assay by incubating the peptides for additional 6 h with B.
subtilis. As depicted in Figure 3C, all peptides lost their activity when pre-incubated in bacterial cell culture supernatant but retained activity when pre-incubated in growth media only. However, all of the peptides were susceptible to fast degradation, and the shorter fragments did not exhibit significant activity. This suggests a very fast mechanism of action for the peptides to exert their antimicrobial effect.

Hemolytic activity and cytotoxic profiles in mammalian cells
In a next set of experiments, we wanted to find out how the peptides would perform in presence of mammalian host cells, as cytotoxic behavior was not uncommon for strong antimicrobial peptides 44 . We examined peptides 3a-3g and 4a-4e, since those seemed to have the most promising overall activities. Firstly, we assessed their hemolytic activity by using human red blood cells (hRBC), which were incubated for 1 h or 24 h, respectively, with the peptides (Figure 4 and Figure S14). Strikingly, after 24 h incubation hRBCs were nearly not affected by the peptides. Only peptides 4d and 4e were slightly more active and exhibited hemolytic activities of up to 15 % when incubated with hRBCs at higher concentrations of about 40 µM.   and MCF-7 (human breast cancer) cells ( Figure 5B and 5C). In contrast to HEK293 cells, for both cancerous cell lines a clear toxic effect was observed when incubating them either with mutants 3d-3g, or with the fluorinated peptides 4c-4e. It is worth mentioning that 4b, containing only natural phenylalanine, exhibited no toxicity to HeLa cells, but was significantly more toxic to MCF-7 cells at higher concentrations (> 20 µM), indicating further selectivity between different cancerous cell lines. In addition, peptides 3a-3c and 4a were nearly non-toxic to both tested cell lines.

Cellular uptake ability in mammalian cells
Since the AMPs of this work were developed from a CPP sequence 28 , we were also interested if they were able to translocate in cells. Therefore, we incubated HEK293 and HeLa cells for 30 min with fluorescently-labeled peptides in a non-toxic concentration of 5 or 10 µM, respectively, and quantified their cellular uptake by flow cytometry.

FIGURE 6
We normalized the overall uptake to the parent CPP sC18, and from this it became clear that all investigated peptides, with the exception of 3a and 3c, showed higher cellular accumulations than the original CPP in both cell lines ( Figure 6 and Figure   S15). However, the uptake of the peptides was quite differently. For the series 3 peptides (which were tested at 10 µM), the highest accumulation in HEK293 cells was achieved by 3d and 3g with 32-and 50-fold increased uptake values compared to sC18. In HeLa cells the uptake was only moderately increased compared to sC18 by nearly 5-to 14-fold. On the other side, all peptides from the 4 th generation (which were tested at 5 µM) showed rather similar uptake values in HeLa and HEK293 cells.
In both cell lines, 4d displayed the highest accumulation with a 21-fold and 18-fold increase in uptake in comparison to sC18, followed by 4c and 4e with a more than 10-fold increase in uptake. As all these peptides having the highest accumulation were also the most hydrophobic ones, we assumed a direct correlation of hydrophobicity to cellular uptake for both cell lines tested. Thus, not the overall net charge was dependent for the high uptake values, but rather a proper integration of potency is also relied on targeting intracellular processes has yet to be elucidated.
Interestingly, 3d, 3e and 3g, accumulated significantly higher in HEK293 than HeLa cells. Only 3g exhibited slight toxic effects when applied in higher concentration to HEK293 cells. In this case, we could therefore not correlate uptake and cytotoxicity.
Instead, this observation let us conclude that particularly 3d and 3g should be investigated further as cell-penetrating peptides in HEK293 cells and related cell lines. All in all, the distinct uptake values might be the result of specific membrane interactions, which are most probably dependent on the different membrane compositions of both cell lines, as well as the distinct peptide sequences used 46 . In spite of this, it has been recently demonstrated that sequence motif and tail length of the hydrophilic basic phase of the helix are important determinants for successful cellular uptake 47 . This would be also in agreement for our peptides, since we supposed helix formation for all of our peptides and have observed increased uptake with increased hydrophobicity. In addition, it is known that CF fluorescence is lowered when in acidic compartments. Therefore, in future we will perform careful studies using fluorescence microscopy that will show if those different fluorescent values might depend on different local distributions of the peptides within the cells.

More insights into the mechanism of action
Our results reported thus far let us conclude that the novel AMPs act very fast against bacterial and cancerous cells and presumably by a membrane lysis process.
To test the mode of action and the potential of membrane disruption when in contact with mammalian cells, we analyzed the activity of lactate dehydrogenase (LDH) released from HeLa cells (Figure 7), since those cells were most affected after peptide treatment. From the third generation we only included peptides 3d-3g, since they appeared to be the most active ones out of this series. As depicted from Figure   7, we observed high outflow of LDH after 1 h incubation of increasing peptide concentrations of up to 40 µM. This was very prominent and more pronounced when cells were incubated with peptides 4b-4e. Again, 3g displayed a high membrane activity comparable to the fluorinated peptides. Electrostatic attraction might be the major reason for this activity observed. Indeed, it was already discussed that the interaction between negatively charged components of cancer cells and the positively Lastly, we deduced in more detail the mechanism of action when in contact with bacterial cells. For this we chose to compare peptides 3g and 4e, since they were the most active from the respective optimization rounds and comprised different lengths and substitutions.
We used scanning electron microscopy (SEM) to find out about alterations in the morphology of the outer bacterial membrane. Therefore, B. subtilis, C. glutamicum and P. fluorescens were chosen as representatives and incubated for 90 min with 4 x MIC of 3g and 4e ( Figure 8). As was nicely seen, all bacteria were characterized by a loss of structural integrity after peptide treatment, assumedly due to the peptides inducing antimicrobial effects on the membrane 49 . As the same observation was made for gram-positive and gram-negative bacteria, we concluded that the peptides affected them in a similar way, even though they have different types of cell walls.
Essentially, we hypothesized that the novel peptides mainly acted by membrane depolarization and deformation processes leading to membrane disruption and finally cell lysis.

Conclusions
The peptides we have created in this work stand out by their greatly enhanced antimicrobial activities compared to the original cell-penetrating peptide sC18. Among all peptides generated, especially 3g and 4b-4e evolved as interesting candidates for further broad-spectrum AMP development. Notably, first studies demonstrated also activity against pathogenic P. aeruginosa. Since quick and direct disruption of the membrane was observed, a bactericidal effect is most likely the way of action for those peptides. That the peptides act by this prompt membrane disintegration, was revealed by investigating their proteolytic stability. Obviously, they were highly prone to degradation, and the resulting fragments were not as active as the longer parent peptides. Indeed, from these observations, it can be hypothesized that the peptides must kill the bacteria quickly before they were inactivated by secreted bacterial proteases. Moreover, preliminary proteome studies revealed no significant changes in protein abundances (data not shown), speaking again in favor of a fast membrane To summarize our findings, we have designed highly active antimicrobial peptides with promising selectivity between bacterial and mammalian cells, as well as also between healthy and cancerous cells. Bearing those intriguing properties, they might be an interesting contribution for further research in these fields.

Author contributions
fluorescens. Bacteria were incubated for 6 h at 37 °C. Data represent mean ± SD of n ≧ 3 performed in triplicate. Negative control (water) was set to 100 % to calculate the relative quantity of living cells. Data for further bacterial strains are found in Figures S2-S4.