Therapeutic index of gramicidin S is strongly modulated by d-phenylalanine analogues at the β-turn

Abstract

Analogues of the cationic antimicrobial peptide gramicidin S (GS), cyclo(Val-Orn-Leu-d-Phe-Pro)₂, with d-Phe residues replaced by different (restricted mobility, mostly) surrogates have been synthesized and used in SAR studies against several pathogenic bacteria. While all d-Phe substitutions are shown by NMR to preserve the overall β-sheet conformation, they entail subtle structural alterations that lead to significant modifications in biological activity. In particular, the analogue incorporating d-Tic (1,2,3,4-tetrahydroisoquinoline-3-carboxylic acid) shows a modest but significant increase in therapeutic index, mostly due to a sharp decrease in hemolytic effect. The fact that NMR data show a shortened distance between the d-Tic aromatic ring and the Orn δ-amino group may help explain the improved antibiotic profile of this analogue.

Introduction

A remarkable increase in bacterial resistance to classical antibiotics has in recent years emerged as a major global health concern and spurred intense efforts towards the development of new antimicrobial chemotherapy approaches. A number of peptides with antimicrobial activity have been identified in a wide variety of organisms^1,2 and recognized as promising candidates against multi-drug resistant bacteria. An attractive feature of these peptides is that, by acting mostly through interaction with plasma membrane phospholipids, their developing resistance is less likely than for other mechanisms of action, since it would involve major changes in phospholipid composition that would in turn require substantial overhauling of membrane-based enzymatic and transport systems. Gramicidin S (GS) [cyclo(Val-Orn-Leu-d-Phe-Pro)₂] (Figure 1), a cyclic decapeptide isolated from Bacillus brevis³ and active against bacteria and fungi,^4,5 is one of the most widely studied antimicrobial peptides. GS adopts a full β-sheet pleated structure^6-8 where the Val, Orn and Leu residues align to form two antiparallel β-strands while d-Phe and Pro form type II′ β-turns. The structure is stabilized by four inter-strand hydrogen bonds between the Leu and Val residues and displays C₂-symmetry. The β-sheet places the hydrophobic Val and Leu side chains on a non-polar face of the GS molecule and across from the cationic Orn side chains, thus creating an amphiphilic structure that is thought to be essential for bioactivity^5,9. Although not fully unveiled, the mechanism of action of GS sets out with a peptide-membrane interaction, leading to the disruption of the lipid bilayer and concomitant enhancement of its permeability¹⁰, eventually resulting in cell death. Unfortunately, GS not only affects bacterial membranes, but also mammalian cells such as erythrocytes. This hemolytic activity hampers its use as a systemic antibiotic and confines it to topical administration⁵.

Figure 1.

β-sheet structure of gramicidin S.

Significant efforts have been devoted to rationalize the biological behavior of GS with the ultimate goal of generating derivatives with an improved therapeutic index. In this respect, both the β-strand^11-13 and β-turn^14-23 regions of the molecule have been the subject of modification. SAR studies on these GS analogues have identified several factors underlying the biological properties of GS: cationic charge,^10,23 amphipathic character^24,25 and β-sheet structure^23,24,26 are believed to be essential for an interaction of the peptide with lipid bilayers with some degree of specificity. In addition, global hydrophobicity provided not only by the β-sheet²⁷, but also by the β-turn^14-16,23,28 appears to be strongly related to the biological properties of GS. The β-turn region of GS can be successfully replaced by a variety of peptidomimetics, provided the surrogates have hydrophobic character and adequate geometry^5,15,29-31.

This paper decribes a GS analogue series generated by replacing the d-Phe residue in both β-turn regions by hitherto unexplored, Phe-related residues. Mutations at these two positions have received considerable attention^14,23,32,33 as particularly promising in the search for GS analogues with improved pharmacological profiles. We have examined the antimicrobial and hemolytic activity of these analogues along with their secondary structures by NMR and found that, aside from the β-sheet conformation, side chain packing appears to play an important role in biological activity. Thus, the analogue with the best antimicrobial-to-hemolytic activity ratio (peptide 6, Fig. 2) in the series displays rather distinctive packing between its cationic (Orn) and aromatic (d-Tic) sites.

Figure 2.

Amino acids selected as d-Phe replacements for GS analogues 1-6.

Results

Peptide design and synthesis

Amino acid residues selected as d-Phe replacements in GS are shown in Fig. 2. All of them have d configuration and include various modifications at the aromatic side chain such as insertion of a methylene group (Hpa, analogue 1), or further aromatic rings in diverse arrangements (analogues 2-5), including one (5) where the fluorenyl system imposes coplanarity of the two phenyl rings, with presumable conformational restriction. In a final analogue (6), we explored the behavior of 1,2,3,4-tetrahydroisoquinoline-3-carboxylic acid (Tic), in which a methylene unit connects the aromatic substituent to the backbone.

Solid-phase Boc chemistry protocols^18,22 were used to prepare GS and the six analogues in Fig. 2. Non-commercially available Boc-d-Flg-OH was obtained in enantiomerically pure form following our previously reported methodology.³⁴ In addition, the side chain of N-Boc ornithine was protected as a formamide as described by Kitagawa.³⁵ For each peptide, the synthetic route in Scheme 1 was followed. Once sequence assembly was complete, the linear decapeptide was released from the polymeric support by TFMSA acidolysis and submitted to cyclization under high-dilution conditions. Deprotection of the Orn side chains by acid hydrolysis furnished the target peptides (GS and analogues 1–6), which were purified by reversed-phase HPLC in satisfactory yields and high purity. All peptides were further characterized by MS and NMR.

Scheme 1.

Synthetic strategy for GS and its analogues^a

^aReagents and conditions: (a) standard solid-phase peptide synthesis (Boc chemistry); (b) TFA/TFMSA/TIS 10:1:1, rt, 90 min; (c) HBTU/HOBt/DIEA (3:3:5 eq), DMF, rt, 1 h; (d) 20% HCl in MeOH, 37°C, 21 h. d-Phe* stands for d-Phe or the corresponding substitute.

NMR structural studies

Peptides 1-6 were investigated in aqueous solution by NMR in search for characteristic GS traits such as β-sheet conformation. Since GS and all analogues each contain two Pro residues, minor NMR signals from the cis conformation were observed, as expected in Pro-containing peptides. Still, the trans-trans rotamers were confirmed in all cases as the major species by the observation of Pro ¹³C_β - ¹³C_γ chemical shift differences (Δδ^Pro = δ^Cβ − δ^Cγ, ppm) in the 5.0–5.9 ppm range characteristic of trans, vs. Δδ^Pro ∼ 10 ppm for cis Pro residues ³⁶.

The first evidence that the analogues adopt the same structure as GS comes from the similarity of their ¹H, ¹³C and ¹⁵N chemical shifts with those of the parent peptide (Table ST1). The significant differences (≥ 0.2 ppm) observed for some protons are explained by anisotropy effects from the aromatic ring currents. Analogues 5 and 6 are those with larger ¹H chemical shift differences relative to GS, and the only ones for which some ¹³C and ¹⁵N chemical shifts also deviate considerably from GS (≥ 1.0 ppm; Table ST1).

Chemical shifts of C_αH protons and C_α carbons, given their dependence on (φ) and ψ backbone torsion angles, are good indicators of secondary structure formation. For GS (Figure 1), Δδ_CαH conformational shifts (Δδ_CαH = δ_CαH^observed − δ_CαH^{random coil}, ppm) expected for β-strand residues Val1/1′, Orn2/2′ and Leu3/3′ are positive (Δδ_CαH = +0.40 ppm on average in β-sheets)^37,38 and Δδ_Cα conformational shifts (Δδ_Cα = δ_Cα^observed − δ_Cα^{random coil}, ppm) negative (−1.6 ppm on average in β-sheets)³⁷. Turn residues were excluded from this analysis as no references are available for random coil values of d-Phe*4/4′. The Δδ_CαH and Δδ_Cα profiles of GS and peptides 1–4 and 6 are very similar and fit the expected pattern, except for Val1/1′ (Figure 3), whose anomalous C_αH chemical shifts can be explained by the strong anisotropy of the aromatic ring, as reported for some β-hairpin peptides³⁹. The larger Δδ_CαH and very negative Δδ_Cα values displayed by Leu3/3′in analogue 6 are probably due to the cyclic nature of d-Tic4/4′ leading to effects equivalent to those described for Pro-preceding residues⁴⁰. In summary, these data confirm that peptides 1–4 and 6 in aqueous solution adopt the same backbone conformation as GS. The resemblance is not so clear-cut in the case of analogue 5, for which Δδ_CαH values for other residues than Val1/1′ also deviate from the expected behavior (Figure 3). Again, deviations are attributable to aromatic ring current effects, which in this particular peptide might differ from the rest, as the phenyl rings in d-Flg (Figure 2) are necessarily coplanar. Further confirmation about the β-sheet conformation of residues 1/1′ to 3/3′ came from their large ³J_CαH-NH coupling constants (8.2–10.0 Hz; Table ST2). Also, the small ³J_CαH-NH coupling constants of residues 4/4′ corroborate their involvement in a turn.

Figure 3.

Histograms showing the Δδ_CαH and Δδ_Cα profiles (Δδ = δ^observed − δ^{random coil}, ppm) for GS (black bars) and its analogues (patterned or colored bars as indicated in the corresponding insets) in aqueous solution at pH 3.0 and 5°C. Lys δ^{random coil} values⁴⁰ were used to calculate the Δδ values for the Orn residues.

Indeed, the strongest evidence about the conformational features of the peptides came from NOE data. The GS β-sheet structure (Figure 1) is characterized by three NOEs involving backbone protons C_αH Orn2 – C_αH Orn2′ (distance in a canonical antiparallel β-sheet: 2.2 Å), NH Val1 – NH Leu3′, and NH Val1′ – NH Leu3 (distance in a canonical antiparallel β-sheet: 3.3 Å). Since GS and peptides 1–6 are symmetrical, the first NOE is not observable, while the last two must overlap into a single cross-peak corresponding to both NOEs. Such a cross-peak is present in the NOESY spectra of GS and peptides 1, 3 and 6 (Table 1), demonstrating that the three analogues have GS-like β-sheet backbones. The failure to detect such a NOE in 4 and 5 precludes its use as a proof of β-sheet backbone structure, but does not necessarily exclude the presence of this conformational trait.

Table 1.

Summary of NOEs observed for GS and peptides 1-6. Observation of one or more NOE involving at least one side chain proton of the indicated residues is shown by a “+”. d-Phe* stands for d-Phe or the corresponding substitute. “Ov” indicates signal overlapping. “w” and “vw” indicates weak and very weak peak intensities, respectively.

NOE	GS d-Phe	1 d-Hpa	2a d-1-Nal	3 d-2-Nal	4 d-Dip	5 d-Flg	6 d-Tic
NH Val1 – NH Leu3′/NH Val1′ – NH Leu3	+	+	----	+(vw)	----	Ov	+
Val1/1′ – Leu3/3′	+	+	----	+	+	+	+
Orn2/2′ – dPhe*4/4′	----	----	----	----	----	----	+
Leu3/3′ – dPhe*4/4′	+	+	+(w)	+	+	+	+(vw)
dPhe*4/4′ – Pro5/5′	+	+	----	+	+	+	----

^aSignal-to-noise ratio is poor in NOESY spectra of peptide 2 as a consequence of its low solubility.

Aside from NOEs involving backbone protons, NOEs between Val and Leu side chain protons were observed for GS and analogues (Table 1 & Figure 4), except for 2, probably due to low solubility. Since Val and Leu side chains pack together on the same face of the GS β-sheet, these NOEs provide additional proof of the peptides adopting GS-like β-sheet structures.

Figure 4.

Selected NOESY spectral regions for peptides 1 (A) and 6 (B) in H₂O/D₂O 9:1 v/v at pH 3 and 5 °C. Relevant sequential and non-sequential NOEs are boxed.

The cross-strand hydrogen bonding network in the β-sheet (NH Val1 – O Leu3′, NH Val1′– O Leu3, NH Leu3 – O Val1′, and NH Leu3′ – O Val1; see Figure 1) can be also examined by NMR. Experimental evidence for the involvement of NH amide protons in intramolecular hydrogen bonds can be obtained from NH/ND exchange rates and from temperature coefficients of the amide protons (Δδ/δT). The latter are easier to measure and have long been recognized as related to solvent exposure. Accessible amide protons usually have Δδ/δT values within the −6.0 to −9.5 ppb K⁻¹ range, while values above −4.5 ppb K⁻¹ indicate low solvent accessibility and hence involvement in intra-molecular hydrogen bonds. As expected, solvent-exposed NH amide protons of Orn2/2′ and d-Phe*4/4′ displayed temperature coefficients below −6.5 ppb K⁻¹, with peptide 5 as single exception (Table 2). More interestingly, the low Δδ/δT absolute values for the NH amide protons of Leu3/3′ in all peptides, as well as for the NH of Val1/1′ in 6, suggested participation in intramolecular hydrogen bonds, as expected for GS. In contrast, Val1/1′ NH amide protons in both GS and peptides 1–5 had Δδ/δT values in the range typical of solvent-accessible protons, with values larger (in absolute value) than the −9.5 ppb K⁻¹ limit for peptides 3–5. Two explanations for this behavior can be given: either (i) hydrogen bonding in Val1/1′ NH is non-existent or weaker than that of the NH of Leu3/3′, or (ii) anisotropy effects from d-Phe4/4′ (or d-Phe*4/4′) affect the Val1/1′ NH chemical shifts in both GS and 1–5. In either case, the different behavior of Val1/1′ vs. Leu3/3′ NH protons would arise from their proximity to the turn region (Figure 1) and the ensuing disruption of the antiparallel β-sheet regularity. Such deviation from canonical β-sheet for Val1/1′ residues would also account for their anomalous Δδ_CαH and Δδ_Cα values (see above; Figure 3).

Table 2.

Temperature coefficients^a (Δδ/δT, ppb.K⁻¹) for amide protons in GS and its analogues.

Peptide	d-Phe*b	Val1,1′(NH)	Orn2,2′(NH)	Leu3,3′(NH)	d-Phe*4,4′(NH)
GS	d-Phe	−7.4	−7.1	−0.8	−8.6
1	d-Hpa	−6.1	−7.8	−1.2	−11.0
2	d-1-Nal	−8.0	−7.0	−1.5	−8.0
3	d-2-Nal	−11.4	−6.6	0.2	−7.6
4	d-Dip	−13.4	−7.3	−1.7	−10.1
5	d-Flg	−23.4	−3.6	0.0	c
6	d-Tic	−1.3	−8.0	−3.9	—

^aMeasured in H₂O/D₂O 9:1 (v/v) at pH 3.0.

^bd-Phe* stands for d-Phe or the corresponding substitute.

^cNot determined; NH protons not observed at 25 °C probably because of fast exchange with water.

To discern between the two alternatives the NH/ND exchange rates, which provide more definitive evidence about solvent protection and are unaffected by aromatic ring currents, were measured. Thus, the amide NH signals of Orn2/2′ and d-Phe4/4′ in GS at 5 °C disappeared shortly after dissolving the sample in D₂O (see Experimental Section), indicating fast solvent exchange. In contrast, NH protons of Val1/1′ and Leu3/3′ were slow-exchanging, their NMR signals remaining after >18 h. This is in agreement with the NH protons of both Val1/1′ and Leu3/3′ being hydrogen-bonded, and thus suggests that the large Δδ/δT absolute values observed for Val1/1′ NH protons are due to magnetic anisotropy. Their variability among GS and the six analogues stems from their distinctive d-Phe*4/4′ residues, each giving rise to characteristic ring current effects. In this regard, the fact that the Δδ/δT coefficient for Val1/1′ in peptide 6 has the expected value for hydrogen-bonded amide protons suggests that the side chain of d-Tic freezes the phenyl ring in an orientation that minimizes its anisotropy effects over the Val1/1′ NH proton.

Analogue 5, for which the Val1/1′ temperature coefficient is the most negative (−23.4 ppb K⁻¹; Table 2), and that of Orn2/2′ less negative than expected for an exposed amide proton, merits further comment (Table 2 & Figure 1). The most likely explanation is that the anisotropy effects of the large, rigid aromatic d-Flg side chain strongly influence the chemical shifts of both NH protons. This is supported by the fact that the remaining analogues show very similar chemical shifts for the NH protons of Val1/1′ and Orn2/2′ (7.52–7.82 and 8.64–8.74 ppm, respectively; Table ST1), while those of 5 are quite different (8.25 and 8.31 ppm, respectively).

Having shown that analogues 1–6 adopt the same β-sheet backbone structure as the parent GS (perhaps a bit distorted for peptide 5), we next investigated side chain packing, which is responsible for the polar/hydrophobic distribution seemingly essential for membrane interaction and hence antimicrobial and hemolytic activity. A qualitative examination of NOEs from d-Phe*4/4′ (Table 1 & Figure 4) indicates that the aromatic rings interact with the adjacent Leu and Pro residues, both in GS and in 1–5. For analogue 6, in contrast, NOE data suggest that the aromatic rings interact with Orn2/2′ and retain the proximity to Leu3/3′ but not to Pro5/5′. This is confirmed by large chemical shift differences (relative to GS) of some protons of Orn2/2′ (0.4–0.6 ppm; Table ST1) and of Pro5/5′ (1.18 and 0.33 ppm for the C_δδ′H protons; Table ST1), while for the side chain protons of Leu3/3′ the differences do not exceed 0.22 ppm. Proximity between the d-Tic and Orn side chains may be explained by a π-cation interaction.

To further visualize the side chain packing of peptides 1–6 relative to GS, structure calculations were performed. Due to the symmetry of GS and analogues, every observed NOE has four possible assignments. However, after our NMR analysis demonstrated that the peptides adopt the GS β-sheet structure (Figure 1), some NOE ambiguities could be solved. Thus the restraints used for structure calculation combined (i) theoretical restraints for the backbone of strand residues Val1/1′, Orn2/2′, and Leu3/3′ derived from those expected for the GS β-sheet (Figure 1; more details in the Experimental Section and the Supporting Information), and (ii) experimental distance restraints derived from the complete set of NOEs observed for each peptide. For every non-intra-residual NOE, all the existing possibilities were taken into account and the NOE assignments compatible with the GS β-sheet were identified by iterative structure calculations. The final model structures obtained for GS and peptides 1–6 (Figures 5 & SF3) are well defined, particularly for 6 (see RMSD values in Table ST3). A general examination of the structures corroborates the main conformational traits deduced from the qualitative analysis of NMR parameters. Concerning the orientation of the aromatic side chain at positions 4/4′, the phenyl rings in 1 are quite disordered and explore a much wider space than in GS, while the aromatic systems of peptides 4 and 5 adopt quite well defined positions. Also noteworthy is the different relative orientation of the two phenyl rings in d-Flg and d-Dip, in the latter case with the rings almost perpendicular to each other, and one in an orientation close to that of the d-Phe ring in GS (Figure SF3). Even more remarkable are the differences between GS and analogue 6 (Figure 5) in the relative orientation of the aromatic rings and of the positively charged Orn side chains, with noticeable consequences on their respective biological activities (see below).

Figure 5.

Model structures for GS (A) and peptide 6 (B) in two different views. Backbone atoms are shown in black. Side chains for aromatic d-Phe and d-Tic are colored in green, for Pro in cyan, for Orn in red, and for Leu and Val in magenta.

Biological activity

Microbicidal activities were determined in solution, since in solid media they tend to be undermeasured, especially for Gram-negatives.⁴ Two Gram-positive (Staphylococcus aureus and Listeria monocytogenes) and a Gram-negative (Acinetobacter baummanii ATCC 19606 and its isogenic strain resistant to polymyxin E (PXE)) bacterial strains were tested. Although GS is mostly active against Gram-positives, the occurrence of cutaneous infection by A. baumannii and the increasing resistance to PXE, the last universal drug active against these bacteria⁴¹, prompted us to include both of them in the panel. Results are listed in Table 3.

Table 3.

Cytotoxic (hemolytic) and antimicrobial activity of GS and analogues

		Erythrocytes	S. aureus		L. monocytogenes		A. baumanni S		A. baumannii Re
Peptide	D-Phe*a	HC50b (μM)	MIC50c (μM)	TId	MIC50 (μM)	TI	MIC50 (μM)	TI	MIC50 (μM)	TI
GS	D-Phe	21.1 (± 2.6)	7.9 (± 0.8)	2.7 (1)	6.7 (± 0.2)	3.1 (1)	10.1 (± 0.4)	2.1 (1)	13.1 (± 0.0)	1.6 (1)
1	D-Hpa	7.0 (± 0.3)	3.5 (± 0.1)	2 (0.7)	2.4 (± 0.0)	2.9 (0.9)	4.6 (± 0.1)	1.5 (0.7)	6.2 (± 0.1)	1.1 (0.7)
2	D-1-Nal	5.0 (± 0.1)	2.9 (± 0.5)	1.7 (0.6)	8.6 (± 0.6)	0.6 (0.2)	35.9 (± 0.0)	0.1 (0.0)	>40	>0.1
3	D-2-Nal	7.1 (± 3.5)	4.6 (± 0.6)	1.5 (0.6)	7.8 (± 0.4)	0.9 (0.3)	39.6 (± 0.0)	0.2 (0.1)	>40	>0.2
4	D-Dip	8.6 (± 1.1)	3.4 (± 0.0)	2.5 (0.9)	7.7(± 0.6)	1.1 (0.4)	9.6 (± 0.9)	0.9 (0.4)	17.8 (± 0.0)	0.5 (0.3)
5	D-Flg	6.2 (± 0.3)	8 (± 0.0)	0.8 (0.3)	9.2(±1.4)	0.7 (0.2)	15.4 (± 1.3)	0.4 (0.2)	>40	>0.1
6	D-Tic	>50	4.9 (± 0.4)	>10.2 (3.8)	8.3 (± 0.5)	>6 (2.0)	>40	>ND	>40	ND

^ad-Phe* stands for d-Phe or the corresponding substitute.

^bHC₅₀: peptide concentration required for 50% lysis of sheep erythrocytes.

^cMIC₅₀: peptide concentration required for 50% inhibition of bacterial growth after 24 h, relative to a control culture.

^dTI (therapeutic index)= HC₅₀/MIC₅₀

^eAn A. baumannii strain resistant to polymyxin E (colistin). ND, not determined. Standard deviations (SD) are shown in parentheses.

Within the Gram-positive group, the tested set of analogues had better and comparable activities, respectively, for Staphylococcus and Listeria than the parental GS. In contrast, against the Gram-negative A. baumannii, activity in the analogue series tended to deteriorate relative to GS, more markedly so for the isogenic PXE-resistant strain. The sole exception was the d-Hpa-containing peptide 1, which performed slightly better than GS against both PXE-sensitive and resistant strains, suggesting it is impervious to modifications in lipopolysacharide (LPS) structure.

The hemolytic effect of the peptides was evaluated on sheep erythrocytes (Table 3) as an indicator of toxicity against mammalian cells. In this assay, five (1-5) out of six analogues turned out to be more toxic than the parent GS. The finding that the d-Flg analogue (5) was more cytotoxic and yet less potent than the parent antibiotic underscores the fact that subtle structural changes in GS may translate into different abilities to damage prokaryotic and eukaryotic cells. An opposite, though similar and more promising, segregation of antibacterial and hemolytic activities was found for 6, which has GS-like MIC₅₀ values for Gram-positives (Table 3) but is essentially harmless towards erythrocytes.

Beyond adequate intrinsic microbicidal activity, a convenient assessment of the clinical potential of an antibiotic is provided by its therapeutic index (TI, defined as HC₅₀/MIC₅₀, see Table 3), where the cytotoxic and microbicidal effects are combined. As found in previous SAR studies on GS ^23,25,42,43, improved TIs are more often due to a decrease in cytotoxicity than to an increase in antibiotic potency. In the present series, the above-mentioned (higher than for GS) toxicity values, coupled with the minute, if any, improvements in microbicidal activity, resulted in TI values slightly lower or clearly worse than that of the parent structure. The sole exception to this pattern was the d-Tic analogue (6), for which a modest but significant improvement in TI was obtained (Table 3).

Discussion

The type II′ β-turn is a structural motif especially suited to induce alignment of its two side sequence segments into an antiparallel β-sheet. The quintessential example of this situation is GS, whose type II′ β-turns have been a testing ground for extensive d-Phe-Pro replacements by a variety of peptidomimetics, with variable success (see Introduction and refs.29^-31,44-46). In comparison, fewer studies have addressed the substitution of either l-Pro^17,47 or d-Phe^33,46,48-50 by appropriate surrogates. The goal of this work was to explore and structurally interpret the fine-tuning of GS antibiotic activity by replacement of d-Phe with a novel set of non-coded aromatic d-amino acids already tested as turn-promoters.

The activity of GS is assumed to be somehow related to the permeabilization of bacterial membranes, though no unified opinion on the nature of such process exists⁵¹. Hydrophobicity has been cited as a relevant parameter^25,42,43, and inspection of the current set of d-Phe analogues (Figure 2) shows considerable variation in the number and size of aromatic side-chains, hence in overall hydrophobicity at this position. The presence of a sizable hydrophobic domain is closely related to hemolytic activity in antimicrobial peptides, as it seemingly promotes a higher partition of the peptide into biological membranes, regardless of their zwitterionic or anionic character^24,25,42. Accordingly, the hemolytic and Gram-positive microbicidal activities run somewhat parallel, provided hydrophobicity does not drastically reduce peptide solubility in aqueous medium⁵². For Gram-negatives, however, the situation changes in that an external barrier, i.e., the outer membrane (OM), must be disrupted prior to the “lethal hit” i.e., permeabilization of the inner membrane (IM). In this scenario, while high hydrophobicity improves the affinity of the antibiotic peptide for LPS (the major component of the OM), excess affinity may sequester the peptide into the OM, hence precluding its crucial interaction with the IM target. Therefore, a rather subtle balance between hydrophobicity and cationic character is desirable for activity against Gram-negatives¹⁸. These considerations broadly agree with the finding that peptides 1-5, more toxic than GS towards erythrocytes (Table 3), are also more hydrophobic than GS by an accepted criterion²³, namely RP-HPLC retention times consistently longer than those of the parent peptide (see Supporting Information, Figure SF1). On the other hand, the d-Tic (6) analogue combines decreased cytotoxicity with higher hydrophobicity (longer HPLC retention time) than GS (Table 3). Likewise, the similar HPLC retention times for peptides 1-4 (10.3-10.6 min, Figure SF1) do not correlate well with their antimicrobial profiles. Thus, while 1 improved on GS against all bacteria tested, Nal-containing analogues 2 and 3, with presumably comparable hydrophobicity, were similar to GS against Gram-positives but practically inactive against Gram-negatives. In conclusion, it appears that, at least for this series, hydrophobicity has limited value in explicating antibiotic performance.

NMR results shed additional light on the structure-activity analysis of the peptides. In particular, they reveal a varying degree of conformational freedom in their side chains that can be reasonably related to differences in antimicrobial activity. For instance, the aromatic ring of the d-Hpa-containing analogue (1), separated from the backbone by an extra methylene group, can plausibly explore a larger conformational space and consequently a wider repertoire of membrane (either prokaryotic or eukaryotic) interaction modes than the d-Phe-containing GS, thus explaining its increased activity towards both bacterial and mammalian cells. In contrast, the naphthalene rings in 2 and 3 have conformational freedoms similar in principle to the phenyl group of GS, but their bulkiness is likely to pose a limitation on their interaction with LPS and ensuing OM disruption in Gram-negative bacteria. For Gram-positives, on the other hand, no comparable tightly packed external barrier exists, which would agree with both the preservation of antibiotic activity in both analogues, as well as their hemolytic properties. Similar considerations on the orientation and/or rotational freedom of the aromatic rings can be applied to the contrasting activities of 4 (d-Dip) and 5 (d-Flg). In the latter analogue, the two aromatic rings are fastened into a bulky planar tricylic system, whose predictable sluggishness may help to explain its significantly worse profile than 4.

By far the most striking result in the present series is the improved therapeutic profile of the d-Tic analogue 6. With an apparent hydrophobicity (based on RP-HPLC retention time) slightly higher than GS, it is basically equipotent to GS against Gram-positives but has considerably lower toxicity, hence a much better TI (Table 3). From the structural point of view, 6 has two distinctive features over the other analogues in the series. First, the severe conformational restriction on the (proline-like) imino acid d-Tic imposed by the cyclization between the phenyl group and the backbone, and borne out by the observation of practically a single conformer by NMR. Interestingly, the hydrogen-bonding interaction between the δ-amino group of Orn and the backbone carbonyl of d-Phe reported for GS and some analogues^53,54 would thus not be present in analogue 6. The absence of this favorable effect would most likely be compensated by the stabilizing contribution that the increased rigidity of d-Tic -compared with d-Phe- imparts to the β-turn structure. As a rigid side chain is conceivably more difficult to insert into lipid bilayers than a flexible one (i.e., that of GS), this could account for the lack of hemolytic effect of 6 and thus its improved antibiotic profile. Second, in 6 the δ-NH₃⁺ groups of Orn are positioned close to the aromatic ring of d-Tic, possibly favored by a π-cation interaction. According to NMR and MD studies on the membrane-bound structure of GS ^43,55,56, the peptide lies flat relative to the bilayer plane, near the interface between polar and apolar regions, so that the δ-NH₃⁺ groups of Orn can interact with phospholipid polar head groups. Now, if a similar contact model is assumed for analogue 6, the aforementioned π-cation interaction between d-Tic and Orn must disappear. The thermodynamic penalty thus incurred would be higher for the interaction with (zwitterionic) eukaryotic than for prokaryotic membranes, where the loss of the d-Tic – Orn interaction would be compensated by electrostatic attraction between Orn δ-NH₃⁺ and phosphate head groups. Again, this may account for the higher relative loss in hemolytic than in bactericidal effect.

Conclusions

Analogues of the antimicrobial peptide GS with different aromatic replacements at both d-Phe residues have been synthesized in order to explore how the conformational freedom and the size of the aromatic moiety modulate both biological and structural properties. NMR studies confirmed that the overall molecular shape of the GS framework is maintained in all analogues (some distortion cannot be discarded for the d-Flg derivative), so that variations in antibiotic profile can be ascribed to the different d-Phe replacements. The results suggest that both bulkiness and orientation of the aromatic system at the 4/4′ position are essential for biological activity, to the point that slight modifications in these parameters bring about significant changes in antibacterial activity as well as in the desirable specificity for bacterial over mammalian cells. An interesting finding is that analogue 6, with d-Tic replacing d-Phe, retains antibiotic potency against Gram-positive bacteria but is hardly cytotoxic. Our NOE data clearly indicate that in this analogue the Orn side chain orientates towards the d-Tic aromatic rings. A plausible explanation for this orientation is the existence of a cation-π interaction between the δ-NH₃⁺ groups of the Orn side chain and the aromatic ring of d-Tic. In any event, and regardless of the nature of the interaction, the specific orientation of the d-Tic aromatic ring relative to the Orn side chain allows explaining the distinct biological behavior exhibited by this peptide. While the increase in therapeutic index relative to GS is only five-fold, it is nonetheless remarkable because, in contrast to other studies^25,42, it has been achieved with minimal alteration of the GS structure, including chirality and cycle size. Modifications of this type may be helpful in understanding the mechanism of action of GS and developing peptide antibiotics with improved pharmaceutical applications.

Experimental Section

Chemicals and instrumentation for peptide synthesis

Fluorenylglycine was prepared as previously reported.³⁴N-Boc ornithine was purchased from Bachem (Bubendorf, Switzerland) and its side-chain amino group was protected as a formamide by reaction with formic acid and 1,1′-oxalyldiimidazole.³⁵ All other amino acids were purchased from Senn Chemicals (Dielsdorf, Switzerland), NeoMPS (Strasbourg, France) or Fluka (Buchs, Switzerland). Chloromethylated Merrifield resin and TFMSA were from Fluka, HATU from GenScript (Piscataway, NJ), and HBTU from Matrix Innovation (Montreal, Quebec). Solvents and other chemicals were from SDS (Peypin, France). Mass determination by the MALDI-TOF technique was done in a Voyager DE-RP spectrometer (Applied Biosystems, Foster City, CA) using 2,5-dihydroxybenzoic acid as a matrix. Analytical RP-HPLC was performed on an LC-2010A workstation (Shimadzu Corporation, Kyoto, Japan) with a Luna C₈ (3 μm, 50 × 4.6 mm) column (Phenomenex B.V., Utrecht, The Netherlands) eluted with a linear 5–95% gradient of CH₃CN (+0.036% TFA, v/v) into H₂O (+0.045% TFA, v/v) over 15 min at 1 mL/min flow rate, with UV detection at 220 nm. Preparative RP-HPLC purification was done on a Phenomenex Luna C₈ column (10 μm, 250 × 10 mm) running linear gradients of CH₃CN (+0.1% TFA, v/v) into H₂O (+0.1% TFA, v/v) as indicated for each peptide, at a flow rate of 5 mL/min. The preparative system included two Shimadzu LC-8A pumps, a Shimadzu SPD-10A detector and a Foxy Jr. fraction collector (Teledyne Isco, Lincoln, NE).

General procedure for peptide synthesis

All peptides were synthesized manually by solid-phase methods on a Boc-Pro-Merrifield resin (0.1 mmol) using Boc chemistry. Boc-amino acids (0.3 mmol) were coupled by HBTU/DIEA (0.3 and 0.6 mmol, respectively) for 30–45 min in DMF. HATU was used instead of HBTU for coupling of the 4/4′ residues. To avoid diketopiperazine formation, the third amino acid of each sequence was incorporated by the in situ neutralization method⁵⁷. Coupling and deprotection reactions were monitored by the nynhidrin⁵⁸ or p-nitrophenylester⁵⁹ colorimetric tests. The linear decapeptide was cleaved from the resin by treatment with TFA/TFMSA/TIS 10:1:1 (v/v/v) (4 mL) at room temperature for 90 min, then precipitated with cold tert-butyl methyl ether. The peptide was redissolved in glacial acetic acid, filtered off the resin and lyophilized. The residue was dissolved in DMF to a final concentration of 2 mg/mL and stirred for 1 h at room temperature in the presence of HBTU/HOBt/DIEA (3:3:5 eq). After solvent removal and lyophilization, the cyclized peptide was deformylated by treatment with 20% hydrochloric acid in methanol at 37°C for 21 h. The solvent was evaporated under reduced pressure, the residue was taken up in glacial acetic acid and lyophilized. Final purification was by preparative reversed-phase HPLC as indicated in each case. HPLC-homogeneous fractions were combined and lyophilized to give white powders of ≥99% HPLC purity (see Supporting Information, Figure SF2).

• cyclo(Val-Orn-Leu-d-Phe-Pro)₂ (GS): RP-HPLC, linear 35–70% CH₃CN gradient into H₂O for 30 min (33 mg, 0.029 mmol, 25% overall yield). [M+H]⁺_calcd = 1141.7; [M+H]⁺_obs = 1141.1.

• cyclo(Val-Orn-Leu-d-Hpa-Pro)₂ (1): RP-HPLC, linear 45–75% CH₃CN gradient into H₂O for 30 min (51 mg, 0.044 mmol, 41% overall yield). [M+H]⁺_calcd = 1169.7; [M+H]⁺_obs = 1169.5.

• cyclo(Val-Orn-Leu-d-1-Nal-Pro)₂ (2): RP-HPLC, linear 45–75% CH₃CN gradient into H₂O for 30 min (34 mg, 0.027 mmol, 26% overall yield). [M+H]⁺_calcd = 1241.7; [M+H]⁺_obs = 1241.8.

• cyclo(Val-Orn-Leu-d-2-Nal-Pro)₂ (3): RP-HPLC, linear 45–75% CH₃CN gradient into H₂O for 30 min (26 mg, 0.021 mmol, 20% overall yield). [M+H]⁺_calcd = 1241.7; [M+H]⁺_obs = 1241.6.

• cyclo(Val-Orn-Leu-d-Dip-Pro)₂ (4): RP-HPLC, linear 50–80% CH₃CN gradient into H₂O for 30 min (22 mg, 0.017 mmol, 15% overall yield). [M+H]⁺_calcd = 1293.8; [M+H]⁺_obs = 1293.6.

• cyclo(Val-Orn-Leu-d-Flg-Pro)₂ (5): RP-HPLC, linear 55–85% CH₃CN gradient into H₂O for 30 min (18 mg, 0.015 mmol, 14% overall yield). [M+H]⁺_calcd = 1289.7; [M+H]⁺_obs = 1289.8).

• cyclo(Val-Orn-Leu-d-Tic-Pro)₂ (6): RP-HPLC, linear 45–65% CH₃CN gradient into H₂O for 30 min (23 mg, 0.020 mmol, 19% overall yield). [M+H]⁺_calcd = 1165.7; [M+H]⁺_obs = 1165.8.

NMR spectroscopy

Samples for NMR experiments were prepared at 1–2 mM peptide concentration in 0.5 mL of H₂O/D₂O 9:1 (v/v) or pure D₂O at pH 3.0. pH was measured with a glass microelectrode and was not corrected for isotope effects. NMR spectra were acquired on a Bruker AV 600 MHz spectrometer equipped with a z-gradient cryoprobe. A methanol sample was used to calibrate the temperature of the NMR probe. One-dimensional (1D) and two-dimensional (2D) spectra were acquired by standard pulse sequences using presaturation of the water signal. Mixing times for 2D TOCSY and NOESY were 60 and 150ms, respectively. The ¹H–¹³C and ¹H–¹⁵N HSQC spectra⁶⁰ at natural ¹³C and ¹⁵N abundance were recorded in D₂O and H₂O/D₂O 9:1 (v/v), respectively. Data were processed using TOPSPIN (Bruker Biospin, Rheinstetten, Germany) software. Sodium 2,2-dimethyl-2-silapentane-5-sulphonate (DSS) was used as an internal reference for ¹H chemical shifts. The ¹³C and ¹⁵N chemical shifts were indirectly calibrated by multiplying the spectrometer frequency that corresponds to 0 ppm in the ¹H spectrum, assigned to internal DSS reference, by 0.25144954 and 0.101329118, respectively.⁶¹

The ¹H-NMR signals of the peptides were assigned by sequential assignment methods.⁶² The ¹³C and ¹⁵N resonances were then assigned following the cross-correlations observed in the HSQC spectra between the proton and the hetero-nucleus to which it is bonded.

To measure the NH/ND exchange rates, a freeze-dried sample of peptide GS was solved in pure D₂O at pH 3.0, and after the 10–15 min required to set up the NMR experiment consecutive 1D and 2D 20ms-TOCSY spectra were recorded at 5 °C.

Structure calculation

Structure calculations were performed by using the CYANA program⁶³ and an annealing strategy. Since the non-proteinogenic amino acids were not included in the standard CYANA libraries, we built them using MOLMOL⁶⁴ and manual optimization (these libraries are available upon request from the authors). For this purpose, X-ray data of compounds containing these amino acids were retrieved from the Cambridge Structural Database⁶⁵. Theoretical constraints for the GS β-sheet incorporated for structure calculation included (φ) and ψ angle restraints, lower and upper-limit distance restraints for the four characteristic cross-strand hydrogen-bonds, and upper-limit distance restraints for the backbone atoms of strand residues (Table ST4). Experimental distance constraints were derived from 2D NOESY spectra recorded in H₂O, also in D₂O for analogue 6. The NOE cross-peaks were integrated by using the automatic integration subroutine of the Sparky program (T.D. Goddard & S.G. Kneller, University of California at San Francisco) and then calibrated and converted to upper-limit distance constraints with CYANA⁶³. Given the symmetrical nature of the peptides, for structure calculations residues were re-numbered from 1 to 10 starting at Leu3 (Figure 1). Lower and upper limit restraints required for peptide backbone cyclization were introduced with CYANA⁶³ (see Supporting Information). For each peptide, a total of 50 conformers were generated and the 20 conformers with the lowest target function were analyzed. Model structures for GS and analogues 1–6 were examined with MOLMOL. A side chain torsion angle was considered as well defined when its root mean square deviation between values in the 20 best calculated structures was less than ±30°.

Antimicrobial activity

Stocks of Staphylococcus aureus CECT 240, Listeria monocytogenes CECT 4032, Acinetobacter baumannii ATCC 19606 and its isogenic colistin-resistant strain 19606R (obtained by continuous growing under increasing colistin concentration) were maintained at -80 °C in freezing medium (65% glycerol, 0.1 M MgSO₄, 25 mM Tris-HCl, pH 8.0). Two days prior to the assay for microbicidal activity, they were thawed and grown in MBH medium (Mueller-Hinton II Broth Cation Adjusted (Becton-Dickinson, Cockeysville, MD) at 37 °C; for 19606R, 64 μg/mL colistin sulfate (Sigma, Madrid, Spain) was included⁴¹. Bacterial cells were harvested at exponential growth phase, washed twice with phosphate buffered saline (PBS, 10 mM Na₂HPO₄, 1 mM KH₂PO₄, 140 mM NaCl, 3 mM KCl, pH 7.0) and resuspended in MBH, at 5 × 10⁵ CFU/mL. Aliquots (100 μL) from this suspension were transferred into a polypropylene 96 well plate, and bacteria were allowed to proliferate for 24 h at 37 °C in the presence of the corresponding analogue concentration. Afterwards, growth was measured by turbidimetry at 600 nm in a model 680 microplate reader (Bio-Rad Laboratories, Hercules. CA). MIC₅₀ was defined as the lowest peptide concentration inhibiting bacterial growth by 50%, relative to untreated control, and was calculated using the SigmaPlot (Systat Software, San Jose, CA) software, v. 9.0.

Hemolytic activity assay

Hemolytic activity of the peptides was determined by triplicate. Defibrinated sheep blood (Biomedics, Madrid, Spain) was centrifuged and washed twice with Hank's medium (136 mM NaCl; 4.2 mM Na₂HPO₄; 4.4 mM KH₂PO₄; 5.4 mM KCl; 4.1 mM NaHCO₃, pH 7.2), supplemented with 20 mM d-glucose (Hank's-Glc). Erythrocytes were resuspended in the same buffer at 2 ×10⁷ erythrocytes/mL and 100 μL-aliquots of the suspension were incubated with the peptides (4 h, 37 °C). The remaining erythrocytes were harvested in a Micro 200 microfuge (A. Hettich GmbH & Co KG, Germany) (14,000 rpm, 5 min, 4 °C), 80 μL of the supernatant were transferred into a 96-well culture microplate, and hemoglobin release was read at 550 nm in a Bio Rad 680 (Hercules, CA) microplate reader. The asymptotic ordinate of the GS supernatant was taken as 100% hemolysis. HC₅₀ values were calculated using SigmaPlot, version 9.0.

Abbreviations

ATCC

American Type Culture Collection

Boc

tert-butoxycarbonyl

CECT

Spanish Type Culture Collection

COSY

correlated spectroscopy

DIEA

N,N-diisopropylethylamine

Dip

β,β-diphenylalanine

DMF

N,N-dimethylformamide

DSS

2,2-dimethyl-2-silapentane-5-sulphonate

Flg

fluorenylglycine

For

formyl

GS

gramicidin S

HATU

2-(7-aza-1H-benzotriazole-1-yl)-1,1,3,3-tetramethyluronium hexafluorophosphate

HBTU

2-(1H-benzotriazole-1-yl)-1,1,3,3-tetramethyluronium hexafluorophosphate

HC₅₀

50% hemolytic concentration

HOBt

N-hydroxybenzotriazole

Hpa

homophenylalanine

HSQC

heteronuclear single quantum coherence spectra

IM

inner membrane

LPS

lipopolysacharide

MALDI-TOF

matrix-assisted laser desorption ionization-time of flight

MIC₅₀

minimal inhibitory concentration

1-Nal

1-naphthylalanine

2-Nal

2-naphthylalanine

NOESY

nuclear Overhauser enhancement spectroscopy

OM

outer membrane

Orn

ornithine

Phg

phenylglycine

PXE

polymyxin E

RMSD

root mean square deviation

SAR

structure-activity relationship

TFA

trifluoroacetic acid

TFMSA

trifluoromethanesulfonic acid

TI

therapeutic index (=HC₅₀/MIC₅₀)

Tic

1,2,3,4-tetrahydroisoquinoline-3-carboxylic acid

TIS

triisopropylsilane

TOCSY

total correlated spectroscopy