Design, Synthesis, and Study of Lactam and Ring-Expanded Analogues of Teixobactin

Abstract

This paper describes the chemical synthesis, X-ray crystallographic structure, and antibiotic activity assay of lactam analogues of teixobactin and explores ring-expanded analogues of teixobactin with β³-homo amino acids. Lactam analogues of teixobactin containing all four stereoisomers of azathreonine at position 8 were synthesized on a solid support from commercially available stereoisomeric threonine derivatives. The threonine stereoisomers are converted to the diastereomeric aza-threonines by mesylation, azide displacement, and reduction during the synthesis. d-aza-Thr₈,Arg₁₀-teixobactin exhibits 2–8-fold greater antibiotic activity than the corresponding macrolactone Arg₁₀-teixobactin. Aza-teixobactin analogues containing other stereoisomers of aza-threonine are inactive. A dramatic 16–128-fold increase in the activity of teixobactin and teixobactin analogues is observed with the inclusion of 0.002% of the mild detergent polysorbate 80 in the MIC assay. The X-ray crystallographic structure of N-Me-d-Gln₄,d-aza-Thr₈,Arg₁₀-teixobactin reveals an amphipathic hydrogen-bonded antiparallel β-sheet dimer that binds chloride anions. In the binding site, the macrolactam amide NH groups of residues 8, 10, and 11, as well as the extra amide NH group of the lactam ring, hydrogen bond to the chloride anion. The teixobactin pharmacophore tolerates ring expansion of the 13-membered ring to 14-,15-, and 16-membered rings containing β³-homo amino acids with retention of partial or full antibiotic activity.

Graphical Abstract

INTRODUCTION

Teixobactin is a new class of peptide antibiotic against Gram-positive bacteria that inhibits cell-wall formation, interrupting both the synthesis of peptidoglycan and the synthesis of teichoic acid, and ultimately causing bacterial cell lysis.^{1 ,2} Teixobactin is thought to bind the highly conserved prenyl-pyrophosphate-saccharide regions of lipid II and related membrane-bound cell-wall precursors, and thus precludes the development of antibiotic resistance.³ Furthermore, these targets are extracellular and represent the bottleneck of peptidoglycan synthesis. Teixobactin exhibits remarkable antibiotic activity against all important Gram-positive pathogens, including methicillin-resistant Staphylococcus aureus (MRSA), drug-resistant Streptococcus pneumonia, and vancomycin-resistant Enterococci (VRE).

Teixobactin is a non-ribosomal undecapeptide containing a linear tail (residues 1–7) and a macrocyclic ring (residues 8–11). It contains four d-amino acids at positions 1, 4, 5, and 8, namely N-Me-d-Phe₁, d-Gln₄, d-allo-Ile₅, and d-Thr₈, as well as seven l-amino acids at positions 2, 3, 6, 7, 9, 10, and 11. The tail contains a unique pattern of d-l-l-d-d-l-l stereochemistry with a hydrophobic–hydrophobic–hydrophilic–hydrophilic–hydrophobic–hydrophobic–hydrophilic pattern of sidechains. The macrocyclic ring is composed of d-Thr₈–Ala₉–allo-End₁₀–Ile₁₁, in which the C-terminal Ile₁₁ and the hydroxy group of d-Thr₈ form an ester bond to close the 13-membered lactone ring. Residue 10 is the rare amino acid allo-enduracididine (allo-End₁₀), a cyclic arginine analogue.

We recently reported the X-ray crystallographic structure of a teixobactin analogue.⁴ The analogue forms hydrogen-bonded antiparallel β-sheet dimers that bind sulfate anions (Figure 1A). In the X-ray crystallographic structure, the three NH groups of the macrolactone ring form three hydrogen bonds to each sulfate anion (Figure 1B). The macrolactone ring oxygen points toward the bound anion but is unable to form a hydrogen bond. The N-terminal ammonium group of the second monomer subunit of the dimer also hydrogen bonds to the sulfate anion. The dimers further assemble to form a double-helix of β-sheet fibrils. The binding of the sulfate anions suggests how teixobactin might bind to the anionic pyrophosphate group of lipid II and related cell-wall precursors, and thus inhibit cell-wall biosynthesis.^4,5

Figure 1.

(A) Antiparallel β-sheet dimer formed by teixobactin. (B) Coordination of an anion by the teixobactin macrolactone ring. (C) Hypothesized coordination of an anion by an azateixobactin macrolactam ring.

The current paper begins by exploring the hypothesis that replacement of the lactone ring oxygen with an NH group will allow the resulting macrolactam ring to better bind anions by forming an additional hydrogen bond to the bound anion and thus lead to enhanced antibiotic activity (Figure 1C).⁶ We report a synthesis of macrolactam derivatives of teixobactin that contain d-aza-threonine at position 8 and find that a lactam derivative of teixobactin is two to eight times more active than the corresponding lactone. We then explore whether the 13-membered macrocyclic ring composed of residues 8–11 is optimal for binding the pyrophosphate group of lipid II and related cell-wall precursors by expanding the ring with β³-homo amino acids and find that the teixobactin pharmacophore tolerates ring expansion to 14-, 15-, and 16-membered rings with retention or partial retention of activity. Through these studies, we further elucidate the role of the macrocyclic ring in the teixobactin pharmacophore.

RESULTS AND DISCUSSION

Solid-Phase Syntheses of Lactam Analogues of Teixobactin Containing Aza-Threonine at Position 8.

We had previously reported a synthetic route to teixobactin analogues through solid-phase peptide synthesis (SPPS) on 2-chlorotrityl chloride resin followed by solution-phase macrolactamization.⁷ In this route, d-Thr₈ is introduced without a protecting group at the hydroxy position. Because d-aza-threonine is not commercially available in a suitably protected form, we envisioned adapting this route by incorporating d-allo-Thr at position 8 and converting it to d-aza-threonine on the solid support by an SN2 displacement reaction with a nitrogen-containing nucleophile.

We tested this approach using the tripeptide Boc-Ala-d-allo-Thr-Gln(Trt) on 2-chlorotrityl chloride resin and found that attempted conversion of the d-allo-Thr residue to the mesylate and then to the azide instead gave elimination to form the corresponding dehydropeptide (Figure S1). We hypothesized that the elimination reaction could be avoided by introducing the azide group before elongating the peptide chain (Figure S2). Because the Fmoc protecting group proved labile to sodium azide, we first replaced the Fmoc group on d-allo-Thr with an Alloc group, to give the dipeptide Alloc-d-allo-Thr-Gln(Trt) on resin. This dipeptide was then converted to the mesylate and then to Alloc-d-azido-Thr-Gln(Trt) on resin by treating with triethylamine and mesyl chloride in dichloromethane, followed by sodium azide in a mixture of 15-crown-5 and DMF. Alloc deprotection with (Ph₃P)₄Pd and phenylsilane liberated the α-amino group of d-azido-threonine for subsequent SPPS. This sequence of steps proceeded cleanly and afforded the tripeptide Boc-Ala-d-azido-Thr-Gln(Trt)-OH in greater than 90% conversion by HPLC analysis (Figures S3–S7).

Reduction of the azide group of the resin-bound Boc-Ala-d-azido-Thr-Gln(Trt) to an amino group proved challenging. Treatment with stannous chloride (SnCl₂, PhSH, DIPEA)⁸ resulted in approximately 60% conversion after four treatments with the reducing cocktail; Staudinger reduction (Ph₃P and H₂O/THF)⁹ stalled at the imino-phosphorane intermediate, and no hydrolysis to the amine was observed, even at high temperatures (50 °C) and in various mixtures of solvents. Saneyoshi et al. reported a derivative of triphenylphosphine, triphenylphosphine-2-carboxamide (Ph₂P-o-C₆H₄CONH₂), which contains an ortho-carboxamide group to facilitate the hydrolysis of the imino-phosphorane intermediate by providing anchimeric assistance.¹⁰ When we used triphenylphosphine-2-carboxamide in our model system, we observed good conversion from the azido group to the amino group (Figures S8–S9).

We applied the conditions that we developed to the synthesis of d-aza-Thr₈,Arg₁₀-teixobactin (2a). The synthesis began by attaching Fmoc-Ala-OH to 2-chlorotrityl resin (Figure 2).¹¹ Fmoc-d-allo-Thr-OH was then introduced by standard Fmoc-based SPPS using HCTU as the coupling reagent. The Fmoc group was removed by treatment with 20% piperidine in DMF, and the N-terminus was protected with allyl chloroformate. The hydroxy group of d-allo-Thr₈ was then converted to a mesyl group with methanesulfonyl chloride and to an azide group with sodium azide. The Alloc protecting group was removed with (Ph₃P)₄Pd and phenylsilane, and residues 7 through 1 were introduced by SPPS. The azide group of d-azido-Thr₈ was reduced to the corresponding amine with triphenylphosphine-2-carboxamide and H₂O. Residues 11 and 10 were then introduced by SPPS.¹¹ Fmoc deprotection followed by cleavage from the resin with 20% hexafluoroisopropanol (HFIP) in CH₂Cl₂ afforded the protected acyclic precursor. Macrolactamization with HBTU and HOBt, followed by global deprotection with trifluoroacetic acid (TFA), RP-HPLC purification, and lyophilization, afforded 13.5 mg of d-aza-Thr₈,Arg₁₀-teixobactin (2a) as the trifluoroacetate salt with ≥95% purity, from a 0.15 mmol scale synthesis.

Figure 2.

Synthesis of d-aza-Thr₈,Arg₁₀-teixobactin (2a).

This synthetic route proved versatile and also allowed us to prepare the three diastereomeric analogues, d-allo-aza-Thr₈,Arg₁₀-teixobactin (2b), l-aza-Thr₈,Arg₁₀-teixobactin (2c), and l-allo-aza-Thr₈,Arg₁₀-teixobactin (2d) (Figure 3). Replacing d-allo-threonine with d-threonine in the synthesis afforded d-allo-aza-threonine and gave diastereomer 2b. Replacing d-allo-threonine with l-allo-threonine in the synthesis afforded l-aza-threonine and gave diastereomer 2c. Replacing d-allo-threonine with l-threonine in the synthesis afforded l-allo-aza-threonine and gave diastereomer 2d.¹² The stereoisomers containing l-aza-threonine and l-allo-aza-threonine (2c and 2d) exhibited an exceptionally strong propensity to form gels, making purification difficult and affording only low yields of these diastereomeric teixobactin analogues after purification by RP-HPLC.

Figure 3.

Diastereomeric azateixobactin analogues 2a–2d.

Minimum Inhibitory Concentration (MIC) Assays of Lactam Analogues of Teixobactin.

We assessed the antibiotic activity of d-aza-Thr₈,Arg₁₀-teixobactin (2a) and stereoisomeric analogues (2b–2d) in minimum inhibitory concentration (MIC) assays against three Gram-positive bacteria (S. aureus, S. epidermidis, and B. subtilis, Table 1). We used teixobactin and vancomycin as positive controls and the Gram-negative bacterium E. coli as a negative control. d-aza-Thr₈,Arg₁₀-teixobactin (2a) exhibits MIC values of 0.5–1 μg/mL across the Gram-positive bacteria and is twice as potent as Arg₁₀-teixobactin (1a).^13,14 The teixobactin analogues containing other stereoisomers of aza-threonine (2b–2d) exhibited no antibiotic activity (MIC >32 μg/mL).

Table 1.

MIC values of teixobactin and teixobactin analogues (μg/mL).

	Staphylococcus aureus ATCC 29213	Staphylococcus epidermidis ATCC 14990	Bacillus subtilis ATCC 6051	Escherichia coli ATCC 10798
Arg10-teixobactin (1a)	2	1	2	>32
d-aza-Thr8,Arg10-teixobactin (2a)	1	0.5	1	>32
d-allo-aza-Thr8,Arg10-teixobactin (2b)	>32	>32	>32	>32
l-aza-Thr8,Arg10-teixobactin (2c)	>32	>32	>32	>32
l-allo-aza-Thr8,Arg10-teixobactin (2d)	>32	>32	>32	>32
teixobactin	0.25	0.25	0.25–0.5	>32
vancomycin	0.25	0.25	0.5	>32

Effect of Polysorbate 80 on Antibiotic Activity.

In the original report of teixobactin, the authors described the use of the mild detergent polysorbate 80 at 0.002% concentration in their MIC assays.¹ Having performed our previous MIC assays without polysorbate 80, we decided to investigate the effect of polysorbate 80 on the MIC values of azateixobactin analogue 2a. When we performed the MIC assay with broth containing 0.002% polysorbate 80, the MIC values of azateixobactin analogue 2a decreased from 0.5–1 μg/mL to 0.008–0.03 μg/mL, a dramatic 16–128-fold increase in antibiotic activity (Table 2). We also observe a similar decrease in the MIC of teixobactin, from 0.25–0.5 μg/mL to <0.008 μg/mL, the lowest concentration we tested. In contrast, we observe only a modest two-fold decrease in the MIC of vancomycin.¹⁵ We observe a similar but smaller increase in the activity of teixobactin analogue 1a with 0.002% polysorbate 80. Thus, azateixobactin analogue 2a proved 4–8 times more active than teixobactin analogue 1a in the presence of 0.002% polysorbate 80. Although the enhanced antibiotic activity of azateixobactin analogue 2a is consistent with a model in which the additional NH group better binds the anionic pyrophosphate group of lipid II and related cell-wall precursors, other mechanisms of enhanced antibiotic activity cannot be precluded.

Table 2.

MIC values of teixobactin and teixobactin analogues (μg/mL) with 0.002% polysorbate 80.

	Staphylococcus aureus ATCC 29213	Staphylococcus epidermidis ATCC 14990	Bacillus subtilis ATCC 6051	Escherichia coli ATCC 10798
Arg10-teixobactin (1a)	0.06	0.13	0.06	>8
d-aza-Thr8,Arg10-teixobactin (2a)	0.008	0.03	0.016	>8
teixobactin	<0.008	<0.008	<0.008	>8
vancomycin	0.125	0.25	0.25	>8

The authors of the original teixobactin report suggest that polysorbate 80 prevents the binding of the antibiotic to plastic surfaces.^1,16 We believe that other effects may also be involved. Having observed that teixobactin and its analogues form gels and amyloid-like fibrils upon addition to buffer or culture media,^4,17 we believe that inclusion of polysorbate 80 in the broth may help solubilize the gels or fibrils, and thus increases the bioavailability and activity of teixobactin and its analogues.

X-ray Crystallographic Structure of N-Me-d-Gln₄, d-aza-Thr₈,Arg₁₀-teixobactin (3a).

To assess the effect of the additional amide NH group on the structure of azateixobactin analogues, we turned to X-ray crystallography. We had previously found that N-methylation of the peptide backbone of d-Gln₄ facilitated crystallization of a teixobactin analogue by attenuating its propensity to form a gel.⁴ In the current study we utilized this finding and synthesized N-Me-d-Gln₄, d-aza-Thr₈,Arg₁₀-teixobactin (3a).¹⁸

We screened N-Me-d-Gln₄, d-aza-Thr₈,Arg₁₀-teixobactin (3a) for crystallization in 864 conditions in a 96-well plate format using crystallization kits from Hampton Research (PEG/Ion, Index, and Crystal Screen). Hexagonal prism -shaped crystals grew in conditions containing polyethylene glycol (PEG) and chloride salts. With further optimization in 24-well plates, a crystallization buffer of 0.16 M CaCl₂, 0.1 M HEPES sodium at pH 7.00, and 24% PEG 400 afforded crystals of azateixobactin analogue 3a suitable for X-ray diffraction. Three X-ray diffraction data sets were acquired at the Advanced Light Source (ALS) synchrotron at a wavelength of 1.77 Å (7000 eV). The data sets were processed using XDS¹⁹ and merged using BLEND.²⁰ The structure was solved by single-wavelength anomalous diffraction (SAD) phasing using the anomalous signal from the chloride anion. The structure was refined with PHENIX²¹ in the P3₂21 space group at 2.10 Å resolution. The asymmetric unit contains one N-Me-d-Gln₄, d-aza-Thr₈,Arg₁₀-teixobactin (3a) molecule, as well as one chloride anion and four ordered water molecules (Figure 4, PDB 6PSL). We refined the N-methyl terminus (N-Me-d-Phe₁) as the free amine, rather than as the ammonium ion, because there is only a single chloride anion in the asymmetric unit, which balances the positive charge of the arginine sidechain. We found no electron density or voids in the lattice that could account for an additional anion.

Figure 4.

X-ray crystallographic structure of N-Me-d-Gln₄, d-aza-Thr₈,Arg₁₀-teixobactin (3a) binding chloride anions (PDB 6PSL). (A and B) Monomer side and top views. (C and D) Dimer side and top views with two bridging water molecules shown as non-bonded spheres. (E) Alignment of the dimer assembly.

The X-ray crystallographic structure reveals an amphipathic hydrogen-bonded antiparallel β-sheet dimer that binds chloride anions (Figure 4). Residues 1–7 form the dimerization interface and create an amphipathic antiparallel β-sheet containing both D and L residues. The macrolactam amide NH groups of residues 8, 10, and 11, as well as the extra amide NH group of the lactam ring, hydrogen bond to the chloride anion. The amide NH group of Ala₉ hydrogen bonds to the hydroxy group of Ser₇.²²

In the antiparallel β-sheet assembly, the sidechains of hydrophobic residues N-Me-d-Phe₁, Ile₂, d-allo-Ile₅, Ile₆, and Ile₁₁, as well as the β-methyl group of d-aza-Thr₈, create a hydrophobic face, and the sidechains of hydrophilic residues Ser₃, N-Me-d-Gln₄, Ser₇, and Arg₁₀, as well as the N-methylamine terminus, create a hydrophilic face. The N-methyl group of N-Me-d-Gln₄ points outward from the dimer, blocking further assembly. The antiparallel β-sheet dimer brings the N-methylamine terminus of each monomer subunit into proximity with the macrolactam ring of the other monomer subunit and allows each methylamine group to hydrogen bond to a chloride anion.

The X-ray crystallographic structure of azateixobactin analogue 3a reveals a mode of antiparallel β-sheet assembly distinct from that which we had previously reported a related teixobactin analogue (N-Me-d-Phe^I₁,N-Me-d-Gln₄,Arg₁₀-teixobactin, PDB 6E00).⁴ In our previous X-ray crystallographic antiparallel β-sheet dimer, residues 1–7 also create an amphipathic dimerization interface (Figure S32). The antiparallel β-sheet dimer assembly also brings each N-terminus into the proximity with the anion binding site and allows each N-terminus to hydrogen bond to the bound anion. However, there are three significant differences between the current antiparallel β-sheet dimer (3a) and our previously reported antiparallel β-sheet dimer: (1) The dimerization interfaces involve opposite edges of the β-sheets. In the current structure, residues 3 and 5 form hydrogen-bonded pairs; in our previously reported structure, residues 2 and 6 form hydrogen-bonded pairs (Figures 4E and S32). (2) The psi and phi angles of Ser₇ differ dramatically between the two structures, thus rotating the macrocycle toward the dimerization interface of each structure. (3) The dimer in the current structure binds two chloride anions, while the dimer in our previously reported structure binds two sulfate anions. Although the details of the crystallographic structures of the previous and current dimers differ, both structures are consistent with the model shown in Figure 1A.

Ring-Expanded Teixobactin Analogues.

The 13-membered macrolactone ring of teixobactin is substantially smaller than that of many other cyclic depsipeptide antibiotics that bind lipid II,²³ including ramoplanin²⁴ (49-membered macrolactone), lysobactin/katanosin B²⁵ (28-membered macrolactone), and plusbacin A3²⁶ (28-membered macrolactone). Our X-ray crystallographic structures of teixobactin analogues bound to chloride and sulfate anions suggest that a larger ring might allow teixobactin analogues to better accommodate the larger pyrophosphate group of lipid II and related cell-wall precursors. Inspired by our X-ray crystallographic structures and the larger macrolactone rings of other antibiotics, we set out to explore the effect of the teixobactin macrolactone ring size upon antibiotic activity. In this section, we expand the 13-membered macrolactone ring to 14-, 15-, and 16-membered macrolactone rings with β³-homo amino acids.

We synthesized seven ring-expanded teixobactin analogues containing one, two, or three β³-homo amino acids at positions 9, 10, and 11 (Figure 5) and assessed their activities in MIC assays (Table 3 and Table S1). Six of the seven ring-expanded teixobactin analogues were active against Gram-positive bacteria, indicating that the teixobactin pharmacophore tolerates ring expansion. β³h-Arg₁₀-teixobactin (5) and β³h-Arg₁₀,β³h-Ile₁₁-teixobactin (9) exhibit comparable activity to Arg₁₀-teixobactin, while the other ring-expanded analogues exhibit diminished activity. Molecular modeling studies suggest that the ring expanded analogues are more flexible than teixobactin or Arg₁₀-teixobactin and that the NH groups of the rings are less well aligned to bind anions (Figures S33–S34). Collectively, the MIC and molecular modeling studies suggest that the 13-membered ring of teixobactin may provide an optimal balance of size and preorganization for binding the pyrophosphate group of lipid II and related cell-wall precursors.

Figure 5.

Arg₁₀-teixobactin and ring-expanded Arg₁₀-teixobactin analogues containing β³-homo amino acids at positions 9, 10, and 11. The ring-expanded analogues contain 14-, 15-, and 16-membered macrolactone rings.

Table 3.

MIC values of ring-expanded teixobactin analogues (μg/mL).

	Staphylococcus aureus ATCC 29213	Staphylococcus epidermidis ATCC 14990	Bacillus subtilis ATCC 6051	Escherichia coli ATCC 10798
Arg10-teixobactin (1a)	2	1	2	>32
β3h-Ala9,Arg10-teixobactin (4)	32	4	8	>32
β3h-Arg10-teixobactin (5)	8	1	2	>32
Arg10,β3h-Ile11-teixobactin (6)	4	2	4	>32
β3h-Ala9, β3h-Arg 10-teixobactin (7)	>32	>32	>32	>32
β3h-Ala9,Arg10,β3h-Ile11-teixobactin (8)	32	16	16	>32
β3h-Arg10,β3h-Ile11-teixobactin (9)	4	1	4	>32
β3h-Ala9,β3h-Arg10,β3h-Ile11-teixobactin (10)	16	2	4	>32
teixobactin	0.25	0.25	0.25–0.5	>32
vancomycin	0.25	0.25	0.5	>32

CONCLUSION

Lactam analogues of teixobactin containing aza-threonine at position 8 are readily prepared by solid-phase synthesis, with conversion of the corresponding diastereomeric threonine analogue to the aza-threonine analogue by mesylation, azide SN2 displacement, and Staudinger reduction with triphenylphosphine-2-carboxamide. Teixobactin analogues containing all four diastereomers of aza-threonine can be prepared by this route. Replacement of the lactone ring oxygen with an NH group increases antibiotic activity, with d-aza-Thr₈,Arg₁₀-teixobactin (2a) exhibiting 2–8-fold greater antibiotic activity than Arg₁₀-teixobactin (1a). Polysorbate 80 exhibits a dramatic effect on the antibiotic activity of teixobactin and teixobactin analogues; d-aza-Thr₈,Arg₁₀-teixobactin exhibits an MIC of 0.008 μg/mL against S. aureus in the presence of 0.002% polysorbate 80.

X-ray crystallography reveals that the additional NH group of azateixobactin analogue 3a hydrogen bonds to a bound chloride anion and supports a model in which azateixobactin analogues achieve enhanced antibiotic activity by better binding to the anionic pyrophosphate group of lipid II and related cell-wall precursors. The X-ray crystallographic structure further supports a model in which the formation of hydrogen-bonded dimers or amyloid-like higher order assemblies is central to antibiotic activity,⁴ with two molecules of teixobactin or a teixobactin analogue coordinating to the bound anion. Although ring-expanded teixobactin analogues were hypothesized to better accommodate the pyrophosphate group of lipid II and related cell-wall precursors, teixobactin analogues containing β³-homo amino acids exhibited no greater antibiotic activity. Collectively, these studies illustrate how chemical synthesis, X-ray crystallography, and antibiotic activity assays may be used together to provide insights into the unique activity of teixobactin. We anticipate that these studies will facilitate the design of teixobactin analogues with improved properties that are useful in the clinic.

EXPERIMENTAL SECTION

General information

Methylene chloride (CH₂Cl₂) was passed through alumina under argon prior to use. Amine-free N,N-dimethylformamide (DMF) was purchased from Alfa Aesar. Fmoc-d-allo-Ile-OH was purchased from Santa Cruz Biotechnology. Fmoc-N-Me-d-Gln(Trt)-OH was purchased from ChemPep. Other protected amino acids were purchased from CHEM-IMPEX. 2-(Diphenylphosphino)benzoic acid was purchased from Arctom chemicals. Preparative reverse-phase HPLC was performed on a Rainin Dynamax instrument equipped with an Agilent Zorbax SB-C18 column. Analytical reverse-phase HPLC was performed on an Agilent 1260 Infinity II instrument equipped with a Phenomonex Aeris PEPTIDE 2.6μ XB-C18 column. HPLC grade acetonitrile (MeCN) and deionized water (18 MΩ) containing 0.1% trifluoroacetic acid (TFA) were used as solvents for both preparative and analytical reverse-phase HPLC. Deionized water (18 MΩ) was obtained from a Barnstead NANOpure Diamond purification system or a ThermoScientific Barnstead GenPure Pro water purification system. Glass solid-phase peptide synthesis vessels with fritted disks and BioRad Polyprep columns were used for solid-phase peptide synthesis. Teixobactin analogues 2a–10 were prepared and studied as the trifluoroacetate salts.

Synthesis of d-aza-Thr₈,Arg₁₀-teixobactin (2a)

Resin Loading.

2-Chlorotrityl chloride resin (300 mg, 1.2 mmol/g) was added to a 10-mL Bio-Rad Poly-Prep chromatography column. The resin was suspended in dry CH₂Cl₂ (5 mL) and allowed to swell for 15 min. The CH₂Cl₂ was drained with a flow of nitrogen. The resin was loaded with a solution of Fmoc-Ala-OH (90 mg, 0.29 mmol) and 2,4,6-collidine (300 μL) in dry CH₂Cl₂ (7 mL) and rocked for 4 h.

Resin Capping.

The solution was drained with a flow of nitrogen, and the resin was washed with dry CH₂Cl₂ (3x). A mixture of CH₂Cl₂ (5 mL), MeOH (0.8 mL), and DIPEA (0.4 mL) was made and poured into the Poly-Prep column containing the resin, and the column was rocked for 1 h to cap any unreacted sites in the resin. The solution was drained with a flow of nitrogen, and the resin was washed with dry CH₂Cl₂ (3x). The resin was washed with MeOH and blown with nitrogen until dry resin was observed.

Loading check.

Approximately 1 mg of the dry resin was weighed out into a scintillation vial and 20% piperidine in DMF (3 mL) was added. The vial was rocked for 30 min. The mixture was filtered using a Pasteur pipet plugged with glass wool. The UV/Vis spectrometer was blanked at 290 nm with a cuvette filled with 20% piperidine in DMF. The absorbance of the filtered mixture was measured [1.4 mg of resin weighed; A₂₉₀ = 1.1234; 0.15 mmol loading].

Fmoc deprotection.

The loaded resin was transferred to a solid-phase peptide hand coupling vessel. The resin was washed with dry CH₂Cl (3x) and then dry DMF (3x). To the reaction vessel, 20% piperidine in dry DMF (5 mL) was added. Using a nitrogen flow to bubble the hand coupling vessel, the reaction was mixed for 20 min. The resin was washed with dry DMF (3x).

Coupling Fmoc-d-allo-Thr-OH with HCTU.

Based on loading, Fmoc-d-allo-Thr-OH (105 mg, 0.30 mmol, 2 equiv) and HCTU (121 mg, 0.30 mmol, 2 equiv) were weighed out and dissolved in 20% collidine in dry DMF. This solution was added to the reaction vessel containing the deprotected peptide on resin. Using a nitrogen flow to bubble the hand coupling vessel, the reaction was mixed for 4 h. The resin was washed with dry DMF (3x).

Fmoc deprotection.

To the reaction vessel, 20% piperidine in dry DMF (5 mL) was added. Using a nitrogen flow to bubble the hand coupling vessel, the reaction was mixed for 20 min. The resin was washed with dry DMF (3x).

Alloc protection.

The resin was transferred to a Poly-Prep column with dry DMF and the solution was drained with a flow of nitrogen. The resin was washed with dry CH₂Cl₂ (3x). To the resin in the Poly-Prep column, dry CH₂Cl₂ (5 mL), DIPEA (38 μL, 0.23 mmol, 1.5 equiv) and allyl chloroformate (23 μL, 0.23 mmol, 1.5 equiv) were added sequentially. The column was then capped and rocked for 1 h. The resin was washed with dry CH₂Cl (3x).

Mesylation.

Dry CH₂Cl₂ (6 mL) was added to the resin in the Poly-Prep column. The column was then capped and rocked in a cold room (4 °C) for 15 min. DIPEA (254 μL, 1.5 mmol, 10 equiv) was added, and the mixture was rocked in a cold room (4) for an additional 15 min. Methanesulfonyl chloride (113 μL, 1.5 mmol, 10 equiv) was added, and mixture was rocked in a cold room (4 °C) for an additional 15 min. The resin was then washed with dry CH₂Cl₂ (3x) and then with dry DMF (3x). The resin was transferred to the hand coupling vessel.

S_N2 with NaN₃.

NaN₃ (474 mg, 7.5 mmol, 50 equiv) was added to the resin in the hand coupling vessel. Dry DMF (1 mL) and 15-crown-5 (1 mL) were added to the hand coupling vessel. [Not all of the NaN₃ dissolves in the solvent mixture.] Plastic tubing with a continuous water flow at 55 °C was wrapped around the hand coupling vessel to provide heating. Using a gentle nitrogen flow, the mixture was bubbled for 12 h at 55 °C. The resin was washed with 10 mL of 20% H₂O in THF (5x) to remove excess NaN₃. The resin was transferred to a Poly-Prep column with dry DMF and then washed with dry CH₂Cl₂ (3x).

Alloc deprotection.

A mixture of CH₂Cl₂ (5 mL), (Ph₃P)₄Pd (16.9 mg, 0.015 mmol, 0.1 equiv) and PhSiH₃ (360 μL, 3 mmol, 20 equiv) was added to the resin in the Poly-Prep column, and the column was rocked for 30 min. The resin was washed with dry CH₂Cl₂ (3x) and then with dry DMF (3x) and transferred to a hand coupling vessel.

Peptide coupling.

The linear peptide was synthesized through the following cycles: i. coupling of amino acid (0.60 mmol, 4 equiv) with HCTU (241 mg, 0.60 mmol, 4 equiv) in 20% (v/v) 2,4,6-collidine in dry DMF (3 mL) for 30 min, ii. resin washing with dry DMF (3x), iii. Fmoc deprotection with 20% (v/v) piperidine in dry DMF (3 mL) for 20 min, and iv. resin washing with dry DMF (3x). For d-to-l and l-to-d amino acid couplings, the reaction time in step i was increased to 1 h. After completing the linear synthesis, the resin was transferred to a 10-mL Bio-Rad Poly-Prep chromatography column. The resin was then washed with dry DMF (3x) and then dry THF (3x).

Azide reduction.

Triphenylphosphine-2-carboxamide²⁷ (223 mg, 0.45 mmol, 5 equiv) in THF (5 mL) was added to the resin in a Poly-Prep column, and the column was rocked for 12 h. The solution was drained with a flow of nitrogen and 20% H₂O in THF (5 mL) was added and rocked for 4 h. The resin was washed with dry DMF (3x) and transferred to a hand coupling vessel using DMF.

Peptide coupling.

The coupling and Fmoc deprotection of Ile₁₁ and Arg₁₀ was performed as described above. After Fmoc deprotection of Arg₁₀, the resin containing branched linear peptide was transferred to 10-mL Bio-Rad Poly-Prep chromatography column using DMF. The resin was washed with dry DMF (3x) and then with dry CH₂Cl₂ (3x).

Cleavage of the branched linear peptide from the resin.

The protected peptide was cleaved from the resin by adding 20% hexafluoroisopropanol in dry CH₂Cl₂ (7 mL) to the column and rocking for 30 min. The filtrate was collected in a round-bottom flask. The resin was washed with a second aliquot of 20% hexafluoroisopropanol (7 mL). The filtrates were combined and concentrated under reduced pressure to afford a clear oil. The oil was placed under vacuum (≤ 100 mTorr) to remove any residual solvents.

Macrolactamization.

To the round-bottom flask containing cleaved peptide, a mixture of HBTU (332 mg, 0.9 mmol, 6 equiv) and HOBT (118 mg, 0.9 mmol, 6 equiv) in dry DMF (50 mL) was added and stirred for 15 min under nitrogen. Diisopropylethylamine (153 μL, 0.9 mmol, 6 equiv) was added to the stirring solution and then stirred for 12 h under nitrogen. The solution was evaporated by rotary evaporator and the residue was dried under vacuum (≤ 100 mTorr) to give a pale-yellow pellet.

Global deprotection.

A solution of TFA (9 mL), H₂O (0.5 mL), and TIPS (0.5 mL) was added to the round-bottom flask containing cyclized peptide and stirred for 1 h under nitrogen. The solution was evaporated by rotary evaporator and the residue was dried under vacuum (≤ 100 mTorr).

Purification.

The globally deprotected peptide was dissolved in approximately 35% CH₃CN in H₂O (10 mL), and the solution was centrifuged at 14,000 rpm for 5 min. The solution was filtered through a 0.20-μm nylon filter. The peptide was purified by reverse-phase HPLC with H₂O/CH₃CN (gradient elution of 20–95% CH₃CN with 0.1% TFA). Pure fractions were identified by analytical HPLC and electrospray ionization (ESI) mass spectrometry and were combined and lyophilized. d-aza-Thr₈,Arg₁₀-teixobactin (2a) was isolated as trifluoroacetic acid (TFA) salt of a 13.5 mg white powder with ≥95% purity.

Synthesis of diastereomeric azateixobactin analogues (2b–2d)

Other diastereomeric azateixobactin analogues (2b–2d) were prepared using the procedures described above. Replacing d-allo-threonine with d-threonine in the synthesis afforded d-allo-aza-threonine and gave diastereomer 2b. Replacing d-allo-threonine with l-allo-threonine in the synthesis afforded l-aza-threonine and gave diastereomer 2c. Replacing d-allo-threonine with l-threonine in the synthesis afforded l-allo-aza-threonine and gave diastereomer 2d. For synthesis of l-allo-aza-Thr₈,Arg₁₀-teixobactin (2d), the azide reduction with triphenylphosphine-2-carboxamide was performed for 24 h instead of 12 h. The stereoisomers containing l-aza-threonine and l-allo-aza-threonine (2c and 2d) exhibited an exceptionally strong propensity to form gels, making purification difficult and affording only low yields of these diastereomeric teixobactin analogues after purification by RP-HPLC.

Synthesis of N-Me-d-Gln₄, d-aza-Thr₈,Arg₁₀-teixobactin (3a)

N-Methylated azateixobactin analogue (3a) was prepared using the procedures described above. Coupling of Fmoc-Ser₃(tBu)-OH after N-Me-d-Gln₄ was performed using 4 equiv Fmoc-Ser₃(tBu)-OH with coupling reagent HATU (4 equiv), HOAt (4 equiv) in 20% (v/v) collidine in dry DMF (5 mL) for 12 h.

Synthesis of ring-expanded teixobactin analogues (4–10)

Ring-expanded teixobactin analogues containing β³-homo amino acids (4–10) were synthesized using the procedures we have previously reported.⁷ Dry DMF was used instead of a mixture of CH₃CN/THF/CH₂Cl₂ for the cyclization step.

HPLC conditions and MS results

Analytical RP-HPLC was performed on a C18 column with an elution gradient of 5–67% CH₃CN over 15 min for teixobactin analogues 5, 8, 9, and 10, and with an elution gradient of 5–100% CH₃CN over 20 min for other analogues.

d-aza-Thr₈,Arg₁₀-teixobactin (2a). MS (ESI-MS): m/z: [M+H]⁺ Calcd for C₅₈H₉₉N₁₆O₁₄ 1243.8; found 1243.3. d-allo-aza-Thr₈,Arg₁₀-teixobactin (2b). MS (ESI-MS): m/z: [M+H]⁺ Calcd for C₅₈H₉₉N₁₆O₁₄ 1243.8; found 1243.4. l-aza-Thr₈,Arg₁₀-teixobactin (2c). MS (ESI-MS): m/z: [M+H]⁺ Calcd for C₅₈H₉₉N₁₆O₁₄ 1243.8; found 1243.9. l-allo-aza-Thr₈,Arg₁₀-teixobactin (2d). MS (MALDI-MS): m/z: [M+H]⁺ Calcd for C₅₈H₉₉N₁₆O₁₄ 1243.8; found 1243.6. N-Me-d-Gln₄, d-aza-Thr₈,Arg₁₀-teixobactin (3a). MS (ESI-MS): m/z: [M+H]⁺ Calcd for C₅₉H₁₀₁N₁₆O₁₄ 1257.8; found 1257.8. β³h-Ala₉,Arg₁₀-teixobactin (4). MS (ESI-MS): m/z: [M+H]⁺ Calcd for C₅₉H₁₀₀N₁₅O₁₅ 1258.8; found 1258.6. β³h-Arg₁₀-teixobactin (5). MS (ESI-MS): m/z: [M+H]⁺ Calcd for C₅₉H₁₀₀N₁₅O₁₅ 1258.8; found 1258.5. Arg₁₀,β³h-Ile₁₁-teixobactin (6). MS (ESI-MS): m/z: [M+H]⁺ Calcd for C₅₉H₁₀₀N₁₅O₁₅ 1258.8; found 1258.6. β³h-Ala₉,β³h-Arg₁₀-teixobactin (7). MS (ESI-MS): m/z: [M+H]⁺ Calcd for C₆₀H₁₀₂N₁₅O₁₅ 1272.8; found 1272.6. β³h-Ala₉,Arg₁₀,β³h-Ile₁₁-teixobactin (8). MS (ESI-MS): m/z: [M+H]⁺ Calcd for C₆₀H₁₀₂N₁₅O₁₅ 1272.8; found 1272.5. β³h-Arg₁₀,β³h-Ile¹¹-teixobactin (9). MS (ESI-MS): m/z: [M+H]⁺ Calcd for C₆₀H₁₀₂N₁₅O₁₅ 1272.8; found 1272.6. β³h-Ala₉,β³h-Arg₁₀,β³h-Ile₁₁-teixobactin (10). MS (ESI-MS): m/z: [M+H]⁺ Calcd for C₆₁H₁₀₄N₁₅O₁₅ 1286.8; found 1286.6.

Minimum inhibitory concentration (MIC) assay of teixobactin analogue

MIC assays of teixobactin and teixobactin analogues (2a–10) were performed using the procedure we have previously reported.^7,17

Preparing the peptide analogue stock.

Solutions of d-aza-Thr₈,Arg₁₀-teixobactin (2a), other teixobactin analogues (2b–10), teixobactin and vancomycin were prepared gravimetrically by dissolving an appropriate amount of peptide in an appropriate volume of sterile DMSO to make 20 mg/mL stock solutions. The stock solutions were stored at −20 °C for subsequent experiments.

Prior to transferring the peptide solution in broth to 96-well plates, the peptide solution was vigorously vortexed with a vortex mixer for ca. 30 seconds to ensure thorough dispersion and/or dissolution of teixobactin and teixobactin analogues, which have a propensity to form gels.

Serial dilution of peptides in a 96-well plate.

In the serial dilution, the first well containing 200-uL peptide solution was mixed with pipette up and down at least ten times before moving to the next well. In the following serial dilutions, each well was also mixed with pipette up and down at least ten times.

Performing the minimum inhibitory concentration (MIC) assays with Mueller-Hinton broth containing 0.002% polysorbate 80.

Mueller-Hinton broth containing 0.002% (v/v) polysorbate 80 was autoclaved and used in the MIC assay. The broth was used to dilute the 20 mg/mL DMSO solution of teixobactin and teixobactin analogues to 16 μg/mL and then to serially dilute the peptide solutions in the 96-well plate to the following concentrations: 16, 8, 4, 2, 1, 0.5, 0.25, 0.13, 0.063, 0.031, and 0.016 μg/mL. Each well contained 100 μL of solution. The broth was used to dilute bacterial suspension OD₆₀₀ of 0.075 and to further dilute the bacterial suspension to 1 × 10⁶ CFU/mL. 100-μL aliquots of the diluted bacteria were added to each well to give final concentration of 5 × 10⁵ CFU/mL and final peptide concentrations of 8, 4, 2, 1, 0.5, 0.25, 0.13, 0.062, 0.031, 0.016, and 0.008 μg/mL.

Crystallization of N-Me-d-Gln₄, d-aza-Thr₈,Arg₁₀-teixobactin (3a)²⁸

N-Me-d-Gln₄,d-aza-Thr₈,Arg₁₀-teixobactin (3a) was dissolved in 0.2 micron syringe filtered NANOpure H₂O (10 mg/mL). Crystallization conditions were screened by screening in a 96-well plate format using three crystallization kits from Hampton Research (PEG/Ion, Index, and Crystal Screen). Each well was loaded with 100 μL of a different mother liquor solution from the kits. The hanging drops were set up using a TTP Labtech Mosquito^® liquid handling instrument. Hanging drops were made by combining an appropriate volume of N-Me-d-Gln₄,d-aza-Thr₈,Arg₁₀-teixobactin (3a) with an appropriate volume of well solution to create three 150-nL hanging drops with 1:1, 1:2, and 2:1 peptide:well solution. Hexagonal prism-shaped crystals grew in all conditions that contained polyethylene glycol (PEG) and chloride salts.

Crystal growth was optimized using conditions containing HEPES Na, PEG 400 and CaCl₂. In the optimization, the HEPES Na (pH 5.5–8.0), CaCl₂, and PEG 400 concentrations were varied across the 4×6 matrix of a Hampton VDX 24-well plate to afford crystals suitable for X-ray diffraction. The hanging drops for these optimizations were prepared on glass slides by combining 1 or 2 μL of teixobactin solution with 1 or 2 μL of well solution in ratios of 1:1, 2:1, and 1:2. Crystals that formed were checked for diffraction using a Rigaku Micromax-007 HF diffractometer with a Cu anode at 1.54 Å. As a result of the optimization, 0.16 M CaCl₂, 0.1 M HEPES Na pH 7.00, and 24% PEG 400 afforded crystals suitable for X-ray diffraction.

X-ray crystallographic data collection, data processing, and structure determination

Data collection was performed with the BOS/B3 software at Advanced Light Source (ALS) using beamline 8.2.2 at a wavelength of 1.771190 Å (7000 eV). The rotation method was employed and three sets of 360 images each were collected at a 1.0° rotation interval (a total of three complete 360° rotations). The three sets were processed with XDS,²⁹ and the resulting datasets were merged with BLEND.³⁰ The structure was solved with SAD phasing implemented in the Hybrid Substructure Search (HySS)³¹ module of the Phenix suite³² using the anomalous signal from the chloride ion. The initial electron density maps were generated using the substructure coordinates as initial positions in Autosol.³³ The structure was then refined with Phenix.refine³⁴ under Phenix using Coot³⁵ for model building. All B-factors were refined isotropically and riding hydrogen atoms coordinates were generated geometrically. The bond length, angles, and torsions restraints for unnatural amino acids (N-Me-d-Gln, d-aza-Thr, and d-allo-Ile) were generated with eLBOW³⁶ under Phenix. We refined the N-methyl terminus (N-Me-d-Phe₁) as the free amine, rather than as the ammonium ion, because there is only a single chloride anion in the asymmetric unit, which balances the positive charge of the arginine sidechain. We found no electron density or voids in the lattice that could account for an additional anion.