Structures of β-hairpin antimicrobial protegrin peptides in lipopolysaccharide membranes: mechanism of gram selectivity obtained from solid-state nuclear magnetic resonance

Abstract

The structural basis for the gram selectivity of two disulfide-bonded β-hairpin antimicrobial peptides (AMPs) is investigated using solid-state nuclear magnetic resonance (NMR) spectroscopy. The hexa-arginine PG-1 exhibits potent activities against both gram-positive and gram-negative bacteria, while a mutant of PG-1 with only three cationic residues maintains gram-positive activity but is 30-fold less active against gram-negative bacteria. We determined the topological structure and lipid interactions of these two peptides in a lipopolysaccharide (LPS)-rich membrane that mimics the outer membrane of gram-negative bacteria and in the POPE/POPG membrane, which mimics the membrane of gram-positive bacteria. (31)P NMR line shapes indicate that both peptides cause less orientational disorder in the LPS-rich membrane than in the POPE/POPG membrane. (13)C chemical shifts and (13)C-(1)H dipolar couplings show that both peptides maintain their β-hairpin conformation in these membranes and are largely immobilized, but the mutant exhibits noticeable intermediate-time scale motion in the LPS membrane at physiological temperature, suggesting shallow insertion. Indeed, (1)H spin diffusion from lipid chains to the peptides shows that PG-1 fully inserts into the LPS-rich membrane whereas the mutant does not. The (13)C-(31)P distances between the most hydrophobically embedded Arg of PG-1 and the lipid (31)P are significantly longer in the LPS membrane than in the POPE/POPG membrane, indicating that PG-1 does not cause toroidal pore defects in the LPS membrane, in contrast to its behavior in the POPE/POPG membrane. Taken together, these data indicate that PG-1 causes transmembrane pores of the barrel-stave type in the LPS membrane, thus allowing further translocation of the peptide into the inner membrane of gram-negative bacteria to kill the cells. In comparison, the less cationic mutant cannot fully cross the LPS membrane because of weaker electrostatic attractions, thus causing weaker antimicrobial activities. Therefore, strong electrostatic attraction between the peptide and the membrane surface, ensured by having a sufficient number of Arg residues, is essential for potent antimicrobial activities against gram-negative bacteria. The data provide a rational basis for controlling gram selectivity of AMPs by adjusting the charge densities.

Fig. 1

Schematics of the different membrane structure of bacteria. (a) Gram-negative bacteria have a lipopolysaccharide (LPS)-rich outer membrane, a thin peptidoglycan layer, and an inner phospholipid membrane. (b) Gram-positive bacteria have a thick peptidoglycan layer and a cytoplasmic membrane. The main phospholipids in both bacteria are phosphatidylethanolamine (PE) and phosphatidylglycerol (PG) (79). LPS is about 10 mol% of the phospholipid amount in E. coli cells (38). The long O-antigen chain of LPS is shortened for simplicity.

Fig. 2

Chemical structures of the peptides and lipids used in this study. (a) Amino acid sequences of disulfide-bonded PG-1 and IB-484. (b) Chemical structure of ReLPS. (c) Chemical structures of DEPE, POPE and POPG.

Fig. 3

³¹P static (left) and MAS (right) spectra of POPE/POPG membranes at 303 K. (a) Without peptide. (b) With 8 mol% IB484.

Fig. 4

Experimental (black) and simulated (red and blue) static ³¹P spectra of ReLPS-containing membranes without and with peptides. The spectra were measured at 303 K and 313 K. (a) DEPE spectra. (b) ReLPS/DEPE spectra. (c) Spectra of ReLPS/DEPE membranes containing 8 mol% PG-1. (d) Spectra of ReLPS/DEPE membranes containing 8 mol% IB484. Best-fit simulations yield the percentages of each component.

Fig. 5

¹³C chemical shifts of PG-1 and IB484 in ReLPS/DEPE membranes. (a) Representative 2D¹³C-¹³C DARR spectrum (40 ms mixing) of Arg3, Val13-labeled IB484 in POPE/POPG membranes at 263 K. The 2D spectrum was measured in 19 hrs with a maximum t₁ evolution time of 6 ms, 218 t₁ slices and 160 scans per slice. (b) 1D ¹³C CP-MAS spectra of ReLPS/DEPE membranes. Asterisks indicate spinning sidebands. (c) ¹³C CP-MAS spectrum (black) and ¹³C DQ filtered spectrum (red) of PG-1 in ReLPS/DEPE membranes. (d) ¹³C CP-MAS spectrum (black) and ¹³C DQ filtered spectrum (red) of IB484 in ReLPS/DEPE membranes. The 1D spectra were measured between 237 and 253 K under 4–5 kHz MAS. The 1D DQ filtered spectra were acquired with 2560 scans.

Fig. 6

C-H order parameters of IB484 and PG-1 in different membranes by 2D DIPSHIFT experiments. (a) IB484 in POPE/POPG membranes at 298 K under 4.5 kHz MAS. (b) IB484 in ReLPS/DEPE membranes at 308 K under 5 kHz MAS. (c) PG-1 in POPE/POPG membranes at 298 K under 4.5 kHz MAS. (d) PG-1 in ReLPS/DEPE membrane at 308 K under 5.0 kHz MAS. Solid and dashed lines are best-fit simulations without and with the T₂ decays, respectively.

Fig. 7

2D ¹³C-detected ¹H spin diffusion spectra of IB484 in the ReLPS/DEPE membrane to determine the peptide’s depth of insertion. (a) Representative 2D spectrum, measured with a 100 ms mixing time. (b) Sum of the Arg3 and Val13 Cα ¹H cross sections as a function of mixing time.

Fig. 8

Analysis of ¹H spin diffusion data of PG-1 and IB484 in two lipid membranes. (a) Buildup curves of IB484 in ReLPS/DEPE (filled circles) and POPE/POPG membranes (open squares), measured at 308 K and 288 K, respectively. The intensity is the sum of the Arg3 and Val13 Cα ¹H cross sections. Top: Spin diffusion from lipid CH₂ protons to the peptide. Bottom: Spin diffusion from water to the peptide. (b) Buildup curves of IB484 (circles) and PG-1 (triangles) in ReLPS/DEPE membranes at 308 K. For the IB484 data, the sum of the Arg3 and Val13 Cα intensity is shown as filled circles, while the Arg3 Cα intensity is shown as open circles.

Fig. 9

¹³C-³¹P REDOR data of PG-1 and IB484. (a) PG-1 Arg4 Cα and Cζ distances to ³¹P in the ReLPS/DEPE membrane (filled symbols). The POPE/POPG data (open symbols) are superimposed (27). (b) IB484 Arg3 Cα, Cζ and Val13 Cα data in the ReLPS/DEPE membrane (filled symbols) versus the POPE/POPG membrane (open symbols). (c) Representative REDOR S₀ and S spectra of IB484 in the POPE/POPG membrane. Mixing time: 10.7 ms. The REDOR experiments were carried out at 233–237 K under 4.0 or 4.5 kHz MAS.

Fig. 10

Structural models of PG-1 and IB484 in ReLPS/DEPE and POPE/POPG membranes. (a) In the POPE/POPG membrane, PG-1 is TM and causes toroidal pores with strong orientational disorder (26, 27). (b) In the ReLPS/DEPE membrane, PG-1 is TM and causes barrel-stave pores with little lipid orientational disorder. (c) In the POPE/POPG membrane, IB484 is TM and causes toroidal pores with moderate lipid disorder. (d) In the ReLPS/DEPE membrane, IB484 is partially inserted without causing lipid disorder. The peptide cannot subsequently cross into the cytoplasmic membrane, thus explaining its weak antimicrobial activity against Gram-negative bacteria.