Nature - USA (2020-08-20)

Nature | Vol 584 | 20 August 2020 | 481

mutant produced only a modest phenotype, highlighting the promi-
nent multipoint backbone-mediated coordination scheme observed
in PbgA (Fig. 3b–d). Thus, only mutations expected to disrupt lipid A
binding along the periplasmic leaflet profoundly affected the outer
membrane barrier, which suggests that the LPS–PbgA interface is an
essential mediator of outer membrane homeostasis in E. coli.

PbgA-derived peptides bind LPS and kill E. coli

We postulated that a peptide derived solely from the IFD sequence
might bind to LPS in vitro. A synthetic, linear peptide encompassing
the lipid A-binding (LAB) motif from PbgA bound to LPS selectively
(wild-type LAB (LABWT); dissociation constant (Kd) of approximately
75 μM) over all major E. coli phospholipids, whereas peptides expected
to destabilize key lipid A-binding determinants (LABΔα7 and LABT213D)
showed no binding (Fig. 3e, Extended Data Fig. 7). We predicted that
the H221W and D225R mutations might promote membrane parti-
tioning and this LABWT+ peptide (209-SYPMTARRFLEKWGLLR-22 5)
had improved affinity for LPS (Kd value of approximately 55 μM) while
maintaining selectivity over phospholipids (Fig. 3e).
Because PMX antibiotics kill Gram-negative bacteria by targeting
lipid A^4 ,^5 , we tested the LAB peptides for antibacterial activity. The LABWT

peptide had no effect on E. coli growth, potentially owing to its large

molecular mass of greater than 2 kDa. For the LABWT+ peptide, we meas-

ured minimal inhibitory concentrations (MICs) of 25–400 μM in chemi-

cally or genetically permeabilized cells (Supplementary Tables 5, 6).

LABΔα7 and LABT213D peptides, which were unable to bind LPS, showed

no effect on cell growth under matched conditions (Supplementary

Table 5). Alanine-scanning and truncation studies ultimately produced

a synthetic peptide (LABv2.0) with an MIC of 200 μM against intact,

wild-type E. coli K-12 (Supplementary Tables 5, 7, 8).

Optimized LAB achieves broad-spectrum activity

Starting from LABv2.0, structure-guided design suggested that T213Dap

((S)-2,3-diaminopropionic acid) should introduce a salt-bridge to

the 1-phospho-GlcNAc, and A214F mutation might improve mem-

brane partitioning and hydrophobic interactions with LPS (Fig. 3b, d).

The resulting LABv2.1 peptide had an MIC of 25 μM against E. coli

K-12 (Table  1 ). Inspection of our LPS–PbgA structure and associated

data led to three predictions about LABv2.1 peptide activity (Fig.  3 ,

Extended Data Fig. 4c). First, consistent with conservation of lipid A

across Gram-negative bacteria^1 , MICs of 12.5–200 μM were obtained

against the clinically relevant pathogens Enterobacter cloacae,

abc

PD

TMD

IFD

Cytoplasm

Innermembrane

LPS LPS

Periplasm

~60 Å

~60 Å

α 6

α 7

α 8

N-terminus

PbgA

IFD

EptA linker

PbgA

D268

R451

G450

T302

E240H453

D452

Zn T280

2+

E. coli

PbgA

Charge d

• 
  
  
  
  • 
    
  
  
  
  

LPS

**

IC^50

(μM)

101

10 –1

10 –3

WT

VectorD268AT302VR451A

Fig. 2 | PbgA structural features. a, PbgA crystal structure in cartoon and
electrostatic representation. TMD, IFD and periplasmic domain (PD) are in
blue, pink and green, respectively, with LPS as green sticks. b, TMD-based
alignment with the PbgA–IFD and EptA linker (PDB code 5FGN) in pink and cyan,
respectively. Note that the EptA periplasmic domain is oriented approximately

180° relative to PbgA. c, Pseudo-hydrolase active site of PbgA (green) and

catalytic site in EptA (cyan). d, Rifampicin sensitivity of UPEC ΔpbgA strains

with plasmids expressing wild-type PbgA or mutants. Data are mean ± s.d. from

n = 6 or more independent experiments per strain. **P < 0.001, Bonferroni

corrected unpaired two-tailed t-test.

ab e

Pseudo-active site

Cytoplasm

Inner

membrane

LPS

~25 Å

Conservation

Low High

IFD

b PbgAPbgA Top view

R216

A214 R215

P211 M212 T213

Y210

Bulk inner

membrame

_ 7

c

4 ′-PO 4 –

α 7

R216

F217

M212

d

4 ′-PO KDO

4 –

LPS

CL

PG

PE

1

(^01)
(^01)
(^01)
0
Response (nm)
LABWT LABWT+ LABT213D
0
(^150100)
(^5025)
10
Ligand (μM)
Kd(LPS)
LABWT 73 ± 13 μM
LABLABWTT213D+ (^) ND52 ± 11μM

---

IC^50
(μM)
101
10 –1
10 –3
10 –5
WTVectorM212AM212M212KET213VT213DR215AR216AR216DR216E
MTR-AVA
Fig. 3 | The periplasmic lipid A-binding motif of PbgA. a, Conservation
analysis calculated across 500 PbgA homologues, surface representation. LPS
(sticks) and approximate membrane boundaries are indicated. b, Close-up
view of the lipid A-binding motif with LPS (green stick representation), water
molecules (blue spheres) and most bonding interactions (yellow dashes),
shown. c, Rifampicin sensitivity of UPEC ΔpbgA strains with plasmids
expressing wild-type PbgA or mutants. Data are mean ± s.d. from n = 5 or more
independent experiments per strain. *P < 0.0041, **P < 0.001, Bonferroni
corrected unpaired two-tailed t-test. MTR-AVA, M212A/T213V/R216A triple
mutant. d, As in b, but a side view. e, Synthetic, biotinylated PbgA-derived lipid
A-binding (LAB) peptides transferred into different concentrations of
detergent solubilized lipids; binding assessed by interferometry
measurements. CL, cardiolipin; PE, phosphatidylethanolamine; PG,
phosphatidylglycerol. Data are mean and s.d. and representative of n = 3
experiments.