Article

A Gram-negative-selective antibiotic that

spares the gut microbiome

[Link] Kristen A. Muñoz1,2, Rebecca J. Ulrich1,2, Archit K. Vasan3,4,5, Matt Sinclair3,4, Po-Chao Wen3,4,5,
Jessica R. Holmes6, Hyang Yeon Lee1,2, Chien-Che Hung7,8, Christopher J. Fields6,
Received: 7 April 2023
Emad Tajkhorshid1,3,4,5, Gee W. Lau9 & Paul J. Hergenrother1,2,4 ✉
Accepted: 1 May 2024

Published online: xx xx xxxx

Infections caused by Gram-negative pathogens are increasingly prevalent and are
Check for updates typically treated with broad-spectrum antibiotics, resulting in disruption of the gut
microbiome and susceptibility to secondary infections1–3. There is a critical need for
antibiotics that are selective both for Gram-negative bacteria over Gram-positive
bacteria, as well as for pathogenic bacteria over commensal bacteria. Here we report
the design and discovery of lolamicin, a Gram-negative-specific antibiotic targeting
the lipoprotein transport system. Lolamicin has activity against a panel of more than
130 multidrug-resistant clinical isolates, shows efficacy in multiple mouse models of
acute pneumonia and septicaemia infection, and spares the gut microbiome in mice,
preventing secondary infection with Clostridioides difficile. The selective killing of
pathogenic Gram-negative bacteria by lolamicin is a consequence of low sequence
homology for the target in pathogenic bacteria versus commensals; this doubly
selective strategy can be a blueprint for the development of other microbiome-
sparing antibiotics.

Perturbation of the gut microbiome resulting from antibiotic treat- expected to induce significant gut dysbiosis. For example, colistin, a rare
ment can lead to increased vulnerability to colonization of oppor- Gram-negative-only antibiotic that is clinically approved, has antibac-
tunistic pathogens such as C. difficile4, as well as increased risk for terial activity against some human gut commensals1 and has shown to
gastrointestinal, renal and haematological abnormalities5,6. The vast cause C. difficile-associated diarrhea and pseudomembranous colitis12.
majority of clinically approved antibiotics kill only Gram-positive In addition to its effects on the microbiome, the neurotoxicity and
bacteria (Gram-positive-only antibiotics) or kill both Gram-positive nephrotoxicity associated with colistin complicate its use in the clinic,
and Gram-negative bacteria (broad-spectrum antibiotics). Notably, and thus colistin is typically utilized only as an antibiotic of last resort13.
surveys that assess large numbers of different antibiotics against a With the rise in difficult-to-treat Gram-negative infections and
variety of commensal bacteria show that essentially all of these drugs increased global emergence of colistin-resistance genes in Entero-
have a deleterious effect on gut commensals, and gut dysbiosis has bacterales species14, the need for novel antibacterial agents (especially
been linked to both Gram-positive-only and broad-spectrum antibi- for Gram-negative ESKAPE pathogens15) is clear. However, the intricate
otics7. For example, fluoroquinolone treatment results in rapid and cellular membrane structure and efflux pumps within Gram-negative
significant decreases in taxonomic richness and diversity8, and other bacteria confound the discovery of such compounds, and as a result
broad-spectrum antibiotic classes (cephalosporins and β-lactams) are no new class of antibiotics active against these microorganisms has
associated with profound losses of intestinal symbionts5,7. Even short been approved by the US Food and Drug Administration since the
exposure to the Gram-positive-only antibiotic clindamycin results broad-spectrum quinolones more than 50 years ago16. The discov-
in long-lasting changes to the gut microbiome, affecting taxonomic ery of antibiotics selective for Gram-negative bacteria is even more
composition and causing selection for resistance genes, detectable arduous; targets in the outer membrane could be exploitable, but
up to two years after dosing9,10. most compounds that engage these targets only have activity against
Currently, the effect of Gram-negative-only antibiotics on microbiome Gram-negative strains whose permeability has been compromised17–19.
perturbation remains unclear as there are vanishingly few of these com- Few compounds with selective activity against wild-type Gram-negative
pounds. Discovery of such compounds is challenging, as the majority bacteria have been identified, with the recent discoveries of the LpxC
of potential antibiotic targets are shared between Gram-positive and inhibitors, arylomycins, and BamA-targeting natural products being
Gram-negative bacteria. As the gut microbiome is composed of a signifi- notable exceptions20–23.
cant fraction of Gram-negative bacteria (up to 47% in some surveys11), Key to developing a Gram-negative-only antibiotic that does not
compounds that indiscriminately kill Gram-negative bacteria would be disturb the microbiome is to target an essential protein or complex

1
Department of Chemistry, University of Illinois at Urbana-Champaign, Urbana, IL, USA. 2Carl R. Woese Institute for Genomic Biology, University of Illinois at Urbana-Champaign, Urbana, IL, USA.
3
Theoretical and Computational Biophysics Group, NIH Center for Macromolecular Modeling and Visualization, Beckman Institute for Advanced Science and Technology, University of Illinois at
Urbana-Champaign, Urbana, IL, USA. 4Department of Biochemistry, University of Illinois at Urbana-Champaign, Urbana, IL, USA. 5Center for Biophysics and Quantitative Biology, University of
Illinois at Urbana-Champaign, Urbana, IL, USA. 6High-Performance Computing in Biology, Roy J. Carver Biotechnology Center, University of Illinois at Urbana-Champaign, Urbana, IL, USA.
7
Veterinary Diagnostic Laboratory, University of Illinois at Urbana-Champaign, Urbana, IL, USA. 8Department of Veterinary Clinical Medicine, University of Illinois at Urbana-Champaign, Urbana,
IL, USA. 9Department of Pathobiology, University of Illinois at Urbana-Champaign, Urbana, IL, USA. ✉e-mail: hergenro@[Link]

Nature | [Link] | 1
Article
O
R3
N H R2 CN
O N N N
HN O O O
HN HN
F N R1
N N N
N

1 2 Hybrid scaffold Lolamicin

29
Fig. 1 | Identification of lolamicin. Previously reported pyridinepyrazole (1) and pyridineimidazole (2) LolCDE inhibitors led to a hybrid scaffold and ultimately
lolamicin.

that is present exclusively in Gram-negative bacteria, but with homol- globularities and numbers of rotatable bonds, but lack a primary amine.
ogy divergence between pathogenic and commensal bacteria. Here Synthesis and accumulation assessment of 1 and 2 confirm that these
we identify lolamicin, a novel Gram-negative-only antibiotic that acts compounds accumulate poorly in E. coli (Extended Data Fig. 1b).
by disrupting the Lol lipoprotein transport system. In addition to its In an attempt to improve whole-cell accumulation and wild-type
Gram-negative spectrum of activity, lolamicin has high specificity for Gram-negative activity, we appended a primary amine to the par-
certain Gram-negative ESKAPE pathogens over commensal bacteria and ent scaffolds. Overlaying the structures of compounds 1 and 2 sug-
is active in mouse models of infection while sparing the gut microbiome gested a hybrid scaffold with the methylpyridine modification (Fig. 1).
and preventing C. difficile infection. Appending amines onto this scaffold revealed that several of these com-
A potential cellular target for the development of pathogen-specific, pounds indeed have increased whole-cell accumulation, but the amine
narrow-spectrum antibiotics is the Lol system, a five-component appeared deleterious to target engagement, as indicated by a signifi-
protein system located in the periplasm that is responsible for lipo- cant reduction in antibiotic activity in the permeability compromised
protein transport between the inner and outer membranes in Gram- E. coli strains (compounds 3–6; Supplementary Table 1). From these
negative bacteria24. As lipoproteins of Gram-positive bacteria are results, it was postulated that high total whole-cell accumulation may
entirely retained and anchored to the cytoplasmic membrane, lipo- not be required for antibacterial activity against a periplasmic target
protein transport is not required, making the Lol system exclusive to such as Lol, as the periplasm comprises only around 7% of the volume
Gram-negative bacteria24. This lipoprotein transport system is cru- of a Gram-negative cell36. Thus, we constructed and assessed additional
cial for the growth of Escherichia coli and is conserved across many compounds that, on the basis of their structure, would be unlikely to
high-priority Gram-negative ESKAPE pathogens. Notably, as detailed accumulate to detectable levels in the whole-cell accumulation assay
later, the Lol system of E. coli has low sequence homology with that of (compounds 7–11; Supplementary Table 1). This iterative approach
human commensal Gram-negative bacteria, suggesting the potential led to the identification of lolamicin (Fig. 1). Lolamicin has marked
for exploiting this protein complex as a means to develop doubly selec- activity in laboratory strains of Gram-negative pathogens including
tive antibiotics: a Gram-negative-selective target with further specific- E. coli, Klebsiella pneumoniae and Enterobacter cloacae, but not in the
ity for Gram-negative pathogens over commensals. With this in mind, wild-type or efflux-deficient strains of Pseudomonas aeruginosa or
we sought to identify a Lol inhibitor with antibiotic activity against Acinetobacter baumannii (Table 1). The whole-cell accumulation of
Gram-negative pathogens and then use this compound to evaluate lolamicin in E. coli was below that of the low-accumulating controls
the potential of pathogen-selective, Gram-negative-only antibiotics (Supplementary Table 1).
as microbiome-sparing drugs. Further assessment of lolamicin against a panel of Gram-positive
aerobic pathogens in addition to a range of Gram-positive and Gram-
negative anaerobic commensal bacteria revealed it to be inactive up
Identification of lolamicin to its aqueous solubility limit (128 µg ml−1; Table 1 and Supplementary
The Lol system comprises five distinct proteins. LolA is a periplasmic Table 2). By contrast, amoxicillin and clindamycin potently kill the
chaperone, LolB is an outer membrane receptor, and proteins C, D and majority of these gut commensal bacteria (Supplementary Table 3).
E form the LolCDE complex, a transporter that facilitates the trafficking This panel of obligate anaerobes includes Bacteroides species, the
of outer membrane-specific lipoproteins from the inner membrane. A most predominant Gram-negative organisms in the human gut37.
series of studies have identified novel Lol inhibitors for each of these This selectivity of lolamicin for E. coli, K. pneumoniae, and E. cloacae
subcomponents17,25–31. Noting the non-essentiality of LolA and LolB over other pathogenic and commensal bacteria (Gram-positive and
and essentiality of LolCDE determined from genomic screens32, we Gram-negative) is consistent with the sequence similarity of the LolCDE
prioritized the optimization of a LolCDE inhibitor. Whole-cell screens complex, with high similarity between E. coli, K. pneumoniae and E. cloa-
performed at AstraZeneca in 2015 identified pyridinepyrazoles29 and cae, and low similarity between this group and other species (Table 1
pyridineimidazoles29 (exemplified by compounds 1 and 2, respectively; and Supplementary Table 4). Lolamicin also exhibited minimal toxicity
Fig. 1) as inhibitors of the LolCDE complex. Although the lack of signifi- towards mammalian cell lines and red blood cells, and showed no shift
cant antimicrobial activity against wild-type Gram-negative pathogens in minimum inhibitory concentration (MIC) value when evaluated in
(Extended Data Fig. 1a), lack of in vivo efficacy, poor solubility and the presence of 50% human serum (Supplementary Tables 2 and 5).
high resistance frequencies prevented these inhibitors from being
advanced as drugs, resistant-mutant studies provided convincing
evidence that these compounds kill E. coli through inhibition of the Activity against clinical isolates
LolCDE complex29. Additionally, compounds 1 and 2 have significant Lolamicin was further evaluated against an extensive panel of
activity in efflux-deficient (E. coli ∆tolC) and permeability-deficient multidrug-resistant clinical isolates of E. coli, K. pneumoniae and
(E. coli ∆rfaC) strains (Extended Data Fig. 1a), suggesting that if com- E. cloacae (Fig. 2a–c). This collection of clinical isolates contains strains
pound entry and accumulation could be facilitated, improved activity with a variety of resistance profiles, including resistance to antibiotics
against pathogenic wild-type Gram-negative strains and clinical isolates such as carbapenems, aminoglycosides, sulfonamides, trimethoprim,
could be possible. In accordance with the previously disclosed eNTRy tetracyclines and colistin (Supplementary Table 6). In these suscepti-
rules for accumulation in Gram-negative bacteria33–35, compounds 1 bility studies in E. coli, lolamicin kills 50% of the strains at 1–2 µg ml−1,
and 2 meet two of the three desired parameters, possessing favourable 90% of the strains at 4 µg ml−1, and all strains at 8 µg ml−1 (Fig. 2a–c), and

2 | Nature | [Link]
Table 1 | Spectrum of lolamicin activity against frequency of 1.4 × 10−6 and lolamicin shows a resistance frequency of
Gram-negative pathogens and gut commensal bacteria 2.2 × 10−8 (Fig. 2d). We then generated resistant mutants to lolamicin in
E. coli BW25113, K. pneumoniae ATCC 27736 and E. cloacae ATCC 29893.
Bacterial strain MIC (µg ml−1) Sequence identity (%) At 8× MIC, lolamicin shows a resistance frequency of 3.4 × 10−7 in wild-
Gram-negative pathogens LolC LolD LolE type E. coli, 1.2 × 10−8 in wild-type K. pneumoniae and 5.2 × 10−7 in wild-
E. coli ∆tolC JW5503 0.008 100% 100% 100% type E. cloacae (Fig. 2e). We sequenced the lol locus from resistant
E. coli ∆rfaC JW3596 0.06 100% 100% 100% colonies, including 21 of the E. coli BW25113 colonies and 21 K. pneu-
moniae ATCC 27736 colonies; the results show 9 different mutations
E. coli BW25113 2 100% 100% 100%
in E. coli and 8 mutations in K. pneumoniae (Supplementary Table 10).
E. coli ATCC 25922 1 100% 100% 100%
Resistance mutations to lolamicin mapped back to individual amino
E. coli AR0349 1 100% 100% 100% acid changes in the LolC and LolE proteins, implicating the lipoprotein
E. coli AR-0549 2 100% 100% 100% transport pathway as the target of lolamicin. Changes in fitness were
Salmonella typhimurium ATCC 1 99% 95% 92% assessed by evaluating growth rates for lolamicin-resistant mutants
14028 compared with wild-type E. coli BW25113 in co-culture competition
K. pneumoniae ATCC 27736 1 90% 95% 87% assays. The predominant mutant LolC-E195K (which represented 7 out
K. pneumoniae ATCC BAA- 4 90% 95% 87% of the 21 mutations in E. coli) demonstrated a significant fitness cost
1705 compared with wild-type E. coli as shown by a 100-fold reduction in
K. pneumoniae AR0040 1 90% 95% 87% growth (Extended Data Fig. 2). LolE-F367S (which represented 3 out
E. cloacae ATCC 29893 4 95% 92% 89% of the 21 mutations) showed morphological changes compared to the
wild-type E. coli BW25113, with mutant colonies appearing smaller than
E. cloacae AR0163 8 95% 92% 89%
wild-type colonies; however there was no significant change in the num-
E. cloacae ATCC BAA-2468 4 95% 92% 89%
ber of colony-forming units (CFU). The remaining seven mutants did
Proteus mirabilis ATCC 35659 >128 61% 68% 63% not show a statistically significant fitness cost (Extended Data Fig. 2).
P. aeruginosa PAO1 >128 37% 57% 37% Finally, we generated time–kill growth curves with strains of E. coli,
P. aeruginosa PAO1 ∆6 >128 37% 57% 37% K. pneumoniae and E. cloacae, revealing that lolamicin shows bactericidal
A. baumannii ATCC 19606 >128 38% 53% 38% effects (as defined by CLSI guidelines) in E. coli (10,000× reduction
in CFUs), bacteriostatic effect against K. pneumoniae (100× reduc-
A. baumannii ATCC 19606 ∆3 32 38% 53% 38%
tion in CFUs), and a time-dependent bactericidal effect in E. cloacae
Stenotrophomonas >128 27% 58% 33%
(bactericidal at early time points with a 3,000× reduction in CFUs
maltophilia ATCC 13637
at 4 h) (Extended Data Fig. 3).
Gram-negative gut commensal bacteria
To further explore the mode of action of lolamicin, we performed
Morganella morganii ATCC >128 64% 70% 62% confocal microscopy to identify the phenotypic signature of lolamicin
25380
treatment compared with other inhibitors of outer membrane tar-
Bacteroides fragilis ATCC >128 24% 44% 23% gets38. Globomycin inhibits LspA, an enzyme that cleaves the signal
25285
peptide from apolipoprotein in the inner membrane before lipoprotein
Bacteroides vulgatus KLE 2303 >128 25% 44% 25%
is further processed for binding to LolCDE and trafficking to the outer
Bacteroides cellulosilyticus >128 24% 44% 24% membrane39. Globomycin induces a cell swelling phenotype that can
KLE 2342
be visualized using confocal microscopy and differs from the pheno-
Faecalibacterium prausnitzii >128 27% 47% 21% type induced by β-lactams, which inhibit peptidoglycan crosslinking29.
ATCC 27768
Lolamicin-treated E. coli BW25113 demonstrated a marked swelling in
Akkermansia muciniphila >128 25% 44% 19% bacterial size consistent with dysfunctional lipoprotein trafficking,
BAA-835
matching globomycin-treated cells (Extended Data Fig. 4a), but distinct
MIC assays were performed in Mueller Hinton, brain heart infusion or Todd Hewitt broth from phenotypic changes induced by β-lactams (Extended Data Fig. 4a).
per Clinical and Laboratory Standards Institute (CLSI) guidelines. MIC assays against gut
Image analysis and measurement of the cell area of DMSO-, lolamicin-
commensal bacteria were performed under anaerobic conditions per CLSI guidelines. All
experiments were performed in biological triplicate. Sequence identity compares the E. coli and globomycin-treated bacteria showed a threefold increase in bac-
(strain K12) LolCDE complex with the complex from other Gram-negative bacteria. Sequence terial size upon inhibition of lipoprotein trafficking (Extended Data
alignment was performed with NCBI Blast-Blosum62 with the parameters ‘GAVLI, FYW, CM, Fig. 4g). In lolamicin-resistant E. coli BW25113-LolC-N265K, in which
ST, KRH, DENQ, P’. Sequence identity was determined with the Sequence Manipulation a mutation in LolC renders lolamicin ineffective, no cell swelling was
Suite ([Link] Activity of lolamicin against
observed for lolamicin-treated cells, but cell swelling was retained in
Gram-positive bacteria is presented in Supplementary Table 2.
globomycin-treated cells (Extended Data Fig. 4b). Similar effects were
observed with the E. coli LolE-D264N mutant (Extended Data Fig. 4c),
and K. pneumoniae and its lolamicin-resistant mutants (Extended Data
markedly outperforms compounds 1 and 2, which have only minimal Fig. 4d–f; quantification in Extended Data Fig. 4g,h).
activity (Fig. 2a–c and Supplementary Tables 7–9). This narrow MIC
range highlights the potential of lolamicin; sequencing lolCDE of an
E. coli strain with an MIC value above the MIC90 (the MIC value at which Characterization of the lolamicin-binding site
growth was inhibited in 90% of isolates) revealed no mutations that With key resistance mutations identified, we used molecular model-
lead to an amino acid change. ling to probe the binding site for lolamicin and better understand the
basis of inhibition of LolCDE. Ensemble docking was performed as an
initial screen for putative lolamicin-binding poses and sites, using a
Mode of action of lolamicin cryo-electron microscopy (cryo-EM) structure of the lipoprotein-bound
We generated spontaneous mutants resistant to lolamicin via the pre-transport conformation of LolCDE40,41 (Protein Data Bank (PDB)
large inoculum method, initially in efflux pump-deficient E. coli ∆tolC ID: 7MDX) (Fig. 3 and Extended Data Fig. 5a,b), which has the most
( JW5503) so that lolamicin could be directly compared to a progeni- accessible lumen among the resolved LolCDE structures. Starting from
tor compound. At 8× MIC in JW5503, compound 1 shows a resistance this wide-open conformation, we generated an apo LolCDE form by

Nature | [Link] | 3
Article
a b c
100 E. coli K. pneumoniae E. cloacae
100 100

Strains Inhibited at MIC (%)

Strains Inhibited at MIC (%)

Strains Inhibited at MIC (%)

Lolamicin Lolamicin Lolamicin

80 1 80 1 80 1

60 60 60

40 40 40

20 20 20

0 0 0
0.25 0.50 1 2 4 8 16 32 0.25 0.50 1 2 4 8 16 32 0.25 0.50 1 2 4 8 16 32
[Compound] (μg ml–1) [Compound] (μg ml–1) [Compound] (μg ml–1)

d e
10–5 Lolamicin 10–4 E. coli BW25113
1
Frequency of resistance

K. pneumoniae ATCC 27736

Frequency of resistance
10–5 E. cloacae ATCC 29893
10–6

10–6
10–7
10–7
10–8
10–8

10–9 10–9
8 16 32 4 8 16
[Compound] (fold change above MIC) [Lolamicin] (fold change above MIC)

Fig. 2 | Antimicrobial assessment and resistance frequency studies of Supplementary Tables 7–9. d, Frequency of E. coli JW5503 ∆tolC resistance to
lolamicin. a–c, Activity of compound 1 and lolamicin against multidrug- compound 1 and lolamicin. Data are mean ± s.e.m., n = 3 biologically independent
resistant clinical isolates of E. coli (a; n = 47), K. pneumoniae (b; n = 61), and samples. e, Frequency of resistance of E. coli BW25113, K. pneumoniae ATCC
E. cloacae (c; n = 18). MIC assays were performed in Mueller Hinton broth per 27736 and E. cloacae ATCC 29893 to lolamicin. Data are mean ± s.e.m., n = 3
CLSI guidelines, in biological triplicate. Resistance genes in clinical isolate biologically independent samples.
panels are listed in Supplementary Table 6. A full listing of this MIC data is in

removing the bound lipoprotein and used this model for extensive the vicinity of TS1 and TS2, suggesting possible functional roles of the
equilibrium molecular dynamics simulations in a realistic lipid bilayer, transient sites (Fig. 3 and Extended Data Fig. 5). Some of the experi-
accumulating a total of 2.5 µs of simulation, sampling a multitude of mentally deleterious mutations such as E195K appear to interfere with
LolCDE conformations. lolamicin binding by participating in salt bridges with nearby residues
The approximate free energy landscape of the resulting confor- that occupy part of BS1, BS2 and TS1 (Extended Data Fig. 5e–g); notably,
mations was projected onto representative reaction coordinates D264 appears to be highly engaged in a salt bridge in several of the
(Extended Data Fig. 5c,d). Protein conformations obtained from the simulation replicates (Extended Data Fig. 5f). N265K appears to have
simulation trajectories were discretized by clustering analysis using the a structural role by increasing the fluctuation of the door bar region,
root mean square deviation (r.m.s.d.) of the transmembrane domains which is expected to destabilize the nearby BS1 (Extended Data Fig. 5h).
as the dissimilarity metric, and a representative structure from each Finally, the F367S mutation, although not directly in the binding site,
cluster was used for lolamicin docking with AutoDock Vina. The result- is positioned at the top of the door bar loop, potentially increasing
ing 10 unique protein–inhibitor poses were individually embedded in the flexibility of the door bar owing to the reduction of steric and π–π
lipid bilayers and each simulated for 100 ns. stacking interactions with nearby residues, thus destabilizing the for-
From these protein–inhibitor simulations, four putative lolamicin- mation of a tight binding pocket for lolamicin (Extended Data Fig. 5i).
binding modes were identified as high-occupancy clusters. Two of Analysis of the contributions of LolCDE residues to lolamicin-binding
these clusters (here named binding sites BS1 and BS2) showed greater interactions reveals a dominant role for hydrophobic interactions
residence time than the other two; the locations of the other two, with nonpolar or aromatic residues (Fig. 3). This finding offers a prob-
shorter-lived clusters are named as transient sites TS1 and TS2 (Fig. 3). able explanation for the decreased efficacy of compounds contain-
Lolamicin within BS1 and BS2 overlaps substantially with the cryo-EM ing primary amines (compounds 3–6) due to the decreased binding
density of the native substrate Lpp lipoprotein at its fatty acid chains, strengths of an amino or ammonium group in the hydrophobic binding
and forms contacts with several residues known to natively interact pocket. To test this hypothesis, a negative control with a primary amine
with lipoprotein40,41 (Fig. 3). Point mutations of many of the residues (compound 3) was docked into the BS1 binding site, and subsequent
lining BS1 and BS2 and their immediate surroundings, such as D264/ simulations captured its spontaneous unbinding (Extended Data
I268/Y366 of LolE, are deleterious to the lipoprotein shuttling and cell Fig. 5b), indicating a reduced binding affinity of compound 3 relative
viability40,41. Similarly, point mutations at these same residues lining BS1 to lolamicin, probably owing to the primary amine group disrupting
and BS2 confer resistance to lolamicin, as observed in the resistance hydrophobic interactions in the binding pocket.
mutation studies (Supplementary Table 10). The locations of these
putative lolamicin-binding sites, combined with the data from resistant
mutants, strongly suggests that lolamicin competitively inhibits lipo- In vivo efficacy of lolamicin
protein trafficking by physically occupying the lipoprotein-binding site. Compound 1 and lolamicin were found to be well-tolerated by mice
Mapping the lolamicin-resistant mutations onto the LolCDE struc- when administered at 100 mg kg−1 twice a day for 3 days via intraperito-
ture, it is apparent that these residues line BS1 and BS2 and are also in neal injection. Additionally, preliminary pharmacokinetic assessment

4 | Nature | [Link]
 a b
Residence
time (ns)
BS1 393
BS2 >449 D264
TS2 F51
TS1 >242
TS2 >101 M267
BS2 M48
LolC LolE
Door bar
BS1 F51
Y366 A47
M48
TS1 I268
LolD LolD G44
M266 M272
V40

Fig. 3 | Major binding and transient binding sites of lolamicin in LolCDE. transmembrane segment 3 extending into the periplasm and bending toward
a, Representative structure of LolCDE (PDB ID: 7MDX) embedded in a realistic LolC, parallel to the inner membrane40. Lolamicin poses from each cluster are
bacterial inner membrane. b, Left, high-occupancy lolamicin-binding modes coloured blue, red, grey and green, and the residence time in each binding site
were obtained via r.m.s.d.-based clustering (conducted in VMD) of the is shown in the table. Right, magnified views of BS2 (top) and BS1 (bottom),
simulation, overlaid with the Lpp lipoprotein (light blue stick representation) highlighting residues with a contact probability of over 20%.
from the cryo-EM structure of LolCDE (PDB ID: 7MDX). The door bar signifies

at this dose suggested the potential of lolamicin as an active antibiotic Fig. 5, we observed marked shifts in bacterial populations following
in vivo (Supplementary Table 11). Using this dosing regimen, compound administration of amoxicillin and clindamycin. At the class level,
1 and lolamicin were first evaluated head-to-head in two infection mod- treatment with clindamycin resulted in large relative decreases in the
els: an acute pneumonia infection model examining bacterial burden, members of Bacilli and Bacteroidia, with an observable increase in
and a septicaemia model analysing overall survival. Mice were infected members of Gammaproteobacteria, detectable at four days and seven
with E. coli AR0349 (a colistin-resistant clinical isolate, where lolamicin days after treatment (Extended Data Fig. 6). Specifically, proportions
has an MIC of 1 µg ml−1 and compound 1 has an MIC of >32 µg ml−1). Mice of Enterobacteriaceae and Tannerellaceae families increased, whereas
treated with lolamicin showed a two-log reduction in bacterial burden those of Bacteroidaceae, Acholeplasmataceae, Lachnospiraceae and
(Fig. 4a). Additionally, 100% of mice burdened with septic infection Muribaculaceae diminished (Fig. 5a). Although used for Gram-positive
were successfully rescued (Fig. 4b); compound 1 fared poorly in these infections, clindamycin treatment resulted in decreases in Bacteroidia,
experiments compared to lolamicin. Analogous experiments showed a class of Gram-negative bacteria. This is not surprising given the anti-
the efficacy of lolamicin against mice infected with colistin-resistant biotic activity demonstrated by clindamycin against Bacteroides spe-
K. pneumoniae (Fig. 4c), carbapenem-resistant K. pneumoniae (Fig. 4d) cies in the in vitro MIC assessments (Supplementary Table 3). These
and colistin-resistant E. cloacae (Fig. 4e,f). results are also consistent with observations with clindamycin in pre-
Given the superiority of lolamicin relative to compound 1 as dem- vious microbiome analyses performed both in mice and humans1,42.
onstrated by these infection models, we turned to the evaluation of Analogous shifts were also detected in mice treated with amoxicillin,
efficacy when given orally, as a prelude to the gut microbiome studies. which showed relative increases in members of Bacteroidia, namely
Pharmacokinetic experiments revealed significant oral bioavailabil- Bacteroidaceae, and enrichment in the members of Gammaproteo-
ity for lolamicin (F = 47%; Supplementary Table 11) and lolamicin was bacteria like Enterobacteriaceae. Amoxicillin treatment also resulted
well-tolerated when administered to mice via oral gavage at 200 mg kg−1 in relative decreases in Lachnospiraceae and Muribaculaceae (Fig. 5a).
twice a day for 3 days, so this dosing regimen was used for infection By contrast, lolamicin treatment did not cause any substantial changes
models. Lolamicin was again evaluated in an acute pneumonia infection in taxonomic composition over the course of the three-day treatment
model and septicaemia model using colistin-resistant E. coli AR0349. or the following 28-day recovery (Fig. 5a and Extended Data Figs. 6–7).
Mice treated with oral lolamicin showed a three-log reduction in bacte- We used alpha diversity measures such as Shannon–Weaver diversity
rial burden when compared to treatment with vehicle (Fig. 4g). Addi- index and observed species analysis as well as the Bray–Curtis compo-
tionally, more than 70% of mice with septic infection were successfully sitional beta diversity measure to examine the population diversity
rescued (Fig. 4h). and richness across all treatment groups (Fig. 5b,c and Extended Data
Fig. 8). The Shannon–Weaver metric reflects the diversity of species
within a given community based on both richness and abundance.
Lolamicin and the gut microbiome The Shannon–Weaver metric was reduced following treatment with
Lolamicin occupies an unusual position in the antibiotic arsenal, as amoxicillin and clindamycin (Fig. 5b). Shannon index values decreased
it displays selective antimicrobial activity for certain pathogenic from 4.2 on day 0 to 2.2 on day 7 in clindamycin-treated mice and
Gram-negative bacteria and has no effect on Gram-positive bacteria decreased from 3.9 to 2.6 in amoxicillin-treated mice (Fig. 5b). There
or on non-pathogenic Gram-negative commensal bacteria. Of course, was also no statistically significant difference in diversity values from
the strains tested in these experiments do not reflect the full diver- the lolamicin-treated group, comparable to the vehicle negative control
sity of the gut microbiome; we therefore examined the effect of lola- at day 0 (Fig. 5b). Given that any decreases in microbiome diversity and
micin on the gut microbiome of mice. We treated healthy mice with richness affect host metabolism, immune response and susceptibil-
either vehicle, amoxicillin (broad-spectrum antibiotic), clindamycin ity to secondary infections2,3, minimal to no perturbations of these
(Gram-positive-only antibiotic), or lolamicin (Gram-negative-only metrics are desirable. Furthermore, observed species (the number of
antibiotic) at doses known to produce antibacterial efficacy in vivo. species in a community) was calculated at different rarefaction depths
DNA was extracted from faecal samples collected at day 0 (before (every 1,000 steps in sample size) to plot rarefaction curves for each
antibiotic treatment), 7 (4 days after ending antibiotic treatment), 10 sample (Fig. 5c). In the clindamycin-treated group, alpha rarefaction
(7 days after ending antibiotic treatment) and 31 (28 days after ending revealed marked depletion in richness, particularly on days 7 and 10,
antibiotic treatment), and the microbiome composition was analysed although reduced richness was still detected at day 31 (Fig. 5c). Similarly,
by full-length 16S ribosomal RNA (rRNA) sequencing. As detailed in decreased richness on days 7 and 10 was observed in the amoxicillin

Nature | [Link] | 5
Article
a E. coli AR-0349 b E. coli AR-0349 c K. pneumoniae AR-0040 d K. pneumoniae BAA-1705
**** ****
8 *** 100 8 NS 100
**** Lolamicin ****

log CFU per lung

log CFU per lung

80 80
1

Survival (%)
Survival (%)
6 6 Lolamicin
60 Vehicle 60 1
40 40 Vehicle
4 4
20 20
**** ****
2 0 **** 2 0 ****
Vehicle 1 Lolamicin 0 24 48 72 96 120 Vehicle 1 Lolamicin 0 24 48 72 96 120 144
Time (h) Time (h)
e E. cloacae AR-0163 f E. cloacae AR-0163 g E. coli AR-0349 h E. coli AR-0349
****
NS 100 100
8 **** 8 **** Lolamicin
Vehicle

log CFU per lung

log CFU per lung

80 Lolamicin 80
Survival (%)

Survival (%)
6 60 1 60
Vehicle 4
40 40
4 ***
20 2 20
**** ****
2 0 0 0
Vehicle 1 Lolamicin 0 24 48 72 96 120 144 Vehicle Lolamicin 0 24 48 72 96 120
Time (h) Time (h)

Fig. 4 | In vivo efficacy studies with lolamicin. a,b, Acute pneumonia E. coli AR0349. In g,h, CD-1 mice were treated with vehicle (20% DMSO, 30% water,
(a; 2.7 × 108 CFU per mouse) and septicaemia (b; 4.2 × 108 CFU per mouse) 50% PEG400) or lolamicin (200 mg kg−1 oral administration; n = 8 biologically
infection models initiated with colistin-resistant E. coli AR0349. c,d, Acute independent mice for pneumonia model, n = 15 biologically independent mice
pneumonia infection model initiated with colistin-resistant K. pneumoniae for septicaemia model) twice daily post-infection for 3 days. a,c,e,g, Data
AR0040 (c; 8.0 × 107 CFU per mouse) and septicaemia infection model initiated are mean ± s.d.; one-way ANOVA with Tukey’s multiple comparisons test.
with carbapenem-resistant K. pneumoniae BAA-1705 (d; 5.8 × 107 CFU per b,d,f,h, Two-tailed log-rank (Mantel–Cox) test with lolamicin as reference.
mouse). e,f, Acute pneumonia (e; 7.2 × 108 CFU per mouse) and septicaemia a, Vehicle vs compound 1, P = 0.0003; vehicle vs lolamicin, P < 0.0001;
(f; 9.0 × 108 CFU per mouse) infection models initiated with colistin-resistant compound 1 vs lolamicin, P < 0.0001. b, Vehicle, P < 0.0001; compound 1,
E. cloacae AR0163. In a–f, CD-1 mice were treated with vehicle (50% DMSO, P < 0.0001. c, Vehicle vs compound 1, P = 0.4392; vehicle vs lolamicin, P < 0.0001;
50% PEG400) or compound (1 or lolamicin, 100 mg kg−1 intraperitoneal injection; compound 1 vs lolamicin, P < 0.0001. d, Vehicle, P < 0.0001; compound 1,
n = 8 biologically independent mice for pneumonia model, n = 15 biologically P < 0.0001. e, Vehicle vs compound 1, P = 0.1407; vehicle vs lolamicin, P < 0.0001;
independent mice for septicaemia model) twice daily post-infection for compound 1 vs lolamicin, P < 0.0001. f, Vehicle, P < 0.0001; compound 1,
3 days. g,h, Acute pneumonia (g; 2.9 × 108 CFU per mouse) and septicaemia P = 0.0007. g, Lolamicin, P < 0.0001. h, Vehicle, P < 0.0001. *P < 0.05, **P < 0.01,
(h; 6.0 × 108 CFU per mouse) infection models initiated with colistin-resistant ***P < 0.001, ****P < 0.0001; NS, not significant.

treatment group. Conversely, in the lolamicin treatment group, species problematic Enterobacteriaceae, but shows no killing of commensal
richness was retained throughout both the treatment and recovery gut bacteria in culture and causes no significant changes to the mouse
periods, which was analogous to the species richness detected in the gut microbiome. Furthermore, the intact gut bacterial community in
vehicle control. lolamicin-treated mice can successfully clear C. difficile colonization,
distinguishing lolamicin from antibiotics that are currently used in the
clinic. Given the known health burden of C. difficile infections (500,000
C. difficile challenge experiment infections and 30,000 deaths annually, with a 35% recurrence rate
As microbiome perturbation resulting from antibiotic treatment has in the USA44), routine use of microbiome-sparing antibiotics would
been implicated in causing secondary C. difficile infections, a study was have a significant positive effect on human health. To advance as a
next performed to determine the relationship between microbiome translational candidate, the frequency of resistance to lolamicin would
compositional changes and C. difficile clearance43. Using the same treat- probably need to be improved through iterative chemical synthesis
ment groups as in the microbiome study, we treated healthy mice with and lead optimization, as has been achieved for other antibiotics45–48
antibiotic and then challenged them with C. difficile to evaluate their (and demonstrated herein on this scaffold; Fig. 2d), and/or it could be
ability to spontaneously clear C. difficile colonization. As lolamicin used in combination with other antibiotics.
treatment did not induce significant levels of perturbation in the micro- The minimal perturbation of the gut microbiome by lolamicin
biome experiment, it was hypothesized that mice treated with lolamicin highlights this doubly selective strategy as promising for developing
would be capable of spontaneously clearing the pathogen, whereas microbiome-sparing antibiotics against Gram-negative pathogens. In
mice treated with clindamycin or amoxicillin would not be able to do a similar manner, this strategy may also be leveraged for the develop-
so. As shown in Fig. 5d, treatment with lolamicin resulted in little to no ment of future LolCDE inhibitors for P. aeruginosa or A. baumannii,
C. difficile colonization, similar to the vehicle control. However, mice as the sequence homologies of the Lol target in these Gram-negative
treated with amoxicillin or clindamycin demonstrated an inability to pathogens is low compared with commensal bacteria. There are other
clear C. difficile, and high-level colonization was observed throughout biological targets that could also be exploited in an analogous fashion,
the duration of the experiment (Fig. 5d). giving rise to other classes of pathogen-specific antibiotics. One such
target for generating narrow-spectrum agents is FabI of the fatty acid
biosynthesis (FAS-II) pathway49. Owing to the presence of various
Discussion FabI isoforms across different bacteria, FabI inhibition should not
In accord with its mode of action and with the sequence homology be lethal for human commensals50,51. Other potential doubly selec-
of its target, lolamicin shows promising antibacterial activity against tive targets include the Gram-negative essential outer membrane

6 | Nature | [Link]
 a b
Vehicle Lolamicin Family
1.00 [Eubacterium]
coprostanoligenes group
Acholeplasmataceae
Akkermansiaceae
Relative abundance

0.75 Alteromonadaceae
Anaerofustaceae
Anaerovoracaceae
0.50 Bacteroidaceae 4
Butyricicoccaceae
Christensenellaceae
Clostridiaceae
0.25 Deferribacteraceae

Diversity value
Defluviitaleaceae
Eggerthellaceae
0 Enterobacteriaceae
Enterococcaceae
Day 0 Day 7 Day 10 Day 31 Day 0 Day 7 Day 10 Day 31 Erysipelatoclostridiaceae 3
Erysipelotrichaceae **
Amoxicillin Clindamycin Lachnospiraceae
1.00 Lactobacillaceae
Monoglobaceae
Moraxellaceae
Relative abundance

0.75 Morganellaceae
Muribaculaceae
Oscillospiraceae 2
Paenibacillaceae
0.50 Peptococcaceae
Peptostreptococcaceae
Pseudomonadaceae
Ruminococcaceae

7
0.25

y
Staphylococcaceae

da

da

da

da

da

da

da

da
Tannerellaceae

in

in

in
cl

cl

illi

illi
ic

ic

yc

yc
UCG−010

hi

hi

ic

ic
m

am

m
Ve

Ve

ox

ox
NA

la

la

da
0

Lo

Lo

d
Am

Am

lin

lin
Day 0 Day 7 Day 10 Day 31 Day 0 Day 7 Day 10 Day 31

C
c Day 0 Day 7 d Vehicle Lolamicin
10 10

log CFU per 0.1 g faeces

600 600
8 8
Species richness

400 400 6 6

Vehicle 4 4
200 Lolamicin 200
Amoxicillin 2 2 NS
Clindamycin
0 0
0 0
0 10,000 20,000 30,000 40,000 0 10,000 20,000 30,000 40,000 1 3 5 1 3 5

Day 10 Day 31 Amoxicillin Clindamycin

10 10
****
log CFU per 0.1 g faeces

600 600 ****

Species richness

8 8

400 400 6 6

4 4
200 200
2 2
0 0
0 0
0 10,000 20,000 30,000 40,000 0 10,000 20,000 30,000 40,000 1 3 5 1 3 5
Sequence sample size Sequence sample size Time (days) Time (days)

Fig. 5 | Lolamicin spares the gut microbiome and prevents C. difficile c, Species richness measured by alpha rarefaction before (day 0) and after (day 7,
colonization. a, Taxonomic analysis showing bacterial population shifts at the day 10 and day 31) antibiotic administration. d, Daily counts of C. difficile in faecal
family level over a 31-day period before (day 0) and after (day 7, day 10 and day 31) samples of mice treated with vehicle, lolamicin, amoxicillin or clindamycin from
antibiotic administration. b, Diversity analysis before (day 0) and after (day 7) the time of challenge (day 0) with 1.2 × 104 spores of C. difficile strain 630 up to
antibiotic administration measured by Shannon index. n = 6 biologically 5 days post-infection. Bold lines represent median counts of treatment groups
independent mice examined over 2 independent experiments (day 0 and day 7). and light lines indicate data for individual mice. Statistical significance between
Unpaired Wilcoxon rank-sum test with vehicle day 0 as reference. Vehicle day 7, day 0 and day 5 by one-way ANOVA with Tukey’s multiple comparisons. CD-1 mice
P = 0.24; lolamicin day 0, P = 0.06; lolamicin day 7, P = 0.24; amoxicillin day 0, (6 per cohort) were treated with vehicle (20% DMSO, 30% water, 50% PEG400)
P = 0.13; amoxicillin day 7, P = 0.08; clindamycin day 0, P = 0.06; clindamycin day 7, or compound (clindamycin, 100 mg kg−1; amoxicillin, 100 mg kg−1; or lolamicin,
P = 0.002. Boxes span from the first to the third quartiles, centre lines represent 200 mg kg−1) twice a day for 3 days via oral gavage. Lolamicin day 5, P = 0.6436;
median values and whiskers show data lying within 1.5× the interquartile range amoxicillin day 5, P < 0.0001; clindamycin day 5, P < 0.0001.
of the top and bottom quartiles. Data points outside whiskers represent outliers.

proteins, BamA and LptD52. BamA, a component of the Bam complex of the inner membrane. MsbA is conserved across Gram-negative
responsible for the insertion of β-barrel proteins into the outer mem- bacteria but its sequence identity diverges across Gram-negative
brane, possesses high sequence similarity among Gram-negative pathogens.
ESKAPE pathogens, but low sequence similarity among gut resi- Another intriguing aspect of the Lol system is its location in the peri-
dents. Similarly, LptD, which has an important role in the assembly plasm. The orthogonal sieving of the outer and inner membranes of
of lipopolysaccharides (LPS) in the outer membrane, has sequence Gram-negative bacteria is believed to make cytoplasmic targets more
divergence among Gram-negative bacteria and human commensals. difficult to reach54, and suggests the potential exploitability of periplas-
A final potential doubly selective target is the Gram-negative essential mic proteins as antibacterial targets, as such antibiotics would only
inner membrane protein MsbA53, an ATP-binding cassette transporter need to penetrate the outer membrane. Currently, there simply are not
that delivers lipid A and lipopolysaccharides to the periplasmic face enough periplasmic-targeted antibiotics with activity against wild-type

Nature | [Link] | 7
Article
(non-permeability-deficient) bacteria to probe this hypothesis, and the 23. Smith, P. A. et al. Optimized arylomycins are a new class of Gram-negative antibiotics.
Nature 561, 189–194 (2018).
majority of antibiotics that do engage periplasmic targets do so in a 24. Tokuda, H. & Matsuyama, S.-I. Sorting of lipoproteins to the outer membrane in E. coli.
covalent manner (for example, arylomycins and beta-lactams), com- Biochim. Biophys. Acta 1693, 5–13 (2004).
plicating this analysis. As a result, there is still much to learn regarding 25. Pathania, R. et al. Chemical genomics in Escherichia coli identifies an inhibitor of bacterial
lipoprotein targeting. Nat. Chem. Biol. 5, 849–856 (2009).
the relationship between accumulation and antibiotics with targets 26. Barker, C. A. et al. Degradation of MAC13243 and studies of the interaction of resulting
in the periplasm, and discovery of additional compounds is needed. thiourea compounds with the lipoprotein targeting chaperone LolA. Bioorg. Med. Chem.
Our resistant-mutant studies and computationally derived binding Lett. 23, 2426–2431 (2013).
27. Hoang, H. H. et al. Outer membrane targeting of Pseudomonas aeruginosa proteins
sites for lolamicin are consistent with a periplasmic mode of action shows variable dependence on the components of Bam and Lol machineries. mBio 2,
for the drug. e00246–00211 (2011).
The intestinal microbiome is central to maintaining host health, 28. Ito, H. et al. A new screening method to identify inhibitors of the Lol (localization of
lipoproteins) system, a novel antibacterial target. Microbiol. Immun. 51, 263–270 (2007).
and its perturbation can result in many deleterious effects, including 29. Nayar, A. S. et al. Novel antibacterial targets and compounds revealed by a high-throughput
C. difficile infection and beyond. Consequently, pathogen-specific cell wall reporter assay. J. Bacteriol. 197, 1726–1734 (2015).
antibiotics such as lolamicin will be critical to minimizing collateral 30. Liu, J. et al. Natural inhibitors targeting the localization of lipoprotein system in Vibrio
parahaemolyticus. Int. J. Mol. Sci. 23, 14352 (2022).
damage to the gut microbiome; this microbiome-sparing effect would 31. Breidenstein, E. B. M. et al. SMT-738: a novel small-molecule inhibitor of bacterial
make such antibiotics superior for patients compared with antibiotics lipoprotein transport targeting Enterobacteriaceae. Antimicrob. Agents Chemother. 68,
in current clinical practice. e0069523 (2024).
32. Grabowicz, M. & Silhavy, T. J. Redefining the essential trafficking pathway for outer
membrane lipoproteins. Proc. Natl Acad. Sci. USA 114, 4769–4774 (2017).
33. Richter, M. F. et al. Predictive compound accumulation rules yield a broad-spectrum
Online content antibiotic. Nature 545, 299–304 (2017).
34. Richter, M. F. & Hergenrother, P. J. The challenge of converting Gram-positive-only
Any methods, additional references, Nature Portfolio reporting summa- compounds into broad-spectrum antibiotics. Ann. NY Acad. Sci. 1435, 18–38 (2019).
ries, source data, extended data, supplementary information, acknowl- 35. Munoz, K. A. & Hergenrother, P. J. Facilitating compound entry as a means to discover
edgements, peer review information; details of author contributions antibiotics for Gram-negative bacteria. Acc. Chem. Res. 54, 1322–1333 (2021).
36. Prochnow, H. et al. Subcellular quantification of uptake in Gram-negative bacteria.
and competing interests; and statements of data and code availability Anal. Chem. 91, 1863–1872 (2019).
are available at [Link] 37. Wexler, H. M. Bacteroides: the good, the bad, and the nitty-gritty. Clin. Microbiol. Rev. 20,
593–621 (2007).
38. Nonejuie, P., Burkart, M., Pogliano, K. & Pogliano, J. Bacterial cytological profiling rapidly
1. Maier, L. et al. Unravelling the collateral damage of antibiotics on gut bacteria. Nature identifies the cellular pathways targeted by antibacterial molecules. Proc. Natl Acad. Sci.
599, 120–124 (2021). USA 110, 16169–16174 (2013).
2. Lynch, S. V. & Pedersen, O. The human intestinal microbiome in health and disease. 39. Grabowicz, M. Lipoproteins and their trafficking to the outer membrane. EcoSal Plus 8,
New Eng. J. Med. 375, 2369–2379 (2016). [Link] (2019).
3. Schubert, A. M., Sinani, H. & Schloss, P. D. Antibiotic-induced alterations of the murine 40. Tang, X. et al. Structural basis for bacterial lipoprotein relocation by the transporter
gut microbiota and subsequent effects on colonization resistance against Clostridium LolCDE. Nat. Struct. Mol. Biol. 28, 347–355 (2021).
difficile. mBio 6, e00974 (2015). 41. Sharma, S. et al. Mechanism of LolCDE as a molecular extruder of bacterial triacylated
4. Owens, R. C. Jr., Donskey, C. J., Gaynes, R. P., Loo, V. G. & Muto, C. A. Antimicrobial- lipoproteins. Nat. Commun. 12, 4687 (2021).
associated risk factors for Clostridium difficile infection. Clin. Infect. Dis. 46, S19–S31 42. Buffie, C. G. et al. Profound alterations of intestinal microbiota following a single dose of
(2008). clindamycin results in sustained susceptibility to Clostridium difficile-induced colitis.
5. Iizumi, T., Battaglia, T., Ruiz, V. & Perez Perez, G. I. Gut microbiome and antibiotics. Infect. Immun. 80, 62–73 (2012).
Arch. Med. Res. 48, 727–734 (2017). 43. Lesniak, N. A., Schubert, A. M., Sinani, H. & Schloss, P. D. Clearance of Clostridioides
6. Poon, S. S. B. et al. Neonatal antibiotics have long term sex-dependent effects on the difficile colonization is associated with antibiotic-specific bacterial changes. mSphere 6,
enteric nervous system. J. Phys. 600, 4303–4323 (2022). e01238-20 (2021).
7. Lange, K., Buerger, M., Stallmach, A. & Bruns, T. Effects of antibiotics on gut microbiota. 44. Feuerstadt, P., Theriault, N. & Tillotson, G. The burden of CDI in the United States:
Digest. Dis. 34, 260–268 (2016). a multifactorial challenge. BMC Infect. Dis. 23, 132 (2023).
8. Gu, S. et al. Effect of the short-term use of fluoroquinolone and β-lactam antibiotics on 45. Garcia Chavez, M. et al. Synthesis of fusidic acid derivatives yields a potent antibiotic with
mouse gut microbiota. Infect. Drug Resist. 13, 4547–4558 (2020). an improved resistance profile. ACS Infect. Dis. 7, 493–505 (2021).
9. Lofmark, S., Jernberg, C., Jansson, J. K. & Edlund, C. Clindamycin-induced enrichment and 46. Oefner, C. et al. Increased hydrophobic interactions of iclaprim with Staphylococcus
long-term persistence of resistant Bacteroides spp. and resistance genes. J. Antimicrob. aureus dihydrofolate reductase are responsible for the increase in affinity and antibacterial
Chemother. 58, 1160–1167 (2006). activity. J. Antimicrob. Chemother. 63, 687–698 (2009).
10. Hertz, F. B. et al. Effects of antibiotics on the intestinal microbiota of mice. Antibiotics 9, 47. Purnapatre, K. P. et al. In vitro and in vivo activities of DS86760016, a novel leucyl-tRNA
191 (2020). synthetase inhibitor for Gram-negative pathogens. Antimicrob. Agents Chemother. 62,
11. Lagier, J. C., Million, M., Hugon, P., Armougom, F. & Raoult, D. Human gut microbiota: e01987-17 (2018).
repertoire and variations. Front. Cell Infect. Microbiol. 2, 136 (2012). 48. Schuster, M. et al. Peptidomimetic antibiotics disrupt the lipopolysaccharide transport
12. Singh, H. Management with colistin. Ind. J. Crit. Care. Med. 14, 161–162 (2010). bridge of drug-resistant Enterobacteriaceae. Sci. Adv. 9, eadg3683 (2023).
13. Falagas, M. E. & Kasiakou, S. K. Colistin: the revival of polymyxins for the management of 49. Rana, P. et al. FabI (enoyl acyl carrier protein reductase)—a potential broad spectrum
multidrug-resistant Gram-negative bacterial infections. Clin. Infect. Dis. 40, 1333–1341 therapeutic target and its inhibitors. Eur. J. Med. Chem. 208, 112757 (2020).
(2005). 50. Parker, E. N. et al. An iterative approach guides discovery of the FabI inhibitor fabimycin, a
14. Chatzidimitriou, M. et al. mcr genes conferring colistin resistance in Enterobacterales; late-stage antibiotic candidate with in vivo efficacy against drug-resistant Gram-negative
a five year overview. Acta Med. Acad. 50, 365–371 (2021). infections. ACS Cent. Sci. 8, 1145–1158 (2022).
15. Rice, L. B. Federal funding for the study of antimicrobial resistance in nosocomial 51. Yao, J. et al. A pathogen-selective antibiotic minimizes disturbance to the microbiome.
pathogens: no ESKAPE. J. Infect. Dis. 197, 1079–1081 (2008). Antimicrob. Agents Chemother. 60, 4264–4273 (2016).
16. Hoffman, P. S. Antibacterial discovery: 21st century challenges. Antibiotics 9, 213–213 52. Konovalova, A., Kahne, D. E. & Silhavy, T. J. Outer membrane biogenesis. Annu. Rev.
(2020). Microbiol. 71, 539–556 (2017).
17. Nickerson, N. N. et al. A novel inhibitor of the LolCDE ABC transporter essential for 53. Ho, H. et al. Structural basis for dual-mode inhibition of the ABC transporter MsbA. Nature
lipoprotein trafficking in Gram-negative bacteria. Antimicrob. Agents Chemother. 62, 557, 196–201 (2018).
e02151–02117 (2018). 54. Silver, L. L. A Gestalt approach to Gram-negative entry. Bioorg. Med. Chem. 24, 6379–6389
18. Zhang, G. et al. Cell-based screen for discovering lipopolysaccharide biogenesis (2016).
inhibitors. Proc. Natl Acad. Sci. USA 115, 6834–6839 (2018).
19. Lehman, K. M. & Grabowicz, M. Countering Gram-negative antibiotic resistance: recent Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
progress in disrupting the outer membrane with novel therapeutics. Antibiotics 8, 163 published maps and institutional affiliations.
(2019).
20. Brown, M. F. et al. Potent inhibitors of LpxC for the treatment of Gram-negative infections. Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this
J. Med. Chem. 55, 914–923 (2012). article under a publishing agreement with the author(s) or other rightsholder(s); author
21. Miller, R. D. et al. A novel antibiotic targeting BamA identified by a computational search. self-archiving of the accepted manuscript version of this article is solely governed by the
Nat. Microbiol. 7, 1661–1672 (2022). terms of such publishing agreement and applicable law.
22. Imai, Y. et al. A new antibiotic selectively kills Gram-negative pathogens. Nature 576,
459–464 (2019). © The Author(s), under exclusive licence to Springer Nature Limited 2024

8 | Nature | [Link]
Methods 8.8 ml of fresh PBS and aliquoted into 10× 1.5 ml Eppendorf tubes (875 µl
each). CFU was determined by a calibration curve. The samples were
Bacterial strains equilibrated at 37 °C with shaking for 5 min, compound was added
Staphylococcus aureus ATCC 29213, E. coli MG1655, E. coli BAA-2340, (final concentration 50 µM), and then samples were incubated at 37 °C
E. coli BAA-2469, E. coli BAA-2471, E. cloacae BAA-2341, E. cloacae BAA- with shaking for 10 min. After incubation, 800 µl of the cultures were
2468, K. pneumoniae BAA-1705, K. pneumoniae BAA-2342, K. pneumo- carefully layered on 700 µl of silicone oil (9:1 AR20/Sigma High Tem-
niae BAA-2470, K. pneumoniae BAA-2472, K. pneumoniae BAA-2473, perature, cooled to −78 °C). Bacteria were pelleted through the oil by
Enterococcus faecium ATCC 19434, Enterococcus faecalis ATCC 19433, centrifuging at 13,000g for 2 min at room temperature (supernatant
M. morganii ATCC 25830, P. mirabilis ATCC 35659, S. maltophilia ATCC remains above the oil); the supernatant and oil were then removed by
13637, S. typhimurium ATCC 14028, Lactobacillus casei ATCC 393, pipetting. To lyse the samples, each pellet was suspended in 200 µl of
Lactobacillus rhamnosus ATCC 53103, Bifidobacterium animalis ssp. water, and then they were subjected to 3 freeze-thaw cycles of 3 min in
animalis ATCC 25527, A. muciniphila ATCC BAA-835 and Clostridium liquid nitrogen followed by 3 min in a water bath at 65 °C. The lysates
leptum ATCC 29065 were obtained from the American Type Culture were pelleted at 13,000g for 2 min at room temperature and the super-
Collection (ATCC). E. coli BW25113, E. coli JW5503 and E. coli JW3596 natant was collected (180 µl). The debris was resuspended in 100 µl of
were obtained from the Keio Collection. PAO1 ∆6, A. baumannii ATCC methanol and pelleted as before. The supernatants were removed and
19606 and A. baumannii 19606 ∆3 were provided by H. Zgurskaya. combined with the previous supernatants collected. Finally, remain-
E. cloacae ATCC 29893 was provided by W. van der Donk. Bacillus cereus ing debris was removed by centrifuging at 20,000g for 10 min at room
ATCC 11778 and Listeria monocytogenes ATCC 19115 were provided by temperature. Supernatants were analysed by liquid chromatography–
A. Salyers. Bacillus anthracis Sterne 7702 was provided by D. Mitchell. tandem mass spectrometry.
Bacillus subtilis PY79 was provided by E. Garner. Staphylococcus Samples were analysed with the 5500 QTRAP LC–MS/MS system
epidermidis NRS 101 was obtained from the Network on Antimicrobial (AB Sciex) with a 1200 series HPLC system (Agilent Technologies)
Resistance in S. aureus (NARSA). Streptococcus pneumoniae strain 1003 including a degasser, an autosampler, and a binary pump. The liquid
and Streptococcus pyogenes strain 1055 were obtained from Cubist. chromatography separation was performed on an Agilent SB-Aq col-
Streptococcus salivarius KLE 2370, Streptococcus parasanguinis KLE umn (4.6 × 50 mm, 5 µm) (Agilent Technologies) with mobile phase
2375, B. cellulosilyticus KLE 2342, B. fragilis ATCC 25285, B. vulgatus KLE A (0.1% formic acid in water) and mobile phase B (0.1% formic acid in
2303, F. prausnitzii ATCC 27768, E. faecalis KLE 2341, Bifidobacterium acetonitrile). The flow rate was 0.3 m /min. The linear gradient was as
longum ATCC BAA-999, Lactobacillus reuteri ATCC 23272 and Blautia follows: 0–3 min, 100% mobile S18 phase A; 10–15 min, 2% mobile phase
producta ATCC 27340 were provided by K. Lewis. Clinical isolates of A; 15.5–21 min, 100% mobile phase A. The autosampler was set at 5 °C.
E. coli, K. pneumoniae and E. cloacae were obtained from the Centers The injection volume was 15 µl. Mass spectra were acquired with both
for Disease Control and Prevention. positive electrospray ionization at the ion spray voltage of 5,500 V and
negative electrospray ionization at the ion spray voltage of −4,500 V.
Antimicrobial susceptibility assays The source temperature was 450 °C. The curtain gas, ion source gas 1,
Susceptibility testing was performed in biological triplicate, using and ion source gas 2 were 33, 50 and 65, respectively. Multiple reaction
the micro-dilution broth method as outlined by the Clinical and monitoring was used to quantify metabolites. Power analysis was not
Laboratory Standards Institute. Aerobic bacteria were cultured with used to determine the number of replicates. Error bars represent the
cation-adjusted Mueller Hinton Broth (Sigma-Aldrich, 90922) or Todd s.e.m. of three biological replicates. Statistical significance of accumu-
Hewitt Broth (Sigma-Aldrich, T1438) in round-bottom 96-well plates lation was determined using a two-sample Welch’s t-test (one-tailed
(Corning, 3788). Susceptibility testing with KLE collection bacteria test, assuming unequal variance) relative to the negative controls.
was determined under anaerobic conditions (Coy anaerobic chamber, All compounds evaluated in biological assays were ≥ 95% pure.
37 °C, 5% H2, 10% CO2, 85% N2). Overnight cultures were grown in brain
heart infusion broth (Fisher Scientific, DF0418-17-7) supplemented with Competition assay with mutants
0.5% yeast extract, 0.1% L-cysteine hydrochloride, and 15 µg ml−1 haemin Overnight cultures of E. coli BW25113 or lolamicin-resistant BW25113
(BHI-ych). Human serum albumin was purchased from Sigma-Aldrich grown in LB broth were diluted to 5 × 105 CFU ml−1 (OD600 = 0.0005).
(A1653). Microsoft 365 Excel was used to calculate MIC. Using this subculture, 100 µl of BW25113 and 100 µl of mutant
were added to 9.8 ml of fresh LB media to reach a concentration of
Time–kill kinetics 5 × 103 CFU ml−1 of each strain. Co-cultures were incubated at 37 °C
An overnight culture of E. coli BW25113, K. pneumoniae ATCC 27736, for 48 h to assess fitness cost of each mutant strain compared to the
or E. cloacae ATCC 29893 was diluted in 5 ml of Mueller Hinton broth wild-type strain. At time = 0, 24 and 48 h, 100 µl of competition culture
to ~5.0 × 105 CFU ml−1 and incubated at 37 °C while shaking for 1 h. The was removed, serially diluted and plated on LB agar and LB agar con-
bacteria were then challenged with drug at 4× MIC and observed over taining 8 µg ml−1 of lolamicin (4× MIC of BW25113). Plates were grown
the course of 8 h. At each time point 100 µl of culture was collected and at 37 °C overnight and colonies enumerated. Colony counts on LB agar
serially diluted in sterile PBS before plating on drug-free agar plates. represented total CFU of both wild-type and mutant strains in competi-
Colonies were counted after incubation at 37 °C overnight and CFU tion culture; colony counts on 8 µg ml−1 lolamicin agar represented the
per ml was calculated. number of lolamicin-resistant CFU; and the difference between total
CFU and lolamicin-resistant CFU represented number of wild-type CFU.
Accumulation assay
The accumulation assay33 was performed in triplicate in batches of ten Cell culture
samples, with each batch containing tetracycline as a positive control. HFF-1 cells (male, newborn) were obtained from the ATCC. HFF-1 cells
E. coli BW25113 was used in these experiments. For each replicate, 2.5 ml were grown in Dulbecco’s Modified Eagle’s minimum essential medium
of an overnight culture of E. coli was diluted into 250 ml fresh Luria Ber- with 15% fetal bovine serum (Gemini Benchmark; 100-106), 100 µg ml−1
tani (LB) broth (Lennox) and grown at 37 °C with shaking to an optical penicillin and 100 µg ml−1 streptomycin. A-549 cells (male, adult) were
density (OD600) of 0.55. The bacteria were pelleted at 3,220g for 10 min at obtained from the ATCC. A-549 cells were grown in Roswell Park Memo-
4 °C and the supernatant was discarded. The pellets were resuspended rial Institute medium with 10% fetal bovine serum (Gemini Benchmark;
in 40 ml of phosphate-buffered saline (PBS) and pelleted as before, 100-106), 100 µg ml−1 penicillin and 100 µg ml−1 streptomycin. All cells
and the supernatant was discarded. The pellets were resuspended in were cultured at 37 °C under a 5% CO2 environment. Cell lines were
Article
authenticated externally by a commercial vendor and inspected visu- the following conditions for 35 cycles: initial denature: 94 °C for 5 min;
ally in house. Cell lines were not tested for Mycoplasma contamination. denature: 94 °C for 30 s; anneal: 72 °C for 2 min. A 10-µl portion of the
Media were prepared by the University of Illinois School of Chemical PCR reaction mixture was analysed by agarose gel to confirm the prod-
Sciences Cell Media Facility. uct. PCR reactions were purified using GeneJET PCR Purification Kit
(Thermo Scientific). PCR amplicons were submitted to the Core DNA
Cell viability Sequencing Facility at the University of Illinois at Urbana-Champaign
HFF-1 cells were seeded (8,000 cells per well) in a 96-well plate (Greiner for Sanger sequencing with the following primers to sequence lolC, lolD
Bio-One, 655180) and allowed to attach overnight. Cells were treated and lolE (E. coli: N3P-149, N3P-150, N3P-151, N3P-152, N3P-153, N3P-154,
with investigational compounds in DMSO. For half-maximum inhibi- N3P-155, N3P-156, N3P-157; K. pneumoniae: N3P-200, N3P-201, N3P-206;
tory concentration (IC50) determination, the concentrations of the N3P-208; N3P-251; N3P-252; N3P-253; N3P-254; N3P-255). See Supple-
tested compounds were 100 nM to 1,000 µM (1% DMSO final; 100 µl mentary Table 12 for sequences of the primers used in this study. All
per well). Raptinal (100 µM) was used as a dead control. On each plate, primers were obtained from Integrated DNA technologies.
at least three technical replicates per compound were performed.
After 24 h post-treatment, cell viability was assessed using the Alamar Confocal microscopy
Blue method. Stock Alamar Blue solution (10 µl 440 µM resazurin Overnight cultures of bacteria were grown in LB broth to early log phase
(Sigma-Aldrich; R7017) in sterile 1× PBS) was added to each well, and the (OD600 = 0.20) then treated with compound at 3× MIC or right below
plate was incubated for 3–4 h. Conversion of Alamar Blue was measured solubility limit at 30 °C for 1.5 h. Treated bacteria were collected and
with a plate reader (SpectraMax M3; Molecular Devices) by fluorescence compounds were washed with PBS to remove excess compound. Bac-
(excitation wavelength: 555 nm; emission wavelength: 585 nm; cut off teria were stained with Hoechst 33342 (2 µg ml−1, Sigma-Aldrich B2261)
570 nm; autogain). Percentage death was determined by normalizing for 20 min and FM-4−64 (1 µg ml−1, Fisher Scientific T13320) for 3 min.
to DMSO-treated cells and Raptinal-treated cells. For IC50 determina- Bacteria were pelleted, washed with PBS twice, and resuspended in
tion, the data were plotted as compound concentration versus the 3.7% formaldehyde and gently agitated on a nutator for 30 min. Cells
percentage of dead cells and fitted to a logistic-dose-response curve were washed with PBS twice, then 3 µl of fixed bacteria were added
using OriginPro 2015 (Origin Lab). The data were generated in tripli- to a coverslip (22 × 40 mm, VWR), covered with an agarose pad, and
cate, and IC50 values were reported as the average of three separate imaged on a Zeiss 710 multiphoton confocal microscope through a
experiments along with s.e.m. values. Microsoft 365 Excel was used 63×/1.4 refractive index oil objective. Images were acquired and pro-
for measuring cell viability. cessed using Zen Black (Zen 2.3) and experiments were performed in
biological triplicate.
Human red blood cell haemolysis
Whole human blood in citrate phosphate dextrose was obtained from Equilibrium apo simulations
BioIVT and stored at 4 °C and used before expiration date. 100 µl of Initial systems were constructed starting from a lipoprotein-bound
whole blood was combined with 500 µl saline (0.9% NaCl) and centri- cryo-EM structure of LolCDE (PDB ID: 7MDX)41. The lipoprotein was
fuged for 5 min at 300g. The supernatant was carefully removed from removed, resulting in an apo structure for the simulation. The apo
the erythrocyte pellet and the liquid was discarded. The pellet was protein was then embedded into a symmetric membrane approximat-
further washed 3× in 500 µl saline. The erythrocyte pellet was resus- ing the E. coli inner membrane with the following lipid abundances:
pended in 800 µl of red blood cell buffer (10 mM Na2HPO4, 150 mM 12.5% POPE, 12.5% QMPE, 43% PMPE, 8% OYPE, 10% PMPG, 9% PYPG
NaCl, 1 mM MgCl2, pH 7.4). To a 0.5-ml Eppendorf tube or a PCR plate and 5% PVCL255.
was added 1.0 µl of 30× compound in DMSO and 19 µl RBC buffer. For The protein–membrane system was then solvated with TIP3P water
negative controls, 1.0 µl DMSO was combined with 19 µl RBC buffer. and buffered in 0.15 M NaCl56. Each step of the membrane building
For positive controls 1.0 µl 30% Triton X-100 was combined with 19 µl process was carried out using the Membrane Builder module of
RBC buffer. Tubes were briefly centrifuged. Next, 10 µl of washed eryth- CHARMM-GUI57. The lipid environment was then shuffled using the
rocyte suspension was added to each tube and sealed. After incuba- Membrane Mixer plugin of VMD to generate five initial systems with
tion at 37 °C for 2 h, the samples were centrifuged for 5 min at 300g varying lipid arrangements58. Each system was slowly equilibrated with
and 20 µl of the supernatant was carefully removed and transferred 1 kcal mol−1 Å−2 restraints on lipid headgroups and protein heavy atoms
to the wells of a clear flat-bottomed 384-well plate. Absorbance was over a 2-ns period in the NVT ensemble. After equilibration each replica
measured at 540 nm. was simulated for 500 ns as an NPT ensemble, resulting in a total of
2.5 µs sampling of apo protein.
Selection of resistant mutants To identify distinct protein conformations, r.m.s.d.-based clustering
Resistant mutants were selected using the large inoculum method. In of the transmembrane domains of LolCDE was performed on the full
brief, E. coli BW25113, K. pneumoniae ATCC 27736, or E. cloacae ATCC dataset using the VMD Cluster utility (3-Å cut-off)59. This yielded 10
29893 (~1.0 × 107 CFU) was plated on 100-mm plates of Luria Bertani agar distinct clusters of apo LolCDE, and one representative conformation
containing 8, 16, 32 or 64 µg ml−1 lolamicin. For E. coli ∆tolC ( JW5503), was used for each cluster. Lolamicin was then docked to each of these
~1.0 × 107 CFU was plated on 100-mm plates of Luria Bertani agar con- protein conformations using AutoDock Vina60.
taining 0.063, 0.13, 0.25 or 0.5 µg ml−1 lolamicin. Colonies were visible
after incubating at 37 °C for 48 h. Resistant colonies were confirmed Inhibitor-bound LolCDE simulations
by growing colonies in selective media with the same concentration The top drug–protein pose in which the drug contacts key experimen-
of lolamicin. tally identified residues in each of the ten protein conformations was
chosen for additional molecular dynamics simulations to examine
Sequencing of LolCDE complex their stability. Each lolamicin–protein system was equilibrated with
The lolC, lolD, and lolE genes were amplified by colony PCR. Colo- 1 kcal mol− Å−2 restraints on lipid headgroups, protein heavy atoms, and
nies were picked and diluted in 100 µl of sterile H2O. PCR reactions the inhibitor over a 2-ns period in the NVT ensemble. Then the 10 sys-
were set up by combining 25 µl MiFi Mix (Bioline), 1 µl 20 µM primer tems were independently simulated each for 100 ns with no restraints
mix (N3P-149 (5′-GGTTATGACCCGAGGTGTAATG-3′) and N3P-150 yielding a total of 1 µs of sampling.
(5′-CTACCAGCTCACACACTGCATC-3′)), 10 µl template, and 14 µl H2O. To identify distinct protein–drug conformations, r.m.s.d.-based
The reaction was performed on a C1000 Themal Cycler (Bio-Rad) under clustering of the drug was performed on the full dataset using the
VMD Cluster utility (2.75 Å cut-off) to generate 10 clusters. From the then vortexed and centrifuged to remove the proteins. Supernatants
clustering we highlighted two putative binding sites (BS1 and BS2) were analysed with the QTRAP 5500 LC–MS/MS system (Sciex) in the
and two transiently bound states (TS1 and TS2) of lolamicin. Each of Metabolomics Laboratory of the Roy J. Carver Biotechnology Center,
these 4 bound states was simulated for an additional 500 ns for further University of Illinois at Urbana-Champaign.
sampling. Next, the pharmacokinetic assessment via oral gavage was deter-
All molecular dynamics simulations were performed using NAMD61,62. mined. Lolamicin was formulated in 20% DMSO, 30% water, and 50%
The force field for all simulations was CHARMM3663,64 and timestep was PEG400. Mice were treated with lolamicin (200 mg kg−1) via oral gav-
set to 2 fs. We used a 12 Å cut-off for short-range, non-bonded interac- age, with three mice per time point (0, 15, 30, 45, 60, 120, 240, 480, and
tions, with a switching starting at 10 Å. Long-range electrostatic interac- 1,440 min). At the specified time points, mice were killed, blood was
tions were calculated with the particle mesh Ewald (PME) method with collected and centrifuged, and the serum was frozen at –80 °C until
grid density of 1 Å−3, and PME interpolation of 6 (ref. 65). The SHAKE analysis similar to before. Software Analyst 1.6.2 was used for data
algorithm was used to maintain rigidity for all bonds with hydrogen acquisition and analysis. The 1200 Series HPLC system (Agilent Tech-
atoms66. Langevin thermostat with damping coefficient of 1.0 ps−1 was nologies) includes a degasser, an autosampler and a binary pump. The
used to maintain temperature constant at 310 K. The Nosé–Hoover liquid chromatography separation was performed on an Agilent Zorbax
Langevin piston barostat with a period of 50 fs and a decay of 25 fs SB-Aq column (4.6 × 50 mm; 5 µm) with mobile phase A (0.1% formic
was used to maintain pressure at 1 atm (refs. 67,68). All simulations acid in water) and mobile phase B (0.1% formic acid in acetonitrile). The
were run using flexible cell allowing dimensions of the periodic cell flow rate was 0.35 ml min−1. The linear gradient was as follows: 0–2 min,
to independently change while maintaining x–y plane aspect ratio. 100% A; 8–12 min, 2% A; 12.1–17 min, 100% A. The autosampler was set
Lennard–Jones and PME forces were updated every timestep. at 10 °C. The injection volume was 5 µl. Mass spectra were acquired
under positive electrospray ionization with a voltage of 5,500 V. The
MTD of compound 1 and lolamicin in mouse source temperature was 450 °C. The curtain gas, ion source gas 1 and
The protocol was approved by the Institutional Animal Care and Use ion source gas 2 were 33, 65 and 60 psi, respectively. Multiple reaction
Committee (IACUC) at the University of Illinois at Urbana-Champaign monitoring was used for quantification with external calibration. The
(protocol number: 19181). In these studies, 10- to 12-week-old female limit of quantification of (signal-to-noise ratio (S/N) = 10) was 1 nM.
C57BL/6 mice purchased from Charles River Laboratories were used. Pharmacokinetic parameters were calculated with a one-compartment
The maximum tolerated dose (MTD) of a single compound via intraperi- model using a nonlinear regression program (Phoenix WinNonlin Ver-
toneal injection was determined first. Compound 1 and lolamicin were sion 8.1; Certara).
formulated in 50% DMSO, 50% PEG400. Compound 1 and lolamicin were
given by intraperitoneal injection (n = 3 for each compound at each Acute pneumonia bacterial burden model
given concentration). All the mice were monitored for signs of toxicity The protocol was approved by the Institutional Animal Care and Use
for 2 weeks. For multiple doses, the compound was given by daily intra- Committee (IACUC) at the University of Illinois at Urbana-Champaign
peritoneal injection for 5 consecutive days and the mice were monitored (protocol number: 20256). In these studies, 6- to 8-week-old male CD-1
for signs of toxicity for 1 month. The MTD was the highest dosage with mice purchased from Charles River Laboratories were used. Mice were
acceptable toxicity (for example, <20% weight loss). Compound 1 and randomly chosen and divided into subsequent groups. No additional
lolamicin were well-tolerated as a single dose of 200 mg kg−1. Further randomization was used to allocate the experimental groups; blind-
analysis showed that compound 1 and lolamicin were well-tolerated ing was not performed for subsequent quantification. First, efficacy
with daily dosing of 200 mg kg−1 for 5 consecutive days. was determined by treatment of lolamicin via intraperitoneal injec-
Next, the MTD of lolamicin via oral gavage was determined. Lola- tion. Acclimated CD-1 mice were infected via intranasal inoculation
micin was formulated in 20% DMSO, 30% water and 50% PEG400. of bacteria: E. coli AR0349 at 2.7 × 108 CFU per mouse, K. pneumoniae
Lolamicin was given via oral gavage. All the mice were monitored for AR0040 at 8 × 107 CFU per mouse, or E. cloacae AR0163 at 7.2 × 108
signs of toxicity for two weeks. For multiple doses, compound was CFU per mouse. Infected mice were treated twice a day for 3 days with
given twice daily orally for three consecutive days and the mice were either vehicle, compound 1, or lolamicin (intraperitoneal; 100 mg kg−1).
monitored for signs of toxicity. The MTD was the highest dosage For E. coli infection, mice were treated with vehicle or compound 4, 8,
with acceptable toxicity (for example, <20% weight loss). Lolamicin 24, 32, 48 and 56 h post-infection (n = 8 per group). MIC = 1 µg ml−1 for
at 200 mg kg−1 was well-tolerated twice a day for 3 days. The MTD of lolamicin. For K. pneumoniae infection, mice were treated with vehicle
lolamicin was used to inform the dosing schedule used in subsequent or compound 4, 10, 21, 29, 43 and 51 h post-infection (n = 8 per group).
efficacy studies. MIC = 1 µg ml−1 for lolamicin. For E. cloacae infection, mice were treated
with vehicle or compound 4, 11, 24, 32, 48 and 56 h post-infection (n = 8
Pharmacokinetic assessment of compound 1 and lolamicin per group). MIC = 8 µg ml−1 for lolamicin. Drugs were formulated in
The protocol was approved by the Institutional Animal Care and Use 50% DMSO, 50% PEG400 from solid immediately before treatment. At
Committee (IACUC) at the University of Illinois at Urbana-Champaign 72 h post-infection, CFU values were determined in the lungs through
(protocol number: 19181). In these studies, 10- to 12-week-old female serial dilutions.
C57BL/6 mice purchased from Charles River Laboratories were used. Next, efficacy was determined by treatment of lolamicin via oral
Mice were randomly chosen and divided into subsequent groups. gavage. Acclimated CD-1 mice were infected via intranasal inoculation
No additional randomization was used to allocate the experimental of bacteria: E. coli AR0349 at 2.9 × 108 CFU per mouse. Infected mice
groups; blinding was not performed for subsequent quantification. were treated twice a day for 3 days with either vehicle or lolamicin (oral
First, the pharmacokinetic assessment via intraperitoneal injection was gavage; 200 mg kg−1). Drugs were formulated in 20% DMSO, 30% water
determined. Compound 1 and lolamicin were formulated in 50% DMSO and 50% PEG400 from solid immediately before treatment. At 72 h
and 50% PEG400. Mice were treated with compound 1 (200 mg kg−1) or post-infection, CFU values were determined in the lungs through serial
lolamicin (200 mg kg−1) via intraperitoneal injection, with three mice dilutions. Statistical significance was determined by one-way ANOVA
per time point (0, 15, 30, 45, 60, 120, 240, 480 and 1,440 min). At the with Tukey’s multiple comparisons test.
specified time points, mice were killed, blood was collected and centri-
fuged, and the serum was frozen at −80 °C until analysis. The proteins Bacterial sepsis survival model
in a 10-µl aliquot of serum were precipitated by the addition of 50 µl The protocol was approved by the Institutional Animal Care and Use
methanol with the addition of 10 µl internal standard. The sample was Committee (IACUC) at the University of Illinois at Urbana-Champaign
Article
(protocol number: 20256). In these studies, 6- to 8-week-old male CD-1 the additional parameters: “errorEstimationFunction=PacBioErrfun”,
mice purchased from Charles River Laboratories were used. Mice were “BAND_SIZE = 32”. Default steps were used to denoise reads, and derep-
randomly chosen and divided into subsequent groups. No additional licate into ASVs, followed by taxonomic assignment using the DADA2
randomization was used to allocate the experimental groups; blinding implementation of the RDP Classifier71 and the Silva v138.1 database72
was not performed for subsequent quantification. For the preparation custom formatted by DADA2 developers73 for the optimization of long
of each inoculum, overnight cultures of clinical isolates were diluted read data. Multiple sequence alignment of the resulting ASV sequences
in Luria Bertani broth and grown to log-phase growth at 37 °C. First, was performed by DECIPHER v.2.22.074 followed by a midpoint-rooted75
efficacy was determined by treatment of lolamicin via intraperitoneal phylogenetic analysis using Fasttree v2.1.0 to produce a maximum
injection. Infection was established via a 100-µl retro-orbital injec- likelihood tree used in data analysis steps. Raw counts, taxonomic
tion of bacteria: E. coli AR0349 at 5.2 × 108 CFU per mouse, K. pneu- assignments, and the phylogenetic tree for the ASVs were imported into
moniae BAA-1705 at 5.8 × 107 CFU per mouse, or E. cloacae AR0163 at R v. 4.2.176 using the package phyloseq v. 1.40.077. ASVs were filtered out
8.95 × 108 CFU per mouse. Mice infected with K. pneumoniae BAA-1705 if they were mitochondrial or chloroplast or did not have an assigned
were treated once a day for 3 days with either compound 1 or lolamicin phylum. We then agglomerated ASVs that had phylogenetic tip lengths
(intraperitoneal injection; 200 mg kg−1; n = 15 per group). Mice infected <0.175, added their counts together, and then performed prevalence
with E. coli AR0349 or E. cloacae AR0163 were treated twice a day for 3 filtering with a threshold of 5% (present in 5 samples), leaving 231 ‘taxa’
days with either compound 1 or lolamicin (intraperitoneal injection; for statistical analysis of alpha diversity (Observed species richness,
100 mg kg−1). Compound 1 was formulated in 50% DMSO, 50% PEG400 Chao1, Shannon–Weaver, Simpson’s, inverse Simpson’s and Faith’s
and lolamicin was formulated in 50% DMSO, 50% PEG400 or 20% DMSO, population diversity estimates) and beta diversity (Bray–Curtis, and
30% water, 50% PEG400 from solid immediately before treatment. For Weighted UniFrac estimates).
survival analyses, a Kaplan–Meier log-rank survival test was performed Alpha diversity analyses were performed using the R packages phy-
using GraphPad Prism [Link]. loseq and vegan v. 2.6-278; a standard R-based Wilcoxon rank-sum test
Next, efficacy was determined by treatment of lolamicin via oral was performed to assess differences between control and treatment
gavage. Infection was established via a 100-µl retro-orbital injection of effects. Beta diversity analyses were performed in R using phyloseq and
bacteria: E. coli AR0349 at 6.0 × 108 CFU per mouse. Mice infected with vegan using relative proportion normalization. PERMANOVA tests were
E. coli AR0349 were treated twice a day for 3 days with either vehicle performed on beta diversity values using the vegan function adonis2.
or lolamicin (oral gavage; 200 mg kg−1; n = 15 per group). Lolamicin Differential abundance analysis was performed using normalization
was formulated in 20% DMSO, 30% water and 50% PEG400 from solid methods developed for phyloseq using DESeq2 v. 1.36.079 on ASVs
immediately before treatment. that were agglomerated to the family and genus levels. R code used
to conduct this analysis can be found at: [Link]
Effect of lolamicin on mouse faecal microbiome hergenrother-16S-mouse-2022Sept and [Link]
The protocol was approved by the Institutional Animal Care and Use zenodo.1098065580.
Committee (IACUC) at the University of Illinois at Urbana-Champaign
(protocol number: 20256). In these studies, four-week-old male CD-1 C. difficile challenge
mice purchased from Charles River Laboratories were used. Mice were The protocol was approved by the Institutional Animal Care and Use
randomly chosen and divided into subsequent groups. No additional Committee (IACUC) at the University of Illinois at Urbana-Champaign
randomization was used to allocate the experimental groups; blinding (protocol number: 20256). In these studies, 6-week-old male CD-1 mice
was not performed for subsequent quantification. Mice were dosed purchased from Charles River Laboratories were used. Mice were ran-
twice a day for 3 days with amoxicillin (100 mg kg−1) (GoldBio), clinda- domly chosen and divided into subsequent groups. No additional ran-
mycin (100 mg kg−1) (GoldBio), or lolamicin (200 mg kg−1) via oral gav- domization was used to allocate the experimental groups; blinding was
age, or were treated with vehicle (20% DMSO, 30% water, 50% PEG400) not performed for subsequent quantification. Mice were dosed twice
(n = 6 per group). Faecal pellets were collected right before treatment a day for 3 days with amoxicillin (100 mg kg−1) (GoldBio), clindamycin
(day 0), 4 days after treatment (day 7), 7 days after treatment (day 10) (100 mg kg−1) (GoldBio) or lolamicin (200 mg kg−1) via oral gavage, or
and 28 days after treatment ended (day 31). DNA was isolated from stool were treated with vehicle (20% DMSO, 30% water, 50% PEG400) (n = 6
pellets using the QIAamp DNA Stool Mini Kit. Sequencing of a stool per group). Mice were challenged with C. difficile strain 630 (1.2 × 104
pellet from each mouse before and after treatment was performed spores) via oral gavage 24 h after antibiotic treatment stopped. Once
by PacBio. 16S amplicons were generated and converted to a PacBio the gavages were completed, the remaining spore solution was seri-
library with the SMRTBell Express Template Prep kit 3.0. The library ally diluted and plated to confirm the spore concentration that was
was sequenced on 1 SMRTcell 8M on a PacBio Sequel IIe using the CCS delivered.
sequencing mode and a 30 h movie time. CCS analysis was done using Faecal samples were collected on the day antibiotic treatment was
SMRTLink V11.0 using the following parameters: ccs –min-passes 3 started, the day of the C. difficile challenge, and the following five days.
–min-rq 0.999. Under anaerobic conditions, faecal samples were serially diluted in
16S and metagenomic data analyses were completed by the anaerobic PBS and plated on CCFA plates. After 24 h of anaerobic incu-
High-Performance Computing in Biology (HPCBio) group at the bation at 37 °C, the number of CFUs was determined.
University of Illinois. PacBio data targeting the entire 16S region
were quality-assessed using FASTQC69 and then processed using the Reporting summary
PacBio-adapted version of the TADA Nextflow-based workflow (https:// Further information on research design is available in the Nature
[Link]/h3abionet/TADA/blob/master/[Link]) that implements Portfolio Reporting Summary linked to this article.
a containerized version of the DADA2 (v1.22, R v4.1.1) workflow70 for
dereplicating and denoising reads to generate amplicon sequence vari-
ants (ASVs). In brief, sequences less than or equal to 1,800 nucleotides Data availability
(nt) in size underwent primer sequence removal using primers F27 All data supporting the findings of this study are available within the
(AGRGTTYGATYMTGGCTCAG) and R1492 (RGYTACCTTGTTACGACTT). paper and the Supplementary Information. Raw sequencing data have
Minimal quality trimming was then performed (in which only sequences been deposited at the NCBI Sequence Read Archive under accession
with a minimum of 1,000 nt, a maximum expected errors (EE) score of 2, PRJNA1101557. Source data are provided for Figs. 2, 4 and 5, Extended
and no unclassified bases were retained). Error estimation included Data Figs. 1–4. Source data are provided with this paper.
 74. Wright, E. S. DECIPHER: harnessing local sequence context to improve protein multiple
Code availability sequence alignment. BMC Bioinformatics 16, 322 (2015).
75. Price, M. N., Dehal, P. S. & Arkin, A. P. FastTree 2—approximately maximum-likelihood
Source code for data analysis can be found at [Link] trees for large alignments. PLoS ONE 5, e9490 (2010).
HPCBio/hergenrother-16S-mouse-2022Sept and [Link] 76. R: A Language and Environment for Statistical Computing, v. 4.2.1 (R Foundation for
Statistical Computing, 2019).
10.5281/zenodo.10980655 (ref. 80). 77. McMurdie, P. J. & Holmes, S. phyloseq: an R package for reproducible interactive analysis
and graphics of microbiome census data. PLoS ONE 8, e61217 (2013).
78. Oksanen, J. et al. vegan: Community ecology package. [Link]
55. Pandit, K. R. & Klauda, J. B. Membrane models of E. coli containing cyclic moieties in the package=vegan (2017).
aliphatic lipid chain. Biochim. Biophys. Acta 1818, 1205–1210 (2012). 79. Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion
56. Jorgensen, W. L., Chandrasekhar, J., Madura, J. D., Impey, R. W. & Klein, M. L. Comparison for RNA-seq data with DESeq2. Genome Biol. 15, 550 (2014).
of simple potential functions for simulating liquid water. J. Chem. Phys. 79, 926–935 80. Holmes, J. HPCBio/hergenrother-16S-mouse-2022Sept: Initial release. [Link]
(1983). 10.5281/zenodo.10980656 (2024).
57. Jo, S., Kim, T., Iyer, V. G. & Im, W. CHARMM-GUI: a web-based graphical user interface for 81. Parker, E. N. et al. Implementation of permeation rules leads to a FabI inhibitor with
CHARMM. J. Comput. Chem. 29, 1859–1865 (2008). activity against Gram-negative pathogens. Nat. Microbiol. 5, 67–75 (2020).
58. Licari, G., Dehghani-Ghahnaviyeh, S. & Tajkhorshid, E. Membrane Mixer: a toolkit for
efficient shuffling of lipids in heterogeneous biological membranes. J. Chem. Inform.
Model. 62, 986–996 (2022). Acknowledgements The authors thank L. Li for LC/MS-MS analysis; L. Dirikolu for the
59. Humphrey, W., Dalke, A. & Schulten, K. VMD: visual molecular dynamics. J. Mol. Graph. pharmacokinetic data analysis; A. Hernandez, C. Wright, M. Band, E. Hogan and J. Drnevich for
14, 27–38 (1996). 33-38. PacBio sequencing; and R. Rajabi-Toustani and the Core Facilities at the Carl Woese Institute
60. Trott, O. & Olson, A. J. AutoDock Vina: improving the speed and accuracy of docking with for Genomic Biology for assistance with confocal imaging. We thank the University of Illinois
a new scoring function, efficient optimization, and multithreading. J. Comput. Chem. 31, and the NIH (AI136773 and P41-104601) for funding this work. K.A.M. was a member of the NIH
455–461 (2010). Chemistry–Biology Interface Training Grant (T32-GM136629). R.J.U. is supported by an NIH
61. Phillips, J. C. et al. Scalable molecular dynamics on CPU and GPU architectures with Ruth Kirschstein Award (F31AI161953) and was a NSF predoctoral fellow. This study used
NAMD. J. Chem. Phys. 153, 044130 (2020). computational resources provided by the Delta Advanced Computing and Data Resource,
62. Phillips, J. C. et al. Scalable molecular dynamics with NAMD. J. Comput. Chem. 26, which is supported by the National Science Foundation (award OAC 2005572) and the State of
1781–1802 (2005). Illinois. Delta is a joint effort of the University of Illinois Urbana-Champaign and its National
63. Hart, K. et al. Optimization of the CHARMM additive force field for DNA: improved Center for Supercomputing Applications. This work used the Anvil system at Purdue University
treatment of the BI/BII conformational equilibrium. J. Chem. Theor. Comput. 8, 348–362 through allocation MCA06N060 from the Advanced Cyberinfrastructure Coordination
(2012). Ecosystem: Services & Support (ACCESS) programme, which is supported by National Science
64. Klauda, J. B. et al. Update of the CHARMM all-atom additive force field for lipids: Foundation grants 2138259, 2138286, 2138307, 2137603 and 2138296.
validation on six lipid types. J. Phys. Chem. B 114, 7830–7843 (2010).
65. Darden, T., York, D. & Pedersen, L. Particle mesh Ewald: an N·log(N) method for Ewald Author contributions K.A.M. and P.J.H. conceived this study. K.A.M. designed and synthesized
sums in large systems. J. Chem. Phys. 98, 10089–10092 (1993). all novel compounds described and utilized in this Article. K.A.M. and R.J.U. designed and
66. Ryckaert, J. P., Ciccotti, G. & Berendsen, H. J. C. Numerical integration of the Cartesian carried out biological experiments. K.A.M. performed aerobic and anaerobic MIC assays,
equations of motion of a system with constraints: molecular dynamics of n-alkanes. frequency of resistance and time–kill kinetics studies. R.J.U. performed confocal microscopy,
J. Comput. Phys. 23, 327–341 (1977). co-culture competition experiments, accumulation and frequency of resistance studies. A.K.V.,
67. Martyna, G. J., Tobias, D. J. & Klein, M. L. Constant-pressure molecular-dynamics M.S. and P.-C.W. conducted molecular dynamics simulations. J.R.H. and C.J.F. performed
algorithms. J. Chem. Phys. 101, 4177–4189 (1994). bioinformatics and statistical analyses on long read 16S microbiome data. H.Y.L., C.-C.H. and
68. Feller, S. E., Zhang, Y. H., Pastor, R. W. & Brooks, B. R. Constant-pressure molecular- G.W.L. conducted studies with mouse models. K.A.M. and P.J.H. wrote the manuscript with
dynamics simulation—the Langevin piston method. J. Chem. Phys. 103, 4613–4621 input from R.J.U., A.K.V., M.S., P.-C.W. and E.T. All authors approved the final draft of the
(1995). manuscript.
69. Andrews, S. Fast QC: a quality control tool for high throughput sequence data. http://
[Link]/projects/fastqc/ (2016). Competing interests The University of Illinois has filed patents on compounds described in
70. Callahan, B. J. et al. DADA2: High-resolution sample inference from Illumina amplicon this Article on which K.A.M. and P.J.H. are named inventors.
data. Nat. Method 13, 581–583 (2016).
71. Lan, Y., Wang, Q., Cole, J. R. & Rosen, G. L. Using the RDP classifier to predict taxonomic Additional information
novelty and reduce the search space for finding novel organisms. PLoS ONE 7, e32491 Supplementary information The online version contains supplementary material available at
(2012). [Link]
72. Quast, C. et al. The SILVA ribosomal RNA gene database project: improved data processing Correspondence and requests for materials should be addressed to Paul J. Hergenrother.
and web-based tools. Nucleic Acids Res. 41, D590–D596 (2013). Peer review information Nature thanks Gerard Wright and the other, anonymous, reviewer(s)
73. McLaren, M. R. Silva SSU taxonomic training data formatted for DADA2. Zenodo https:// for their contribution to the peer review of this work.
[Link]/10.5281/zenodo.3986799 (2020). Reprints and permissions information is available at [Link]
Article
a. MIC (µg/mL)
Strain 1 2
E. coli ATCC 25922 8 8
E. coli AR-0549 > 32 16
K. pneumoniae ATCC 27736 > 32 16
K. pneumoniae ATCC BAA-1705 > 32 > 128
K. pneumoniae AR-0040 > 32 32
E. cloacae ATCC 29893 > 32 > 128
E. cloacae AR-0163 > 32 64
E. cloacae BAA-2468 > 32 32
E. coli ΔtolC JW5503 0.06 0.03
E. coli ΔrfaC JW3596 1 0.25

b.
O

Accumulation (nmol/1012 CFUs)

H
N
O N 500
HN O

F N 400
N N
300

1 2 200

Rotatable bonds: 5 3 100

Globularity: 0.113 0.116
Primary Amine: No No 0
Accumulation: 286 ± 106 34 ± 7 1 2

Extended Data Fig. 1 | Previously identified LolCDE inhibitors. (a) Antibiotic Accumulation determined via previously reported accumulation assay33 and
assessment of compounds 1 and 2 against a panel of gram-negative pathogens reported in nmol per 1012 colony-forming units (CFUs). Line indicates average
performed as part of this study. MICs were performed in Mueller Hinton broth accumulation of low accumulating control antibiotics novobiocin, fusidic acid,
per CLSI guidelines and are reported in µg/mL. All experiments were performed erythromycin, and rifampicin. Data shown represents the average of three
in biological triplicate. (b) eNTRy rule parameters (Rotatable bonds, Globularity, independent experiments with standard deviation of the mean.
Functional Group) of compounds 1 and 2 calculated using eNTRyway33,81.
Extended Data Fig. 2 | See next page for caption.
Article
Extended Data Fig. 2 | Growth competition of lolamicin resistant mutants in biological triplicate. Measurements were compared using a two-sample
generated in E. coli BW25113 as compared to wild-type (WT) E. coli BW25113. Welch’s t-test (one-tailed test, assuming unequal variance). NS, not significant
Fitness of E. coli isolates that harbor the mutations in lolC (a) or lolE (b) conferring (P ≥ 0.05); * (P < 0.05); ** (P ≤ 0.01). LolC-E195K t = 24 hr (P = 0.01), t = 48 hr
resistance to lolamicin at concentrations 32-fold above the MIC (64 µg/mL) (P = 0.03); LolC-E255D t = 24 hr (P = 0.87), t = 48 hr (P = 0.05); LolC-Q258P
were evaluated for growth in culture relative to the parental strain. Bacteria t = 24 hr (P = 0.20), t = 48 hr (P = 0.35); LolC-M262I t = 24 hr (P = 0.44), t = 48 hr
were cultured in cation-adjusted Mueller Hinton broth at a starting density (P = 0.18); LolC-N265k t = 24 hr (P = 0.79), t = 48 hr (P = 0.55); LolE-D264N t = 24 hr
of 5 × 103 CFUs/mL and grown for 48 h at 37 °C. At time = 0 hr, 24 hr, and 48 hr, (P = 0.57), t = 48 hr (P = 0.24); LolE-L199P t = 24 hr (P = 0.22), t = 48 hr (P = 0.02);
cultures were serially diluted and plated on LB agar and LB agar containing LolE-I206N t = 24 hr (P = 0.16), t = 48 hr (P = 0.42); LolE-F367S t = 24 hr (P = 0.14),
8 µg/mL lolamicin to quantify number of wild-type and lolamicin resistant t = 48 hr (P = 0.04). ǂLolE F367S displayed morphological changes and smaller
mutants. Each competition between mutant and the parental strain was assessed colonies compared to wild-type E. coli BW25113.
 a.
E. coli BW25113

10 9
Untreated control

Log(CFU/mL)
10 7 Lolamicin (4X MIC)
Ciprofloxacin (4X MIC)
10 5

10 3

10 1
0 2 4 6 8 10
Time (h)

b.
K. pneumoniae ATCC 27736

10 9 Untreated control
Lolamicin (4X MIC)
Log(CFU/mL)

10 7 Ciprofloxacin (4X MIC)

Tetracycline (4X MIC)
10 5

10 3

10 1
0 2 4 6 8 10
Time (h)

c.

E. cloacae ATCC 29893

10 9
Untreated control
Log(CFU/mL)

10 7 Lolamicin (4X MIC)

Ciprofloxacin (4X MIC)
10 5

10 3

10 1
0 2 4 6 8 10
Time (h)
Extended Data Fig. 3 | Time-kill kinetics of lolamicin against gram-negative K. pneumoniae ATCC 27736 growth. (c) The effect of lolamicin and ciprofloxacin
pathogens. (a) The effect of lolamicin and ciprofloxacin on E. coli BW25113 on E. cloacae ATCC 29893 growth. All experiments were performed in biological
growth. (b) The effect of lolamicin, tetracycline, and ciprofloxacin on triplicate and are represented as mean ± s. e. m.
Article

Extended Data Fig. 4 | Lolamicin induces cell swelling in wild-type E. coli mutants was quantified. Length and width were measured in ImageJ and cell area
and K. pneumoniae but not in lolamicin-resistant mutants. Confocal calculated using the area formula for an ellipse (A = π*ab where a = ½ length and
microscopy of (a) E. coli BW25113; lolamicin-resistant mutants, (b) BW25113 b = ½ width). Measurements were compared using two-sample Welch’s t-test
LolC-N265K, (c) BW25113 LolE-D264N; (d) K. pneumoniae ATCC 27736; and (one-tailed test, assuming unequal variance). NS, not significant (P > 0.05);
lolamicin-resistant mutants, (e) LolC-Q258L and (f) LolE-V59L. Scale bar is *** (P < 0.0005). E. coli BW25113: lolamicin (P = 3.44 × 10 −18), globomycin
10 µm. Antibiotics were tested at the following concentrations (3X MIC or just (P = 1.00 × 10 −15); E. coli BW25113 LolC-N265K: lolamicin (P = 0.28), globomycin
below the solubility limit for lolamicin in resistant mutants): E. coli—DMSO 2%; (P = 4.16 × 10 −10); E. coli BW25113 LolE-D264N: lolamicin (P = 0.09), globomycin
lolamicin 8 µg/mL for E. coli BW25113, 64 µg/mL for resistant strains; globomycin (P = 4.44 × 10 −12). K. pneumoniae ATCC 27736: lolamicin (P = 3.59 × 10 −17),
24 µg/mL; mecillinam 0.4 µg/mL; aztreonam 0.1 µg/mL. K. pneumoniae— globomycin (P = 3.22 × 10 −16); K. pneumoniae ATCC 27736 LolC-Q258L: lolamicin
DMSO 2%; lolamicin 3 µg/mL for K. pneumoniae ATCC 27736 or 64 µg/mL for (P = 0.22), globomycin (P = 1.26 × 10 −8); K. pneumoniae ATCC 27736 LolE-V59L:
resistant strains; globomycin 64 µg/mL; mecillinam 3 µg/mL; aztreonam lolamicin (P = 0.24), globomycin (P = 2.59 × 10 −11).
1.5 µg/mL. Cell size (n = 25) in E. coli (g) and K. pneumoniae (h) and resistant
Extended Data Fig. 5 | See next page for caption.
Article
Extended Data Fig. 5 | Conformational landscape of LolCDE. (a) High- final (opaque purple) conformations of LolC loop demonstrate change in
frequency residue/lolamicin contact probability. (b) Probability density of binding pocket shape upon mutation. Donor-acceptor heavy atom distance
lolamicin’s center of mass locations projected onto two reaction coordinates: of 2.69 Å is highlighted for final conformation. Density plot of salt-bridge
distances to BS1 and BS2 (blue). Overlaid color traces show compound 3 distance between LolC E/K195-LolE D264 from WT and mutant simulations
unbinding in five simulation replicates, highlighting consistent and immediate highlights salt bridge formation with E195K peak density. (g) Lolamicin bound
unbinding from BS1. (c) Reaction coordinates (RCs) used to project free energy in BS1. Conformational sampling of lolamicin rendered as a density with a
landscape of transporter. Two orientation-based RCs, opening of the TMD single conformer (stick). Mutation sensitive LolE residues D264 and I268 (density
periplasmic region (α) and opening of the TMD intracellular region (β), and two and stick), and distant Q198 (stick) shown. (h) LolE-N265K mutation relative to
distance-based RCs, distance between the nucleotide binding domains (d NBD), lipid bilayer. Also highlighted is LolC-G357 for which the N/K265-door bar
and distance between periplasmic domains (d PD), were used for conformation distance was measured. Wider distribution for mutant state demonstrates
projections. (d) Free energy landscape projected onto RCs. Red dots indicate instability of nearby binding pocket. (i) Root-mean squared fluctuation (RMSF)
position of starting Cryo-EM structure (7MDX). (e) Lolamicin-resistant values from replica simulations of LolE residues 357 to 377 calculated from
mutations overlap with binding pocket for lipoprotein substrates of LolCDE. multiple simulation replicas for WT and F367S mutants. F/S367 is marked with a
Predicted luminal tunnel displayed as gray surface. (f) Molecular rendering of dashed line. Increased RMSF for mutant is apparent.
LolC K195-LolE D264 salt bridge interaction. Initial (transparent purple) and
Extended Data Fig. 6 | Bacterial composition of mouse fecal microbiota treated with vehicle (20% DMSO, 30% water, 50% PEG400, n = 6 biologically
obtained by full-length 16 s rRNA sequencing at the Class level. Taxonomic independent animals) or compound (clindamycin, 100 mg/kg; amoxicillin,
analysis showing bacterial population shifts over a 31-day period before (Day 0) 100 mg/kg; or lolamicin, 200 mg/kg; n = 6 biologically independent animals
and after (Day 7, Day 10, and Day 31) administration of antibiotic. CD-1 mice were for each compound) twice a day for three days via oral gavage.
Article

Extended Data Fig. 7 | Bacterial composition of mouse fecal microbiota treated with vehicle (20% DMSO, 30% water, 50% PEG400, n = 6 biologically
obtained by full-length 16 s rRNA sequencing at the Order level. Taxonomic independent animals) or compound (clindamycin, 100 mg/kg; amoxicillin,
analysis showing bacterial population shifts over a 31-day period before (Day 0) 100 mg/kg; or lolamicin, 200 mg/kg; n = 6 biologically independent animals
and after (Day 7, Day 10, and Day 31) administration of antibiotic. CD-1 mice were for each compound) twice a day for three days via oral gavage.
Extended Data Fig. 8 | PCoA ordination of Bray-Curtis dissimilarity values treated with vehicle (20% DMSO, 30% water, 50% PEG400, n = 6 biologically
for samples before (day 0) and after (day 7, day 10) lolamicin, amoxicillin, independent animals) or compound (clindamycin, 100 mg/kg; amoxicillin,
and clindamycin administration. Vehicle-treated samples are included as 100 mg/kg; or lolamicin, 200 mg/kg; n = 6 biologically independent animals
negative controls. Samples are grouped by day and color coded as shown in the for each compound) twice a day for three days via oral gavage.
key. Points represent individual samples for each grouping. CD-1 mice were
 nature portfolio | reporting summary
Corresponding author(s): Paul J. Hergenrother
Last updated by author(s): Apr 18, 2024

Reporting Summary
Nature Portfolio wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency
in reporting. For further information on Nature Portfolio policies, see our Editorial Policies and the Editorial Policy Checklist.
Please do not complete any field with "not applicable" or n/a. Refer to the help text for what text to use if an item is not relevant to your study.
For final submission: please carefully check your responses for accuracy; you will not be able to make changes later.

Statistics
For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.
n/a Confirmed
The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement
A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly
The statistical test(s) used AND whether they are one- or two-sided
Only common tests should be described solely by name; describe more complex techniques in the Methods section.

A description of all covariates tested

A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons
A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient)
AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals)

For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted
Give P values as exact values whenever suitable.

For Bayesian analysis, information on the choice of priors and Markov chain Monte Carlo settings
For hierarchical and complex designs, identification of the appropriate level for tests and full reporting of outcomes
Estimates of effect sizes (e.g. Cohen's d, Pearson's r), indicating how they were calculated
Our web collection on statistics for biologists contains articles on many of the points above.

Software and code

Policy information about availability of computer code
Data collection ImageJ (Fiji) was used to measure the pixels/area of bacteria from confocal imaging.
Autodock Vina was used for docking studies with lolamicin and LolCDE PDB 7MDX.
CHARMM-GUI was used with with membrane builder module and membrane mixer plugin of VMD for simulations.
Molecular dynamic simulations performed in NAMD.

Data analysis Custom code:

Source code for data analysis used in microbiome study is available on Github ([Link]
mouse-2022Sept) and Zenodo (HTML: [Link] DOI: 10.5281/zenodo.10980656)

Commercial software used for data analysis:

Phoenix WinNonlin Version 8.1 was used for data analysis for PK studies
GraphPad Prism 10.2.0. was used for data analysis for MIC panels, compound accumulation in bacteria, MTD, efficacy studies, resistance
March 2021

mutants studies, time-kill kinetics, competition studies, plotting confocal measurements, and Clostridioides difficile challenge model.
Excel (Microsoft 365) was used for MIC determination and mammalian cell toxicity (percent cell death).
OriginPro 2020 (OriginLab, Northampton, MA) was used for data analysis in determining mammalian cell toxicity (IC50).
MNova was used for NMR analysis.
For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors and
reviewers. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Portfolio guidelines for submitting code & software for further information.

1
 nature portfolio | reporting summary
Data
Policy information about availability of data
All manuscripts must include a data availability statement. This statement should provide the following information, where applicable:
- Accession codes, unique identifiers, or web links for publicly available datasets
- A description of any restrictions on data availability
- For clinical datasets or third party data, please ensure that the statement adheres to our policy

Source code for data analysis can be found at Github ([Link] and Zenodo (HTML: [Link]
doi/10.5281/zenodo.10980655; DOI: 10.5281/zenodo.10980656). Raw sequencing data from microbiome experiments can be found at https://
[Link]/sra/PRJNA1101557. The authors declare that all other data supporting the findings of this study are available within the paper and the
Supplementary Information.

Human research participants

Policy information about studies involving human research participants and Sex and Gender in Research.

Reporting on sex and gender not applicable

Population characteristics not applicable

Recruitment not applicable

Ethics oversight not applicable

Note that full information on the approval of the study protocol must also be provided in the manuscript.

Field-specific reporting
Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.

Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see [Link]/documents/[Link]

Life sciences study design

All studies must disclose on these points even when the disclosure is negative.

Sample size No statistical methods were used to predetermine samples size for MTD. Sample size was selected based on literature precedence and
sample sizes needed to achieve statistically significant results. For bacterial burden and mortality studies, sample size was calculated based on
80% power to detect a difference in the means between the compound-treated and vehicle-treated groups, based on a t-test with a two-
sided 0.05 level of significance and with expected outcome difference of 50% between these two groups.

Data exclusions No data exclusions.

Replication We have three or more biological replicates as indicated in the manuscript for MIC determination, mammalian cell toxicity (percent cell death
and IC50), compound accumulation studies, fitness of resistance mutants, competition studies, confocal microscopy, and spontaneous
frequency of resistance in bacteria. For MTD and PK studies three mice were used per group. For survival studies, 15 mice were used for each
group. For bacterial burden studies, 8 mice were used per group. For microbiome studies and C. difficile challenge studies, 6 mice per group
were used. All replicates were reported and were successful.

Randomization Mice were randomly assigned to treatment or vehicle groups.

Blinding Investigators were not blinded as it was not relevant to treatment outcomes or data analysis for MTD, bacterial burden, survival, microbiome,
or C. difficile challenge experiments.
March 2021

Reporting for specific materials, systems and methods

We require information from authors about some types of materials, experimental systems and methods used in many studies. Here, indicate whether each material,
system or method listed is relevant to your study. If you are not sure if a list item applies to your research, read the appropriate section before selecting a response.

2
 Materials & experimental systems Methods

nature portfolio | reporting summary

n/a Involved in the study n/a Involved in the study
Antibodies ChIP-seq
Eukaryotic cell lines Flow cytometry
Palaeontology and archaeology MRI-based neuroimaging
Animals and other organisms
Clinical data
Dual use research of concern

Eukaryotic cell lines

Policy information about cell lines and Sex and Gender in Research
Cell line source(s) HFF-1 and A-549 cells were obtained from ATCC.

Authentication Cell lines were authenticated externally by commercial vendor and were inspected visually in-house.

Mycoplasma contamination Cell lines were not verified to be free of mycoplasma contamination.

Commonly misidentified lines No commonly misidentified species used.

(See ICLAC register)

Animals and other research organisms

Policy information about studies involving animals; ARRIVE guidelines recommended for reporting animal research, and Sex and Gender in
Research

Laboratory animals For MTD and PK studies, 10- to 12-week-old female C57BL/6 mice purchased from Charles River were used. For efficacy studies and
Clostridioides difficile challenge, six- to eight-week-old male CD-1 mice purchased from Charles River were used. For microbiome
study, four-week-old male CD-1 mice purchased from Charles River Laboratories were used. These mice were housed in positively
ventilated microisolator cages equipped with automatic recirculating water, situated in a room featuring laminar, high-efficiency
particle accumulation-filtered air. The animals were provided with autoclaved food, water, and bedding". The animal room operates
on an alternating 12-hours light/dark cycle, with ambient temperature set to 68 C (range 67 C - 71 C) and humidity at 28% (range
17% - 59%)

Wild animals This study did not involve wild animals.

Reporting on sex Sex-based analysis was not performed. This study was not intended to test for sex differences.

Field-collected samples This study did not involve samples collected from the field.

Ethics oversight The protocols were approved by the IACUC at the University of Illinois at Urbana-Champaign. (MTD and PK protocol number: 19181;
efficacy studies, microbiome study, and Clostridioides challenge protocol number: 20256).
Note that full information on the approval of the study protocol must also be provided in the manuscript.

March 2021

This checklist template is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give
appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in
the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by
3
statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit [Link]