1

Bacterial Survival Guide: Metabolic Pathways Leading to Antibiotic Tolerance in Staphylococcus aureus

by Eliza Zalis

Bachelor of Arts, Providence College

A dissertation submitted to

The Faculty of the College of Science of Northeastern university

in partial fulfillment of the requirements for the degree of Doctor of Philosophy

April 10th, 2019

Dissertation directed by

Kim Lewis University Distinguished Professor of Biology

2

Acknowledgements

I owe a debt of gratitude to my advisor Kim Lewis for his mentorship over the past five years. I

entered his lab with no microbiology or research experience and he allowed me to do

microbiology research anyway. I remain grateful for his trust and have benefitted from his insight

and leadership. Kim has shown me the importance of addressing questions that are both

important and interesting, which is a habit I hope to practice for the rest of my life.

I thank the members of my dissertation committee, including Veronica Godoy, Win Chai, Eddie

Geisinger, and Tim van Opijnen for their guidance. Their advice has shaped this project and it has

been a pleasure to learn from them.

I am also grateful for each member of the Lewis Lab, past and present. I have spent five years

doing research alongside excellent scientists and I thank them for their support and inspiration.

My labmates have helped me learn how to tackle big questions by focusing on the specifics while

still maintaining broad perspective. I thank in a special way Sarah Rowe, Brian Conlon, Austin

Nuxoll, Autumn Brown-Gandt, Yeva Yue Shan, Pooja Balani, Bijaya Sharma, and Phil Strandwitz

for their discussion and advice during my first year in lab. I also thank the Persister SistersTM,

especially Yeva Shan, David Cameron, Austin Nuxoll, Sylvie Manuse, Nadja Leimer, Jeff Quiqley,

Gabriel Fox, Samantha Nicolau, and Michael Gates for their insight and friendship. My labmates

have been incredible colleagues but have also become great friends. It has been a true joy to

work with them.

3

I thank the teachers who have guided my education, from kindergarten to high school biology

and through my undergraduate and graduate study. I have pursued science because of their

instruction and encouragement.

I thank the Biology Department of Northeastern University for creating an environment of

productive learning. I am grateful for the friends I have made in the department and the support

of my fellow PhD students.

I have enjoyed serving on the board of Graduate Women in Science and Engineering at

Northeastern and the Boston Bacterial Meeting Organizing Committee. These experiences have

offered opportunities for service and outreach and have enriched my experience as a scientist in

Boston. Through these groups I have met new collaborators, mentors, and friends and I have

enjoyed every part of getting to know these people. I thank them for their dedication to

community engagement and science communication.

I also thank my friends outside of lab. They make my life better and I am grateful for their humor,

understanding, and love.

I am the person I am today because of my family. I am grateful for their endless support and love.

Any success I ever achieve is because of them. I thank my parents Sharon and Roy for encouraging

curiosity and creativity. I thank my siblings Eva, Joe, and Gretchen for their perspective and

humor. My grandmother Agnes has been a role model in education and in life and I am grateful

for her example. I thank my aunts, uncles, and cousins for their discussions about science and

beyond.

4

Abstract of Dissertation

Persisters are rare cells in a bacterial population that are able to tolerate antibiotic treatment [1,

2]. Antibiotic tolerance is distinct from antibiotic resistance. Although both tolerance and

resistance result in antibiotic treatment failure, they do so by very different mechanisms which

must be understood separately in order to develop effective treatments for bacterial infections.

Antibiotic resistance occurs when bacteria acquire genetic mutations that allow them to grow in

the presence of antibiotics. Tolerance occurs when bacterial cells undergo a phenotypic switch

to a non-growing state [3]. Most antibiotics kill bacteria by binding to targets and interrupting

active cellular processes. Non-growing cells have fewer active targets for antibiotics to corrupt,

rendering antibiotics less effective. In clinical practice, antibiotics prescribed to treat bacterial

infections are often ineffective in completely eradicating a bacterial population. Surviving

persister cells can eventually resume growth, causing relapsing and recurrent infections.

Staphylococcus aureus is a notorious human pathogen responsible for pneumonia, endocarditis,

osteomyelitis, toxic shock syndrome, skin and soft tissue infections, and infection of implanted

devices. This research uncovers the mechanism by which S. aureus forms persister cells. We find

that low ATP levels lead to increased antibiotic tolerance. S. aureus is known to rely on the TCA

cycle and amino acids derived from host tissues to fuel growth during infection [4]. We show that

defects in the TCA cycle and catabolism of amino acids leads to low intracellular ATP and

increased antibiotic tolerance. We propose that within a population, natural fluctuation in gene

expression leads to phenotypic heterogeneity. We hypothesize that cells with low levels of

metabolic gene expression compared to the bulk of the population generate less ATP and are

5

better able to survive antibiotic treatment. We sort individual cells with low levels of TCA gene

expression and show that they are indeed more tolerant to antibiotic treatment than the rest of

the population. We conclude that stochastic gene expression leads to heterogeneity in

phenotypic antibiotic susceptibility and gives rise to persister cells.

We next investigate persister cell resuscitation. We knew that low TCA cycle activity and low ATP

levels in S. aureus lead to high tolerance. We therefore expected that resuscitation from the

persister state would involve cellular metabolism. We conduct a broad screen of a library of S.

aureus mutants to see what genes are involved in growth resumption following antibiotic

treatment. We identify 262 genes implicated in resuscitation including genes involved in carbon

and nitrogen metabolism, nucleotide synthesis, and translation.

We also seek to understand the prevalence of metabolic gene defects in clinical isolates of S.

aureus. We assemble a collection of isolates from patients with endocarditis, osteomyelitis, skin

and soft tissue infections, and atopic dermatitis. We identify isolates with high tolerance

compared to wild type strains. We perform whole genome sequencing of each isolate and

compare sequences to a reference strain to identify variants. We expected that antibiotic usage

would select for mutations that facilitate tolerance and hypothesized that TCA cycle mutations

would be prevalent in clinical isolates. We indeed found multiple high-impact nucleotide

polymorphisms in TCA genes in S. aureus clinical isolates.

This work describes the mechanism of antibiotic tolerance in S. aureus. Defects in the TCA cycle

cause low intracellular ATP levels and high tolerance to multiple classes of antibiotics. We identify

mutations in TCA cycle genes in clinical isolates of S. aureus. We demonstrate that natural

6

fluctuation in TCA cycle gene expression yields heterogeneity in phenotypic antibiotic

susceptibility.

7

Table of Contents

Acknowledgements 2

Abstract of Dissertation 4

Table of Contents 7

List of Figures 11

List of Tables 12

Glossary of Terms 12

Chapter 1. Introduction 15

1.1 Persister Cells 15

1.2 Antibiotic Tolerance versus Resistance 15

1.3 Chronic and Recurrent Infection 16

1.4 Staphylococcus aureus: Commensal and Pathogen 17

Chapter 2. Persister formation in Staphylococcus aureus is associated with ATP depletion 20

2.1 Abstract 20

2.2 Results 21

2.3 Materials and Methods 33

2.3.1 Bacterial strains and growth conditions 33

2.3.2 Strain Constructions 34

2.3.3 Persister Assays 35

8

2.3.4 Arsenate and rifampicin persister assays 35

2.3.5 Flow cytometry and FACS analysis using gfp reporters 35

2.3.6 Proteomic analysis 37

2.3.7 Real-Time qRT-PCR 37

2.3.8 ATP Assays 38

2.4 Contributions 38

2.5 Supplemental Information 39

Chapter 3. Stochastic variation in expression of the TCA cycle produces persister cells 45

3.1 Abstract 46

3.2 Introduction 47

3.3 Results 47

3.4 Discussion 55

3.5 Materials and Methods 57

3.5.1 Bacterial strains, culture conditions, and strain construction 57

3.5.2 Proteomic sample preparation 57

3.5.3 Proteomics and data analysis 58

3.5.4 Persister assays 59

3.5.5 ATP quantification of bulk culture 59

3.5.6 Construction of S. aureus HG003 expressing QUEEN2m 59

3.5.7 Microscopy 60

3.5.8 Single -cell ATP quantification using QUEEN 60

3.5.9 FACS analysis using GFP reporters 61

9

3.6 Acknowledgements 61

3.7 Supplemental Information 62

Chapter 4. Persister Resuscitation 70

4.1 Abstract 70

4.2 Introduction 70

4.3 Results 71

4.3.1 Growth resumption 71

4.3.2 NTML screen 73

4.4 Discussion 81

4.5 Material and Methods 82

4.5.1 Strains and culture conditions 83

4.5.2 Resuscitation screen 83

4.5.3 Data analysis 84

Chapter 5. Clinical isolates of S. aureus harbor TCA cycle mutations 84

5.1 Abstract 85

5.2 Introduction 85

5.3 Results 86

5.4 Methods 92

5.4.1 S. aureus isolate sources 92

5.4.2 Whole genome sequencing and read mapping 92

5.4.3 MIC and persister experiments 93

10

5.5 Discussion 93

5.6 Supplemental Information 96

Chapter 6. Dissertation conclusion and future directions 101

6.1 Summary 101

6.2 Ongoing research and future directions 103

6.2.1 Persister resuscitation 103

6.2.2 Noise-quenching 104

6.2.3 Eradicating persister cells 105

References Cited 106

11

List of Figures

Figure 1: Model of Persistence.

Figure 2.1: Toxin-antitoxin modules and stringent response do not control persister formation in

S. aureus.

Figure 2.2: Activation of stationary markers is heterogeneous.

Figure 2.3: Persister sorting using stationary markers Pcap5A and ParcA.

Figure 2.4: Reduction in ATP induces persister formation and expression of stationary phase

markers.

Figure 3.1: TCA cycle enzyme abundance increases in stationary phase.

Figure 3.2: TCA cycle mutants have lower ATP levels than wild type.

Figure 3.3: S. aureus mutants lacking functional late TCA cycle genes exhibits increased antibiotic

tolerance.

Figure 3.4: Fluorescence-activated cell sorting after antibiotic treatment yields enrichment for

persister cells in populations expressing relatively low levels of TCA cycle genes.

Figure 4.1: Antibiotic killing and growth resumption after antibiotic inactivation.

Figure 3.2: Post-treatment persister resuscitation screen output

Figure 5.1: Variant type classified as deletion, insertion, nonsynonymous SNP, or synonymous

SNP.

12

Figure 5.2: Biological subsystems implicated in genome variance analysis.

Figure 5.3: High-impact variants in TCA cycle genes.

List of Tables

Table 4.1: S. aureus mutants with significantly faster resuscitation after ciprofloxacin treatment

compared to the plate average.

Table 4.2: S. aureus mutants with significantly slower resuscitation after ciprofloxacin treatment

compared to the plate average.

Table 5.1: S. aureus clinical isolate source information and diagnosis.

Glossary of Terms

ADEP: acyldepsipeptide

Amp: ampicillin

Ars: arsenate

ATP: adenosine triphosphate

Cam: chloramphenicol

CA-MRSA: community-associated methicillin-resistant Staphylococcus aureus

CCCP: Carbonyl cyanide m-chlorophenylhydrazine

13

CFU: colony forming unit

Cip: ciprofloxacin

Ctrl: control

Exp: Exponential

FACS: fluorescence-activated cell sorting

Gent: gentamicin

GFP: green fluorescent protein

KEGG: Kyoto Encyclopedia of Genes and Genomes

LBB/A: Lysogeny Broth/Agar

MHB/A: Mueller-Hinton Borth/Agar

MIC: Minimum Inhibitory Concentration

MOPS: 3-(N-morpholino) propanesulfonic acid

MRSA: Methicillin-resistant Staphylococcus aureus

MSSA: Methicillin-sensitive Staphylococcus aureus

NTML: Nebraska transposon mutant library

OD: optical density

Ox: oxacillin

14

PATRIC: Pathosystems Resource Integration Center

PBS: phosphate buffered saline

PCR: Polymerase chain reaction

QUEEN: quantitative evaluator of cellular energy

RAST: Rapid Annotations using Subsystems Technology

Rpm: rotations per minute

Stat: Stationary

TA: Toxin-antitoxin

TCA: tricarboxylic acid cycle

Tn-seq: Transposon sequencing

TSB/A: Tryptic Soy Broth/Agar

Unk: unkown

WT: wild type

15

Chapter 1: Introduction

1.1 Persister Cells

Persister cells are rare bacterial cells within a population that are able to tolerate antibiotic

treatment. Upon exposure to antibiotics, they survive but do not grow. Persister cells were

first described in 1944 by Joseph Bigger. Bigger found that treating a population of bacteria

with penicillin killed most of the cells, but a small portion were able to survive the antibiotic

treatment [1]. He observed resumed growth of the bacteria when penicillin was no longer

present in the culture. Bigger proposed that his failure to sterilize a culture in vitro might have

clinical implications; he proposed that these persister cells might contribute to antibiotic

treatment failure in patients with bacterial infections. This discovery of persister cells went

largely unexplored for fifty years, but in recent years has experienced a resurgence in the

fields of microbiology and infectious disease. Multiple microbial species have been shown to

produce persister cells, including Staphylococcus aureus, Escherichia coli, Candida albicans,

Pseudomonas aeruginosa, Mycobacterium tuberculosis, Salmonella Typhimurium, and

Borrelia burgdorferi [5-10]. Persister cells typically account for a small fraction of a

community but allow a population of bacteria to tolerate treatment with multiple types of

antibiotics.

1.2 Antibiotic Tolerance and Resistance

Antibiotic resistance is an important problem but it does not account for all cases of antibiotic

treatment failure. A patient can suffer from an infection caused by a pathogen that is drug-

susceptible by laboratory tests, but still experience treatment failure. The phenomenon of

resistance is separate from antibiotic tolerance. Resistance occurs when genetic mutations

16

develop in a bacterial population that enable cells to grow even in the presence of that

antibiotic. Efflux pumps and genetic mutations that alter the structure of an antibiotic’s

molecular target can confer resistance [11]. Persister cells are phenotypic variants within a

population. These antibiotic tolerant cells are genetically susceptible to antibiotics but are

not killed by antibiotics. Figure 1 illustrates the phenomenon of antibiotic tolerance. It has

recently been shown that antibiotic tolerance facilitates the development of resistance. Since

tolerant cells are more likely to survive antibiotic treatment, they are more likely to develop

resistance mutations [12]. Antibiotic resistance and tolerance are fundamentally distinct but

result in the same outcome of antibiotic treatment failure.

1.3 Chronic Infection

The clinical implications of antibiotic tolerance are manifold. Persister cells that tolerate

antibiotic treatment can resume growth after antibiotics are no longer present in the

environment. In the clinical setting, resumed bacterial growth means repopulation of the

infection site and recurrent infection. A patient with a recurrent and relapsing infection would

be subject to multiple hospital visits and rounds of treatment.

Antibiotic tolerance has been implicated in multiple cases of infectious disease. Cystic fibrosis

is one example of persister cells in chronic infection. Patients with cystic fibrosis suffer from

Pseudomonas aeruginosa infections and undergo years of antibiotic treatment, but

treatment is not effective in eradicating the infection. When researchers tested the persister

levels of longitudinal isolates taken from the lungs of cystic fibrosis patients, they found that

samples collected during later stages of disease progression exhibited higher levels of

persister cells than those taken earlier [7], suggesting that selection for genetic mutations

17

that promote tolerance occurs in vivo. Additionally, Mycobacterium tuberculosis, Candida

albicans, and Escherichia coli have been found to exhibit tolerance in cases of recurrent

infection [5, 13, 14].

Finding new treatments for chronic infections would decrease costs in health care and

improve patient outcomes. The aggregate cost of hospital-acquired infections in the United

States is estimated to be more than $15 billion each year [15, 16]. Staphylococcus aureus is

one of the most common human pathogens and is notorious for causing recalcitrant

infections, often associated with the formation of biofilms [17]. Hospital-associated

infections are frequently caused by S. aureus. Often, patients experience chronic and

recurrent S. aureus infections while battling other diseases that require hospital stays. Elderly

and immunocompromised individuals are especially likely to develop serious S. aureus

infections and often require complicated and costly treatment [15, 16].

1.4 Staphylococcus aureus: Commensal and Pathogen

Staphylococcus aureus is a human commensal and a pathogen. An estimated 30% of humans

are colonized asymptomatically, but S. aureus can cause endocarditis, osteomyelitis,

bacteremia, toxic shock syndrome, and pneumonia, in addition to skin and soft tissue

infections. Colonization is a risk factor for developing a S. aureus infection [18], as is

hospitalization [19, 20]. Staphylococcus aureus causes more than ten thousand deaths in the

U.S. each year [21] and even non-fatal cases impose a significant burden on health care

facilities. S. aureus is widely acknowledged as one of the most pernicious agents of healthcare

associated infections [22-24]. Patients often suffer from relapsing infection, especially in

18

cases of indwelling medical devices. S. aureus is a notorious biofilm-former, capable growing

on cardiovascular implants, joint replacement devices, and catheters [25]. Biofilms contain

cells embedded in an exopolymeric matrix and provide protection from immune factors and

phagocytosis [26]. S. aureus biofilms are heterogeneous. Nutrients and signaling molecules

diffuse throughout the geography of the biofilm, promoting diverse phenotypes [27]. Cells

can disperse from a biofilm to establish infection at remote body sites.

Staphylococcus aureus employs various strategies to survive the host immune response.

Pathogenic S. aureus produces a number of toxins and virulence factors that interfere with

neutrophil activation, chemotaxis, and adhesion to epithelial cells [28, 29]. S. aureus also

evades targeted production of antimicrobial peptides and reactive oxygen species [30]. One

of the major weapons against invading S. aureus is the production of nitric oxide within

activated phagocytes. Unlike other Staphylococcal species, S. aureus is able to produce NAD+

to maintain glycolytic flux and redox balance by fermenting glucose via lactate

dehydrogenase and can therefore survive nitric oxide stress. [31-34] Increasing investigation

of the microenvironment of S. aureus infection has improved our understanding of the

complicated battle between pathogen and host.

19

Figure 1: Model of Persistence. After addition of an antibiotic to a culture of

growing bacteria, most of the population will die (represented at 5 hours), but

persisters survive in the presence of antibiotics. The “persister plateau” represents

the relatively stable population of persisters, which can persist for long persiod of

treatment. If a population acquires resistance to an antibiotic, the population will

grow in the presence of the antibiotic (dashed grey line).

20

Chapter 2: Persister formation in Staphylococcus aureus is associated with ATP depletion

This chapter includes work published in the following published article. [35]

Conlon, B. P., Rowe, S. E., Gandt, A. B., Nuxoll, A. S., Donegan, N. P., Zalis, E. A., ... & Lewis, K.

(2016). Persister formation in Staphylococcus aureus is associated with ATP depletion. Nature

microbiology, 1(5), 16051.

Antibiotic resistance is a major human health problem[36]. However, most pathogens that cause

hard to treat chronic infections are not drug resistant[2, 17, 37]. There is mounting evidence that

drug-tolerant persister cells contribute to this phenomenon [5, 7, 10, 13, 38]. Persister cells are

phenotypic variants that survive lethal doses of antibiotics and are genetically identical to their

drug susceptible kin. The mechanism of persister formation has been extensively studied in the

closely related Gram-negative organisms Escherichia coli and Salmonella Typhimurium [2, 39,

40]. In E. coli, isolated persisters express toxin/antitoxin (TA) modules[41], most of which code

for mRNA endonucleases called interferases [42]. While deletion of individual interferases has no

phenotype, a knockout of ten TAs produced a decrease in persisters in both a growing culture

and in stationary phase4. A small fraction of persisters forms in E. coli when cells stochastically

express the HipA toxin12. HipA is a protein kinase17 which phosphorylates glutamyl aminoacyl-

tRNA synthetase, inhibiting protein synthesis18,19. Selection for increased drug tolerance in

vitro led to the identification of a hipA7 mutant allele that produces up to 1000-fold more

persisters than the wild type6. We recently identified hipA7 strains among patients with chronic

urinary tract infections12. Similarly, hip mutants are common among isolates of P.

aeruginosa from patients with cystic fibrosis11, and from patients with chronic Candida

21

albicans infections20. In S. Typhimurium, TA modules are responsible for a sharp increase in

persisters when the pathogen infects macrophages9. These findings provide a link between

persisters and clinical manifestation of disease.

Little is known about the mechanism of persister formation in Gram-positive species. We first

sought to examine the role of TAs in persister formation in S. aureus. There are three known type

II TAs in S. aureus- mazEF, and relBE homologues axe1/txe1, and axe2/txe2. These are the only

three TAs predicted in S. aureus 8325, the parental strain of HG001 and HG003. An additional

phage associated toxin-antitoxin was identified in S. aureus Newman using the TAfinder too but

overexpression of the potential toxin did not inhibit growth (Supplemental Figure 2.1). Therefore

we continued with analysis of the three active type II TAs. The toxins from all three modules are

RNA endonucleases[43]. We constructed a triple knockout in the TAs (Δ3TA), and examined the

strain’s ability to form persisters. Ciprofloxacin causes a characteristic biphasic killing of wild

type S. aureus with a subpopulation of surviving persisters (Figure 2.1). Unexpectedly, knockout

of all TAs had no effect on the level of persisters in exponentially growing or stationary phase

cells (Figure 2.1a). A similar result was obtained with oxacillin, vancomycin and rifampicin (Figure

2.1b). It remains possible that these TAs or as yet unannotated TAs play a role in persister

22

formation under a specific environmental condition, but we see no evidence of a role for the TAs

we examined, in persister formation under regular growth conditions.

Figure 2.1: Toxin-antitoxin modules and stringent response do not control persister formation

in S. aureus. The contribution of (A-B) toxin-antitoxin modules, mazEF, axe1-txe1 and axe2-txe2,

in strain Newman and (C) the stringent response element rsh in strain HG001and d, the stringent

response regulator, codY, in strain SH1000 to persister formation in S. aureus. Strains were grown

for 4 hours to mid-exponential phase (exp) or overnight to stationary (stat) phase in MHB and

challenged with either ciprofloxacin (cip), vancomycin (vanc), oxacillin (ox) or rifampicin (rif) (10×

MIC). Aliquots were removed at indicated time points, washed and plated to enumerate

survivors. All experiments were performed in biological triplicates. Standard deviations (SD) are

indicated.

A B

C D

23

It is known that S. aureus exhibits complete tolerance to many antibiotics at stationary state

which is another important distinction between this pathogen and E. coli[38, 44]. It appears

that S. aureus cells in a stationary state exhibit antibiotic tolerance similar to persisters. We

reasoned that persisters in exponential phase may be cells that have entered the stationary

phase early. To examine this we used two reporters of the stationary phase. The promoter of the

capsular polysaccharide operon, Pcap5A, has been shown to be activated in the stationary

phase[45, 46]. Increase in relative fluorescence of a strain carrying Pcap5A-GFP over time in a

growing culture confirmed the suitability of this promoter as a marker of the stationary phase

(Figure 2.2A-B). The promoter of the arginine deiminase pathway, ParcA, was used as a second

It is known that S. aureus exhibits complete tolerance to many antibiotics at stationary state

which is another important distinction between this pathogen and E. coli[38, 44]. It appears

that S. aureus cells in a stationary state exhibit antibiotic tolerance similar to persisters. We

reasoned that persisters in exponential phase may be cells that have entered the stationary

phase early. To examine this we used two reporters of the stationary phase. The promoter of the

capsular polysaccharide operon, Pcap5A, has been shown to be activated in the stationary phase

[45, 46]. Increase in relative fluorescence of a strain carrying Pcap5A-GFP over time in a growing

culture confirmed the suitability of this promoter as a marker of the stationary phase (Figure

2.2A-B). The promoter of the arginine deiminase pathway, ParcA, was used as a second marker,

since proteomic analysis showed that the ArcA protein accumulates specifically in the stationary

phase, increasing in abundance 10.5-fold relative to exponential phase. Analysis of ParcA fused

to gfp confirmed that this promoter is activated specifically in stationary phase (Supplemental

Figure 2.2). Real-time qRT-PCR analysis showed that transcript levels of cap5A and arcA increase

24

3.88 and 25.38-fold in stationary phase, respectively. These promoters were inserted upstream

of gfpuvr in plasmid pALC1434 to yield Pcap5A::gfp and ParcA::gfp.

Figure 2.2: Activation of stationary markers is heterogeneous. (A) Growth (OD600) and (B) GFP

expression of HG003 Pcap5A::gfp over time. The blue lines represent entrance into stationary

A B

C D

E

E

25

phase. Distribution of GFP signal in (C) Pcap5A::gfp and d, ParcA::gfp at hourly intervals. The cut-

off for the bright fraction is represented by a blue line. This cut-off represents the level of

expression in a stationary phase culture. (E) A subpopulation of stationary phase cells, defined as

cells with stationary phase levels of expression of ParcA and Pcap5A, is always present and

increases with population density. The blue line represents an estimation of the entrance into

stationary phase. All experiments were performed in biological triplicates. SD are indicated. C-D

are representative of one replicate.

Flow cytometry was then used to track cells expressing high levels of the stationary phase

markers (termed bright) at hourly intervals from early exponential to stationary phase (Figure

2.2C-D). We found that a subpopulation of cells expresses stationary markers in early exponential

phase, and their frequency increases with the rise in the density of the population (Figure 2.2E).

This suggests that stationary phase does not initiate in a uniform manner but is a heterogeneous

process.

We next sought to determine if the subpopulation of stationary phase cells in a growing culture

were in fact persisters. For this, we employed Fluorescence-Activated Cell Sorting (FACS). S.

aureus HG003 Pcap5A::gfp or HG003 ParcA::gfp were grown to mid-exponential or stationary

phase and analyzed by FACS (Figure 2.3A-B). In order to examine whether the bright cells were

persisters, the exponential phase culture was exposed to a lethal dose of ciprofloxacin (10× MIC)

for 24h. The culture was then re-analyzed by FACS, and cells were gated into bright,

middle and dim populations based on expression of Pcap5A::gfp, or ParcA::gfp (Figure 2.3A-B).

Cells were then sorted onto MH agar in 96 spots to enumerate survivors from each population

26

(32 spots for each population: bright, middle, dim). The lethal dose of ciprofloxacin causes ~3 logs

of killing in the total culture, so cells were sorted onto MH agar plates at 1, 10, 100, 1000, and

5000 per spot to achieve viable counts for each population (representative plate, Figure 2.3C-D).

The bright population had 100–1000 fold more survivors than the middle and dim populations

27

with both markers. We chose to compare only the middle and bright fractions for quantification

as the dim fraction had < 100% sorting efficiency (Figure 2.3E-F).

A B

C D

E

F

28

Figure 2.3: Persister sorting using stationary markers Pcap5A and ParcA. Expression of (A)

Pcap5A::gfp or b, ParcA::gfp in exponential following ciprofloxacin challenge (grey peak) and

stationary phase (green peak) measured by FACS. Exponential phase cells were gated into 3

populations depending on expression of GFP - dim (pink peak), middle (orange peak) or bright

(red peak - cells expressing stationary phase levels of reporter in exponential phase). (C-D) Cells

were sorted based on dim, middle or bright GFP expression onto MHA plates at 1000 events/spot

for both Pcap5A::gfp and ParcA::gfp. Representative plates are shown. Survivors from each

population of HG003 or Δcap5A harboring (E), Pcap5A::gfp and (F), ParcA::gfp were counted

following incubation overnight at 37°C. The asterisks indicate statistical significance between

middle and bright populations, determined using Student’s t-test (P ** < 0.005 or P***<0.0005).

All experiments were performed in biological triplicates. SD are indicated. A-D are representative

of one replicate.

To determine if expression of capsular polysaccharide contributes to ciprofloxacin tolerance, we

transformed plasmid Pcap5A::gfp into a cap5A mutant strain and repeated the cell sorting

experiment. Disrupting the cap5A gene did not alter the expression profiles of

Pcap5A::gfp (Supplemental Figure 2.3). Similarly, the bright cells in a cap5A mutant also

exhibited a 100-fold enrichment for cells tolerant to ciprofloxacin in exponential phase compared

to the middle fraction showing that entry into stationary phase rather than levels of the CapA

protein affect persister formation (Figure 2.3E). We also examined persister formation in

an arcA mutant and found it to be similar to the wild-type strain (Supplemental Figure 2.4). As a

control for stationary phase reporters, we repeated the experiment using a promoter that is also

expressed in exponential phase (Pspa::gfp). In this case, the bright population had no enrichment

29

of persisters compared to the middle of the population (Supplemental Figure 2.3). This shows

that expression of a stationary marker, rather than expression of GFP per se, determines whether

a cell is a persister.

We wanted to further examine any potential role for the stringent response and tested

expression of the persister markers in the rshsyn mutant background.

Neither cap5A nor arcA promoter activity were significantly affected by mutation

of rshsyn (Supplemental Figure 2.2). We reasoned that a decrease in the energy level of the cell

in stationary phase could lead to antibiotic tolerance. Killing by bactericidal antibiotics results

from corrupting active targets1. Aminoglycosides kill by causing mistranslation, which leads to

the production of toxic peptides [47]; fluoroquinolones inhibit the re-ligation step of DNA gyrase

and topoisomerase, causing double strand breaks [48], and β-lactams lead to a futile cycle of

peptidoglycan synthesis and autolysis [49]. A decrease in ATP would decrease the activity of ATP

dependent antibiotic targets such as gyrase, topoisomerase, and RNA polymerase, leading to

antibiotic tolerance, and ATP has previously been suggested to impact survival to antibiotics [50-

53].

We examined ATP levels of an exponential and stationary phase population and indeed found

that ATP levels decrease significantly in the stationary phase (Figure 2.4A). We then found that

emulating stationary phase ATP levels in an exponential phase population by decreasing it with

arsenate resulted in a 325-fold induction in persister formation (Figure 2.4B). ATP levels are

lowered by arsenate as it forms a rapidly-hydrolysable ADP-As, producing a futile cycle [54].

Interestingly, we found that stationary phase-specific promoters were also activated in response

30

to arsenate (Figure 2.4C). Hence, these promoters are activated in the stationary phase as ATP

levels in the cells drop. The Pcap5A and ParcA promoters then enable single-cell detection of

ATP, linking a decrease in the energy level to antibiotic tolerance in individual persisters. It was

clear that cells with reduced ATP levels are antibiotic tolerant and express markers of this

phenotypic state. What remained unclear was whether a transcriptional response was necessary

for persister formation. To examine this, we again induced persister formation with arsenate,

however, we also included a 15 minute pre-incubation with rifampicin at 1× MIC, which was

sufficient to inhibit induction of stationary markers (Figure 2.4C) but did not cause cell death

(Supplemental Figure 2.5). Inhibition of transcription did not impede persister induction (Figure

2.4D) (Supplemental Figure 2.5). This shows that although a specific transcriptional response that

includes expression of Pcap5A and ParcA is induced in response to low ATP, this response is not

required for antibiotic tolerance. Rather, tolerance of both stationary populations and persisters

can be explained by a drop in ATP which will result in a decrease in the activity of drug targets.

To further test whether ATP levels determine persister formation, we examined killing in a

medium where ATP concentration is expected to increase. Supplementing TSB medium with

31

glucose increased ATP significantly and resulted in a 100 fold reduction in persisters

(Supplemental Figure 2.6).

Figure 2.4: Reduction in ATP induces persister formation and expression of stationary phase

markers. (A) Titering arsenate to produce stationary phase levels of ATP. Arsenate was added to

an exponential phase population of S. aureus for 15 minutes before measuring ATP. (B) Decrease

of ATP results in a 325-fold induction of persisters in exponential phase. On the x-axis, “–“

indicates cell count before addition of ciprofloxacin and “+” represents cell count after 24 hour

incubation in 10 × MIC of ciprofloxacin (0.4 µg/ml) C. Pcap5A::gfp and ParcA::gfp are induced by

depletion of ATP. (D) inhibition of transcriptional response by addition of 0.1 µg/ml rifampicin

15 minutes prior to ATP depletion (30 minutes of 1 mM arsenate), with results represented as

A B

C D

32

the Log % survival after 24 hours ciprofloxacin treatment. All experiments were performed in

biological triplicates. SD are indicated.

Promoters of arcA and cap5A are induced when ATP drops in stationary phase or in the presence

of arsenate. Cells expressing these markers are highly enriched for persisters. Low ATP can lead

to tolerance of a stationary culture, and explains antibiotic tolerance of a persister sub-

population. This work links the phenomena of population-wide tolerance and persister cell

tolerance. A growing population contains cells that enter into stationary state early, and these

become antibiotic tolerant persisters. Persisters form as cells lose ATP. The entrance into

stationary state is stochastic, with the frequency of persisters increasing with cell density. Our

measurements of ATP in single persister cells by FACS have been performed with two different

reporters, ParcA-GFP and Pcap5A-GFP. Both are ATP sensors, but the detection requires

transcription and translation of GFP. To establish direct causality, it would be interesting to

perform single cell detection of ATP in persisters more directly, such as with a FRET-based sensor

[55], once it is adapted to S. aureus.

Interestingly, tolerance to clinically relevant daptomycin was also observed in stationary phase

[56]. Also, a recent study shows that altered levels of inorganic phosphate and polyphosphate in

daptomycin tolerant cells, which could also be related to depletion of ATP39. A recent study

shows that population heterogeneity and capsular polysaccharide expressing sub-populations

also occur in vivo in persistent carriers of S. aureus [46]. The role of ATP levels in recalcitrance

of S. aureus infection should be examined and ATP levels of cells during infection may be an

important determinant of the outcome of infection.

33

Understanding how persisters form will improve our ability to control chronic infections. We

recently identified a compound capable of killing persisters, acyldepsipeptide (ADEP4). ADEP4

targets ClpP and converts it into a non-specific protease, which forces both growing and dormant

cells to self-digest [57]. Importantly, ADEP4 dissociates the protease from its ATP-dependent

chaperones and the dysregulated proteolysis does not require ATP. In combination with

rifampicin (to decrease resistance development), ADEP4 eradicated a biofilm both in vitro and in

a mouse model of a chronic S. aureus infection. This shows that persisters can be killed by a

compound which does not require an ATP-dependent target. In this regard, it is interesting to

note that stationary cells of S. aureus exhibit considerable tolerance to daptomycin, a

membrane-acting antibiotic [56, 57]. Why dormant cells would be tolerant to this compound is

an interesting problem that remains to be solved.

This study suggests that a new mechanism of persister formation, loss of energy leading to drug

tolerance, operates in S. aureus. It is possible that this is a general mechanism of tolerance which

governs persister formation in other bacteria as well.

2.3 Materials and Methods

2.3.1 Bacterial strains and growth conditions

S. aureus were cultured in Mueller-Hinton broth (MHB) or Tryptic Soy Broth (TSB) with or without

added glucose. TSB and TSB without glucose was buffered to pH 7.0 using 100mM MOPS. Bacteria

were routinely grown at 37°C at 225 rpm Media were supplemented with chloramphenicol 10

34

µg/ml to maintain plasmids where necessary. MSSA strains Newman, SH1000, and HG001 were

used to analyze the role of TA modules and stringent response as mutations of interest had

previously been constructed and characterized in these backgrounds [43, 58, 59]. The model

strain HG003 was used for all subsequent experiments. For E. coli experiments, growth of the

overexpression strain was compared to an empty vector control in a plate reader over 16 hours

at 37°C in LB medium supplemented with 0.2% arabinose.

2.3.2 Strain Constructions

For construction of reporter plasmids, primers Pcap5A_1 (gcgcgaattctctatctgataataatcatc)

Pcap5A_2 (ggcctctagactaatgtactttccattatt), Pspa_1 (gcaggaattctttccgaaattaaacctcagc) Pspa_2

(gcagtctagaattaataccccctgtatgta) and ParcA_1 (gcgcgaattcaaaatgtatattttgaccca) and ParcA_2

(ggcctctagatctatttcctccttttatct) flanked by restriction sites EcoRI and XbaI were used to amplify

predicted promoter sequences of cap5A, spa and arcA, respectively. The promoter regions were

cloned upstream of gfpuvr into the EcoRI and XbaI sites of plasmid pALC1434 [60]. A Newman

strain was created containing deletions for all three known type II toxin-antitoxin systems

(Newman ΔTA3). Using Newman ∆mazEF(ALC4072) as a starting strain,

the axe1/txe1 and axe2/txe2 operons were deleted by sequential allelic exchange using the

pMAD plasmids pALC6480 and pALC648143, respectively. Deletion of these genes was verified by

PCR analysis and chromosomal DNA sequencing. For hypothetical toxin overexpression, the

primers Ptox_1 gcgcgaattcatggaagaaactttaa and Ptox_2 gcgcggtaccttatgcaatttaaaaa were used to

amplify the toxin and the fragment was digested with EcoRI and KpnI and cloned into the pBAD33

35

vector upstream of an arabinose inducible promoter, digested with the same restriction

enzymes.

2.3.3 Persister Assays

Strains were grown to mid-exponential or stationary phase (~16h) in MHB in 14 ml round bottom

snap-cap culture tubes. Cells were plated for CFU counts and challenged with antibiotics

ciprofloxacin, rifampicin, vancomycin, gentamicin or oxacillin (4.0 µg/ml, 0.4 µg/ml, 10 µg/ml and

1.5 µg/ml respectively) At indicated times, an aliquot of cells was removed, washed with 1% NaCl,

and plated to enumerate survivors. All experiments were performed in biological triplicates.

Averages and standard deviations are representative of three biological replicates. Rifampicin

resistant mutants arise spontaneously at the frequency of ~2.3 × 10−8. Rifampicin killing in

exponential phase selected for the proliferation of rifampicin resistant mutants, which had

repopulated the exponential phase cultures by 24h (Supplemental Figure 2.5). For this reason,

levels of persisters tolerant to rifampicin were examined in stationary phase only.

2.3.4 Arsenate and rifampicin persister assays

Strains were grown to mid-exponential phase in MHB media. Where indicated, rifampicin 0.01

µg/ml was added for 15 minutes and/or arsenate 1 mM for 30 minutes prior to ciprofloxacin

challenge for 24h (10× MIC).

2.3.5 Flow cytometry and FACS analysis using gfp reporters

Fluorescent protein level was analyzed with a BD Aria II flow cytometer (BD Biosciences) with a

70-micron nozzle. Cell population was detected using forward and side scatter parameters (FSC

36

and SSC), and fluorescence was analyzed with emitting laser of 488 nm and bandpass filter of

525/15 nm, using a FACS ARIA II (Becton Dickinson, CA). Strains harboring plasmids Pcap5A::gfp,

ParcA::gfp or Pspa::gfp were grown to mid-exponential and stationary phase in MHB containing

10 µg/ml chloramphenicol. For growth curve construction, the population was gated so that over

90% of the stationary phase population were designated ‘bright’. These gates were applied to all

timepoints. At each timepoint, cfu was measured and the number of stationary phase cells was

calculated by multiplying the percentage of cells in the bright fraction by the total cell number.

An overnight culture was sub-cultured 1:100 into fresh MHB and grown for 3 hours. 300 µl of this

was added to 3 ml of MHB to begin the growth curve. This sub-culturing step removed any carry-

over of stationary phase cells from the stationary phase culture. For FACS analysis of persisters,

strains were exposed to ciprofloxacin for 24h. Before the challenge, an aliquot of the culture was

diluted and plated for cfu. Challenged cells were washed and plated to enumerate survivors. Cells

pre- and post- antibiotic challenge were analyzed by FACS. A gate was drawn based on stationary

phase expression of Pcap5A::gfp or ParcA::gfp. Exponential phase cells expressing stationary

phase levels of Pcap5A::gfp or ParcA::gfp were termed ‘bright’. Two gates were drawn within the

exponential phase Pcap5A::gfp expression peak and termed ‘middle’ and ‘dim’ respectively. To

calculate the percent survival of each population following antibiotic challenge, first, we

calculated the sorting efficiency from each population prior to antibiotic challenge. Events (cells)

from each population were sorted in a 96-well format with 32 spots for each population; dim,

middle and bright. 1 event per spot (for 32 spots) and colonies were counted following

incubation. For the middle and bright fractions we achieved 100% sorting efficiency (32 colonies),

however the sorting efficiency for the dim fraction was lower, ~90% or 29 colonies. This indicated

37

that not all events in the dim fraction were cells. For this reason we chose to focus on the

differences between the bright population and the middle or bulk of the population. Following

antibiotic challenge, cells (events) from each population were sorted onto MH agar plates in a

96-well format at 1, 10, 100, 1000, 5000 per spot (32 spots / population) to enumerate survivors.

A similar method was applied for all reporters. Ciprofloxacin treatment did not affect expression

of any reporters used in this study. A control experiment was performed where samples were

sonicated for 5 minutes in a sonicating water bath prior to cell sorting. Sonication had no impact

on the sorting results, confirming that cell aggregation was not influencing FACS experiments.

Cells were analyzed and sorted using FACS-Diva software. Figures were generated using FlowJo

software. Experiments were performed in triplicate. Error bars represent the standard deviations

of the means, and statistical significance was determined by the Student’s t test.

2.3.6 Proteomic analysis

Biological duplicates were grown in MHB and harvested in the mid-exponential and stationary

phase of growth. Samples were labelled and fractionated and mass spectrometry was performed

as previously described [38].

2.3.7 Real-Time qRT-PCR

RNA was isolated from exponential phase population after 4 hours of growth and stationary

phase after 16 hours of growth using a QIAGEN® RNA purification kit. Samples were treated with

Turbo DNase and RNA integrity was confirmed on a bioanalyzer. Reverse transcriptase was used

to convert to cDNA as per manufacturer’s instructions. Serial 10-fold dilutions of genomic DNA

were used to construct standard curves for each set of primers. qRT-PCR was performed using

38

SYBR® green enzyme. Fold change was calculated based on the cycle number required to achieve

a predesignated quantity of signal normalized to a 16S rRNA control.

2.3.8 ATP Assays

ATP levels of stationary and mid-exponential cultures with the addition of various concentrations

of arsenate were measured using a Promega BacTiter Glo kit according to the manufacturer’s

instructions.

2.4 Contributions

We would like to thank Dr. Christiane Wolz for the gift of the HG001, HG001 rshsyn and triple

mutant rshsyn, relP, relQ strains. We thank Dr. Richard Lee and Dr. Michael LaFleur for critical

discussions. This work was supported by NIH grant R01AI110578 to KL and by a Charles A. King

fellowship to BC.

39

2.5 Supplemental Information

Supplemental Figure 2. 1. Overexpression of the hypothetical phage associated toxin has no

effect on growth of E. coli. The gene encoding the hypothetical toxin, NWMN_0265 from S.

aureus Newman was overexpressed in E. coli MG1655 in vector pBAD33. The toxin was cloned

downstream of an arabinose inducible promoter and grown in the presence of 0.2% inducer

and growth was compared to that of an empty vector control at 37°C in Luria Bertoni (LB) broth.

Data represent biological triplicates. Error bars represent standard deviation.

A B

40

Supplemental Figure 2.2. Promoter activity in a rshsyn mutant background. (A) Promoters of

cap5A and arcA are specifically induced upon the onset of stationary phase with GFP expression

increasing as (B) growth ceases at the onset of stationary phase. Expression of Pcap5A and

ParcA is not affected by mutation of rshsyn. The blue line represents an estimation of the

entrance to stationary phase state. Data is an average of 3 biological replicates.

Supplemental Figure 2.3. Pcap5A activity is not affected by mutation of cap5A and cells

expressing Pspa::gfp in exponential phase are not enriched for persisters. (A) Pcap5A::gfp

expression in HG003 wild type (grey peak) and the cap5A mutant (blue peak). Strains were

grown to mid- exponential phase and analyzed by FACS. (B) S. aureus HG003 Pspa::gfp

expression in exponential phase following ciprofloxacin treatment (grey peak) and stationary

A B

C

41

phase (green peak) measured by FACS. Exponential phase cells treated with ciprofloxacin

were gated to dim (purple peak), middle (orange peak) and bright (red peak) expression of

GFP. (C) Survivors from each population were sorted onto MHA plates and enumerated

following incubation overnight at 37ºC. (A) and (B) are representative experiments. (C) is the

average of 3 biological replicates and error bars represent standard deviation.

Supplemental Figure 2.4. Mutation of arcA does not have an impact on persister formation

in S. aureus HG003. Wild-type and ∆arcA cells were grown to mid-exponential phase and

challenged with 10 x MIC of ciprofloxacin. Cfus were recorded at 24 and 48 hours. Data is an

average of 3 biological replicates and error bars represent standard deviation.

42

Supplemental Figure 2.5. Arsenate protects against killing by ciprofloxacin. Killing by

ciprofloxacin (4.0 µg/ml) in the presence of rifampicin (0.01 µg/ml) and/or arsenic acid (1 mM).

Rifampicin was added where indicated 15 minutes before the start of the experiment. Arsenate

was added, where indicated 30 minutes before the start of the experiment. Ciprofloxacin was

added where indicated at t = 0. Survivors were enumerated after 16 and 24h exposure. Data is

averaged from 3 biological replicates and error bars represent standard deviation.

43

Supplemental Figure 2.6. Increased ATP results in fewer persisters. (A) ATP levels were

measured after 3 hours of growth in TSB and TSB without glucose. (B) Survival of cells in TSB and

TSB without glucose after treatment with ciprofloxacin at 10 x MIC. Results are the average of 3

biological replicates and error bars represent standard deviation.

Supplemental Figure 2.7. Rifampicin resistance emerges in exponential phase. S. aureus

HG003 was grown to mid-exponential phase and rifampicin was added at t=0 to 10 x MIC (0.4

µg/ml). Cfus were counted at various timepoints. After an initial decline, the culture rebounded

over 24 hours due to a high frequency of resistance. Results are an average of 3 biological

replicates and error bars represent standard deviations.

44

Chapter 3: Stochastic variation in expression of the TCA cycle produces persister cells

This chapter contains work from the following manuscript, which has been submitted for publication:

Stochastic variation in expression of the TCA cycle produces persister cells

Eliza A. Zalis*, Austin S. Nuxoll*, Sylvie Manuse, Geremy Clair, Lauren C. Radlinski, Brian P. Conlon, Kim

Lewis

3.1 Abstract

Chronic bacterial infections are difficult to eradicate, though they are caused primarily by drug-

susceptible pathogens [2]. Antibiotic-tolerant persisters largely account for this paradox. In spite of their

significance in recalcitrance of chronic infections, the mechanism of persister formation is poorly

understood. We previously reported that a decrease in ATP levels leads to drug tolerance in Escherichia

coli, Pseudomonas aeruginosa, and Staphylococcus aureus [35, 61, 62]. We reasoned that stochastic

fluctuation in the expression of TCA cycle enzymes can produce cells with low energy levels. S. aureus

knockouts in glutamate dehydrogenase, 2-oxoketoglutarate dehydrogenase, succinyl CoA synthetase,

and fumarase have low ATP and exhibit increased tolerance to fluoroquinolone, aminoglycoside, and -

lactam antibiotics. FACS analysis of TCA genes shows a broad Gaussian distribution in a population, with

over three orders of magnitude difference in the levels of expression between individual cells. Sorted

cells will low levels of TCA enzyme expression have an increased tolerance to antibiotic treatment. These

findings suggest that fluctuation in expression of energy generating components serve as a mechanism

of persister formation.

3.2 Introduction

Persister cells are rare phenotypic variants that are able to survive antibiotic treatment [2]. Unlike

resistant bacteria, which have specific mechanisms to prevent antibiotics from binding to their targets,

45

persisters evade antibiotic killing by entering a tolerant non-growing state. Persisters have been

implicated in chronic infections in multiple species [5, 7, 10, 17] and growing evidence suggests that

persister cells are responsible for many cases of antibiotic treatment failure [17].

Toxin-antitoxin (TA) modules and the stringent response have been proposed as mechanisms of

antibiotic tolerance, based primarily on studies of E. coli. However, these findings have recently been

challenged [62, 63]. Similarly, we reported that knocking out all TA modules and (p)ppGpp synthases in

S. aureus had no effect on persister formation [35]. In search of an alternative mechanism, we found

that persister cells in a growing population express stationary cell markers, cap5A and arcA, coding for

capsular polysaccharide synthesis and arginine deiminase, respectively. Importantly, expression of

cap5A and arcA was induced by treatment with arsenate which depletes ATP through a futile cycle: ADP-

As – ADP + As (spontaneous hydrolysis) – ADP-As. This suggests that these markers actually respond to

ATP decrease, and the rare stationary-like cells in a growing population have low energy levels. These

cells had a considerably higher antibiotic tolerance after being sorted out from a growing population,

showing that they are persisters. Importantly, dropping ATP to stationary levels with arsenate treatment

in a growing culture recapitulates the persister level of a stationary population, showing that low energy

is sufficient for tolerance. If low ATP results in tolerance, then high ATP should have the opposite effect.

Indeed, supplementing TSB medium with glucose increased ATP significantly and resulted in a 100 fold

reduction in persisters. We made similar observations linking ATP and persisters in a study of E. coli [62].

Based on these findings, we proposed a “low energy” mechanism of persister formation. This hypothesis

provides a satisfactory explanation for the mechanism of drug tolerance. Bactericidal compounds kill by

corrupting active targets, and when ATP is low, cells become tolerant to antibiotics. In a stationary

population of S. aureus, ATP levels are indeed low. The entire population is highly tolerant to antibiotics

and is equivalent to persisters. However, how ATP levels may decrease in rare cells of a growing

46

population remained unknown. We reasoned that random fluctuations in the levels of energy-

generating components could lead to low-energy cells. Here we show that cells in a growing population

that have low levels of expression of TCA cycle enzymes are tolerant to killing by antibiotics.

3.3 Results

At late stages of growth, a population of S. aureus exhausts glucose and expression of TCA cycle

enzymes is upregulated [64]. These conditions emulate the nutrient environment during S. aureus

infection [64]. Proteome analysis confirms that this is the case in a stationary phase culture, where the

level of TCA cycle enzymes increases, while glycolytic enzyme abundance decreases. High levels of

enzymes responsible for incorporating amino acids in TCA cycle metabolism are also observed in

stationary phase (Figure 1). Stationary phase is also limited by oxygen which, combined with the lack of

glycolytic substrates would account for previously observed low ATP and high tolerance to antibiotics

[35]. In a late exponentially growing population where oxygen is available but glucose has been largely

exhausted, fluctuation in the levels of TCA cycle enzymes could then lead to a drug tolerant state. We

first sought to examine this in a model experiment by testing antibiotic tolerance of mutants with

knockouts in TCA cycle enzymes.

Figure 1: TCA enzyme abundance increases in late growth phase. (A) Heat map shows enrichment

analysis for proteins in exponential and stationary phase. (B) Map shows major steps in central

metabolism for which significant changes in enzyme abundance were detected between exponential

and stationary phase. Blue indicates decreased abundance in stationary phase; red indicates increased

abundance in stationary phase. Glutamate-catabolizing and TCA cycle enzymes were detected in higher

abundance in stationary phase, when glucose levels are known to be low. Four biological replicates were

analyzed to determine relative abundance for each condition. Gene Ontology and KEGG identifiers were

47

(A) (B)

extracted from UniprotKB. Protein abundance significance was determined using Student’s T-test.

Fisher’s exact tests were performed in R to identify the ontology groups enriched in the proteins

differentially expressed in the two conditions tested.

48

A B

E F

G H Oxacillin

Gentamicin

Ciprofloxacin

49

Figure 2: S. aureus lacking functional late TCA cycle genes exhibits increased antibiotic tolerance.

Antibiotic killing of mutants in TCA cycle genes gltA, sucA, sucC, fumC, and gudB (citrate synthase, 2-

oxogluterate dehydrogenase, succinyl coenzyme A synthetase, fumarase, and glutamate

dehydrogenase) over time compared to wild type (clear symbols) after 10X MIC antibiotic challenge in

TSB medium. Bar graphs represent percent survival of each strain after 96 or 120 hours, as indicated. (A-

B) Ciprofloxacin treatment. (E-F) Gentamicin treatment. (G-H) Oxacillin treatment. Error bars indicate

SEM. Asterisks indicate significance between a mutant and wild type as determined by Sidak’s multiple

comparisons test (*P<0.05, **P<0.01, ***P<0.001, ****P<0.0001). Experiments were performed in

biological triplicate.

We constructed mutants lacking functional enzymes gltA (citrate synthase), gudB (glutamate

dehydrogenase), sucA (2-oxoglutarate dehydrogenase), sucC (succinyl-CoA synthetase), and fumC

(fumarate hydratase) by transducing insertions from the Nebraska Transposon Mutant Library into the

MSSA strain HG003, which is susceptible to antibiotics [65]. Strains were grown to late exponential

phase and challenged with 10x MIC of ciprofloxacin, gentamicin, or oxacillin, representing the main

classes of bactericidal antibiotics – fluoroquinolones, aminoglycosides, and -lactams. The TCA cycle and

amino acid catabolism are important for S. aureus growth in vivo, so we also investigated glutamate

dehydrogenase. Glutamate dehydrogenase fuels the TCA cycle in glucose-deplete conditions by

converting glutamate derived from the abundant proline in collagen into 2-oxoglutarate [4]. All TCA

cycle mutants as well as gudB have significantly higher survival than wild type upon treatment with

antibiotics, showing a multidrug tolerant phenotype characteristic of persisters (Figure 2). The highest

level of persisters is observed with sucA and fumC mutants, where nearly 10% of the population is

persisters.

50

We reasoned that strains with high persister levels should have low ATP. We measured ATP levels with

luciferase and observe that gudB, sucA, sucC, and fumC mutants indeed have significantly lower ATP

than wild type in late exponential phase (Figure 3B).

A B

citrate

isocitrate

2-oxoketoglutarate

succinyl-CoA

succinate

fumarate

malate

oxaloacetate

glutamate

gudB

C

51

Figure 3: TCA cycle mutants strains have lower ATP levels than wild type. (A) TCA cycle model, mutants

highlighted (B) ATP concentration per CFU of all strains. Cultures were grown to late exponential phase

and ATP was measured in the bulk population by luminescence assay. Mutants gudB, sucA, sucC, and

fumC have significantly lower ATP than wild type. Error bars represent standard error. Statistical

significance was determined using ANOVA followed by Sidak’s multiple comparisons test (*P<0.05,

****P<0.0001). (C) Frequency distribution of QUEEN signal for each TCA mutant and wild type strain.

Wild type S. aureus and TCA mutants were grown to early stationary phase in TSB without glucose at

30°C. Strains were grown with chloramphenicol 10 µg/ml to maintain the plasmid and QUEEN

expression was induced with 0.03% xylose. Single cells were analyzed by flow cytometry. Post-

acquisition analysis was performed in FlowJo Software. Ratio values were calculated by dividing

intensities from excitations at 405nm and 408nm (405ex/488ex).

In order to probe the heterogeneity of the population, we sought to examine ATP at a single cell level.

For this, we adopted the QUEEN ATP sensor [66]. QUEEN contains GFP fused to an ATP-binding subunit

of Bacillus PS3 F0F1 ATP synthase. The sensor absorbs at 405 nm and 488 nm and emits at 513 nm. At

higher levels of ATP, there is increased fluorescence from the 405 nm excitation, and decreased

fluorescence from the 488 nm excitation. A ratio between the two emission signals reports ATP

concentration. This ratio does not depend on the amount of the reporter, eliminating errors due to

variation in QUEEN levels among cells. We first cloned QUEEN in plasmid pEPSA5 under a xylose

promoter, but the fluorescence signal was weak. We then codon-optimized a QUEEN construct for

expression in S. aureus, which yielded an improved fluorescence signal (Supplemental Figure 1).

Using FACS analysis, we monitored ATP in single cells of wild type S. aureus and the gltA, gudB, sucA,

sucC, and fumC mutants. The frequency distribution of the ratios is shown in Figure 3C. A wide range of

52

ATP distribution among cells is evident both in the wild type population and in the TCA cycle mutants

(Figure 3C). As expected, the distribution of ATP in mutant populations is shifted to lower levels

compared to the wild type, consistent with an increase in persisters. Persisters tend to wake up during

sorting, which makes sorting prior to antibiotic treatment problematic [35, 62]. Sorting based on QUEEN

signal after antibiotic treatment would not report the ATP status of cells immediately prior to treatment.

Given that knockouts in TCA cycle had lower ATP, we used reporters of their expression to identify low

energy cells. Adding antibiotic prior to sorting provides a snapshot of a protein level at that point in

time.

We cloned the promoter regions of TCA cycle genes upstream of gfpuvr in plasmid pALC1434 [67] yielding

PgltA::gfp, PsucA::gfp, PsucC::gfp, and PfumC::gfp. These strains were grown to late exponential phase

and challenged with ciprofloxacin at 10x MIC. After 24 hours of antibiotic treatment, cells were analyzed

by FACS (Figure 5). Interestingly, there was a broad distribution of expression for each of these genes in

the population. Populations of low, intermediate, and high (“dim, middle, and bright”) TCA gene

expression were gated (Figure 5A-D) and sorted onto agar plates. Surviving cells formed colonies and we

quantified survival of each gated fraction of cells compared to the bulk of the population. We observed

a significant enrichment in persister cells in the dim fractions. In the case of sucA, sucC and fumC, there

was a close to 100 fold difference in persisters between the dim and bright fractions (Figure 4). In order

to test for a possible correlation between persister levels and a general decrease in transcription, we

performed an identical experiment with a strain containing a reporter of the constitutively expressed

sarR. There was no difference in persister levels among the dim, middle and bright populations of the

control. Taken together, these results suggest that random fluctuations in the levels of TCA cycle

enzymes cause a decrease in the energy level, producing drug-tolerant persisters.

53

Figure 5: Sorting of cells with low expression of TCA cycle enzymes enriches in drug tolerant persisters.

(A-E), GFP expression of PgltA::gfp, PsucA::gfp, PsucC::gfp, PfumC::gfp, PsarR::gfp. (F) Percent survival of

each gated fraction and the bulk of the unsorted population. Cells with low expression (Dim) of sucA, sucC,

and fumC exhibited significantly increased survival compared to cells with relatively high expression

(Bright). Sorting on the basis of constitutively expressed PsarR::gfp yields no significant enrichment in any

fraction. Asterisks indicate statistical significance as determined by two-way ANOVA multiple comparisons

(*P<0.05). Graph represents the mean of five biological replicates. A representative plate is shown in

Supplemental Figure 2.

Br i

gh

t

Mid

dle

Dim

Bu

lk

Br i

gh

t

Mid

dle

Dim

Bu

lk

Br i

gh

t

Mid

dle

Dim

Bu

lk

Br i

gh

t

Mid

dle

Dim

Bu

lk

Br i

gh

t

Mid

dle

Dim

Bu

lk

-4

-3

-2

-1

0

Lo

g1

0 %

su

rv

iva

l

P s u c A ::G F P P s u c C ::G F P P fu m C ::G F P P s a rR ::G F PP g ltA ::G F P

* **

Br i

gh

t

Mid

dle

Dim

Bu

lk

Br i

gh

t

Mid

dle

Dim

Bu

lk

Br i

gh

t

Mid

dle

Dim

Bu

lk

Br i

gh

t

Mid

dle

Dim

Bu

lk

Br i

gh

t

Mid

dle

Dim

Bu

lk

-4

-3

-2

-1

0

Lo

g1

0 %

su

rv

iva

l

P s u c A ::G F P P s u c C ::G F P P fu m C ::G F P P s a rR ::G F PP g ltA ::G F P

* **

Br i

gh

t

Mid

dle

Dim

Bu

lk

Br i

gh

t

Mid

dle

Dim

Bu

lk

Br i

gh

t

Mid

dle

Dim

Bu

lk

Br i

gh

t

Mid

dle

Dim

Bu

lk

Br i

gh

t

Mid

dle

Dim

Bu

lk

-4

-3

-2

-1

0

Lo

g1

0 %

su

rv

iva

l

P s u c A ::G F P P s u c C ::G F P P fu m C ::G F P P s a rR ::G F PP g ltA ::G F P

* **

Br i

gh

t

Mid

dle

Dim

Bu

lk

Br i

gh

t

Mid

dle

Dim

Bu

lk

Br i

gh

t

Mid

dle

Dim

Bu

lk

Br i

gh

t

Mid

dle

Dim

Bu

lk

Br i

gh

t

Mid

dle

Dim

Bu

lk

-4

-3

-2

-1

0L

og

10

% s

urv

iva

lP s u c A ::G F P P s u c C ::G F P P fu m C ::G F P P s a rR ::G F PP g ltA ::G F P

* **

Br i

gh

t

Mid

dle

Dim

Bu

lk

Br i

gh

t

Mid

dle

Dim

Bu

lk

Br i

gh

t

Mid

dle

Dim

Bu

lk

Br i

gh

t

Mid

dle

Dim

Bu

lk

Br i

gh

t

Mid

dle

Dim

Bu

lk

-4

-3

-2

-1

0

Lo

g1

0 %

su

rv

iva

l

P s u c A ::G F P P s u c C ::G F P P fu m C ::G F P P s a rR ::G F PP g ltA ::G F P

* **

A B

C D

E

F

54

3.4 Discussion

Persisters were originally discovered in a study of S. aureus in 1944 [1], but understanding the

mechanism of their formation proved to be unusually challenging. This no doubt is due to the small size

of their population, and a temporary phenotype. Given that bactericidal antibiotics act by corrupting

their targets, we proposed that tolerance is caused by target inactivation [35, 62]. This concept is broad

enough to cover the two types of emerging persister formation mechanisms, specific and generalized.

Specialized toxins such as HipA and TisB govern specific mechanisms of persister formation. Selection for

high-persister mutants led to the identification of a gain-of-function allele in the hipA mutant in E. coli

[68] [68], and subsequent studies determined that it is a kinase [69] that inhibits translation by

phosphorylating glu-tRNA synthase [70, 71]. However, deletion of the hipBA locus has no phenotype,

and it does not appear that HipA plays a role in persister formation of wild type E. coli. At the same time,

we found that hipA7 high persister cells are present in E. coli isolates from patients with urinary tract

infection, a result of in vivo selection for drug tolerance [5]. Another E. coli toxin, TisB, provides an

example of a persister formation mechanism operating in wild type cells. Induced by fluoroquinolones

through the SOS DNA-damage response, TisB is an endogenous antimicrobial peptide that causes

tolerance by decreasing the pmf and ATP [52, 72]. These specific mechanisms however do not explain

how persisters are formed under regular growth conditions. For a while, the idea that RNA

endonuclease TAs such as RelBA or MazEF constitute the main mechanism of persister formation in

bacteria became the standard model [40], but several recent studies failed to find a connection between

these interferase toxins and persisters in E. coli [35, 62, 63]. In particular, a knockout of 10 interferase

TAs had no effect of persister formation [63].

In search of a general mechanism of persister formation, we identified a link between low energy,

specifically low ATP, and drug tolerance in both S. aureus and E. coli [35, 62]. Selectively decreasing the

55

level of ATP by arsenate is sufficient to produce persisters, and low energy cells sorted from a

population by monitoring transcriptional or translational ATP markers are tolerant to antibiotics [35, 62].

If ATP is low, target activity is diminished, providing a simple mechanism for drug tolerance. In the

current study, we sought to identify components that are responsible for producing persisters. Energy

generating components are a logical choice to consider, and not surprisingly we observe, in agreement

with previous observations [73] that knocking out TCA cycle components increases drug tolerance and

lowers ATP levels. The critical question however is whether natural fluctuation in expression of TCA

enzymes is sufficient to produce persisters. We find that sorting cells with low levels of expression of

several TCA cycle enzymes gltA, sucA, sucC, and fumC enriches in drug-tolerant persisters. Interestingly,

FACS analysis shows that noise in expression of these enzymes is considerable, over three orders of

magnitude, and largely follows a typical Gaussian distribution. This noise leads to formation of rare cells

with low levels of enzyme expression, decreased ATP, and drug tolerance.

In retrospect, the low energy hypothesis of persister formation is quite obvious – indeed, the simplest

way to inactivate all antibiotic targets is by lowering ATP, and the mechanism of the specialized TisB

toxin provides a precedent for this. However, noise in an energy generating component sufficient to

produce a drop in ATP is counterintuitive. It is commonly accepted that noise results from fluctuation in

sparsely expressed components. In a classical example, E. coli cells only have about 10 molecules of LacI

on average, and noise in expression produces rare cells with no repressor, resulting in full expression of

the lac operon in the absence of inducer [74]. By contrast, TCA cycle enzymes are abundant, and the

considerable level of noise in their expression we observe is unexpected.

56

The current study is the first step towards identifying components that can lead to a low energy state

and drug tolerance. Future studies will show how widely spread among bacteria is this general

mechanism of persister formation.

3.5 Methods

3.5.1 Bacterial strains, culture conditions, and strain construction

S. aureus strains were grown in either tryptic soy broth (TSB) (MP Biomedicals, USA), TSB without

glucose (Becton, Dickinson, and Company, USA), or Mueller Hinton Broth (MHB) as indicated. Cultures

were grown at 37oC shaking at 225 rpm. Strains encoding the QUEEN construct were grown on TSA

plates with chloramphenicol 10 µg/ml at 30oC or in TSB without glucose shaking at 225 rpm at 30oC with

chloramphenicol 10 µg/ml to maintain the pEPSA5 plasmid. QUEEN expression was induced with 0.03%

xylose. High concentrations of xylose led to a growth defect. The MSSA strain HG003 was used for these

studies and mutations for all TCA cycle genes in this background were transduced from the Nebraska

Transposon Mutant Library (NTML) using bacteriophage 80 or φ11. Mutations were subsequently

confirmed by amplifying from the beginning or end of the gene of interest to the transposon insertion as

previously described [65]. For construction of gfp reporters, promoter regions of gltA, sucAB, sucC, or

fumC were cloned upstream of gfp into the EcoRI and XbaI sites of pALC1434[67]. PsucAB was amplified

with 5’-gggcccgaattcgaaacctcatcaattcgaacaa-3’ and 5’-gggccctctagatttacaccctccacaaaaatgttgaaa-3’.

Escherichia coli DH5 was used to propagate plasmids. DH5 strains were grown in LB Broth, Miller

(Fisher BioReagents, USA) and ampicillin 100 µg/ml was used to maintain plasmids where necessary.

3.5.2 Proteomic sample preparation

Four replicates were prepared from either exponential or stationary phase cells. Cells were grown for 4

or 24 hours, respectively, after overnight cultures were diluted 1:1000 in 100 mL MHB. Cells were grown

in 500 mL baffled flasks at 37oC shaking at 225 rmp. Cells were pelleted at 5000 x g for 7 minutes at 4 oC,

57

washed twice in PBS, and pelleted again. Pellets were resuspended in 500 µL PBS, transferred to a 2 mL

tube, and washed a final time, then flash frozen with liquid nitrogen and stored at -80 oC. Bacterial

pellets were resuspended in 100 mM ammonium bicarbonate and lysed by vortexing 5 times for one

minute then with 0.1 mm zirconia/silica beads with resting periods of 30 s on ice. Samples were then

digested with trypsin as previously described [75] and desalted using C18 SPE cartridges (Discovery C18,

1 mL, 50 mg, Sulpelco). Peptide concentrations were measured by BCA assay (Thermo Scientific).

3.5.3 Proteomics and data analysis

For each sample 0.5 μg of peptides were separated using a 200 minute gradient on a Waters

nanoEquityTM UPLC system (Millford, MA) coupled with a QExactive HF (Thermo Fisher Scientific). MS

scans were recorded at a resolution of 35,000. The Top 12 ions from the survey scan were selected by a

quadrupole mass filter for high energy collision dissociation and mass analyzed by the Orbitrap. A

window of 2 m/z was used for the isolation of ions and collision energy of 28%. MS/MS spectra were

recorded with a resolution of 17,500. resulting data were processed using the MaxQuant V1.5.2.8. [76].

Proteins were identified with at least 2 peptides of lenght higher than 6 residues.

The RefSeq Staphylococcus aureus NCTC8325 database was used for the search (July 2017: 2,768

sequences). Match between run and MaxLFQ were used for quantification, other parameters were

conserved by default. Proteins only identified by site, reverse hits, and contaminants were removed.

Only protein groups with a measured LFQ intensity in at least 60% of one sample type were conserved

for further quantification. LFQ intensities were log2 transformed and median normalized. Statistics (e.g.

Student’s T.test, scaling, etc.) and normalization steps were performed in R using the packages Stat. The

mass spectrometry proteomics data has been deposited to ProteomeXchange Consortium with the

dataset identifier PXD013151.

58

3.5.4 Persister assays

Overnight cultures were diluted 1:1000 in 2 mL TSB (Fisher, MP Biomedicals) in a 14 mL capped culture

tube (VWR International), grown to late exponential phase and starting CFU were plated. Cultures were

challenged with ciprofloxacin, oxacillin, or gentamicin (4, 1.25, and 100 µg ml–1, respectively). To

enumerate survivors over time, 100 µL of culture was removed, pelleted by centrifugation, washed

with 1% NaCl, serial diluted and plated on TSA and CFU were counted after 24 h regrowth on TSA to

enumerate survivors. Experiments were performed in biological triplicates.

3.5.5 ATP quantification of bulk culture

ATP levels were measured using the Promega BacTiter-Glo Microbial Cell Viability Assay according to the

manufacturer’s instructions. A working volume of 100 µL was used. Aliguots from tubes were removed,

pelleted, and resuspended in 1% NaCl before reading luminescence. Experiments were performed in

biological triplicates.

3.5.6 Construction of S. aureus HG003 expressing QUEEN2m

pEPSA5-QUEEN2m

pEPSA5-QUEEN2m was created by amplifying QUEEN2m from pRsetB-his7-QUEEN2m [66] with primers

Q2m-F 5’-CGAGCTGAATTCTAGGGAGAGGTTTTAAACATGAAAACTGTGAAAGTGAATATAAC-3’ and Q2m-R

5’-CGAGCTGGTACCTCACTTCATTTCCGCAACGCTC-3’, and cloning it into the EcoRI/KpnI sites of pEPSA5

downstream of a xylose promoter [77]. Restrictions sites are underlined, and a RBS from sarA gene has

been added in Q2m-F primer (here in bold).

pEPSA5-QUEEN2m(opt)

59

QUEEN2m(opt) was codon-optimized from the original sequence by using JCat tool [78] and DNA

synthesized (Genewiz) with the same RBS used in pEPSA5-QUEEN2m in a custom plasmid. A fragment

containing QUEEN2m(opt) and its RBS was excised from this plasmid with EcoRI/KpnI and cloned into

the EcoRI/KpnI sites of pEPSA5 [77].

Those two plasmids have been transformed into S. aureus RN4220 and amplified from this background

before to be transformed in S. aureus HG003 background. Only the optimized version was transformed

into mutant strains.

3.5.7 Microscopy

S. aureus HG003-pEPSA5-QUEEN2m and HG003-pEPSA5-QUEEN2m(opt) were cultured at 30C in TSB

without glucose (complemented with chloramphenicol 10 µg/mL and xylose 0.03% for the pEPSA5

maintenance and the induction of QUEEN2m expression, respectively) to stationary phase, inoculated

into fresh medium of the same composition at 1:100, and cultured for 4.5 hours at 30°C. Samples were

washed with PBS, placed on top of a PBS agarose pad 1%, and observed under a ZEISS LSM 710 confocal

microscope using 63x oil immersion objective lens. The two fluorescent signals 405ex and 488ex were

sequentially collected. DIC image was recorded alongside the 405ex acquisition. Images were acquired

by Zen Software at a resolution of 1024 x 1024 and lane average of 8, and processed with the Fiji

software [79].

3.4.8 Single-cell ATP quantification using QUEEN

Flow cytometry was carried out using an Attune flow cytometer. Spatially separated violet and blue

lasers were used for excitation at 405 nm and 488 nm, respectively, to calculate the ratio of emission

at513 nm produced by each of these excitation wavelengths. To prepare samples, -80 oC stocks were

grown on TSA plates with chloramphenicol 10 µg/ml at 30oC. Single colonies were selected and

60

overnight cultures were inoculated in TSB without glucose plus chloramphenicol 10 µg/ml shaking at

225 rpm at 30 oC. Tubes were inoculated 24 hours later from overnight cultures and grown 25 hours

(HG003 empty vector, HG003 WT, gltA, sucA, and fumC mutants) or 26 hours (gudB and sucC mutants)

to reach comparable CFU/mL. Xylose 0.03% was used to induce expression of QUEEN. Strains used:

HG003 pEPSA5 empty vector, HG003 pEPSA5-QUEEN2m, gltA:: pEPSA5-QUEEN2m, gudB::

pEPSA5-QUEEN2m, sucA:: pEPSA5-QUEEN2m, sucC:: pEPSA5-QUEEN2m, and fumC::

pEPSA5-QUEEN2m. Three biological replicates were analyzed for each strain.

3.4.9 FACS analysis using GFP reporters

Cell sorting was carried out using a BD FACSAria II with a 70 micron nozzle. Briefly, 1:1000 dilutions of

overnight cultures were grown 5 hours to late exponential phase and challenged with ciprofloxacin (10x

MIC) for 24 hr. After 24 h, cells were diluted 1:100 and sonicated as previously described [80] to

disperse aggregates of S. aureus cells. FACS DIVA software was used in sorting setup; the initial

population of cells was gated by size using forward and side scatter (FSC and SSC) then on the basis of

gfp fluorescence (GFP-A). Gates were set to include the brightest 5% and dimmest 5%, in addition to a

middle ~30% of the population. Cells were sorted from each population onto TSA plates. Plates were

incubated at 37 oC for 24 h and colonies were counted. Percent survival was calculated from dim,

middle, and bright GFP fractions.

Acknowledgements

We would like to thank Hiromi Imamura for the QUEEN-2m coding plasmid. We thank Sarah Rowe and

David Cameron for critical discussions. This work was supported by NIH grant R01-AI110578-01

61

62

gene

symb

ol

refSeq ID Description UniprotID Locus pvalue

log2

(stat/ex

p)

Pathway

E.C.

numbe

r

ptsG YP_50130

5.1

PTS system

glucose-specific

transporter subunit

IIABC

PTU3C_STA

A8

SAOUHSC_02

848

0.0005

18 -0.7 Glycolysis

2.7.1.1

99

glcA YP_49875

4.1

PTS system

glucose-specific

protein

PTG3C_STA

A8

SAOUHSC_00

155

0.0002

18 -0.7 Glycolysis

2.7.1.1

99

ptbA YP_49995

5.1

PTS system

transporter subunit

IIA

Q2FYL0_STA

A8

SAOUHSC_01

430

0.0042

13 0.5 Glycolysis

2.7.1.1

99

pgi YP_49945

3.1

Glucose-6-

phosphate

isomerase

G6PI_STAA8 SAOUHSC_00

900

0.0001

44 -1.0 Glycolysis 5.3.1.9

pfkA YP_50031

2.1

6-

phosphofructokina

se

PFKA_STAA8 SAOUHSC_01

807

0.0000

41 -1.5 Glycolysis

2.7.1.1

1

FBPa

se

YP_50128

0.1

hypothetical

protein

SAOUHSC_02822

F16PC_STAA

8

SAOUHSC_02

822

0.9283

60 0.0 Glycolysis

3.1.3.1

1

fdaB YP_50137

9.1

fructose-1,6-

bisphosphate

aldolase

ALF1_STAA8 SAOUHSC_02

926

0.0065

61 -0.4 Glycolysis

4.1.2.1

3

fbaA YP_50084

2.1

fructose-

bisphosphate

aldolase

Q2FWD3_ST

AA8

SAOUHSC_02

366

0.0067

95 -0.4 Glycolysis

4.1.2.1

3

tpiA YP_49935

3.1

triosephosphate

isomerase TPIS_STAA8

SAOUHSC_00

797

0.0003

36 -1.0 Glycolysis 5.3.1.1

gap YP_49935

1.1

glyceraldehyde-3-

phosphate

dehydrogenase

Q2G032_STA

A8

SAOUHSC_00

795

0.0000

01 -1.3 Glycolysis

1.2.1.1

2

gapB YP_50029

8.1

glyceraldehyde 3-

phosphate

dehydrogenase 2

Q2FXP2_STA

A8

SAOUHSC_01

794

0.0508

71 -0.7 Glycolysis

1.2.1.1

2

pgk YP_49935

2.1

phosphoglycerate

kinase PGK_STAA8

SAOUHSC_00

796

0.0001

96 -1.1 Glycolysis 2.7.2.3

63

gpmA YP_50116

5.1

phosphoglyceromu

tase

GPMA_STAA

8

SAOUHSC_02

703

0.0006

28 -0.9 Glycolysis

5.4.2.1

1

pgmB YP_49894

9.1

phosphoglycerate

mutase family

protein

Q2G101_STA

A8

SAOUHSC_00

359

0.0949

29 -9.8 Glycolysis

5.4.2.1

1

pgm YP_49935

2.1

phosphoglycerate

kinase PGK_STAA8

SAOUHSC_00

796

0.0001

96 -1.1 Glycolysis

5.4.2.1

1

eno YP_49935

5.1

phosphopyruvate

hydratase ENO_STAA8

SAOUHSC_00

799

0.0000

14 -0.7 Glycolysis

4.2.1.1

1

pykA YP_50031

1.1 pyruvate kinase KPYK_STAA8

SAOUHSC_01

806

0.0001

70 -1.2 Glycolysis

2.7.1.4

0

lctE YP_49880

3.1

L-lactate

dehydrogenase LDH1_STAA8

SAOUHSC_00

206

0.0483

59 -11.5

Pyruvate/ace

tate

metabolism

1.1.1.2

7

ldh2 YP_50137

4.1

L-lactate

dehydrogenase LDH2_STAA8

SAOUHSC_02

922

0.0004

52 -0.6

Pyruvate/ace

tate

metabolism

1.1.1.2

7

ddh YP_50128

9.1

D-lactate

dehydrogenase

Q2FVA3_STA

A8

SAOUHSC_02

830

0.0003

00 -1.4

Pyruvate/ace

tate

metabolism

1.1.1.2

8

pycA YP_49961

0.1

pyruvate

carboxylase

Q2G2C1_ST

AA8

SAOUHSC_01

064

0.3906

06 -0.2

Pyruvate/ace

tate

metabolism

pdhA YP_49958

9.1

pyruvate

dehydrogenase

complex, E1

component subunit

alpha

Q2FZG4_STA

A8

SAOUHSC_01

040

0.0016

04 -1.2

Pyruvate/ace

tate

metabolism

1.2.4.1

pdhB YP_49959

0.1

pyruvate

dehydrogenase

complex, E1

component subunit

beta

Q2G2A5_STA

A8

SAOUHSC_01

041

0.0000

66 -0.8

Pyruvate/ace

tate

metabolism

1.2.4.1

pdhC YP_49959

1.1

branched-chain

alpha-keto acid

dehydrogenase

subunit E2

Q2G2A4_STA

A8

SAOUHSC_01

042

0.0000

75 -1.1

Pyruvate/ace

tate

metabolism

2.3.1.1

2

64

pdhD YP_49959

2.1

dihydrolipoamide

dehydrogenase

Q2G2A3_STA

A8

SAOUHSC_01

043

0.0005

17 -0.7

Pyruvate/ace

tate

metabolism

1.8.1.4

lpdA YP_50012

9.1

dihydrolipoamide

dehydrogenase

Q2FY51_STA

A8

SAOUHSC_01

614

0.0163

88 -0.8

Pyruvate/ace

tate

metabolism

1.8.1.4

pflB YP_49878

4.1

formate

acetyltransferase PFLB_STAA8

SAOUHSC_00

187

0.0084

47 -1.8

Pyruvate/ace

tate

metabolism

2.3.1.5

4

porA YP_49979

9.1

hypothetical

protein

SAOUHSC_01266

Q2FZ05_STA

A8

SAOUHSC_01

266

0.0152

38 -0.5

Pyruvate/ace

tate

metabolism

1.2.7.1

1

porB YP_49980

0.1

2-oxoglutarate

ferredoxin

oxidoreductase

subunit beta

Q2FZ04_STA

A8

SAOUHSC_01

267

0.0505

90 -0.3

Pyruvate/ace

tate

metabolism

1.2.7.1

1

eutD YP_49914

2.1

phosphotransacety

lase

Q2G0J0_STA

A8

SAOUHSC_00

574

0.0003

71 -1.2

Pyruvate/ace

tate

metabolism

2.3.1.8

acsA YP_50035

1.1

acetyl-CoA

synthetase

Q2G294_STA

A8

SAOUHSC_01

846

0.0024

62 1.5

Pyruvate/ace

tate

metabolism

6.2.1.1

AcsA2 YP_50138

2.1

acetyl-CoA

synthetase

Q2FV14_STA

A8

SAOUHSC_02

929

0.0043

96 -0.4

Pyruvate/ace

tate

metabolism

6.2.1.1

ackA YP_50032

5.1 acetate kinase

ACKA_STAA

8

SAOUHSC_01

820

0.0000

91 -1.2

Pyruvate/ace

tate

metabolism

2.7.2.1

acyP YP_49993

3.1 acylphosphatase

ACYP_STAA

8

SAOUHSC_01

406

0.0000

00 +inf

Pyruvate/ace

tate

metabolism

3.6.1.7

CidC YP_50130

6.1 pyruvate oxidase

Q2FV86_STA

A8

SAOUHSC_02

849

0.0000

04 -1.4

Pyruvate/ace

tate

metabolism

aldH YP_50063

2.1

aldehyde

dehydrogenase

Q2FWX9_ST

AA8

SAOUHSC_02

142

0.0139

02 0.9

Pyruvate/ace

tate

metabolism

1.2.1.3

65

aldA YP_50083

9.1

aldehyde

dehydrogenase ALD1_STAA8

SAOUHSC_02

363

0.9509

16 0.0

Pyruvate/ace

tate

metabolism

1.2.1.3

aldA2 YP_49873

2.1

aldehyde

dehydrogenase ALDA_STAA8

SAOUHSC_00

132

0.0280

68 0.4

Pyruvate/ace

tate

metabolism

1.2.1.3

adh1 YP_49917

1.1

alcohol

dehydrogenase ADH_STAA8

SAOUHSC_00

608

0.0011

73 -1.4

Pyruvate/ace

tate

metabolism

1.1.1.1

pckA YP_50041

1.1

phosphoenolpyruv

ate carboxykinase

PCKA_STAA

8

SAOUHSC_01

910

0.0065

73 0.9 TCA

4.1.1.4

9

pycA YP_49961

0.1

pyruvate

carboxylase

Q2G2C1_ST

AA8

SAOUHSC_01

064

0.3906

06 -0.2 TCA 6.4.1.1

citZ YP_50030

7.1

hypothetical

protein

SAOUHSC_01802

Q2FXN3_STA

A8

SAOUHSC_01

802

0.0055

31 0.8 TCA 2.3.3.1

citB YP_49987

5.1

aconitate

hydratase

Q2FYS9_STA

A8

SAOUHSC_01

347

0.0004

06 0.7 TCA 4.2.1.3

citC YP_50030

6.1

isocitrate

dehydrogenase

Q2FXN4_STA

A8

SAOUHSC_01

801

0.0227

20 0.4 TCA

1.1.1.4

2

sucA YP_49994

4.1

2-oxoglutarate

dehydrogenase E1

component

ODO1_STAA

8

SAOUHSC_01

418

0.0176

69 0.4 TCA 1.2.4.2

pdhD YP_49959

2.1

dihydrolipoamide

dehydrogenase

Q2G2A3_STA

A8

SAOUHSC_01

043

0.0005

17 -0.7 TCA 1.8.1.4

lpdA YP_50012

9.1

dihydrolipoamide

dehydrogenase

Q2FY51_STA

A8

SAOUHSC_01

614

0.0163

88 -0.8 TCA 1.8.1.4

odhB YP_49994

3.1

dihydrolipoamide

succinyltransferase

ODO2_STAA

8

SAOUHSC_01

416

0.0073

55 0.6 TCA

2.3.1.6

1

porA YP_49979

9.1

hypothetical

protein

SAOUHSC_01266

Q2FZ05_STA

A8

SAOUHSC_01

266

0.0152

38 -0.5 TCA 1.2.7.3

porB YP_49980

0.1

2-oxoglutarate

ferredoxin

oxidoreductase

subunit beta

Q2FZ04_STA

A8

SAOUHSC_01

267

0.0505

90 -0.3 TCA 1.2.7.3

66

sucD YP_49975

4.1

succinyl-CoA

synthetase subunit

alpha

Q2FZ36_STA

A8

SAOUHSC_01

218

0.0669

54 0.3 TCA 6.2.1.5

sucC YP_49975

3.1

succinyl-CoA

synthetase subunit

beta

SUCC_STAA

8

SAOUHSC_01

216

0.2682

54 0.2 TCA 6.2.1.5

sdhC YP_49964

7.1

succinate

dehydrogenase

cytochrome b-558

subunit

Q2FZC9_STA

A8

SAOUHSC_01

103

0.0091

17 1.1 TCA 1.3.5.4

sdhA YP_49964

8.1

succinate

dehydrogenase

flavoprotein

subunit

Q2FZC8_STA

A8

SAOUHSC_01

104

0.0002

64 0.8 TCA 1.3.5.4

sdhB YP_49964

9.1

succinate

dehydrogenase

iron-sulfur subunit

Q2FZC7_STA

A8

SAOUHSC_01

105

0.0014

24 0.5 TCA 1.3.5.4

fumC YP_50048

0.1

fumarate

hydratase

Q2FX94_STA

A8

SAOUHSC_01

983

0.0000

45 0.7 TCA 4.2.1.2

mqo1 YP_50110

9.1

malate:quinone

oxidoreductase

Q2FVQ5_ST

AA8

SAOUHSC_02

647

0.0026

38 0.6 TCA 1.1.5.4

mqo2 YP_50138

0.1

malate:quinone

oxidoreductase

Q2FV16_STA

A8 SAOUHSC_27

0.0196

18 -0.9 TCA 1.1.5.4

Supplementary Table 3.1: All enzymes represented in the map in Supplementary Figure 3.1.

P-value was determined by ANOVA.

67

Supplementary Figure 3.2: Optimization of QUEEN2m for expression in S. aureus.

Representative images of exponential phase cultures of S. aureus HG003-pEPSA5-QUEEN2m

and HG003-pEPSA5-QUEEN2m(opt) cultured at 30C in TSB without glucose (complemented

with chloramphenicol 10 µg/mL and xylose 0.03% for the pEPSA5 maintenance and the

induction of QUEEN2m expression, respectively). Samples were washed with PBS, placed on

the top of a PBS agarose pad 1%, and observed under the microscope. The two fluorescent

signals 405ex (false-coloured here in magenta) and 488ex (false-coloured here in green)

were sequentially collected one after the other. Scale bar, 5 µm.

68

Supplementary Figure 3.4: Representative photograph of colonies formed by surviving cells after antibiotic challenge and sorting. FACS of PsucA::gfp gated into dim, middle, and bright fractions is shown. Here, 1600 cells were sorted per fraction.

Dim Middle Bright

Supplemental Figure 3.3: ATP concentrations of wild type S. aureus and TCA cycle mutants. ATP concentrations in single cells was determined by FACS using the optimized QUEEN2m sensor. 50,000 events were analyzed for each sample. Experiment was performed in biological triplicate.

69

Chapter 4: Persister Resuscitation

4.1 Abstract

We have shown that low ATP induces persister cell formation [35]. Within a population of S.

aureus, individual cells with low TCA cycle activity are better able to survive antibiotic treatment

than the bulk of the population. It is likely that multiple factors contribute to maintenance of the

persister state, but this area is largely unexplored. Tolerance is a transient phenotype and

persister cells are able to exit the dormant state and resume growth [3]. This is believed to cause

chronic and relapsing infections. It is possible that persister resuscitation is stochastic or that

growth resumption is triggered by environmental cues [81-83]. We conducted a broad screen for

genes involved in persister resuscitation using the Nebraska Transposon Mutant Library [65] and

identify S. aureus mutants with significantly altered resuscitation kinetics following antibiotic

treatment. We identified 46 mutants with significantly faster resuscitation and 116 mutants with

significantly slower resuscitation compared to the bulk of the tested mutants.

4.2 Introduction

Most pathogens that cause difficult-to-treat infections are not resistant to antibiotics [17, 37]. In

cases of chronic and relapsing infection patients are typically prescribed antibiotics, which are

effective in killing most bacterial cells. Persister cells tolerate antibiotic treatment. When

antibiotics are no longer present in the environment, cells eventually resume growth and can

repopulate the infection site [84]. Neither the cause nor mechanism of persister resuscitation are

understood.

70

The phenomenon of resuscitation from the persister state is well-documented [3, 11, 12] but the

mechanism by which cells wake up from the persister state is poorly understood. Some groups

have proposed a microbial “scout” hypothesis, suggesting stochastic resumed growth among a

population of dormant cells [81]. Other groups have proposed that signaling molecules serve as

wake-up calls for dormant bacteria [82]. Others have observed different resuscitation kinetics

depending on the type of antibiotic treatment [83]. It is possible that bacteria have evolved

multiple strategies for emerging from dormant or tolerant states. Given that low ATP is

associated with persister formation and entrance into the tolerant non-growing state, we

hypothesized that metabolic genes would be active in resuscitation from the persister state.

We conduct a broad screen of nearly 2000 S. aureus mutants and identify candidate genes

involved in resuscitation. We treat growing cells with a lethal dose of ciprofloxacin and allow

surviving persisters to recover. We identify mutants that are fast or slow to resuscitate after

antibiotic treatment.

4.3 Results

4.3.1. Growth resumption

To demonstrate the principle of resuscitation following antibiotic treatment in vitro, we added

penicillinase to E.coli after ampicillin treatment and enumerated surviving cells every hour. We

hypothesized that penicillinase-mediated inactivation of ampicillin would permit bacterial

growth. E. coli MG1655 cells were grown to exponential phase and treated with ampicillin. The

addition of penicillinase after 1 hour of ampicillin killing resulted in rapid resumption of growth

(Figure 4.1). After three hours of permissive growth conditions, the penicillinase-treated cultures

71

reached nearly the same cellular density as the growth controls that had not been treated with

ampicillin. In vitro, removing antibiotics from growth conditions resulted in rapid bacterial

regrowth.

Figure 4.1: Antibiotic killing and growth resumption after antibiotic inactivation. E. coli

cultures were grown to exponential phase and treated with ampicillin at 0h where indicated

(red and purple symbols). After 1 hour, penicillinase was added where indicated (purple

symbols), resulting in resumed growth evident in the increase of CFU. Experiment was

performed in biological triplicate. Error bars represent standard error.

72

We sought to investigate the mechanisms of resuscitation from the persister state. S. aureus is

known to persist, sometimes for months or years, in a dormant state. Very little is known about

the emergence from dormancy. One recent study characterized the E. coli proteome during

growth resumption following antibiotic treatment [85] by quantifying the incorporation of stable

isotope labelled amino acids to peptides following treatment. Those results identified the

proteins that were synthesized by E.coli following tolerance induced by overproduction of the

membrane-depolarizing toxin TisB. Toxin-antitoxin systems have not been found to induce

persistence in S. aureus [35] and it is unclear how conserved resuscitation mechanisms are

conserved between species. We took a genetic approach and designed a library screen to identify

mutants with improved or decreased ability to resume growth after antibiotic treatment.

4.3.2. NTML screen

The Nebraska Transposon Mutant Library is a mariner-based transposon library in the USA300

strain background and contains roughly 2000 transposon mutants [65]. To screen this large

collection, we designed a 96-well plate-based assay to find mutants that had especially fast or

slow resuscitation after antibiotic treatment. Briefly, one mutant was inoculated per well in MHB

and grown overnight until the strains reached stationary phase. Mutants were inoculated into

fresh media in a new plate and strains were grown 3 hours to exponential phase then treated

with ciprofloxacin for 24 hours. Cells were pelleted and washed, then resuspended and diluted

into fresh MHB and post-treatment growth resumption was monitored in a plate reader for 21

hours. The OD600 of each well was recorded every 30 minutes. Output from this phase of the

experiment for a representative plate is shown in Figure 4.2. In all experiments, starting OD was

similar for all mutants, suggesting that a comparable number of cells were in the starting

73

inoculum. This experimental procedure was repeated until each mutant in the NTML had been

screen for resuscitation kinetics.

For each strain, the amount of time required to triple the starting OD was calculated. Strains were

classified as “fast” resuscitators if the time required to triple starting OD was more than 1.5

standard deviations higher than the mean for that plate. Strains were classified as “slow”

resuscitators if the time required to triple starting OD was less than 1.5 standard deviations lower

than the mean for that plate. Wherever possible, the emergence of each plate from stationary

phase was also monitored and the same analysis was performed. These mutants simply have a

long lag time before beginning exponential growth or a general growth defect and are not true

candidates for genes involved in resuscitation. They were therefore excluded from the list of hits

of genes implicated in persister resuscitation. Figure 4.2 shows a representative screening plate

of resuscitation following ciprofloxacin treatment. Fast resuscitators (here, SAUSA300_1208 and

SAUSA300_0585) are shown in green. Slow resuscitators (SAUSA300_0504, SAUSA300_2026,

SAUSA300_2024, SAUSA300_2393, SAUSA300_0444, SAUSA300_0844, SAUSA300_1473,

SAUSA300_1908, SAUSA300_1469) are shown in red.

74

Resuscitation candidate genes are listed in Table 4.1 and Table 4.2. Table 4.1 lists the NTML

mutants that resuscitated faster after ciprofloxacin treatment. These genes are expected to

impede resuscitation or potentially maintain a persister state. Table 4.2 lists the mutants that

resuscitated slowly relative to the bulk of the mutants screened. These genes are expected to

Figure 4.2: Post-treatment persister resuscitation screen output sample. One screening plate

(96 mutants) of the NTML are shown here. One mutant was inoculated per well. Cells were

grown to stationary phase and diluted into fresh media in a new plate. (A) Emergence from

stationary phase was monitored by measuring OD600 every 30 minutes. Following 24 hours of

ciprofloxacin treatment, cells were washed with PBS and inoculated into fresh MHB then grown

for 20 hours. (B) OD600 was measured every 30 minutes. Fast resuscitators were defined at

mutants that tripled their starting OD faster than the majority of mutants (T3OD >1.5 SDbulk) and

are colored in green. Slow resuscitators (T3OD <1.5 SDbulk) are colored in red. Mutants with the

same fast or slow phenotype during growth resumption out of stationary phase were excluded

from the list of candidate resuscitation genes.

A

B

75

facilitate resuscitation. Interestingly, the mutant purM which encodes

phosphoribosylaminoimidazole synthetase was slow to resuscitate in our screen. PurM was

recently shown to be highly synthesized during regrowth after antibiotic treatment [85],

suggesting that it is indeed important in the post-antibiotic resuscitation.

Gene Locus ID

putative surface protein SAUSA300_0883

DNA internalization-related competence protein ComEC/Rec2 SAUSA300_1547

phosphate transporter family protein SAUSA300_0650

phiSLT ORF401-like protein, integrase SAUSA300_1438

putative cobalamin synthesis protein SAUSA300_0424

putative membrane protein SAUSA300_2304

putative membrane protein SAUSA300_0881

lactose phosphotransferase system repressor SAUSA300_2156

3-hydroxyacyl-CoA dehydrogenase SAUSA300_0226

formate/nitrite transporter family protein SAUSA300_2349

putative membrane protein SAUSA300_1809

riboflavin synthase, alpha subunit SAUSA300_1714

conserved hypothetical protein SAUSA300_1485

conserved hypothetical protein SAUSA300_0872

iron compound ABC transporter, permease protein SirB SAUSA300_0116

acetyltransferase, GNAT family family SAUSA300_0943

putative glycosyl transferase SAUSA300_2583

FAD/NAD(P)-binding Rossmann fold superfamily protein SAUSA300_0394

galactose-6-phosphate isomerase subunit LacA SAUSA300_2155

capsular polysaccharide biosynthesis protein Cap5H SAUSA300_0159

hypothetical protein SAUSA300_0232

4-oxalocrotonate tautomerase SAUSA300_1258

hypothetical protein SAUSA300_2585

76

F0F1 ATP synthase subunit beta SAUSA300_2058

hypothetical protein SAUSA300_1486

manganese transport protein MntH SAUSA300_1005

serine protease SplC splC SAUSA300_1756

oligoendopeptidase F, PepF SAUSA300_0902

Oye family NADH-dependent flavin oxidoreductase SAUSA300_0322

oligopeptide ABC transporter, permease protein SAUSA300_2410

tandem lipoprotein SAUSA300_0411

Na+/H+ antiporter family protein SAUSA300_2273

thiol peroxidase, tpx SAUSA300_1659

phiPV083 ORF027-like protein SAUSA300_1952

aspartate kinase SAUSA300_1225

conserved hypothetical protein SAUSA300_1084

transcription antiterminator, glcT SAUSA300_1253

carbamoyl phosphate synthase large subunit carB SAUSA300_1096

amidohydrolase SAUSA300_0534

conserved hypothetical protein SAUSA300_0059

ATP-dependent RNA helicase, DEAD/DEAH box family SAUSA300_1518

tRNA modification GTPase TrmE SAUSA300_2646

glycerol kinase glpA SAUSA300_1192

hypothetical protein SAUSA300_2593

hypothetical protein SAUSA300_111

hypothetical protein SAUSA300_1208

hypothetical protein SAUSA300_0585

Table 4.1: S. aureus mutants with significantly faster resuscitation after ciprofloxacin treatment

compared to the plate average.

Gene Locus ID

conserved hypothetical protein SAUSA300_1560

putative tetracycline resistance protein SAUSA300_0139

excinuclease ABC, A subunit SAUSA300_0742

77

conserved hypothetical protein SAUSA300_0097

putative membrane protein SAUSA300_0351

isopropylmalate synthase-related protein SAUSA300_0879

multidrug resistance protein SAUSA300_2360

putative lipoprotein SAUSA300_2403

conserved hypothetical phage protein SAUSA300_1936

fibronectin binding protein A SAUSA300_2441

staphylococcal tandem lipoprotein SAUSA300_2428

methylenetetrahydrofolate dehydrogenase SAUSA300_0965

transporter gate domain protein SAUSA300_2520

conserved hypothetical protein SAUSA300_1902

conserved hypothetical protein SAUSA300_0595

aminotransferase, class V SAUSA300_1662

glycerol-3-phosphate dehydrogenase SAUSA300_1193

conserved hypothetical protein SAUSA300_0847

conserved hypothetical protein SAUSA300_2311

thymidine kinase SAUSA300_2073

lipoic acid synthetase SAUSA300_0829

putative exonuclease SAUSA300_1970

2,3-bisphosphoglycerate-dependent phosphoglycerate mutase SAUSA300_2362

branched-chain amino acid aminotransferase SAUSA300_0539

deoxyribose-phosphate aldolase SAUSA300_0140

DNA translocase FtsK SAUSA300_1169

putative membrane protein SAUSA300_0279

putative membrane protein SAUSA300_0917

conserved hypothetical protein SAUSA300_1871

phytoene dehydrogenase SAUSA300_2501

phosphoribosylformylglycinamidine synthase I SAUSA300_0970

putative D-isomer specific 2-hydroxyacid dehydrogenase SAUSA300_0179

ABC transporter, permease protein SAUSA300_2307

adenylosuccinate lyase SAUSA300_1889

78

conserved hypothetical protein SAUSA300_1537

conserved hypothetical protein SAUSA300_0780

succinyl-CoA synthetase, beta subunit SAUSA300_1138

amidophosphoribosyltransferase SAUSA300_0972

ATP synthase F1, alpha subunit SAUSA300_2060

4-diphosphocytidyl-2C-methyl-D-erythritol kinase SAUSA300_0472

UTP-glucose-1-phosphate uridylyltransferase SAUSA300_2439

cmp-binding-factor 1 SAUSA300_1791

conserved hypothetical protein SAUSA300_0655

5'-methylthioadenosine/S-adenosylhomocysteine nucleosidase SAUSA300_1558

quinol oxidase, subunit I SAUSA300_0962

magnesium transporter SAUSA300_0910

phosphoribosylaminoimidazole carboxylase, ATPase subunit SAUSA300_0967

fructose specific permease SAUSA300_0685

cytochrome oxidase assembly protein SAUSA300_1015

putative competence protein SAUSA300_0901

alpha,alpha-phosphotrehalase SAUSA300_0968

rod shape-determining protein MreD SAUSA300_1604

tyrosine recombinase xerC SAUSA300_1145

tRNA delta(2)-isopentenylpyrophosphate transferase SAUSA300_1195

alkyl hydroperoxide reductase subunit C SAUSA300_0380

ATP-dependent Clp protease proteolytic subunit SAUSA300_0752

ribosomal RNA small subunit methyltransferase B SAUSA300_1110

phosphoribosylaminoimidazole-succinocarboxamide synthase SAUSA300_0968

Na+/H+ antiporter family protein SAUSA300_0846

L-serine dehydratase, iron-sulfur-dependent, beta subunit SAUSA300_2470

multi drug resistance protein, norA SAUSA300_0680

hypothetical protein SAUSA300_2402

phosphoribosylaminoimidazole synthetase purM SAUSA300_0973

RNA polymerase sigma factor SigB rpoF SAUSA300_2022

cobalt transporter ATP-binding subunit cbiO SAUSA300_2176

79

putative lipoprotein SAUSA300_2355

hypothetical protein SAUSA300_0373

acetyl-CoA c-acetyltransferase, VraB SAUSA300_0560

arginine repressor, ArgR SAUSA300_0066

permease, LctP SAUSA300_0112

glyoxalase family protein SAUSA300_0338

phiSLT ORF527-like protein SAUSA300_1391

hypothetical protein SAUSA300_1484

hypothetical protein SAUSA300_1722

prephenate dehydrogenase SAUSA300_1260

hypothetical protein SAUSA300_0775

hypothetical protein SAUSA300_1543

glyceraldehyde 3-phosphate dehydrogenase 2, gap SAUSA300_1633

ATP-dependent DNA helicase RecG SAUSA300_1120

malate:quinone-oxidoreductase, mqo SAUSA300_2541

Leukocidin/Hemolysin toxin family protein SAUSA300_1974

phosphate starvation-induced protein, PhoH family SAUSA300_1531

gamma-hemolysin component A SAUSA300_2365

conserved hypothetical protein SAUSA300_0356

conserved hypothetical protein SAUSA300_0317

geranyltranstransferase SAUSA300_1470

anti-sigma-B factor, serine-protein kinase rsbW SAUSA300_2023

respiratory nitrate reductase, subunit delta narJ SAUSA300_2341

hypothetical protein SAUSA300_1745

ATP-binding Mrp/Nbp35 family protein SAUSA300_2125

glycine cleavage system protein H gcvH SAUSA300_0791

quinol oxidase, subunit III qoxC SAUSA300_0961

transcriptional repressor CodY SAUSA300_1148

dihydrolipoamide dehydrogenase lpdA SAUSA300_0996

mannose-6-phosphate isomerase manA SAUSA300_2096

30S ribosomal protein S1 rpsA SAUSA300_1365

80

branched-chain alpha-keto acid dehydrogenase subunit E2 SAUSA300_099

hypothetical protein SAUSA300_1213

hypothetical protein SAUSA300_1899

ABC transporter permease SAUSA300_2358

hypothetical protein SAUSA300_0841

hypothetical protein SAUSA300_1739

PemK family protein SAUSA300_2026

anti-sigma-B factor, antagonist rsbV SAUSA300_2024

glycine betaine/carnitine/choline ABC transporter ATP-binding protein opuCa SAUSA300_2393

LysR family regulatory protein gltC SAUSA300_0444

hypothetical protein SAUSA300_0844

transcription antitermination protein NusB SAUSA300_1473

hypothetical protein SAUSA300_1908

arginine repressor argR SAUSA300_1469

Table 4.2: S. aureus mutants with significantly faster resuscitation after ciprofloxacin treatment

compared to the plate average.

4.3 Discussion

In total we identified 46 mutants with significantly faster resuscitation and 116 mutants with

significantly slower resuscitation compared to the bulk of the tested mutants. Mutants with a

fast resuscitation phenotype (Table 4.1) represent genes that are expected to play a role in

maintaining the persister state or impede growth resumption following antibiotic treatment.

Mutants with a slow resuscitation phenotype (Table 4.2) represent genes that are expected to

promote resuscitation from the persister state.

81

Bacterial populations resuming growth after any period of slow or no growth undergo a lag

phase. The duration of lag phase depends on many environmental factors, including prior

stresses such as nutrient deprivation or growth inhibitors [86]. Diluting cells in a stationary state

into fresh media results in a short lag, then rapid exponential growth. It is likely that many of the

cellular pathways active during growth resumption from stationary phase are also active during

recovery from antibiotic treatment. We performed a counter-screen to identify mutants that

have simply have naturally slow or fast growth phenotypes. We focused on genes that play a role

specifically in growth resumption following antibiotic treatment.

We have shown that stochasticity in gene expression results in tolerant subpopulations; it is

possible that stochastic gene expression also launches individual cells into regrowth when

conditions are suitable. Our results show that mutants in metabolic genes are generally slower

to resume growth following antibiotic treatment. Possibly persister cells that resume growth

when conditions are favorable simply exhibit a fortunate combination of gene expression and

timing. The pathways necessary to resume growth after antibiotic treatment are not yet fully

understood, but it is likely that some of the processes necessary for recovery from the persister

state depend on the type of antibiotic used. For example, cells resuming growth after treatment

with DNA-damaging drugs likely need to activate DNA repair and nucleotide synthesis pathways.

Indeed, we see evidence of this in our screen; a purM mutant has a delayed resuscitation

phenotype. Resuscitation after treatment with cell-wall acting antibiotics might require increased

peptidoglycan synthesis. Further research is needed to describe the mechanism of persister cell

resuscitation.

82

4.5 Material and Methods

4.5.1 Strains and culture conditions

E. coli strain MG1655 was used for the initial proof of principle resuscitation experiment. Cultures

were grown in 14mL culture tubes at 37°C in 3mL LB shaking 220rpm. Overnight cultures were

used to inoculate in fresh LB and cultures were grown for 3 hours. Ampicillin 100g/mL was used

where indicated and penicillinase was added where indicated either at 0 hours (ctrl) or after 1

hour of killing with ampicillin. Time points were taken every hour and aliquots were taken from

cultures and diluted in PBS. Dilutions were plated on LBA for cfu.

4.5.2 Resuscitation Screen

The Nebraska Transposon Mutant Library was used to conduct the resuscitation screen [65].

Experimental screening plates (96-well, flat-bottom clear) containing 100uL of MHB per well

were stamped from the 384-well library plates. Plates were covered with a breathable membrane

and incubated 16 hours sharing at 220rpm at 37°C. Overnight plates were used to inoculate a

fresh 96-well plate 1:100. For full growth curves to monitor the emergence from stationary

phase, growth was monitored using a plate reader. For resuscitation experiments, overnight

plates were used to inoculate plates with fresh media and plates were incubated at 37°C for 3

hours. Mutants in each well were treated with ciprofloxacin 320ug/mL and incubated at 37°C for

24 hours. After treatment, plates were centrifuged for 10 minutes at 4000 rpm and media was

removed. Cells were resuspended in 100uL MHB. Resuspended cells were diluted 1:10 into fresh

MHB and growth was monitored in a plate reader. OD was recorded every 30 minutes.

83

4.5.3 Data analysis

Output OD600 values from each plate were recorded, with time points every 30 minutes. For each

well, the time required for the mutant to triple its starting OD was calculated. The mean of all 96

well was calculated, and those mutants with T3OD outside of 1.5 standard deviations from the

mean were considered screen hits. Fast resuscitators were T3OD >1.5 SDbulk and slow resuscitators

were T3OD <1.5 SDbulk.

Chapter 5. Clinical isolates of Staphylococcus aureus

5.1 Abstract

High-persister clinical strains of Staphylococcus aureus have been characterized in multiple

species [5, 7, 87]. We measure persister levels of isolates from patients with endocarditis,

osteomyelitis, skin and soft tissue infections, and atopic dermatitis. Isolates from patients

exhibited universally high persister levels compared to wild type lab strains. We perform whole

genome sequencing of isolates using Illumina Hi-Seq. Variation analysis was performed and high

confidence mutations were annotated and identified as insertions, deletions, or synonymous or

nonsynonymous single nucleotide polymorphisms (SNPs). All mutations predicted to be nonsilent

were mapped to a biological subsystem, resulting in a list of biological systems that are

potentially involved in antibiotic tolerance. The TCA cycle is a major driver of ATP generation for

S. aureus in vivo, so we expected that mutations in TCA cycle enzymes might contribute to high

antibiotic tolerance [4]. We perform multiple sequence alignments of all encoded TCA cycle

84

enzymes for each clinical isolate and find several amino acid substitutions in TCA cycle genes gltA,

acnA, icd, sucA, sucB, sucD, sucC, sdhA, sdhB, and fumC.

5.2 Introduction

Although an estimated 30% of humans are colonized with commensal Staphylococcus aureus, it

can cause a wide range of infections including bacteremia, endocarditis, osteomyelitis,

meningitis, toxic shock syndrome, and skin and soft tissue infections [20]. Colonization increases

an individual’s risk for infection. In a study of bacteremia, blood isolates were identical to

colonizing nasal strains in 82% of patients [88]. Methicillin-resistant S. aureus (MRSA) is the most

widely-known example of antibiotic-resistant S. aureus, but additional cases of resistant S. aureus

have emerged in recent years. The spread of vancomycin intermediate S. aureus (VISA) also adds

to the concerns about S. aureus infection [20, 89]. Although improvements in infection control

procedures have been effective in reducing the spread of resistant S. aureus, it remains a major

public health problem.

Community-associated S. aureus strains are more likely to be susceptible to antibiotics but the

sheer number of community-associated methicillin-resistant S. aureus (CA-MRSA) infections has

increased in recent years [90].

In the early 2000s, there was a sharp increase in the number of cases of MRSA infections that

were not associated with hospitalization [91, 92]. This spread was eventually attributed to the

USA300 North American clone. This clone has an expanded set of virulence genes and multiple

antibiotic resistance cassettes, which give strains an exceptional ability to establish and maintain

skin and soft tissue infections. USA300 strains are resistant to penicillin and often oxacillin and

85

erythromycin. Many strains exhibit decreased susceptibility to fluoroquinolones. The virulence

genes encoded by USA300 clones include lukS-PV/lukF-PV, sek, and seq [93]. The USA300 clone

has spread across North America, Europe, and Asia, making it one of the most widespread CA-

MRSA clones [93-96]. It remains a global health problem.

The majority of MRSA infections – approximately 80% - are healthcare-associated [25]. Invasive

surgery, medical device insertion, and immunodepression all increase the likelihood of

developing a healthcare associated S. aureus infection [97]. Infections on indwelling devices often

form biofilms, which further complicates treatment. Biofilm-associated infection are difficult to

treat and often persist over time. The epidemiology of S. aureus transmission and infection is not

well understood but increases in hospital contagion surveillance has resulted in more information

on the spread of the pathogen.

5.3 Results

The decreasing cost of DNA sequencing has improved our understanding of the epidemiology

and transmission of S. aureus [92, 98]. Very little work has been done using genomics to

understand host adaptation or the progression of chronic infection [99]. We use whole genome

sequencing to identify pathways and biological subsystems that harbor a high number of

mutations in clinical isolates of S. aureus.

The genetic basis of persister formation has been explored in several species [5, 52, 100-102].

Many of these investigations use Tn-seq, which allows for high-throughput in vitro screening of

a transposon library for genes involved in antibiotic tolerance. In a Tn-seq experiment, a

transposon mutant library is exposed to antibiotics and the survivors are cultured. Deep

86

sequencing of the surviving mutants allows for identification of mutants which did not survive

treatment; these mutations are in potential persister genes. Using a transposon library with

dense coverage means that each gene is represented with multiple insertions per coding or

promoter region. These experiments yield a robust list of genes contributing to antibiotic

tolerance. Tn-seq experiments in E. coli have shown that multiple pathways contribute to

persister formation. Genes involved in flagellar structure, amino acid metabolism, and the TCA

cycle were involved in tolerance to aminoglycosides [100]. Most Tn-seq experiments have been

conducted using E. coli mutant libraries but S. aureus has been studied recently [102].

We reasoned that mutations in metabolic pathways that impact ATP generation would affect

phenotypic susceptibility to antibiotics. Resistance mutations alter a strain’s inherent antibiotic

susceptibility by modifying the antibiotic target or enabling cells to pump out antibiotics [11].

High persister mutations permit the cells to tolerate treatment, but not to grow in the presence

of antibiotics. We expected that high persister strains of S. aureus would have mutations in

metabolic genes.

S. aureus isolates were taken from patients with pneumonia, osteomyelitis, or atopic dermatitis.

Clinical details of the isolates can be found in Table 5.1. We tested these patient isolates for

persister levels. Briefly, we determined the MIC of each strain to vancomycin, moxifloxacin, and

doxycycline (Supplemental Figure 5.1). We then treated isolates with 10x MIC and calculated

survival after 48 hours of treatment (Supplemental Figure 5.2). Some isolates had known

antibiotic resistance, and we excluded those drugs from consideration for experiments. Clinical

strains displayed generally increased antibiotic tolerance compared to a reference non-infecting

87

strain from a matched clonal complex. Bearing in mind our previous finding that low ATP leads

to antibiotic tolerance, we wondered whether there were genetically encoded defects in

metabolic pathways that might contribute to increased persister levels in these clinical strains.

Table 5.1: S. aureus clinical isolate source information and diagnosis. Includes isolate number,

isolation source, clonal complex, indication of paired/non-paired isolates and hospital diagnosis.

Key source CC Paired Diagnosis

1 SPUTUM 5 A hospital-acquired pneumonia

8 SPUTUM 5 A hospital-acquired pneumonia

11 SPUTUM 5 hospital-acquired pneumonia

25 WOUND 5 osteomyelytis

26 WOUND 5 osteomyelytis

27 WOUND 5 osteomyelytis

30 NARES 5 colonization

15 NARES 8 H colonization

16 BLOOD 8 H colonization

21 NARES 8 colonization

48 SKIN 5 atopic dermatitis

64 SKIN 5 atopic dermatitis

31 SKIN 5 atopic dermatitis

24 SKIN 209 atopic dermatitis

23 SKIN 22 atopic dermatitis

44 SKIN unk atopic dermatitis

43 SKIN unk atopic dermatitis

40 SKIN 188 atopic dermatitis

88

Variance analysis was performed in parallel for ST5 and ST8. Variants were classified as either

deletions, insertions, nonsynonymous SNP, or synonymous SNP. For ST5 strains, 2369 total

variants were called. For ST8, 3047 total variants were called. Figure 5.1 shows the breakdown

by variant type. Synonymous SNPs were excluded from analysis since they were predicted to be

low-impact variants not affecting the amino acid code.

Identifiers of genes impacted by variants were classified by biological subsystem. Total variants

(InDels or nonsynonymous SNPs) per biological subsystem were quantified. Biological

subsystems are used to categorize proteins with functionally related roles (i.e. enzymes in single

biological pathway). Impacted subsystems are represented in Figure 5.2. Since previous work has

Figure 5.1: Variant type classified as deletion, insertion, nonsynonymous SNP, or

synonymous SNP. Isolates were divided between clonal complex groups for variance

analysis to prevent inclusion of variants that are simply a result of clonal evolutionary

divergence. Isolates from ST5 (A) and ST8 (B) are shown here.

A

B

89

shown that low levels of ATP induce antibiotic tolerance, we were specifically interested in

metabolic subsystems. We suspected that variants in genes encoding metabolic enzymes would

be prevalent in S. aureus isolates. Indeed, genes involved in pyruvate metabolism and the TCA

cycle harbored a high number of variants in clinical strains.

We suspected that mutations in TCA cycle genes would be detected in clinical isolates. Using the

Pathosystems Resource Integration Center, we compared variants in the TCA cycle between each

isolate. We found that the clinical isolates harbored a high number of high-impact mutations in

genes encoding TCA cycle enzymes. We considered variants predicted to alter amino acid

sequence with each coding region. We performed multiple sequence alignment of amino acid

sequence for TCA cycle proteins using the reference genome USA100. The heat map in Figure 5.2

shows the number of high-impact mutations in each TCA cycle gene. Specific amino acid

substitutions are described in Supplemental Table 5.3. Interestingly, we observed a high number

of mutations in late TCA cycle genes, especially fumC. FumC is responsible for converting

fumarate to malate. This step of the TCA cycle occurs after the glutamate portal into the TCA

cycle and mutation in fumC is likely to decrease ATP generation in cells using glutamate to drive

ATP generation via the TCA cycle. Future research is needed to determine the impact of these

mutations on ATP generation.

90

Figure 5.2: Biological subsystems implicated in genome variance analysis. Paired reads were mapped

against either S. aureus N315 (PATRIC Genome ID 158879.11) for ST5 (A) or S. aureus USA300_FPR3757

(PATRIC Genome ID 451515.3) for ST8 (B). Paired reads were aligned using BWA-mem and Freebayes

was used to call variants. Variants affecting amino acid sequence were grouped into biological

subsystems using Pathosystems Resource Integration Center (PATRIC) identifiers. Subsystems are listed

from those with the highest number of variants to lowest, with an arbitrary cutoff of 7

variants/subsystem.

A

B

91

5.4 Methods

5.4.1 S. aureus isolate sources

Isolates were collected from patients diagnosed with pneumonia, osteomyelitis, or atopic

dermatitis. Strains were provided by Bo Shopsin at NYU Langone Medical Center and Karen Acker

Primary cultures were grown in the laboratory and secondary stocks were made and frozen at -

80°C. Secondary stocks were used for all experiments.

5.4.2 Whole genome sequencing and read mapping

Whole genome sequencing was performed using Illumina Hi-Seq 4000. Three PCR amplification

cycles were performed in the DNA library preparation. Using the Pathosystems Resource

Integration Center (PATRIC) platform, paired-end reads were mapped against a reference

genome. Reads from samples in ST5 were mapped against Staphylococcus aureus N315 (PATRIC

Genome ID 158879.11) Reads from ST8 were mapped against Staphylococcus aureus

USA300_FPR3757 (PATRIC Genome ID 451515.3). Variance analysis was performed using

Burrows-Wheeler Aligner (BWA) with minimal exact matches. FreeBayes was used to call SNPs.

High confidence variants were used to compare mutations detected in clinical isolates compared

to reference strains.

Multiple sequence alignments as previously described [103, 104]. Coding sequences for specific

enzymes were aligned and amino acid sequences were compared. Variants were identified com

pared to reference genome USA100 of known clonal complex 5 (PATRIC Genome ID 1280.19099

). Comparisons were based on Rapid Annotations using Subsystems Technology (RAST) annotati

ons [105]. KEGG pathways [106] were used to identify protein families for comparison.

92

5.4.3 MIC assay and persister experiments

MIC assays were performed via broth microdilution assay in 96-well microtiter plates. Isolates

were grown in TSB with 100mM MOPS pH7. MICs were determined for each isolate for

moxifloxacin (Acros Organics, ThermoFisher Scientific, USA), vancomycin (Sigma-Aldrich, USA)

and doxycycline (ThermoFisher Scientific, USA).

For persister assays, strains were diluted 1:100 and grown to late exponential phase. Cultures

were grown in 2mL TSB shaking 220rpm at 37oC. Antibiotics were added at 10x MIC as indicated.

100uL aliquots were removed and washed with PBS. CFU was plated on TSA and recorded to

enumerate survivors.

5.5 Discussion

As previous work has shown, multiple factors can contribute to antibiotic tolerance [100]. Specific

gain-of-function mutations in hipA7 in E.coli leads to increased tolerance [5]. Null mutations in

the carB gene, encoding carbomyl phosphate synthetase leads to decreased tolerance in multiple

species [61]. In vitro work has shown that antibiotic treatment drives the evolution of antibiotic

tolerance [107]. Selection for tolerance mutations result in antibiotic treatment failure.

Sequencing has become more common in diagnosing infectious disease, but genetic screening

for mutations that enable bacteria to tolerate treatment are not part of standard care in most

settings.

Mutations that facilitate antibiotic tolerance are not necessarily advantageous under all

conditions. On the contrary, expression of toxins like hipA7 or mutations in important metabolic

genes often lead to growth defects or fitness disadvantages under some conditions. However,

93

we suspect that these mutations are adaptive. Situations of antibiotic challenge select for cells

with less active cellular processes. These tolerance mutations have been shown to precede the

development of resistance mutations in vitro [12] and in vivo [7].

Since antibiotic application drives the acquisition of tolerance mutations, it is useful to study

clinical bacterial isolates. Isolates are taken from patients during their course of treatment,

sometimes over the course of years. High-persister mutants of Pseudomonas aeruginosa,

Mycobacterium tuberculosis, Candida albicans, and Escherichia coli have been isolated from

patients with cystic fibrosis, tuberculosis, candidiasis, and urinary tract infections [5, 7, 8, 13]. We

test isolates of S. aureus from multiple patients for antibiotic tolerance and perform whole

genome sequencing to identify potential tolerance mutations. Our isolates represent multiple

types of infections and distinct clonal lineages. We show that metabolic pathways harbor high-

impact mutations in clinical isolates. Other groups have shown that mutations in the TCA cycle

are frequently found in clinical isolates [108]. Our results support this conclusion, but more

research on larger collections of clinical isolates is needed to draw conclusions about antibiotic

selection for persister mutations. We suspect that isolates from district sites of infection will

accumulate mutations in different metabolic pathways. For example, conversion of proline in

collagen to glutamate then 2-oxoglutarate is expected to drive ATP generation via the TCA cycle

[4] in a staphylococcal abscess.

94

Figure 5.3: High-impact variants in TCA cycle genes. Mutations affecting amino acid sequence 9f isolates (ID: isolate number, vertical axis) were identified per TCA cycle gene. Comparative pathways were analyzed in PATRIC, based on RAST annotation and multiple sequence alignment of TCA cycle genes.

95

5.6 Supplemental Information

moxifloxacin vancomycin doxycycline

1 16 4 0.125

8 4 1 0.125

11 1 8 0.25

15 4 8 4

16 1 0.125 0.125

21 1 8 0.25

25 4 4 0.25

26 8 8 0.25

27 4 8 0.125

30 16 4 0.25

PE001 0.0625 4 0.0078

PE002 0.25 4 0.25

PE003 0.125 4 0.25

PE004 0.25 8 0.25

PE005 0.125 8 0.25

PE007 0.0625 8 0.0078

PE008 0.125 8 0.25

PE010 0.0625 8 0.0078

HG WT 1 4 0.125

USA300 WT 8 4 0.25

96

Supplemental Figure 5.1: MIC for clinical isolates and wild type lab strains. Concentration

units are g/mL. MIC values were determined using broth dilution method in microtiter plates.

Figure 4.2: Percent survival after treatment with vancomycin. Clinical isolates were grown to

late exponential phase and treated with 10x MIC vancomycin. Blue represents S. aureus ST5

and orange represents ST8. CFU was counted after 72h treatment. Experiment was performed

in biological triplicate. Bars represent SEM.

97

gltA acnA icd sucA sucB sucD sucC sdhA sdhB fumC

1

8

11

15 S152N

L369F

E534D P176S

L56I

I122S

E177D

I211T

16 S152N

L369F

E534D P176S

L56I

I122S

E177D

I211T

21

P55S

E534D

P176S

L56I

I122S

E177D

I211T

25

D278N

26

27

98

30

D278N

48 A37T

64

N259K

R262K

S296A T203I

E134K

E177D

31

24 H88Y

N259K

E134K

E177D

23

Q863H N259K

A267S

G914R

E172K M365T

S411A

D435E

L440M

D488N

E3D

S5P

N8D

I122S

E134K

E177D

44

N259K E534D P176S E197K

S296A

L56I

I122S

E177D

I211T

43

E534D P176S

L56I

I122S

E177D

I211T

40

V609I

H550Y P166T S296A

E134K

99

5.7 Contributions

We thank Paul Planet, Karen Acker, and Bo Shopsin for providing clinical isolates and sequencing

these strains.

P802S P176S E177D

Supplemental Table 5.3: Amino acid substitutions in TCA cycle genes. Specific amino acid

substitutions in clinical S. aureus isolates. RAST annotation was used to identify TCA cycle

subsystem genes. MSA was performed based on Freebayes variant calling. Isolate numbers are

listed in the left-hand column with amino acid changes identified per gene. Only TCA cycle genes

are shown. USA100 was used as a reference genome.

100

Chapter 6. Conclusions and Future Directions

6.1 Summary

This work describes a general mechanism of persister formation mediated by low ATP. This

mechanism was first characterized in S. aureus and appears to be conserved among multiple

species [35, 61, 62]. It was previously believed that toxin-antitoxin systems mediated antibiotic

tolerance, but we showed that deletion of the known TA systems in S. aureus does not impact

tolerance. Rather, ATP depletion leads to increased tolerance.

We sought to understand the mechanism by which cells naturally enter the tolerant persister

state. We investigated the role of the TCA cycle in persister formation. We knew that the TCA

cycle and amino acid flux into the TCA cycle were major drivers of cellular growth in S. aureus in

vivo [4, 109] and expected that TCA cycle defects would impact ATP generation and therefore

antibiotic tolerance. We identified in vitro conditions that would mimic the metabolic state at an

infection site, where glucose concentrations are believed to be low and cells must rely on

secondary carbon sources. We assembled the full proteome of S. aureus during exponential

growth and stationary phase and showed that enzymes involved in the TCA cycle and amino acid

catabolism were more abundant during stationary phase than during exponential phase. We

therefore focused on the transition phase from exponential to stationary growth to investigate

the role of the TCA cycle in persister formation. We constructed mutants lacking functional TCA

components gltA (citrate synthase), gudB (glutamate dehydrogenase), sucA (2-oxoglutarate

dehydrogenase), sucC (succinyl-CoA synthetase), and fumC (fumarate hydratase). We studied the

gudB mutant in addition to the canonical TCA cycle genes because it catalyzes the conversion of

101

glutamate to 2-oxoglutarate. This point of entry into the TCA cycle is important in vivo, where

proline derived from collagen is converted to glutamate, which enters the TCA cycle as 2-

oxoglutarate, thus fueling energy generation. We found that these TCA cycle mutants had

decreased intracellular ATP compared to wild type S. aureus. Since ATP concentrations change

dynamically in single cells, we optimized the ATP biosensor QUEEN for expression in S. aureus to

quantify ATP in single cells with minimal perturbation. We used flow cytometry to quantify ATP

and found that TCA cycle defects caused a population shift where more single cells exhibited low

intracellular ATP. This confirmed our low-ATP phenotype in TCA mutants with a novel method.

These mutants also exhibited increased antibiotic tolerance after treatment with ciprofloxacin,

gentamicin, and oxacillin. These phenotypes could have been due to pleiotropic effects of

mutating major metabolic genes. We wanted to understand the role of the TCA cycle in native

persister cells within a population. We reasoned that natural fluctuation in gene expression gives

rise to phenotypic heterogeneity. We hypothesized that within a population, individual cells

expressing relatively low levels of TCA cycle enzymes would have low ATP and therefore be more

tolerant to antibiotic treatment. We made reporter strains using GFP to monitor TCA gene

expression and sorted single cells after antibiotic treatment. We found that the fraction of the

population expressing low TCA cycle genes were enriched for persister cells.

Persister cells are only interesting because they can resume growth. This is believed to cause

infection relapse. We demonstrated persister resuscitation in vitro. We also sought to identify

pathways involved in persister resuscitation and conducted a screen of nearly 2000 transposon

mutants to find genes implicated in resuscitation. We identified 46 mutants with significantly

faster resuscitation and 116 mutants with significantly slower resuscitation compared to the bulk

102

of the mutant screened. Many of these mutants were defective in metabolic pathways and

nucleotide repair. They offer a promising list of candidate genes for ongoing and future research.

We wondered whether there were genetic signatures of antibiotic tolerance in clinical isolates of

S. aureus. Persister mutations have been identified in other pathogenic species [5, 7, 14]. We test

a collection of clinical isolates for persister levels and find high persister levels in all isolates

compared to wild type lab strains. We performed whole genome sequencing of clinical isolates.

We focused specifically on the TCA cycle and found that mutations in the TCA cycle were

surprisingly common among clinical isolates from multiple types of infections. Interestingly, fumC

appeared to be the TCA cycle gene harboring the most high-impact variants. FumC is a late TCA

cycle enzyme that is responsible for converting fumarate to malate. Any defects in the TCA cycle

that occur in steps after the incorporation of glutamate are likely to decrease ATP generation.

6.2 Ongoing research and future directions

6.2.1 Persister resuscitation

Many groups are working to fill the gaps in our understanding of persister formation and

importantly, the role of persister cells in chronic infection. The phenomenon of resuscitation

from the persister state is well-documented [3, 11, 12] but the mechanism by which cells wake

up from a tolerant state is poorly understood. Some groups have proposed a microbial “scout”

hypothesis, suggesting that stochastic growth resumption among a population of dormant cells

causes persister resuscitation [81]. Other groups have proposed that signaling molecules serve

as wake-up calls for dormant cells [82]. Others have observed different resuscitation kinetics

103

depending on the type of antibiotic treatment [83]. It is possible that bacteria have evolved

multiple strategies for emerging from dormant or tolerant states, but these strategies are not

well described.

This work shows that metabolic defects enable cells to survive antibiotic treatment. Resumed

metabolic activity and ATP generation are therefore likely to correspond with persister

resuscitation. Recent work used incorporation of isotope-labelled amino acids to identify

proteins that are synthesized during post-antibiotic recovery [85]. This characterization of the

proteome during post-treatment gives clues about the processes involved in resuscitation.

Nutrient uptake and nucleotide repair emerged from that work and from our screen as important

processes in persister resuscitation. PurM stands out as one candidate resuscitation factor in S.

aureus. Future work will explore specific pathways involved in the emergence from dormancy.

6.2.2. Noise-quenching

Biological populations are naturally heterogeneous. Bacterial phenotypic heterogeneity can be

influenced by many factors including chemical gradients of signaling factors, access to nutrients,

population density, and noisy gene expression [110]. This work shows that natural stochasticity

in metabolic gene expression results in subpopulations that are more tolerant to antibiotics. This

phenotypic heterogeneity can be manipulated by noise quenching. Overexpression of metabolic

genes is expected to increase metabolic flux and increase ATP levels. Bacterial metabolism is

complex and highly regulated and thus quenching noisy gene expression requires identification

of bottlenecks in ATP synthesis. This work identified the TCA cycle and amino acid metabolism as

potential targets for noise-quenching experiments. Ongoing research is aimed at controlling

104

carbon input to direct metabolic flux through known pathways then overexpressing bottleneck

genes to maximize ATP production. We predict that this will result in improved killing by

antibiotics and eradication of persister cells. Further research will be needed to identify the

bottlenecks in bacterial metabolism in vivo. Preliminary evidence suggests that proline and

glutamate catabolism, which this work identified as important for ATP generation in vitro, is

required for bacterial growth in vivo. The S. aureus genome encodes multiple proteases that can

be used to degrade host proteins [111]. It is also known that host-derived proteases are induced

during staphylococcal abscess formation [112, 113]. It is likely that S. aureus uses its own and

host-derived proteases to break down collagen to liberate proline, which is then converted to

arginine and glutamate [4]. Glutamate flux into the TCA cycle serves as a potential noise-

quenching target that could spike intracellular ATP and increase susceptibility to antibiotic

treatment.

6.2.3. Eradicating persister cells

The goal of studying antibiotic tolerance is to eradicate bacterial infections. Infectious disease is

a major cause of death worldwide. The spread of antibiotic-resistant bacteria is expected to

contribute to increasing cases of fatal infections. New antibiotics and careful and treatment are

needed to combat this crisis. Treatment failure due to tolerant persister cells must also be

addressed in the future of infectious disease treatment. This work shows that bacteria with low

ATP can survive antibiotic treatment. Antibiotics with ATP-independent targets offer a potential

solution to the problem of antibiotic tolerance. One such antibiotic is aceyldepsipeptide 4

(ADEP4), which is capable of killing persister cells. ADEP4 activates the ClpP protease, causing

105

unregulated extensive proteolysis. ADEP4 dissociates ClpP from its ATP-dependent chaperones,

resulting in ATP-independent killing by proteolysis. ADEP4 is eradicates S. aureus biofilms in vitro

and in a mouse model [38]. ADEP4 is effective in killing persister cells because it does not rely on

ATP to cause cell death. Compounds that are effective in killing non-replicating bacteria have

been also been identified in M. tuberculosis [114]. New antibiotics that kill in an ATP-independent

manner offer promise for eradicating dormant persister populations.

In addition to discovering new antibiotics, adjuvant therapies offer another option for treatment

of chronic or recurring infections. Metabolite adjuvants have been proposed as an effective way

to eliminate persister populations. Other groups have observed metabolite-mediated eradication

of aminoglycoside-tolerant populations. Although this phenomenon appears restricted to

aminoglycosides, metabolic stimulation was effective in potentiating aminoglycoside killing in a

mouse model and has possible for clinical impact [115, 116]. This is a promising avenue for

treating recurrent infections because metabolites to be used in combination with existing

antibiotics face fewer regulatory hurdles compared to entirely new antibiotics. Fumarate, for

example, is a potentially appealing tobramycin adjuvant and has already been approved by the

Food and Drug Administration for treating asthma [117]. Further research is needed to identify

metabolic stimulants that are effective in combination with existing antibiotics.

106

References Cited

1. Bigger, J., Treatment of Staphyloeoeeal Infections with Penicillin by Intermittent Sterilisation. Lancet, 1944: p. 497-500.

2. Lewis, K., Persister cells. Annual review of microbiology, 2010. 64: p. 357-372. 3. Balaban, N.Q., et al., Bacterial persistence as a phenotypic switch. Science, 2004. 305(5690): p.

1622-1625. 4. Halsey, C.R., et al., Amino acid catabolism in Staphylococcus aureus and the function of carbon

catabolite repression. MBio, 2017. 8(1): p. e01434-16. 5. Schumacher, M.A., et al., HipBA–promoter structures reveal the basis of heritable multidrug

tolerance. Nature, 2015. 524(7563): p. 59. 6. LaFleur, M.D., C.A. Kumamoto, and K. Lewis, Candida albicans biofilms produce antifungal-

tolerant persister cells. Antimicrobial agents and chemotherapy, 2006. 50(11): p. 3839-3846. 7. Mulcahy, L.R., et al., Emergence of Pseudomonas aeruginosa strains producing high levels of

persister cells in patients with cystic fibrosis. Journal of bacteriology, 2010. 192(23): p. 6191-6199.

8. Keren, I., et al., Characterization and transcriptome analysis of Mycobacterium tuberculosis persisters. MBio, 2011. 2(3): p. e00100-11.

9. Sharma, B., et al., Borrelia burgdorferi, the causative agent of Lyme disease, forms drug-tolerant persister cells. Antimicrobial agents and chemotherapy, 2015: p. AAC. 00864-15.

10. Helaine, S., et al., Internalization of Salmonella by macrophages induces formation of nonreplicating persisters. Science, 2014. 343(6167): p. 204-208.

11. Blair, J.M., et al., Molecular mechanisms of antibiotic resistance. Nature reviews microbiology, 2015. 13(1): p. 42.

12. Levin-Reisman, I., et al., Antibiotic tolerance facilitates the evolution of resistance. Science, 2017: p. eaaj2191.

13. LaFleur, M.D., Q. Qi, and K. Lewis, Patients with long-term oral carriage harbor high-persister mutants of Candida albicans. Antimicrobial agents and chemotherapy, 2010. 54(1): p. 39-44.

14. Torrey, H.L., et al., High persister mutants in Mycobacterium tuberculosis. PLoS One, 2016. 11(5): p. e0155127.

15. Niederman, M.S., Impact of antibiotic resistance on clinical outcomes and the cost of care. Critical care medicine, 2001. 29(4): p. N114-N120.

16. Capitano, B., et al., Cost effect of managing methicillin‐resistant Staphylococcus aureus in a long‐term care facility. Journal of the American Geriatrics Society, 2003. 51(1): p. 10-16.

17. Conlon, B.P., Staphylococcus aureus chronic and relapsing infections: Evidence of a role for persister cells: An investigation of persister cells, their formation and their role in S. aureus disease. Bioessays, 2014. 36(10): p. 991-996.

18. Wertheim, H.F., et al., The role of nasal carriage in Staphylococcus aureus infections. The Lancet infectious diseases, 2005. 5(12): p. 751-762.

19. Warshawsky, B., et al., Hospital-and community-based surveillance of methicillin-resistant Staphylococcus aureus: previous hospitalization is the major risk factor. Infection Control & Hospital Epidemiology, 2000. 21(11): p. 724-727.

20. Tong, S.Y., et al., Staphylococcus aureus infections: epidemiology, pathophysiology, clinical manifestations, and management. Clinical microbiology reviews, 2015. 28(3): p. 603-661.

21. Dantes, R., et al., National burden of invasive methicillin-resistant Staphylococcus aureus infections, United States, 2011. JAMA internal medicine, 2013. 173(21): p. 1970-1978.

107

22. Von Eiff, C., C. Heilmann, and G. Peters, New aspects in the molecular basis of polymer-associated infections due to staphylococci. European Journal of Clinical Microbiology and Infectious Diseases, 1999. 18(12): p. 843-846.

23. Von Eiff, C., et al., Infections associated with medical devices. Drugs, 2005. 65(2): p. 179-214. 24. Götz, F., Staphylococcus and biofilms. Molecular microbiology, 2002. 43(6): p. 1367-1378. 25. Percival, S.L., et al., Healthcare-associated infections, medical devices and biofilms: risk,

tolerance and control. Journal of medical microbiology, 2015. 64(4): p. 323-334. 26. Domenech, M., et al., Biofilm formation avoids complement immunity and phagocytosis of

Streptococcus pneumoniae. Infection and immunity, 2013: p. IAI. 00491-13. 27. Paharik, A.E. and A.R. Horswill, The staphylococcal biofilm: adhesins, regulation, and host

response. Microbiology spectrum, 2016. 4(2). 28. de Oliveira, S., E.E. Rosowski, and A. Huttenlocher, Neutrophil migration in infection and wound

repair: going forward in reverse. Nature Reviews Immunology, 2016. 16(6): p. 378. 29. Kolaczkowska, E. and P. Kubes, Neutrophil recruitment and function in health and inflammation.

Nature Reviews Immunology, 2013. 13(3): p. 159. 30. Guerra, F.E., et al., Epic immune battles of history: neutrophils vs. Staphylococcus aureus.

Frontiers in cellular and infection microbiology, 2017. 7: p. 286. 31. Crooke, A.K., et al., CcpA-independent glucose regulation of lactate dehydrogenase 1 in

Staphylococcus aureus. PLoS One, 2013. 8(1): p. e54293. 32. Fuller, J.R., et al., Identification of a lactate-quinone oxidoreductase in Staphylococcus aureus

that is essential for virulence. Frontiers in cellular and infection microbiology, 2011. 1: p. 19. 33. Richardson, A.R., S.J. Libby, and F.C. Fang, A nitric oxide–inducible lactate dehydrogenase

enables Staphylococcus aureus to resist innate immunity. Science, 2008. 319(5870): p. 1672-1676.

34. Vitko, N.P., N.A. Spahich, and A.R. Richardson, Glycolytic dependency of high-level nitric oxide resistance and virulence in Staphylococcus aureus. MBio, 2015. 6(2): p. e00045-15.

35. Conlon, B.P., et al., Persister formation in Staphylococcus aureus is associated with ATP depletion. Nature microbiology, 2016. 1(5): p. 16051.

36. Hede, K., Antibiotic resistance: an infectious arms race. Nature, 2014. 509(7498): p. S2-S3. 37. Monack, D.M., A. Mueller, and S. Falkow, Persistent bacterial infections: the interface of the

pathogen and the host immune system. Nature Reviews Microbiology, 2004. 2(9): p. 747. 38. Conlon, B.P., et al., Activated ClpP kills persisters and eradicates a chronic biofilm infection.

Nature, 2013. 503(7476): p. 365. 39. Balaban, N.Q., et al., A problem of persistence: still more questions than answers? Nature

Reviews Microbiology, 2013. 11(8): p. 587. 40. Maisonneuve, E. and K. Gerdes, Molecular mechanisms underlying bacterial persisters. Cell,

2014. 157(3): p. 539-548. 41. Keren, I., et al., Persister cells and tolerance to antimicrobials. FEMS microbiology letters, 2004.

230(1): p. 13-18. 42. Gerdes, K., S.K. Christensen, and A. Løbner-Olesen, Prokaryotic toxin–antitoxin stress response

loci. Nature Reviews Microbiology, 2005. 3(5): p. 371. 43. Donegan, N.P., et al., Proteolytic regulation of toxin-antitoxin systems by ClpPC in

Staphylococcus aureus. Journal of bacteriology, 2010. 192(5): p. 1416-1422. 44. Shan, Y., et al., Genetic basis of persister tolerance to aminoglycosides in Escherichia coli. mBio 6:

e00078-15. 2015. 45. Beenken, K.E., et al., Global gene expression in Staphylococcus aureus biofilms. Journal of

bacteriology, 2004. 186(14): p. 4665-4684.

108

46. George, S.E., et al., Phenotypic heterogeneity and temporal expression of the capsular polysaccharide in S taphylococcus aureus. Molecular microbiology, 2015. 98(6): p. 1073-1088.

47. Davis, B.D., L. Chen, and P.C. Tai, Misread protein creates membrane channels: an essential step in the bactericidal action of aminoglycosides. Proceedings of the National Academy of Sciences, 1986. 83(16): p. 6164-6168.

48. Malik, M., X. Zhao, and K. Drlica, Lethal fragmentation of bacterial chromosomes mediated by DNA gyrase and quinolones. Molecular microbiology, 2006. 61(3): p. 810-825.

49. Cho, H., T. Uehara, and T.G. Bernhardt, Beta-lactam antibiotics induce a lethal malfunctioning of the bacterial cell wall synthesis machinery. Cell, 2014. 159(6): p. 1300-1311.

50. Kwan, B.W., et al., Arrested protein synthesis increases persister-like cell formation. Antimicrobial agents and chemotherapy, 2013: p. AAC. 02135-12.

51. Steinbrecher, T., et al., Peptide-lipid interactions of the stress-response peptide TisB that induces bacterial persistence. Biophysical journal, 2012. 103(7): p. 1460-1469.

52. Dörr, T., M. Vulić, and K. Lewis, Ciprofloxacin causes persister formation by inducing the TisB toxin in Escherichia coli. PLoS biology, 2010. 8(2): p. e1000317.

53. Wang, X., et al., Type II toxin/antitoxin MqsR/MqsA controls type V toxin/antitoxin GhoT/GhoS. Environmental microbiology, 2013. 15(6): p. 1734-1744.

54. Moore, S.A., D. Moennich, and M. Gresser, Synthesis and hydrolysis of ADP-arsenate by beef heart submitochondrial particles. Journal of Biological Chemistry, 1983. 258(10): p. 6266-6271.

55. Imamura, H., et al., Visualization of ATP levels inside single living cells with fluorescence resonance energy transfer-based genetically encoded indicators. Proceedings of the National Academy of Sciences, 2009. 106(37): p. 15651-15656.

56. Lechner, S., K. Lewis, and R. Bertram, Staphylococcus aureus persisters tolerant to bactericidal antibiotics. J Mol Microbiol Biotechnol, 2012. 22(4): p. 235-44.

57. Mechler, L., et al., A novel point mutation promotes growth phase-dependent daptomycin tolerance in Staphylococcus aureus. Antimicrob Agents Chemother, 2015. 59(9): p. 5366-76.

58. Geiger, T., et al., Role of the (p) ppGpp synthase RSH, a RelA/SpoT homolog, in stringent response and virulence of Staphylococcus aureus. Infection and immunity, 2010. 78(5): p. 1873-1883.

59. Geiger, T., et al., Two small (p) ppGpp synthases in Staphylococcus aureus mediate tolerance against cell envelope stress conditions. Journal of bacteriology, 2014. 196(4): p. 894-902.

60. Cheung, A.L., C.C. Nast, and A.S. Bayer, Selective activation of sar promoters with the use of green fluorescent protein transcriptional fusions as the detection system in the rabbit endocarditis model. Infect Immun, 1998. 66(12): p. 5988-93.

61. Cameron, D.R., et al., A genetic determinant of persister cell formation in bacterial pathogens. Journal of Bacteriology, 2018. 200(17): p. e00303-18.

62. Shan, Y., et al., ATP-dependent persister formation in Escherichia coli. MBio, 2017. 8(1): p. e02267-16.

63. Harms, A., et al., Prophages and growth dynamics confound experimental results with antibiotic-tolerant persister cells. MBio, 2017. 8(6): p. e01964-17.

64. Somerville, G.A., et al., Staphylococcus aureus aconitase inactivation unexpectedly inhibits post-exponential-phase growth and enhances stationary-phase survival. Infection and immunity, 2002. 70(11): p. 6373-6382.

65. Fey, P.D., et al., A genetic resource for rapid and comprehensive phenotype screening of nonessential Staphylococcus aureus genes. MBio, 2013. 4(1): p. e00537-12.

66. Yaginuma, H., et al., Diversity in ATP concentrations in a single bacterial cell population revealed by quantitative single-cell imaging. Scientific reports, 2014. 4: p. 6522.

109

67. Cheung, A., C.C. Nast, and A. Bayer, Selective activation of sar promoters with the use of green fluorescent protein transcriptional fusions as the detection system in the rabbit endocarditis model. Infection and immunity, 1998. 66(12): p. 5988-5993.

68. Moyed, H.S. and K.P. Bertrand, hipA, a newly recognized gene of Escherichia coli K-12 that affects frequency of persistence after inhibition of murein synthesis. Journal of bacteriology, 1983. 155(2): p. 768-775.

69. Correia, F.F., et al., Kinase activity of overexpressed HipA is required for growth arrest and multidrug tolerance in Escherichia coli. Journal of bacteriology, 2006. 188(24): p. 8360-8367.

70. Kaspy, I., et al., HipA-mediated antibiotic persistence via phosphorylation of the glutamyl-tRNA-synthetase. Nature communications, 2013. 4: p. 3001.

71. Germain, E., et al., Molecular mechanism of bacterial persistence by HipA. Molecular cell, 2013. 52(2): p. 248-254.

72. Berghoff, B.A., et al., Two regulatory RNA elements affect T is B‐dependent depolarization and persister formation. Molecular microbiology, 2017. 103(6): p. 1020-1033.

73. Wang, Y., et al., Inactivation of TCA cycle enhances Staphylococcus aureus persister cell formation in stationary phase. Scientific reports, 2018. 8(1): p. 10849.

74. Liebeke, M., et al., A metabolomics and proteomics study of the adaptation of Staphylococcus aureus to glucose starvation. Molecular BioSystems, 2011. 7(4): p. 1241-1253.

75. Kyle, J.E., et al., Cell type-resolved human lung lipidome reveals cellular cooperation in lung function. Scientific reports, 2018. 8(1): p. 13455.

76. Cox, J., et al., Accurate proteome-wide label-free quantification by delayed normalization and maximal peptide ratio extraction, termed MaxLFQ. Molecular & cellular proteomics, 2014. 13(9): p. 2513-2526.

77. Forsyth, R.A., et al., A genome‐wide strategy for the identification of essential genes in Staphylococcus aureus. Molecular microbiology, 2002. 43(6): p. 1387-1400.

78. Grote, A., et al., JCat: a novel tool to adapt codon usage of a target gene to its potential expression host. Nucleic acids research, 2005. 33(suppl_2): p. W526-W531.

79. Schindelin, J., et al., Fiji: an open-source platform for biological-image analysis. Nature methods, 2012. 9(7): p. 676.

80. Haaber, J., et al., Simple method for correct enumeration of Staphylococcus aureus. Journal of microbiological methods, 2016. 125: p. 58-63.

81. Buerger, S., et al., Microbial scout hypothesis, stochastic exit from dormancy, and the nature of slow growers. Appl. Environ. Microbiol., 2012. 78(9): p. 3221-3228.

82. Dworkin, J. and I.M. Shah, Exit from dormancy in microbial organisms. Nature Reviews Microbiology, 2010. 8(12): p. 890.

83. Lechner, S., et al., Interplay between population dynamics and drug tolerance of Staphylococcus aureus persister cells. Journal of molecular microbiology and biotechnology, 2012. 22(6): p. 381-391.

84. Lewis, K., Multidrug tolerance of biofilms and persister cells, in Bacterial Biofilms. 2008, Springer. p. 107-131.

85. Spanka, D.-T., et al., High-throughput proteomics identifies proteins with importance to postantibiotic recovery in depolarized persister cells. Frontiers in Microbiology, 2019. 10: p. 378.

86. Jõers, A., N. Kaldalu, and T. Tenson, The frequency of persisters in Escherichia coli reflects the kinetics of awakening from dormancy. Journal of bacteriology, 2010. 192(13): p. 3379-3384.

87. Barth Jr, V.C., et al., Heterogeneous persister cells formation in Acinetobacter baumannii. PLoS One, 2013. 8(12): p. e84361.

110

88. Von Eiff, C., et al., Nasal carriage as a source of Staphylococcus aureus bacteremia. New England Journal of Medicine, 2001. 344(1): p. 11-16.

89. Foster, T.J., Antibiotic resistance in Staphylococcus aureus. Current status and future prospects. FEMS microbiology reviews, 2017. 41(3): p. 430-449.

90. Naimi, T.S., et al., Comparison of community-and health care–associated methicillin-resistant Staphylococcus aureus infection. Jama, 2003. 290(22): p. 2976-2984.

91. Chambers, H.F., The changing epidemiology of Staphylococcus aureus? Emerging infectious diseases, 2001. 7(2): p. 178.

92. Monaco, M., et al., Worldwide epidemiology and antibiotic resistance of Staphylococcus aureus, in Staphylococcus aureus. 2016, Springer. p. 21-56.

93. Glaser, P., et al., Demography and intercontinental spread of the USA300 community-acquired methicillin-resistant Staphylococcus aureus lineage. MBio, 2016. 7(1): p. e02183-15.

94. Uehara, Y., et al., Regional outbreak of community-associated methicillin-resistant Staphylococcus aureus ST834 in Japanese children. BMC infectious diseases, 2019. 19(1): p. 35.

95. Mercado-Lubo, R., et al., A Salmonella enterica serovar typhimurium succinate dehydrogenase/fumarate reductase double mutant is avirulent and immunogenic in BALB/c mice. Infection and immunity, 2008. 76(3): p. 1128-1134.

96. Auguet, O.T., et al., Evidence for community transmission of community-associated but not health-care-associated methicillin-resistant Staphylococcus aureus strains linked to social and material deprivation: spatial analysis of cross-sectional data. PLoS medicine, 2016. 13(1): p. e1001944.

97. Kallen, A.J., et al., Health care–associated invasive MRSA infections, 2005-2008. Jama, 2010. 304(6): p. 641-647.

98. Popovich, K.J., et al., Genomic and epidemiological evidence for community origins of hospital-onset methicillin-resistant Staphylococcus aureus bloodstream infections. The Journal of infectious diseases, 2017. 215(11): p. 1640-1647.

99. Sheppard, S.K., D.S. Guttman, and J.R. Fitzgerald, Population genomics of bacterial host adaptation. Nature Reviews Genetics, 2018: p. 1.

100. Shan, Y., et al., Genetic basis of persister tolerance to aminoglycosides in Escherichia coli. MBio, 2015. 6(2).

101. Hu, Y. and A.R. Coates, Transposon mutagenesis identifies genes which control antimicrobial drug tolerance in stationary-phase Escherichia coli. FEMS microbiology letters, 2005. 243(1): p. 117-124.

102. Wang, W., et al., Transposon mutagenesis identifies novel genes associated with Staphylococcus aureus persister formation. Frontiers in microbiology, 2015. 6: p. 1437.

103. Wattam, A.R., et al., PATRIC, the bacterial bioinformatics database and analysis resource. Nucleic acids research, 2013. 42(D1): p. D581-D591.

104. Li, H., et al., The sequence alignment/map format and SAMtools. Bioinformatics, 2009. 25(16): p. 2078-2079.

105. Aziz, R.K., et al., The RAST Server: rapid annotations using subsystems technology. BMC genomics, 2008. 9(1): p. 75.

106. Kanehisa, M. and S. Goto, KEGG: kyoto encyclopedia of genes and genomes. Nucleic acids research, 2000. 28(1): p. 27-30.

107. Van den Bergh, B., et al., Frequency of antibiotic application drives rapid evolutionary adaptation of Escherichia coli persistence. Nature microbiology, 2016. 1(5): p. 16020.

108. Acker, K.P., et al., Metabolic Adaptation Drives Staphylococcus aureus Colonization and Infection of the Skin. Staphylococcus aureus, 2019.

111

109. Zhu, Y., et al., Staphylococcus aureus biofilm metabolism and the influence of arginine on polysaccharide intercellular adhesin synthesis, biofilm formation, and pathogenesis. Infection and immunity, 2007. 75(9): p. 4219-4226.

110. Stewart, P.S. and M.J. Franklin, Physiological heterogeneity in biofilms. Nature Reviews Microbiology, 2008. 6(3): p. 199.

111. Dubin, G., Extracellular proteases of Staphylococcus spp. Biological chemistry, 2002. 383(7-8): p. 1075-1086.

112. Parks, W.C., C.L. Wilson, and Y.S. López-Boado, Matrix metalloproteinases as modulators of inflammation and innate immunity. Nature Reviews Immunology, 2004. 4(8): p. 617.

113. Wang, Y.Y., et al., Peptidoglycan of Staphyloccus aureus induces enhanced levels of matrix metalloproteinase-9 in human blood originating from neutrophils. Shock, 2005. 24(3): p. 214-218.

114. Bryk, R., et al., Selective killing of nonreplicating mycobacteria. Cell host & microbe, 2008. 3(3): p. 137-145.

115. Allison, K.R., M.P. Brynildsen, and J.J. Collins, Metabolite-enabled eradication of bacterial persisters by aminoglycosides. Nature, 2011. 473(7346): p. 216.

116. Collins, J.J., K.R. Allison, and M.P. Brynildsen, Proton-motive force stimulation to potentiate aminoglycoside antibiotics against persistent bacteria. 2019, Google Patents.

117. Meylan, S., et al., Carbon sources tune antibiotic susceptibility in Pseudomonas aeruginosa via tricarboxylic acid cycle control. Cell chemical biology, 2017. 24(2): p. 195-206.