Rosalind AllenSchool of Physics and Astronomy, University of Edinburgh

QLSB II, Como, June 21st 2016

How do antibiotics work?…. and can physicists help?

Antibiotics: molecules that inhibit bacteriaAlexander Fleming 1928

Wikipedia

Y. G. Song, Infect. Chemother. 2012, 44, 263-268

Antibiotics have revolutionised global health

Leading causes of US deaths 1900: pneumonia, tuberculosis, diarrhoea

1997: heart disease, cancer, strokeSource: www.cdc.gov

But there is a looming crisis of antibiotic resistance

Evolution.berkeley.edu www.cdc.gov

What to do about this?

• Understand how antibiotics work and how resistance evolves

can we develop smart strategies to avoid resistance?

• Discover new antibiotics eg screen environmental samples for new compounds

• Reduce antibiotic use eg improve diagnostics to distinguish bacterial and viral infections

What can physicists do to help?

• Help design tools for better diagnosiseg chips to detect DNA of bacterial pathogens

• Help improve basic understandingsimple lab model systemsmathematical and computational models

R. J. Allen and B. WaclawAntibiotic resistance: a physicist’s viewArxiv 1605.06086

How do antibiotics work?

Target processes that differ between bacteria and human cells

• Cell Wall Synthesis beta-lactams, vancomycin

• Protein Synthesisaminoglycosides, tetracyclines, chloramphenicol, macrolides

• Nucleic Acid Synthesisquinolones, metronidazole, rifampicin

• Cell Membranepolymyxins

• Metabolism sulfonamides, trimethoprim

Bactericidal drugs kill bacteriaBacteriostatic drugs stop bacterial growth

www.tnmanning.com

How to quantify antibiotic efficacy?Minimum inhibitory concentration (MIC): concentration that prevents visible growth of bacteria

Small MIC-> high efficacy

Vads.vetmed.vt.edu

IC50

IC50: concentration needed to halve the growth rate

Small IC50-> high

efficacy

From lab assays to clinical usePharmacokinetics: predict antibiotic concentration in the human body

www.biologicaltestcenter.com

Pharmacodynamics: what concentration is needed to treat an infection?

Time-dependent drugs:what matters is time above MICeg penicillins, cephalosporins

Concentration-dependent drugs:what matters is concentrationpeak/MIC or AUC:MICeg quinolones, aminoglycosides

Also need to avoid antibiotic resistance

“After more than 50 years of study, the shape of drug concentration-time curve that is needed at the site of infection for optimum antimicrobial effects is still not known”D. Greenwood in “Antimicrobial chemotherapy”, 4th Ed.

Real infections are complicated

Urinary tract infection

• Bacteria stick to bladder wall

• Damage epithelium, trigger immune response

• Colonise and damage kidneys

• Eventually spread to bloodstream

A. L. Flores-Mireles et al, Nature Reviews Microbiology 13, 269-294 (2015)

But simple models can help

H. Kuwahara et al, Plos computational biology 6 e1000723 (2010)

e.g. Urinary tract infectionE. coli switch stochastically between fibriated and non-fimbriated statesFimbriated bacteria stick to wallsBut also activate immune system

Statistical physics model:• Population grows• Switches between states A and

B• Environmental catastrophe

wipes out A cells, triggered by population

What is the optimum switching rate?P. Visco et al Biophysical Journal 98, 1099-1108 (2010)

Frac

tion

of c

ells

in A

stat

etime

See alsoM. Thattai & A. van Oudenaarden Genetics 167, 523-530 (2004)E. Kussell & S. Leibler Science 309, 2075-2078 (2005)

More detailed example: how does growth rate affect antibiotic efficacy?Virulent infections: fast-growing bacteriaChronic infections: slow-growing bacteria

Do antibiotics work differently for virulent versus chronic infections?

Growth-dependent bacterial sensitivity to ribosome-targeting antibioticsP. Greulich, M. Scott, M. R. Evans & R. J. Allen, Mol. Syst. Biol. 11, 796 (2015)

Philip GreulichMatt Scott

Martin Evans

A simple test: grow E. coli bacteria in the lab on different nutrientsDo fast-growing bacteria respond better or worse to antibiotics than slow-growing bacteria?

6 growth media

4 antibiotics: tetracycline, chloramphenicol, streptomycin, kanamycinAll target the ribosome; cell’s protein synthesis machinery

Result: some antibiotics work better on fast-growing cells

Tetracycline Chloramphenicol

Kanamycin

But others work better on slow-growing cells

Streptomycin

why?

A simple modelRibosomes are needed to make new ribosomes Ribosomes are

needed for growth

• Antibiotic crosses membrane; net inflow rate J.

• Antibiotic binds ribosomes at rate kon, unbinds at rate koff

• Cell grows at rate l, diluting cell contents

• New ribosomes are synthesized at rate s

l and s depend on the ribosome concentration!

Model variablesa(t): intracellular antibiotic concentrationru(t): free (unbound) ribosome concentrationrb(t): antibiotic-bound ribosome concentrationModel equations

Constraints: Free ribosomes are needed for growth l = l(ru)Ribosome synthesis rate is regulated s =s(l)

Dilution due to growth

Antibiotic-ribosome binding

Ribosome synthesis

Antibiotic inflow

Constraint 1: ribosomes are needed for growth

M. Scott, et al Science (2010) 330, 1099Constraints can be obtained from experimental data

Constraint 2: up-regulation of ribosome synthesis

Steady-state growth, synthesis balances dilution

Result: cubic equation linking growth rate and antibiotic concentration

Measures the reversibility of membrane transport and ribosome binding

One key parameter

Good fits to experimental data

Simple prediction for the IC50

Large l0*: IC50 decreases with nutrient richness: Fast-growing cells are more susceptibleSmall l0*: IC50 increases with nutrient richness:Fast-growing cells are less susceptible

Scaled drug-free growth rate

Scal

ed su

scep

tibilit

y

Outcomes:It’s all about reversibilityLink molecular mechanism to whole-cell physiology

Related work:

Rebecca BrouwersLinking mechanism to physiology for cell-wall targeting antibiotics

Dan TaylorHow do bacteria respond to antibiotics in small populations?

What about antibiotic resistance?

Antibiotic resistance

Emergence of bacterial strains that are not inhibited by antibiotic• Gain of a degrading enzyme e.g. beta-lactamases

• Alteration of the bacterial target e.g. changes in ribosome structure

• Change in permeability or transport e.g. increased expression of efflux pumps

Can happen by• Gain of extra DNA (eg plasmids by horizontal gene transfer)• Mutations in genome• Changes in gene expression

www.reactgroup.org

How does an infection become antibiotic resistant?

An individual bacterium arises that is resistante.g. through genetic mutation

It proliferates in competition with sensitive bacteriatypically wins in presence of antibiotic, loses otherwise

It spreads beyond the initial infectione.g. to other people

Usually a multistep process, several mutations

Pathways to drug resistance

D. M. Weinreich et al, Science 312, 111-114 (2006)

Usually several mutations needed for clinically relevant antibiotic resistance

Does evolution always follow the same pathway?Example: Weinreich et al (2006)Construct all combinations of 5 mutations in a b-lactamase enzyme

Measure MIC of all mutants

Attempt to infer possible evolutionary pathways

-> Only a few are feasible

Morbidostat: a smart device for tracking evolutionary pathways in time

E. Toprak et al, Nature Genet. 44, 101-105 (2011)E. Toprak et al, Nature Protocols 8, 555-567 (2013)

Grow bacteria at constant volumeAdd nutrients, remove wasteIf growth rate is positive, add drug-> maintains constant selection for resistance

Trimethoprim: stepped trajectories, mutations only in target protein (dihydrofolate reductase)

Chloramphenicol: smooth trajectories, many mutations involved (translation, transcription, transport)

But real infections can be spatially structured

How does a spatial drug gradient affect evolution of resistance?

Qiucen Zhang et al. Science 2011;333:1764-1767

Experiments in microfluidic “death galaxy” (Bob Austin’s group):E. coli resistance to ciprofloxacin emerges much faster in a drug gradient

Our simulations: bacterial population invades a drug gradient

-> model by chain of connected microhabitats

• Population well-mixed within habitats

• Migration between habitats• Mutation between genotypes• Growth rate depends on local drug

concentrationgenotype m cannot grow if c>bm

• Exponential drug gradient

Microhabitat i

Genotype m

Philip Greulich

Bartek Waclaw

Result: population expands in a series of waves

P. Greulich , B. Waclaw & R. J. Allen, PRL 109, 088101 (2012)

Why? • Strong selection at the wave front• No need to compete with neighbours• Very steep gradient: fronts too narrow to produce mutants

Steepness of gradient

Tim

e to

full

resis

tanc

epo

pula

tion

dens

ity

Time to resistance depends on steepness of gradient

Experiments: Bartek Waclaw

Track evolution in drug gradientsDirectly mimic the model

Preliminary results (E. coli in ciprofloxacin)• We do see evolution of resistance• Mutation rate depends on drug

concentration

ConclusionAntibiotic resistance: how can we help?

Try to understand the basics • how antibiotics work• how resistance evolves

Using simple experimental and mathematical models

Can we connect it to “real biology”?

It remains to be seen….

Postdoctoral positions in the physics of antibiotic resistance

University of Edinburgh soft matter, biological and statistical physics group

www.vacancies.ed.ac.uk ref number 036372 closing date 1st July 2016

Enquiries to Rosalind Allen or Bartek Waclaw rosalind.allen@ed.ac.uk bwaclaw@staffmail.ed.ac.uk

e.g. biofilm infections: bacteria colonising surfaces

L. Hall-Stoodley et al, Nat Rev Microbiol. 2, 95 (2004)

gum disease

catheter contaminat

ion

implant contaminat

ion

R. J. Broomfield et al, J. Medical Microbiology 58, 1367-1375 (2009)

How relevant is this to real infections?

responsible for chronic infectionsbacteria experience different chemical environments

How does antibiotic resistance evolve in these infections?

no antibiotic

with antibiotic

How does biofilm structure affect evolution of antibiotic resistance?

Courtesy of R. McKenzie & G. Melaugh

How does antibiotic change biofilm structure?Can we predict rate of resistance evolution?

Computer simulations

• Simulation tracks individual bacteria• Bacteria interact via physical forces• Bacteria consume nutrient, grow and

divide• Nutrients and drugs diffuse from

above

www.sharklet.com

Can we design smart surfaces to avoid resistant biofilms?