Transcript E3_Drug_2011 - MicrobialEvolution.org

The Problem of Drug Resistance
E3: Lecture 11
Early Antibiotic Research
• Joseph Lister (1827-1912)
- In 1871, he noted that samples of urine with mold
did not permit bacterial growth.
- He also pioneered the introduction of an
antiseptic (phenol) before and during surgery,
drastically reducing the rate of infection.
Joseph Lister
• Ernest Duchesne (1874-1912)
- In 1897, he demonstrated that E. coli was killed
when cultured with Penicillium glaucum.
- He also showed that injection of this mold into
animals infected with typhoid bacilli prevented the
advent of the disease.
Ernest Duchesne
• Alexander Flemming (1881-1955)
- In 1928, noticed a mold contaminant on a
bacterial plate that had been sitting out in the lab.
- He isolated the Penicillium notatum and
demonstrated that it had antimicrobial effects on
Gram-positive pathogens (that cause scarlet fever,
pneumonia, gonorrhea, meningitis, diphteria)
Alexander Flemming
Earlier Antibiotic “Research”
• Microbes have produced antibiotics for a number of
reasons:
- As anticompetitor compounds (e.g., bacteriocins)
- As predatory compounds (e.g., lysing enzymes)
- As quorum-sensing molecules (e.g., nisin)
• Such antibiotics can affect many other species
(broad spectrum) or affect only a few species (narrow
spectrum).
• Most of our antibiotics are derivatives of natural
microbial products.
• We are taking an ancient form of chemical warfare
with the human body as the battlefield.
The Problem of Drug Resistance
Lecture Outline
• Antibiotics & resistance
• Costs: Reversion & Compensation
• Antibiotics & Adaptive Landscapes
• Predicting Resistance
• Summary
The Problem of Drug Resistance
Lecture Outline
• Antibiotics & resistance
• Costs: Reversion & Compensation
• Antibiotics & Adaptive Landscapes
• Predicting Resistance
• Summary
Resistance in the Intensive Care Unit
National Nosocomial Infections Surveillance System Report, 2003
Klebsiella pneumoniae
Pseudomonas aeruginosa
10 %
23 %
28 %
52 %
Enterococcus sp.
Staphylococcus aureus
Resistance to Resistance is Futile
• With sustained use of an antibiotic, resistant
strains appear and spread.
• As a new antibiotic is introduced, the
evolutionary challenge is on– and microbes
generally answer this challenge.
• Facilitating spread is the appearance of
resistance genes on mobile genetic
elements, such as plasmids (this can lead to
transfer between species).
• Often resistance is found in commensal
bacteria, which in some cases can serve as a
genetic reservoir for pathogenic species.
• Particularly troubling is the generation of
multi-drug resistant strains of pathogenic
bacteria.
Resistance Matters
Causes of death, USA
Cause
Deaths/year
HIV/AIDS
18,000
Influenza
37,000
Breast cancer
40,000
Hospital-acquired infection
90,000
• In human disease, drug-resistant bacteria can lead to:
- Increased risk of mortality
- Increased length of hospital stay
- Increased use of other drugs (which can be expensive and lead to complications)
- Foci for the spread of all these problems to other patients
• Thus, drug-resistance causes large financial and health costs.
The Problem of Drug Resistance
Lecture Outline
• Antibiotics & resistance
• Costs: Reversion & Compensation
• Antibiotics & Adaptive Landscapes
• Predicting Resistance
• Summary
Case Study: Tuberculosis
• Tuberculosis is an infectious disease caused by
the bacterium Mycobacterium tuberculosis. The
disease generally progresses in the lungs causing
tissue destruction and necrosis.
• The global incidence of TB has been on the rise,
reaching highest densities in Africa, the Middle
East and Asia.
Mycobacterium
tuberculosis
• Particularly alarming is the rise of drug-resistant
and multi-drug resistant strains of TB (e.g.,
bacteria resistant to isoniazid and rifampicin).
• The origin of resistance in
TB will be affected by the
intensity of antibiotic usage
in a given region.
• The maintenance of
resistance will depend, in
part, on its fitness effects
and evolutionary options of
various TB strains.
0.3%
0.2%
0.1%
0.05%
TB infection
Resistance in Tuberculosis
• Gagneux, Davis Long and colleagues explored the
fitness cost of drug resistance in TB in vitro.
• From a fully grown culture of a clinical isolate, they
exposed the bacteria to the antibiotic rifampicin.
• Resistant colonies were isolated and genotyped
(generally single base changes in the rpoB gene)
Sebastien
Gagneux
Clara Davis
Long
• These authors wanted to gauge the costs (if any) of
antibiotic resistance.
• The antibiotic resistant strain and antibiotic-sensitive
ancestor were placed (at roughly equal starting
frequency) in a liquid broth and competed for a growth
period (one month!)
• With knowledge about the starting densities and final
densities of the resistant strain (R) and the ancestor (A),
a relative fitness measure can be computed:
 R final 

w( R, A)   ln
 Rinitial 
 Afinal 
 ln

 Ainitial 
vs.
Costs of Resistance
• Gagneux et al. find that rifampicin
resistance is costly in TB.
• Furthermore, they find that the cost of
resistance can depend on the genetic
background of the strain.
• Such costs have been found in many
other cases:
- Streptomycin resistance in E. coli
- Fusidic acid resistance in S. aureus
- Fusidic acid resistance in S. typhimurium
- Colicin resistance in E. coli
- Rifampicin resistance in E. coli
• Most of these costs were measured in
vitro. But…
?
=
• By looking in paired patient isolates,
these authors found that resistance was
often costly, but not always…
Reversion and Compensation
• Schrag, Perrot & Levin (1997) selected for
streptomycin-resistant bacteria.
• However, the streptomycin resistant strain did not
revert to sensitivity. It remained resistant to strep.
• Further, the initial cost of streptomycin resistance
was ameliorated– there was compensation.
• They discovered second site mutations (in rpsL).
• Through genetic manipulation, they constructed all
combinations of base changes and found a rugged
landscape!
Take 5 minutes to talk about the following:
Why do you think this landscape is rugged? If landscapes of
pathogenic bacteria generally resemble the one found by
Schrag et al., how would that affect your decision about
length and strength of antibiotic treatments?
24 hrs.
• The initial resistant mutant was costly.
24 hrs.
• These authors then evolved this bacteria for 180
generations in the absence of streptomycin.
Eyeing the Landscape
• Let’s extend the landscape metaphor
sensitive
wild-type
further:
-Imagine that in the absence of the antibiotic,
the population (mostly) resides on a
sensitive wild-type peak.
-Then an antibiotic is applied.
-The sensitive peak drops out.
-Any resistant mutants are immediately
selected– now the population (mostly)
resides on a resistant position.
-Assume that the antibiotic is removed.
-Now, the sensitive peak reappears.
-If there are many ways to compensate, then
the evolutionary trajectory can take several
possible paths.
-It is possible that reversion occurs; it is
possible compensation occurs.
antibiotic
absent
resistant-types
sensitive
wild-type
antibiotic
present
resistant-types
• Some relevant questions:
-Is the landscape actually rugged?
sensitive
-Are compensatory peaks higher or lower
wild-type
than reversion peaks?
-How many ways are there to become
resistant? How many ways to compensate?
?
?
?
?
compensated
resistants
The Problem of Drug Resistance
Lecture Outline
• Antibiotics & resistance
• Costs: Reversion & Compensation
• Antibiotics & Adaptive Landscapes
• Predicting Resistance
• Summary
Evolution Done Wright
Wright’s Shifting Balance Theory
Two basic assumptions:
1.
Some genetic epistasis leading to distinct “peaks” in the landscape
2.
A metapopulation of semi-isolated sparsely populated
subpopulations
a. Migration is low between subpopulations, but present
b. Genetic drift occurs within demes
Phase 1:
Subpopulations drift over the adaptive landscape
Phase 2:
Selection drives subpopulations to new peaks
Phase 3:
Competition between subpopulations where the most fit “pulls”
the metapopulation onto its adaptive peak
“[Fisher’s] theory is one of complete and direct control by natural selection while
I attribute greatest immediate importance to the effects of incomplete isolation”
(Wright, 1931 as quoted in Provine, 1986)
Resistance in “the Balance”
Phase 1: Demes drift over the adaptive landscape
new environment changes landscape
Phase 2: Selection drives demes to new peaks
Phase 3: Interdemic competition where the most fit deme “pulls” the metapopulation onto its
adaptive peak
subpopulation 1
subpopulation 2
subpopulation 3
subpopulation 4
subpopulation 5
subpopulation 6
Population structure allows a more thorough exploration of the adaptive landscape and
thus the ascent of a higher peak globally.
Extending the Metaphor
The Role of Population Structure
Simulating Population Structure
• The population need not be structured into discrete
subpopulations for the discovery of higher peaks.
Global Neighborhood
• Imagine that individual genotypes live on a lattice
and can reproduce with either global or local dispersal.
• If the adaptive landscape is rugged and the
population starts off in a valley, then the following
predictions can be made:
-Under global dispersal, a new mutation that improves
fitness quickly takes over the population. It is likely that
this selective sweep moves the population to a suboptimal peak.
-Under local dispersal, a new mutation that improves
fitness slowly spreads through the population. It is
possible that an even better mutation may be
discovered during this selective creep.
average fitness
High fitness
local
global
Low fitness
time
Local Neighborhood
Antibiotics as a Test Case
“It would clearly be desirable…to conduct selection experiments in subdivided and
mass populations, making sure that each selection regime is replicated so that any
treatment effects can be discerned.”
(Coyne et al., 1997)
1) Obtain several rifampicin resistant strains in E. coli
2) Place each strain in an environment with structure (local dispersal) and no structure
(global dispersal) without rifampicin
3) Track the average fitness in each treatment
Media
+ Rifampicin
…
calculate
fitness
…
calculate
fitness
Slow and Steady Wins the Race
Theoretical Predictions:
average fitness
local
global
time
Lab Results:
3
2
*
BB495
*
Static
Mixed
1
0
Day 9
Day 33
fitness relative to ancestor
fitness relative to ancestor
BB492
8
*
6
4
2
*
0
Day 9
Day 33
Hypothesis:
A structured antibiotic resistant population is more likely to find higher fitness mutations
(be they reversions or compensations)
Static
Mixed
Not a Creature Was “Stirring…”
• Another group of researchers allowed antibiotic-resistant bacteria to evolve over many
generations in both a flask (in vitro) and in a mouse (in vivo). Their results were the
following:
Organism
Antibiotic
Fraction of
Fraction of
Replicates Reverting Replicates Reverting
(Flask)
(Rat/Mouse)
Staphylococcus
aureus
Fusidic acid
4/28
(14%)
4/12
(33%)
Nagaev et
al., 2001
Salmonella
typhimurium
Fusidic acid
2/28
(7%)
14/25
(56%)
Björkman et
al., 2000
Take 2 minutes to talk about the following:
Propose some alternative hypotheses to explain the
differences between the rates of reversion in vitro versus
in vivo. How would you experimentally distinguish your
hypotheses?
Reference

?

The Problem of Drug Resistance
Lecture Outline
• Antibiotics & resistance
• Costs: Reversion & Compensation
• Antibiotics & Adaptive Landscapes
• Predicting Resistance
• Summary
Predicting Evolution
• Miriam Barlow and Barry Hall are developing
Error
prone
PCR
a technique to predict the likely paths of
antibiotic resistance.
• They start with a gene that currently does not
confer high resistance (e.g., TEM-1 b
lactamase does not hydrolyze modern
cephalosporins).
• Next, they produce many mutant versions of
the gene through error prone PCR.
Miriam
Barlow
Introduce
mutations
into cells
• Then, they introduce these mutant genes into
bacterial cells.
• Next, they grow up this population of mutants
in a gradient of antibiotic and select cells from
the highest drug concentration.
• They isolate the resistance gene and begin
the process anew. After a few rounds of this
cycle, they sometimes have a gene that
confers high levels of resistance.
Select
cells
antibiotic concentration
The Barlow-Hall Method
• How well does this in vitro method predict natural
antibiotic resistant mutations?
• Using this method on TEM alleles, the four most
common amino acid substitutions found in vitro (E104K,
R164S, G238S, and E240K) were also the four most
common amino acid substitutions found in naturally
occurring extended-spectrum TEM alleles.
• How does knowledge of the likely evolutionary paths
help us?
• As we design new drugs, knowing the mutations likely to
generate resistance may help us discern tradeoffs
between resistance to different drugs.
• That is, we may be able to map those regions of genome
space (a HUGE space) that engender resistance to drug A
and those regions that engender resistance to drug B. If
these regions do not overlap, then we may be able to
“trap” our bacterium by simultaneous drug use.
•Some evidence of tradeoffs between cefepime and
cefuroxime.
types
resistant
to drug A
types
resistant
to drug B
The Problem of Drug Resistance
Lecture Outline
• Antibiotics & resistance
• Costs: Reversion & Compensation
• Antibiotics & Adaptive Landscapes
• Predicting Resistance
• Summary
Summary
• Antibiotic resistance is a serious public health problem. Multi-drug
resistance is particularly worrying.
• Understanding the ecological and evolutionary consequences of
resistance (e.g., fitness costs, probability of reversion versus
compensation) can actually inform epidemiological thinking (e.g., in the
case of TB).
• Antibiotic resistance also serves as an ideal testing ground to explore
critical issues within evolutionary biology (issues that concerned Wright
and Fisher), including the shape of adaptive landscapes, the role of
population structure, and the constraints on evolutionary trajectories.
• One exciting area (from both practical and academic perspectives) is
the prospect of predicting specific evolutionary trajectories engendering
drug resistance and using this information to design more effective
treatment.