LETTERS
Experimental evolution of bet hedging
Hubertus J. E. Beaumont
1,2
{, Jenna Gallie
1
, Christian Kost
1
{, Gayle C. Ferguson
1
& Paul B. Rainey
1
Bet hedging—stochastic switching between phenotypic states
1–3
—
is a canonical example of an evolutionary adaptation that facilitates
persistence in the face of fluctuating environmental conditions.
Although bet hedging is found in organisms ranging from bacteria
to humans
4–10
, direct evidence for an adaptive origin of this beha-
viour is lacking
11
. Here we report the de novo evolution of bet
hedging in experimental bacterial populations. Bacteria were sub-
jected to an environment that continually favourednew phenotypic
states. Initially, our regime drove the successive evolution of novel
phenotypes by mutation and selection; however, in two (of 12)
replicates this trend was broken by the evolution of bet-hedging
genotypes that persisted because of rapid stochastic phenotype
switching. Genome re-sequencing of one of these switching types
revealed ninemutations that distinguished it from the ancestor. The
final mutation was both necessary and sufficient for rapid pheno-
type switching; nonetheless, the evolution of bet hedging was con-
tingent upon earlier mutations that altered the relative fitness effect
of the final mutation. These findings capture the adaptive evolution
of bet hedging in the simplest of organisms, and suggest that risk-
spreading strategies may have been among the earliest evolutionary
solutions to life in fluctuating environments.
Life exists in ever-changing environments, but surviving under
fluctuating conditions poses challenges. One solution is the evolution
of mechanisms that allow modulation of phenotype in response to
specific environmental cues. An alternative solution is stochastic
phenotype switching, a strategy based on bet hedging, rather than
direct environmental sensing
4–10
.
The general prediction from theory is that fluctuating selection
generated by unpredictable environments can favour the evolution
of bet hedging
1–4,12–14
. Under such conditions, a strategy that generates
random variation in fitness-related traits among individuals within a
population can enhance long-term fitness by increasing the likelihood
that a subset of individuals expresses a phenotype that will be adaptive
in a future environment
3,15,16
. However, the outcome of adaptive
evolution under fluctuating selection is also shaped by factors such
as the frequency of environmental change
12
, the capacity of a given
population to respond to fluctuations by mutation and selection (that
is, evolvability), the presence of suitable environmental cues
13
and the
cost–benefit balance of different strategies
12
.
Here we report the de novo evolution of bet hedging in experimental
bacterial populations. Our populations experienced repeated bouts of
selection in two contrasting environments; they also experienced fluc-
tuating selection wrought by imposition of an exclusion rule and
population bottleneck. Applied at the point of transfer between envi-
ronments, the exclusion rule assigned a fitness of zero to the type that
was common in the current environment; imposition of the bottleneck
meant that only a single phenotypically distinct type was selected from
among the survivors to found the next bout of selection. The exclusion
rule imposed strong selection for phenotypic innovation, whereas the
bottleneck negated the cost of bet hedging—that is, the generation of
types maladapted to the prevailing conditions—by eliminating com-
petition with conspecifics. A natural analogue of this mode of fluctuat-
ing selection is imposed by the host immune system on invading
microorganisms. Indeed, many pathogens have evolved bet-hedging
strategies based on stochastic antigen switching
8
.
When the ancestral genotype of Pseudomonas fluorescens SBW25 is
grown in static broth microcosms, it rapidly diversifies into a range of
niche specialist genotypes by mutation and selection, which each form
distinct colonies on agar plates
17
. In contrast, diversification is con-
strained inshakenmicrocosms,w hichfavourgenotypeswith an ancestral
colony type
18
. We exploited the pleiotropic correlation between niche
specialization and colony morphology to realize our selection regime.
The expected evolutionary response was repeated evolution, fixation and
extinction of genotypes with novel colony morphologies.
Twelve replicate selection lines were founded with the ancestral
genotype and subjected to 16 rounds of alternating selection in static
and shaken microcosms. During each round, populations were pro-
pagated by serial dilution until the emergence of cells that formed
colonies with a heritable morphology different from that of their
immediate ancestor. Selection for different colonies was open-ended
rather than for an a priori defined morphology. Once detected, cells
derived from a single individual of this new type were transferred to
the opposing environment for the next round of selection (Fig. 1a
and Methods). Thus, we imposed selection for a high growth rate in
static and shaken microcosms, and simultaneously fluctuating selec-
tion for colony innovation. Each lineage was associated with a cog-
nate control line that was under stabilizing selection for the ancestral
colony morphology, but otherwise treated identically (see Methods).
This selection regime indeed drove the repeated evolution of new
colony morphologies. However, here we concentrate on the evolu-
tion of genotypes capable of rapid, stochastic colony-morphology
switching (hereafter referred to as colony switching), which emerged
in two of 12 replicates (1B
4
and 6B
4
, Supplementary Note 1). Each of
these switching genotypes formed distinct translucent and opaque
colonies (Fig. 1b). Colony switching persisted in both lines for seven
additional rounds of selection, after which the experiment was ended.
None of the control lines gave rise to colony switching.
To obtain insight into the mechanisms by which switching types
persisted, we re-imposed our selection regime on the genotype 1B
4
(Fig. 1b), but this time ignored the colony variants it generated
through colony switching. Without exception, 1B
4
was driven to
extinction by the evolution of new genotypes with novel colony
morphologies, some of which did not switch (Supplementary Note 2).
This shows that persistence was attributable to bet hedging—that is, the
capacity to generate at high-frequency colonies that were phenotypically
distinct from those selected in the previous round—rather than to an
intrinsic growth rate advantage over non-switching types. This finding
also draws attention to the importance of the bottleneck: in its absence,
1
New Zealand Institute for Advanced Study and Allan Wilson Centre for Molecular Ecology & Evolution, Massey University, Private Bag 102904, North Shore Mail Centre, North Shore
City 0745, Auckland, New Zealand.
2
Institute of Biology Leiden, Leiden University, PO Box 9505, 2300 RA Leiden, The Netherlands. {Present addresses: Institute of Biology Leiden,
Leiden University, PO Box 9505, 2300 RA Leiden, The Netherlands (H.J.E.B.); Department of Bioorganic Chemistry, Max Planck Institute for Chemical Ecology, 07745 Jena, Germany
(C.K.).
Vol 462
|
5 November 2009
|
doi:10.1038/nature08504
90
Macmillan Publishers Limited. All rights reserved
©2009
1B
4
would have been eliminated by non-switching genotypes with faster
growth rates.
The serendipitous discovery that centrifugation of 1B
4
cells
resulted in two discrete fractions prompted microscopic examina-
tion (Supplementary Method 2), which revealed that 1B
4
produced
both capsulated (Cap
1
) and non-capsulated (Cap
2
) cells (Fig. 1c).
Opaque colonies contained a higher proportion of Cap
1
cells than
translucent colonies (Fig. 2a). When cell suspensions were plated
from either colony type, both gave rise to a mixture of opaque and
translucent colonies; however, cell suspensions made from opaque
colonies produced a higher proportion of opaque colonies than those
derived from translucent colonies (Fig. 2b). Together, this suggests
that the founding cell of a colony may determine the morphology of
the latter by biasing the ratio of Cap
1
and Cap
2
cells. This causal link
was corroborated by transposon mutagenesis
19
, which also indicated
that the capsules consist of colanic acid
20
, a previously described
capsule polysaccharide (see Supplementary Method 1 and Sup-
plementary Table 1). Colony switching in the second bet-hedging
genotype also involved Cap
1
and Cap
2
cells.
A likely mechanism of colony switching is reversible on–off switch-
ing of capsule production. Alternatively, colony switching may
involve distinct non-switching Cap
1
and Cap
2
genotypes that form
mixed colonies upon plating owing to chance co-localization, or
mixed cell-type aggregates. We distinguished between these possibi-
lities by determining whether populations that were passed through
two sequential single-cell bottlenecks produced both Cap
1
and Cap
2
cells after each bottleneck (see Supplementary Method 2 and
Supplementary Fig. 1). All replicate populations generated both cell
types after each bottleneck. The probability of observing this pattern
under the null hypotheses of no reversible switching is exceedingly low
(one-tailed Fisher’s exact test, P 5 0.0004; see Supplementary Note 3),
indicating that 1B
4
switches reversibly between Cap
1
and Cap
2
.
Although bet hedging facilitated the long-term persistence of 1B
4
,
the evolutionary emergence of this genotype required it to reach a
detectable frequency within a selection round. To examine if 1B
4
owed
its emergence to a higher fitness than its immediate ancestor (1A
4
), we
competed these two genotypes in static microcosms, the environment
in which 1B
4
emerged. This indicated that 1B
4
was indeed more fit
than 1A
4
(one-sample t-test, n 5 8, P 5 0.0002; Fig. 4).
Repeated single-cell bottlenecks can drive the fixation of random
deleterious mutations, causing a decline in fitness. To test if this had
occurred during our experiment, we measured the fitness of all geno-
types in the 1B
4
lineage relative to the original ancestor in both static
and shaken microcosms (Fig. 3a). This revealed no evidence for a
decrease in fitness. Interestingly, the results indicate non-transitive
fitness interactions between some consecutive genotypes (for
example, 1A
4
is more fit than 1B
4
relative to 1A
0
, but less fit than
1B
4
during direct competition).
Using whole-genome re-sequencing
21,22
, the entire 6.7-megabase-
pair genome of 1B
4
(ref. 23) was analysed to unravel its mutational
history. Nine mutations separating 1B
4
from the original ancestor
were identified, confirmed by Sanger sequencing, and ordered by
inspection of the affected loci in the preceding genotypes (Fig. 3b).
With the exception of the final mutation, all mutations involved non-
synonymous changes at loci previously demonstrated to be muta-
tional targets in the evolution of wrinkly spreader types
18,24–26
(Fig. 3b). The final mutation was a single non-synonymous nucleotide
change in carB (Arg674Cys, Fig. 3b), which encodes the large subunit
of carbamoylphosphate synthetase (CarAB, EC 6.3.5.5), a central
enzyme of the pyrimidine and arginine biosynthetic pathways
27
.
To examine the causal connection between the carB mutation and
colony switching, we introduced this mutation in the immediate
ancestor (1A
4
) by allelic replacement. The engineered genotype dis-
played colony switching (Fig. 3g). Conversely, reversion of the carB
mutation in 1B
4
to wild type abolished colony switching (Fig. 3h).
This demonstrates that the carB mutation is sufficient and necessary
to cause stochastic colony morphology switching in 1A
4
.
Although the exact mechanism of colony switching remains to be
elucidated, two lines of evidence suggest that switching might be
controlled epigenetically, rather than by a mutable locus
8
. First,
Cap
1
and Cap
2
cells were identical at the carB locus. Second, neither
transposon mutagenesis nor genome re-sequencing showed evidence
of the involvement of mutable loci (Supplementary Method 1).
If the carB mutation is the sole cause of the evolution of colony
switching from 1A
4
, it must confer not only colony switching but also
the requisite high fitness in static microcosms in this genetic back-
ground. To assess this, we introduced the mutant carB allele in 1A
4
a
b
c
Round 1 Round 2
Figure 1
|
Colony morphology evolution. a, Populations were propagated in
static or shaken (red arrow) microcosms, and periodically screened for cells
that form novel colony types, a prerequisite for passage to the opposing
environment (the exclusion rule). Propagation between environments
occurred by transfer of cells taken from a single, numerically dominant, novel
colony (the bottleneck).
b, From left: translucent, sectored and opaque
colonies of a bet-hedging genotype (1B
4
). Sectored colonies, which resulted
from stochastic capsule-expression switching during formation of the colony,
were counted as opaque. Scale bar, approximately 2 mm.
c, Capsulated and
non-capsulated cells of 1B
4
(phase-contrast light microscopy with negative
capsule staining). The proportions of capsulated cells in colonies produced by
the original ancestor (1A
0
) and immediate ancestor (1A
4
)of1B
4
were three
orders of magnitude lower (n 5 5, 500 cells examined per colony, binomial
95% confidence intervals 0.0010–0.0037 and 0.0028–0.0057, respectively).
Scale bar, approximately 10 mm.
Opaque Translucent
Proportion of Cap
+
cells
0.0
0.2
0.4
0.6
0.8
1.0
0.0
0.2
0.4
0.6
0.8
1.0
Proportion of derived
opaque colonies
Colony morphology
Opaque Translucent
Colony morphology
a
b
Figure 2
|
Link between colony and cell morphology. a, Mean proportion of
capsulated cells in opaque and translucent colonies of bet-hedging genotype
1B
4
(n 5 6, 100 cells examined per colony). Opaque colonies contained
proportionally more capsulated cells than translucent colonies (analysis of
deviance, F(1,10) 5 61.795, P , 0.0001). Error bars, one standard deviation.
b, Mean proportion of opaque colonies formed by cell suspensions prepared
from the opaque and translucent colonies in
a (n 5 6, between 200 and 700
colonies examined). Cells derived from opaque colonies formed
proportionally more opaque colonies (analysis of deviance,
F(1,10) 5 11.836, P 5 0.0063). Error bars, one standard deviation.
NATURE
|
Vol 462
|
5 November 2009 LETTERS
91
Macmillan Publishers Limited. All rights reserved
©2009
and competed the resulting genotype against 1A
4
. This showed that
the carB mutation increases the fitness of 1A
4
(one-sample t-test,
n 5 9, P 5 0.0013; Fig. 4). Moreover, 1B
4
with a wild-type carB allele
had a lower fitness relative to 1B
4
, confirming that the carB mutation
was essential for the increased fitness of this genotype (one-sample
t-test, n 5 9, P 5 0.0004; Fig. 4). Together, these results indicate that
the fitness effect of the carB mutation is sufficient to explain the
emergence of 1B
4
to a detectable frequency.
Colony switching was caused by a single point mutation but none-
theless took nine rounds of selection to evolve. This led us to question
the importance of the previously fixed mutations
28
. To study this, one
round of the evolutionary experiment was repeated from both 1A
4
and
the original ancestor (1A
0
). Colony switching evolved from 1A
4
(3 from
36 replicates), but never from the ancestral genotype (0 from 138 repli-
cates), indicating that these genotypes differed in their capacity to give
rise to colony switching (two-tailed Fisher’s exact test, P 5 0.0083).
The reliance on previously fixed mutations might stem from epi-
static interactions essential for the carB mutation to cause colony
switching or to confer the requisite fitness benefit in static microcosms.
We distinguished between these hypotheses by introducing the carB
mutation in 1A
0
. In this background it did cause colony switching
(Fig. 3f) but appeared not to confer a significant fitness increase
(Fig. 4). The latter was confirmed by a direct statistical comparison
with the effect of the carB mutation in 1A
4
(analysis of covariance, the
larger variance of the 1A
0
data is explained by a covariate), which
identified a significant epistatic interaction (analysis of deviance,
F(1,16) 5 8.536, P 5 0.01), and indicated that the carB mutation is
beneficial in 1A
4
but deleterious in 1A
0
(Supplementary Note 4). We
conclude that the evolutionary history of 1A
4
‘set the stage’ for the
evolution of stochastic colony morphology switching by altering the
relative fitness effect of the carB mutation.
Owing to the historical nature of the evolutionary process, the
origins of most adaptive phenotypes are obscure
11
. Here we have
provided a mechanistic account of the adaptive evolution of a wide-
spread trait
4–10
. Bet hedging arose as an adaptation to fluctuating
selection imposed by an exclusion rule and bottleneck, two popu-
lation processes that are likely to play a key role in the evolution of
stochastic phenotype switching in nature. Insight into the underlying
molecular details reveals how evolution tinkered with central meta-
bolism to generate a strategy that could reasonably—one might
think—have taken tens of thousands of generations to evolve. The
rapid and repeatable evolution of bet hedging during our experiment
suggests it may have been among the earliest evolutionary solutions
to life in variable environments, perhaps even preceding the evolu-
tion of environmentally responsive mechanisms of gene regulation.
METHODS SUMMARY
P. fluorescens SBW25 (ref. 23) was grown in glass microcosms containing liquid
medium. During each selection round, populations were propagated by transfer
of a sample to a fresh microcosm. Parallel with each transfer, populations were
checked for the presence of cells that formed new colony types. Identification of a
colony different from that with which the selection round had been started
marked the end of a selection round. Cells from these colonies were stored,
and used to found the next selection round in the opposing environment.
This procedure was repeated 15 times. Control lines were continually selected
in shaken microcosms and always propagated by transfer of cells from colonies of
the ancestral type. Relative fitness was measured in competition assays, using
colony morphology or a neutral marker
29
to distinguish genotypes, and
expressed as the ratio of Malthusian parameters
30
.
Full Methods and any associated references are available in the online version of
the paper at www.nature.com/nature.
Received 7 July; accepted 15 September 2009.
1. Cohen, D. Optimizing reproduction in a randomly varying environment. J. Theor.
Biol. 12, 119
–
129 (1966).
c
b
a
ed
f
g
h
1.5
1
0.5
Relative tness
2
0
Mutations
1B
0
1A
1
1B
1
1A
2
1B
2
1A
3
1B
3
1A
4
1B
4
1A
0
>>>>>>>>>
ShakenShaken
StaticStatic
Environment
Locus
*
*
*
*
*
*
*
*
*
*
*
*
carB C2020T
mwsR G2778A
mwsR G2383A
mwsR C3094G
mwsR ΔC2553
awsX Δ229–261
awsR A1141C
wspF 157insG
wssA 164insA
1A
0
1A
0
mut 1A
4
mut 1B
4
WT
1A
4
1B
4
Genotype
Figure 3
|
Fitness, mutational history and effects of the carB mutation in a
bet-hedging lineage. a
, Fitness of each genotype in a bet-hedging lineage
(1B
4
) measured relative to the original ancestor (1A
0
) in shaken (dark circles)
and static (light circles) microcosms. Light bars, genotypes that evolved in
static microcosms (1B
n
); dark bars, genotypes that evolved in shaken
microcosms (1A
n
). Fitness of 1B
4
in static microcosms founded with
predominantly capsulated (yellow circles, Cap
1
proportion 5 0.688, 95%
confidence interval 0.417–0.872) or non-capsulated cells (purple circles,
Cap
1
proportion 5 0.013, 95% confidence interval 0.004–0.031) did not
differ significantly. Asterisks, significant deviation from 1 (dashed line;
P # 0.05, one-sample t-tests, n 5 3). Error bars, one standard deviation.
b, Mutations identified by genome re-sequencing of 1B
4
and ordered through
analysis of the preceding genotypes. Each round of selection caused a genetic
change readily explained by existing knowledge of the genetic causes of the
WS phenotype
18,24–26
. c–h, Colonies of (c) original ancestral genotype,
(
d) immediate ancestor of bet-hedging genotype, (e) bet-hedging genotype,
(
f) original ancestral genotype with mutated carB,(g) immediate ancestor of
bet-hedging genotype with mutated carB and (
h) bet-hedging genotype with
wild-type carB. Colonies were photographed at the same magnification.
Competition
Relative tness
0.6
0.8
1.0
1.2
1.4
1.6
**
***
1B
4
versus
1A
4
1A
4
mut
versus
1A
4
1B
4
WT
versus
1B
4
1A
0
mut
versus
1A
0
***
Figure 4
|
Relative fitness of evolved and engineered strains. Box plot of
the fitness of a bet-hedging genotype relative to its immediate ancestor (1B
4
versus 1A
4
), and the fitness effects of mutated carB in the immediate
ancestor of the bet-hedging genotype (1A
4
mut versus 1A
4
), wild-type carB in
1B
4
(1B
4
WT versus 1B
4
) and mutated carB in the original ancestor (1A
0
mut
versus 1A
0
). All fitness assays were performed in static microcosms. Values
greater than 1 (dashed line) indicate a higher relative fitness of the first
competitor. Key: median (horizontal lines in boxes), interquartile range
(boxes), 90th and 10th percentiles (vertical bars), significant deviation from 1
(***P # 0.001, **P # 0.01; see text for statistics).
LETTERS NATURE
|
Vol 462
|
5 November 2009
92
Macmillan Publishers Limited. All rights reserved
©2009
2. Seger, J. & Brockmann, H. J. in Oxford Surveys in Evolutionary Biology, Vol. 4 (eds
Harvey, P. & Partridge, L.) 182
–
211 (Oxford Univ. Press, 1987).
3. Slatkin, M. Hedging one’s evolutionary bets. Nature 250, 704
–
705 (1974).
4. Ackermann, M. et al. Self-destructive cooperation mediated by phenotypic noise.
Nature 454, 987
–
990 (2008).
5. Danforth, B. N. Emergence dynamics and bet hedging in a desert bee, Perdita
portalis. Proc. Natl Acad. Sci. USA 266, 1985
–
1994 (1999).
6. Friedenberg, N. A. Experimental evolution of dispersal in spatiotemporally
variable microcosms. Ecol. Lett. 6, 953
–
959 (2003).
7. Hairston, N. G. & Munns, W. R. The timing of copepod diapause as an evolutionary
stable strategy. Am. Nat. 123, 733
–
751 (1984).
8. Moxon, E. R., Rainey, P. B., Nowak, M. A. & Lenski, R. E. Adaptive evolution of
highly mutable loci in pathogenic bacteria. Curr. Biol. 4, 24
–
33 (1994).
9. Tonegawa, S. Somatic generation of antibody diversity. Nature 302, 575
–
581
(1983).
10. Venable, D. L. Bet hedging in a guild of desert annuals. Ecology 88, 1086
–
1090
(2007).
11. Gould, S. J. & Lewontin, R. C. The spandrels of San Marco and the Panglossian
paradigm: a critique of the adaptationist programme. Proc. R. Soc. Lond. B 205,
581
–
598 (1979).
12. Kussell, E. & Leibler, S. Phenotypic diversity, population growth, and information in
fluctuating environments. Science 309, 2075
–
2078 (2005).
13. Wolf, D. M., Vazirani, V. V. & Arkin, A. P. Diversity in times of adversity:
probabilistic strategies in microbial survival games. J. Theor. Biol. 234, 227
–
253
(2005).
14. Meyers, L. A. & Bull, J. J. Fighting change with change: adaptive variation in an
uncertain world. Trends Ecol. Evol. 17, 551
–
557 (2002).
15. Acar, M., Mettetal, J. T. & van Oudenaarden, A. Stochastic switching as a survival
strategy in fluctuating environments. Nature Genet. 40, 471
–
475 (2008).
16. Balaban, N. Q., Merrin, J., Chait, R., Kowalik, L. & Leibler, S. Bacterial persistence as
a phenotypic switch. Science 305, 1622
–
1625 (2004).
17. Rainey, P. B. & Travisano, M. Adaptive radiation in a heterogeneous environment.
Nature 394, 69
–
72 (1998).
18. Spiers, A. J., Kahn, S. G., Bohannon, J., Travisano, M. & Rainey, P. B. Adaptive
divergence in experimental populations of Pseudomonas fluorescens. I. Genetic and
phenotypic bases of wrinkly spreader fitness. Genetics 161, 33
–
46 (2002).
19. Giddens, S. R. et al. Mutational activation of niche-specific genes provides insight
into regulatory networks and bacterial function in a complex environment. Proc.
Natl Acad. Sci. USA 104, 18247
–
18252 (2007).
20. Stevenson, G., Andrianopoulos, K., Hobbs, M. & Reeves, P. R. Organization of the
Escherichia coli K-12 gene cluster responsible for production of the extracellular
polysaccharide colanic acid. J. Bacteriol. 178, 4885
–
4893 (1996).
21. Albert, T. J. et al. Mutation discovery in bacterial genomes: metronidazole
resistance in Helicobacter pylori. Nature Methods 2, 951
–
953 (2005).
22. Bentley, D. R. et al. Accurate whole human genome sequencing using reversible
terminator chemistry. Nature 456, 53
–
59 (2008).
23. Silby, M. W. et al. Genomic and genetic analyses of diversity and plant interactions
of Pseudomonas fluorescens. Genome Biol. 10, R51 (2009).
24. Bantinaki, E. et al. Adaptive divergence in experimental populations of
Pseudomonas fluorescens. III. Mutational origins of wrinkly spreader diversity.
Genetics 176, 441
–
453 (2007).
25. Goymer, P. et al. Adaptive divergence in experimental populations of Pseudomonas
fluorescens. II. Role of the GGDEF regulator WspR in evolution and development of
the wrinkly spreader phenotype. Genetics 173, 515
–
526 (2006).
26. McDonald, M. J., Gehrig, S. M., Meintjes, P. L., Zhang, X. X. & Rainey, P. B. Adaptive
divergence in experimental populations of Pseudomonas fluorescens. IV. Genetic
constraints guide evolutionary trajectories in a parallel adaptive radiation.
Genetics (in the press).
27. Mergeay, M., Gigot, D., Beckman, J., Glansdorff, N. & Pierard, A. Physiology and
genetics of carbamoylphosphate synthesis in Escherichia coli K-12. Mol. Gen.
Genet. 133, 299
–
316 (1974).
28. Blount, Z. D., Borland, C. Z. & Lenski, R. E. Historical contingency and the evolution
of a key innovation in an experimental population of Escherichia coli. Proc. Natl
Acad. Sci. USA 105, 7899
–
7906 (2008).
29. Zhang, X. X. & Rainey, P. B. Construction and validation of a neutrally-marked
strain of Pseudomonas fluorescens SBW25. J. Microbiol. Methods 71, 78
–
81 (2007).
30. Lenski, R. E., Rose, M. R., Simpson, S. C. & Tadler, S. C. Long-term experimental
evolution in Escherichia coli. I. Adaptation and divergence during 2,000
generations. Am. Nat. 138, 1315
–
1341 (1991).
Supplementary Information is linked to the online version of the paper at
www.nature.com/nature.
Acknowledgements We thank T. F. Cooper, M. R. Goddard and D. Refardt for
discussion; M. Gray, A. Hurman and G. E. M. Lamers for technical assistance;
M. Ackermann, P. M. Brakefield, T. Fukami, S. Rossell and B. J. Zwaan for comments
on the manuscript; T. J. M. Van Dooren for statistical advice; F. Bertels for
computational analysi s of Solexa data; and E. Libby for theoretical insight. This
work was supported by the Marsden Fund Council from government funding
administered by the Royal Society of New Zealand. H.J.E.B. is supported by a Veni
Fellowship from The Netherlands Organisation for Scientific Research (NWO). J.G.
was supported by a Bright Futures Scholarship from the New Zealand Foundation
for Research, Science and Technology. C.K. was supported by a Feodor Lynen
Fellowship from the Alexander von Humboldt Foundation, Germany. G.C.F. is
supported by a Postdoctoral Fellowship from the New Zealand Foundation for
Research, Science and Technology.
Author Contributions H.J.E.B. and P.B.R. conceived the research and wrote the
manuscript. H.J.E.B. conducted the main selection experiment, examined the
relation between cell and colony morphology, investigated reversible switching,
performed genome analysis and contributed to allelic replacements. J.G.
determined Cap
1
cell proportions, performed transposon mutagenesis and
integration-site identification, confirmed and ordered the mutations, performed
allelic replacements, and contributed to fitness assays. C.K. and G.C.F. performed
fitness assays. All authors commented on the manuscript
Author Information Reprints and permissions information is available at
www.nature.com/reprints. Correspondence and requests for materials should be
addressed to H.J.E.B. (h.j.e.beaumont@biology.leidenuniv.nl).
NATURE
|
Vol 462
|
5 November 2009 LETTERS
93
Macmillan Publishers Limited. All rights reserved
©2009
METHODS
Strain and medium. P. fluorescens SBW25 (ref. 23) was grown at 28 uC in 25-ml
glass microcosms containing 6 ml King’s Medium B with loose caps, and on
King’s Medium B–agar plates for 48 h.
Selection regime. Populations were founded with approximately 10
7
cells from
280 uC glycerol stocks prepared from cells derived from a single colony. For the
first round, populations were grown in static microcosms and propagated by
transfer of a mixed sample (6 ml) to a fresh microcosm at 72-h intervals of approxi-
mately ten generations. Parallel with each transfer, samples were spread onto plates
to screen for colonies with a different morphology (approximately 1,000 screened,
500 minimum). Fuzzy spreader genotypes
17
did not respond to selection (H.J.E.B.,
unpublished observations) and were omitted. Upondetection of new colony types,
a single colony of the numerically dominant new type was streaked to single
colonies on a control plate to confirm heritability of colony morphology, and to
ensure a single-cell bottleneck. Emergence of a colony with a different morphology
marked the end of a round. Cells of these colonies were taken from control plates,
grown in a shaken microcosm (16 h) and stored at 280 uC. Fifteen additional
selection rounds (alternately shaken at 150 r.p.m. and static) were performed, each
foundedwiththe cellsfrom the colony selected in the previous round. Control lines
were continually selected in shakenmicrocosms (for the same number of rounds as
the cognate selection lines) and maintained under stabilizing selection for the
ancestral colony type by continual selection of a numerically dominant smooth
17
colony.
Fitness assays. Cells taken from single colonies (grown from 280 uC stocks)
were pre-conditioned by cultivation in shaken microcosms for 24 h.
Competitions were inoculated with approximately 5 3 10
6
cells of each compe-
titor, and incubated for 72 h. Competitor frequencies were determined by plat-
ing at 0 h and 72 h. Genotypes were distinguished by colony morphology or a
neutral marker
29
. Relative fitness was expressed as the ratio of Malthusian para-
meters
30
.
doi:10.1038/nature08504
Macmillan Publishers Limited. All rights reserved
©2009
