Bet-hedging and epigenetic inheritance
in bacterial cell development
Jan-Willem Veening*†, Eric J. Stewart‡§¶, Thomas W. Berngruber?, Franc ¸ois Taddei‡§, Oscar P. Kuipers*,**,
and Leendert W. Hamoen†
*Molecular Genetics Group, Groningen Biomolecular Sciences and Biotechnology Institute, and?Theoretical Biology Group, Centre for Ecological and
Evolutionary Studies, University of Groningen, 9751 NN, Haren, The Netherlands;‡Institut National de la Sante ´ et de la Recherche Me ´dicale, U571, Universite ´
Paris, 75015 Paris, France;§Faculty of Medicine, Paris Descartes University, F-75015 Paris, France; and†Institute for Cell and Molecular Biosciences, Newcastle
University, Newcastle upon Tyne NE2 4HH, United Kingdom
Edited by Richard M. Losick, Harvard University, Cambridge, MA, and approved January 10, 2008 (received for review January 19, 2007)
Upon nutritional limitation, the bacterium Bacillus subtilis has the
capability to enter the irreversible process of sporulation. This
developmental process is bistable, and only a subpopulation of
cells actually differentiates into endospores. Why a cell decides to
sporulate or not to do so is poorly understood. Here, through the
use of time-lapse microscopy, we follow the growth, division, and
differentiation of individual cells to identify elements of cell
history and ancestry that could affect this decision process. These
analyses show that during microcolony development, B. subtilis
uses a bet-hedging strategy whereby some cells sporulate while
others use alternative metabolites to continue growth, providing
the latter subpopulation with a reproductive advantage. We dem-
onstrate that B. subtilis is subject to aging. Nevertheless, the age
of the cell plays no role in the decision of its fate. However, the
physiological state of the cell’s ancestor (more than two genera-
tions removed) does affect the outcome of cellular differentiation.
We show that this epigenetic inheritance is based on positive
feedback within the sporulation phosphorelay. The extended in-
tergenerational ‘‘memory’’ caused by this autostimulatory net-
work may be important for the development of multicellular
structures such as fruiting bodies and biofilms.
aging ? Bacillus subtilis ? bistability ? sporulation
feedback-based switches can generate bistability; i.e., the occur-
rence of two distinct subpopulations that exhibit different pheno-
types within the isogenic population (2). Maturation in developing
oocytes in Xenopus embryos is governed by a bistable switch, and
the Hedgehog network, responsible for cellular differentiation in a
diversity of eukaryotes, involves bistable switching as well (3, 4).
These types of regulatory switches are also found in single-celled
organisms such as yeasts and bacteria, where they lead to pheno-
typic variability within the isogenic population (2). Based on
mathematical modeling and synthetic gene-regulatory networks, it
was shown that stochasticity in gene expression (referred to as
noise), when amplified by positive feedback, can be the generator
of a bistable response (5). However, intrinsic physiological param-
eters, such as the cell cycle and cell age, are known contributors to
phenotypic variability as well (6). Whether the outcome of a
bistable cellular differentiation process is influenced by such phys-
iological parameters or whether it is purely a stochastic phenom-
enon is unknown. It is also unclear how far in advance of the
appearance of the phenotypic change such decisions are made.
Here, we use Bacillus subtilis, a model organism for studying
bacterial cell developmental processes, to address these questions.
When nutrient sources are dwindling, B. subtilis cells can sporu-
late by forming a highly resistant endospore at one cell pole, which
is later released by lysis of the mother cell (7). It was shown that the
complex positive-feedback architecture of the sporulation signal
transduction cascade is pivotal for this developmental program to
behave as a (unidirectional) bistable switch (8, 9). This bistability is
based on interlinked feedback loops (1). Importantly, such
exemplified by the presence of two distinct subpopulations within
the isogenic culture: sporulating and nonsporulating cells. It is
generally assumed that the decision to sporulate is stochastic in
nature, although direct experimental evidence for this assumption
has never been provided (10, 11). Moreover, there are several
potential nonstochastic phenomena that may play a role in the
decision to sporulate, such as the cell history, cell cycle timing, and
an experimental procedure using quantitative time-lapse micros-
copy, which allows us to follow the outgrowth of a single B. subtilis
cell into a sporulating microcolony.
Sporulation in a Microcolony. Microscopic observation of a growing
microcolony has been used to examine cell aging in Escherichia coli
(12). Here, we adopted this technique to study the origins of the
sporulation decision in B. subtilis. The medium and growth condi-
tions were adjusted so that single cells grew into sporulating
microcolonies of approximately a few hundred cells [for more
example of a sporulating microcolony is shown in Fig. 1A. Growth
of B. subtilis within such microcolonies follows a classical pattern
with exponential growth followed by a period of reduced growth
rate, termed the diauxic shift, after which growth ceases, and the
first endospores become visible (Fig. 1B). This secondary growth
phase is probably caused by the utilization of overflow metabolites,
such as acetoin, after the disappearance of glucose (13). After the
cessation of the secondary growth phase, there is a period of ?10
h during which there is no cell growth. However, there is consid-
this period of apparent dormancy, part of the remaining cells
resume growth, probably by using the nutrients released from the
lysed cells, and a new round of growth and sporulation takes place
(Fig. 1B, SI Fig. 6, and SI Movies 1–3). Spores formed during the
growth of the microcolony were never observed to germinate
during this new growth period.
To visualize the physiological state of individual cells conve-
niently in a single graph, we calculated the growth rate of each cell
as an exponential fit to the length measurements of the cell at all
L.W.H. designed research; J.-W.V. and E.J.S. performed research; J.-W.V., E.J.S., F.T., and
O.P.K. contributed new reagents/analytic tools; J.-W.V., E.J.S., T.W.B., and L.W.H. analyzed
data; and J.-W.V., E.J.S., F.T., O.P.K., and L.W.H. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
Freely available online through the PNAS open access option.
¶Present address: Department of Biology, Northeastern University, Boston, MA 02115.
**To whom correspondence should be addressed. E-mail: email@example.com.
This article contains supporting information online at www.pnas.org/cgi/content/full/
© 2008 by The National Academy of Sciences of the USA
March 18, 2008 ?
vol. 105 ?
no. 11 ?
only grow in length and not in width, the exponential length
increase is a good measurement of growth rate. It yielded a single
value for the growth rate of each cell, which was then plotted at the
during the diauxic shift, many cells do not use the overflow
metabolites to resume cellular growth.
Three morphologically identifiable cell fates that arise after the
lysing cells, and those that survive as vegetative cells. Using custom
software (see SI Materials and Methods), we traced the history and
lineage of every individual cell, enabling us to determine the point
in the lineage at which the fate of a newborn cell is fixed. This point
was visualized by plotting the different fates on a birth time vs.
growth rate graph, similar to Fig. 1C. Every birth point shown
indicates the birth of a cell for which it is certain that either this cell
or all of its descendents will follow a specific viable fate. For clarity,
later second round of growth and spore formation. The resulting
graph (Fig. 1D) reveals an intriguing insight into the differentiation
at the end of the exponential growth phase, but most surprising is
the separation in cell fate decisions that occurs during the diauxic
exclusively have lysis as a cell fate only appear later and are
distributed between the two viable cell fate paths [i.e., sporulation
and diauxic growth (SI Fig.7)]. Clearly, the belief that the only cells
that contribute to the perpetuation of a B. subtilis population that
has exhausted the primary nutrients are those that sporulate needs
revision. It appears that the alternative to sporulation is to use
secondary metabolites to continue growth as long as possible. In
fact, under these conditions, this alternative path results in a
population (in terms of potentially viable cells and spores) than the
initial, early sporulation path (SI Table 1).
Lysis Does Not Depend on the Sporulation Killing Factors.Sporulating
cells express the skf and sdp operons whose gene products are
responsible for the export of killing factors that induce lysis (14).
suggested that the sporulating cells use the nutrients released by
their killed siblings and thereby delay or prevent full commitment
under high-cell-density conditions as in dense colonies on agar
plates. However, under the growth conditions in our work, lysis
within microcolonies did not depend on the presence of these
killing factors (SI Movie 4).
The question remains as to what determines the decision to
sporulate or to follow the diauxic growth path. We considered the
possibility that the aging of the cell plays a role. For instance, in
yeast, aging is a driving force in generating phenotypic variation,
resulting in subpopulations of cells with different resistances to
oxidative stress (15). Rod-shaped bacteria, such as B. subtilis and E.
coli, divide by binary fission whereby cell division takes place at
midcell (16). The newly formed division septum will become a new
cell pole after cytokinesis has completed. As a result of this
symmetrical cell division, two morphologically identical ‘‘progeny’’
cells are formed; however, these two cells differ with respect to the
(A) Still frames (phase contrast) of the outgrowth of a
nifications. (B) Log of microcolony biomass (black line)
and average growth rates (gray triangles) are plotted
from the birth point of the cell until the next cell
division event or cell fate decision (see SI Materials and
Methods). Biomass was calculated as a function of cell
this cell during its life is represented on the y axis (AU).
(D) Cell fates plotted onto the birth points of C. Every
it is certain that either this cell or all of its descendents
will follow a specific fate. Blue circles show the growth
rates of individual cells committed to spore formation;
green triangles show diauxic growth fate cells; and red
triangles indicate lysing cells.
Typical B. subtilis microcolony development.
www.pnas.org?cgi?doi?10.1073?pnas.0700463105Veening et al.
pole suffers from aging and exhibits a decreased growth rate, less
determine whether cell aging could play a role in B. subtilis cell
differentiation, it was necessary first to establish the presence of
(during the primary, glucose-based phase of growth) were deter-
mined based on cell pole age (inspecting a total of 1,080 cells). As
effect during vegetative growth. Despite the presence of aging,
committed to sporulation show that, on average, sporulating cells
are not significantly older or younger than non-spore formers
(expected average pole age of 106 sporulating cells from two
and Methods). Similarly, cells that ultimately lysed were tested for
age (expected average pole age of 729 lysing cells from two
age of cells that initially commit to diauxic growth were also
analyzed, with no significant bias in pole age found (expected
average pole age of 284 diauxic growth cells from two independent
films, 1.00; actual age, 1.03; P ? 0.71). Taken together, these data
demonstrate that cell age is not a factor that determines the fate of
B. subtilis cells.
A replication checkpoint system in B. subtilis ensures that sporu-
lation does not initiate when cells are not in their correct cell cycle
(17). Therefore, we looked for cell cycle-related physiological
parameters that might distinguish spore formers from non-spore
formers, and we closely examined and compared birth times,
growth rates, and cell length of cells demonstrating these two fates
(SI Fig. 9). These analyses did not reveal any apparent differences
in any of the measured parameters between the two cell types.
Because we could not identify any age- or cell cycle-related bias
toward spore formation, this analysis strengthens the supposition
that the initial decision to sporulate is stochastic in its origin.
Previous work has shown that in liquid media, the endospore is
preferably formed at the old pole of the cell (18). An analysis of the
(49%), indicating that there is no significant preference for spores
to form at the old or new pole within microcolonies on solid media.
Cell Lineage History Influences Cell Fate. As shown in Fig. 1D, cells
that will sporulate do not grow during the diauxic shift, which
suggests that the timing of activation of the main sporulation
transcription regulator, Spo0A, differs substantially between cells.
Spo0A is directly responsible for the initiation of sporulation; but
to become active, Spo0A needs to be phosphorylated, which is
achieved through a complex phosphorelay involving multiple phos-
photransferases and phosphatases on which environmental signals
act (19). Using standard light microscopy, only the late stages of
sporulation can be followed. To visualize the induction of the
sporulation cascade at an earlier point during development, we
fused the gene encoding green fluorescent protein (gfp) to the
Spo0A?P-inducible spoIIA promoter (spoIIA-gfp). When micro-
colonies reached the stationary growth phase, part of the popula-
tion begins to express GFP (Fig. 3A and SI Movies 1 and 2). These
(blue circles, showing rejuvenation) or old pole cells (red circles, showing
aging). Error bars represent the SEM. Trend lines are shown in green (the
actual progressions may not be linear). See also SI Fig. 8 and ref. 12.
Aging in B. subtilis. Effects of consecutive divisions as a new pole cells
clarity, the fluorescence of only the first 13 h of the microcolony existence, before the later second round of growth and spore formation, is depicted. The SEM
when the microcolony consisted of 408 cells.
Lineage-related expression of cell fates. (A) Average gfp expression from the spoIIA promoter of subpopulations of cells with different cell fates. For
Veening et al.
March 18, 2008 ?
vol. 105 ?
no. 11 ?
cells are also the first to differentiate into spores (SI Fig. 10).
Eventually, all cells within the microcolony will express GFP to
some level (Fig. 3A). Lysing cells also show a range of induction
levels of GFP, confirming our observation from the growth rate
graph in Fig. 1 that lysing cells arise in both the sporulation and
diauxic-growth pathways (SI Fig. 11). Again, no apparent differ-
spoIIA-gfp at an early stage and cells that did not.
A striking observation of the time-lapse films was that fluores-
cent cells appear to group in families, seen as long lineages of cells
oriented end-to-end. These strings of fluorescent cells are often
of GFP (Fig. 3B). When we examined a strain that carries a
families of fluorescent and nonfluorescent cells were also observed
(SI Movie 5). Sporulating cells also show spatial clustering, albeit
weak, as demonstrated by a randomized nearest-neighbor analysis
(see SI Materials and Methods).
microcolony, we examined whether this spatial correlation might
arise from an underlying lineage correlation. To assess whether
spoIIA-inducing and sporulating cells are closely related or ran-
domly distributed within the genealogy, we used parsimony recon-
struction of ancestral cell states. Parsimony reconstruction esti-
mates the history of cell fates by minimizing the total number of
changes in the fate decisions that have occurred over the history of
the tree (for a more detailed description, see ref. 21). In this way,
spoIIA-gfp-expressing and sporulating cells can be estimated. As
shown in Fig. 4A, expression of GFP from the spoIIA promoter is
highly lineage-dependent and can often be traced back for more
than four cell divisions within the microcolony lineage tree (for
higher resolution, see SI Fig. 12).
To check that the observed lineage dependence in fluorescence
was not generated merely by dilution of already-expressed GFP
among the offspring, we performed an extensive analysis on the
fluorescence profiles throughout the lineage from cells that cluster
together (see SI Fig. 13). Indeed, spoIIA-gfp activation at the time
point used for the parsimony reconstruction (669 min) occurs
been established, indicating that the decision to sporulate occurs
upstream of the spoIIA promoter and GFP induction (SI Fig. 13).
We conclude that the signal to activate Spo0A propagated from
a common ancestor, and because these cell fates are not caused by
genetic mutations, the cell state inherited from the ancestor must
be epigenetic. Analysis of parsimony reconstruction using actual
spore formation as a phenotype also showed clustering within the
lineage tree and can often be traced back more than two cell
14). As a control, we tested whether these clustered trees could be
generated by a random process alone. To do so, we assessed with
a statistical test for phylogenetic signal (20) whether closely related
cells are more likely to express the same phenotype. The mean
squared error (MSE) of independent ‘‘phylogenetic contrasts’’ for
the observed and randomized trees were calculated as described in
ref. 20 (SI Fig. 15) (for spore formation: observed MSE ? 0.102,
mean MSE of permuted data ? 0.179, P ? 0.001; for GFP
expression: observed MSE ? 337, mean MSE of permuted data ?
847, P ? 0.001). This analysis demonstrated that the phylogenetic
indicates that families of cells are more likely to share the same
phenotype in a differentiating B. subtilis colony.
Epigenetic Inheritance Depends on the Phosphorelay. For epigenetic
inheritance to occur, the sporulation signal must be preserved over
multiple generations. Therefore, it is likely that the signal is
reinforced and maintained during growth and division. The auto-
several positive-feedback loops that could be responsible for this
phenomenon (2). For example, Spo0A?P binds to its own pro-
moter to stimulate transcription, and expression of kinA, encoding
the primary kinase of the phosphorelay, is indirectly activated by
spo0A transcription is responsible for the families of spoIIA-GFP-
expressing cells. To do so, we constructed a strain where the native
promoter of spo0A was replaced by an IPTG-inducible promoter
(Pspac) (Fig. 5A). Previous work had shown that wild-type levels of
Spo0A are achieved with this promoter when a concentration of
on an agarose patch containing medium with 40 ?M IPTG clearly
shows two subpopulations in fluorescence distribution and the
presence of subfamilies (Fig. 5C, SI Fig. 16, and SI Movie 6). Even
?M IPTG, spoIIA-gfp expression is bimodal, but now more cells
essential for bistable expression (and sporulation) or for epigenetic
the phosphorelay in this process, the constitutively active mutant
variant of Spo0A (Sad67) was tested. This mutant protein does not
require phosphorylation by the phosphorelay to activate sporula-
tion gene expression (23, 24). Thus, a strain harboring the sad67
gene under the control of the spac promoter lacks any putative
positive-feedback loop acting on spo0A transcription or Spo0A
phosphorylation (Fig. 5B). As shown in Fig. 5D, SI Fig. 16, and SI
Movie 8, in the presence of 40 ?M IPTG, all cells gradually and
synchronously express spoIIA-gfp, and subfamilies are no longer
established. These data suggest that the phosphorelay is important
for both bistable induction of spo0A and the propagation of the
Spo0A activation signal. In fact, by removing a major phosphatase
(RapA) of the phosphorelay (25), the number and size of subfam-
ilies that express spoIIA-gfp increase (SI Fig. 17 and SI Movie 9).
Artificial induction of another phosphorelay phosphatase (Rap60)
(26) gave the opposite effect and resulted in strikingly smaller
subfamilies (SI Fig. 17 and SI Movie 9).
By following a growing B. subtilis microcolony in detail, we came
across some unexpected findings. The total growth rate of a
microcolony follows a classic curve with an exponential phase
followed by a diauxic phase. Surprisingly, it appeared that the
diauxic growth phase is exclusively used by nonsporulating cells
(Fig. 1). Because spore formation is a long and energy-intensive
process, it is plausible that sporulating cells use the overflow
the fluorescence character (GFP) onto the true lineage tree of 356 cells. Red
tips, cells with fluorescence value above 40 fluorescence units. Black tips, cells
below this threshold. (B) Parsimony mapping of fate information onto the
true lineage tree of 531 cells. Every end point in the tree represents one
offspring cell; red tips, spore-forming cells. Detailed parsimony character
mappings are shown in SI Fig. 12.
Inheritance in the decision to sporulate. (A) Parsimony mapping of
www.pnas.org?cgi?doi?10.1073?pnas.0700463105Veening et al.
metabolites to complete this cell fate. The induction of overflow
metabolism can be monitored by following the expression of the
acetoin catabolic pathway (acoA-L), which is exclusively acti-
vated when the growth medium is exhausted for glucose (13).
Indeed, we found that all cells in a microcolony express this
SI Movie 10). That the sporulating cells do not continue to grow
is likely caused by the elevated concentrations of Spo0A?P in
these cells because high levels of Spo0A?P inhibit symmetrical
(vegetative) cell division (27).
Cells that do not sporulate when nutrients become limiting are
not lost. By following the diauxic growth fate, their numbers
increase, and they may sporulate later by using nutrients released
by cells that have lysed. Moreover, these ‘‘diauxic growers’’ are
of spore formation and germination (7). Thus, each of these
pathways is a form of cell specialization. One benefit of the
simultaneous presence of both pathways is the optimal use of
resources in the long run. As mother cells lyse to release the
endospores, they also release cellular components that could be
used as a source of nutrients. In other words, heterogeneity in the
timing of spore formation allows utilization of these resources that
would otherwise be lost. Maybe even more important is that these
two fates have different reproductive potential under different
environments. Because the future is not predictable, the clonal
population profits by being prepared for a variety of future envi-
ronments, a form of bet-hedging (28, 29).
The B. subtilis laboratory strain commonly used, and also used in
this work, is a poor sporulator compared with some natural isolates
(10). Because this strain has been propagated in the laboratory for
many decades, it might be that it is evolutionary optimized to
colonize rapidly from vegetative (diauxic) growing cells. Although
this condition is potentially a response to human-imposed condi-
tions, natural environments with similar fluctuating nutrient levels
may result in the evolution of similar distributions of the bistable
A third cell fate in a maturing B. subtilis colony is cell lysis, and
it may represent a failure to complete the preferred differentiation
pathway. Under our experimental setup, the sporulation killing
factors are not responsible for lysis, indicating a role for other
intrinsic or extrinsic causes.
Cellular aging is a universal phenomenon that also affects
bacteria, as has been shown for the proteobacteria E. coli and
Caulobacter crescentus (12, 30). Here, we demonstrate aging in a
bacterium outside this phylogenetic group (Fig. 2). Older B. subtilis
cells show a clear reduction in growth rate, but cell pole age did not
appear to play a determining role in the choice to sporulate, follow
diauxic growth, or lyse. We also looked at other differences that
might bias cells to a specific fate, such as birth time, cell length, or
growth rate (markers for cell cycle timing and physiological state).
Again, no relation with the separate fates could be identified (SI
Fig. 9). Previous work has shown that sporulation can only occur
It will be interesting to see the importance of the replication state
on the final bifurcation of the differentiation pathways.
When we examined GFP reporter strains that are specific for
Spo0A activation (PspoIIA-gfp and PabrB-gfp), we noticed that GFP-
expressing cells were often clustered (Fig. 3). Such groups of cells
appeared to be phylogenetically related, suggesting that the signal
to activate Spo0A is epigenetically inherited (Fig. 4 and SI Fig. 12).
oocytes and the white-opaque phenotype in Candida albicans, for
module (34, 35); and by using artificial bistable gene regulatory
circuits, epigenetic inheritance of semistable states has recently
been described in Saccharomyces cerevisiae (36, 37). In the case of
B. subtilis sporulation, the phosphorelay provides an excellent
candidate for such a positive-feedback loop, and by testing several
mutants we have shown that both sporulation bistability and
subfamily formation depend on this signal transduction complex
(Fig. 5 and SI Fig. 17). This finding indicates that the sporulation
signal is ‘‘memorized’’ by the positive-feedback architecture of the
Spo0A phosphorylation circuitry, which includes autophosphory-
lating kinases (e.g., KinA and KinB) that are directly and/or
indirectly transcriptionally activated by Spo0A?P. The autostimu-
latory architecture of the phosphorelay seems to be responsible for
IIA/spo0A/sad67, respectively. (C and D) Time-lapse microscopy of strains IIA/?0A/spo0A (Pspac-spo0A) and IIA/spo0A/sad67 (Pspac-sad67). Membranes are stained
by FM5–95 (red), and GFP is depicted in green.
Phosphorelay is required to generate subfamilies. (A and B) Schematic diagram of the regulatory cascade present in strains IIA/?0A/spo0A and
Veening et al.
March 18, 2008 ?
vol. 105 ?
no. 11 ?
both bistable differentiation and epigenetic inheritance. However, Download full-text
the mere presence of positive-feedback regulation does not auto-
matically result in a memory response. For instance, for the
development of genetic competence, another bistable differentia-
tion process in B. subtilis that depends on positive feedback, it was
found that cells were not significantly more likely to become
competent if their sibling became competent (38).
The importance of epigenetic inheritance of cell fates on the
population level may be based on the effect it has on neighboring
cells. In bacterial colonies, which are sessile communities of cells,
epigenetic inheritance affects those cells that are spatially grouped,
in contrast to cells within planktonic cultures. The formation of
biofilms requires systematic cell differentiation, and in B. subtilis,
multicellular structure formation and sporulation are coordinated
and intertwined by the action of Spo0A (11, 39, 40). We propose
Materials and Methods
Strains. The construction of B. subtilis strains abrB-gfp, IIA-gfp, IIA-gfp/?rapA,
strains acoA-gfp, IIA/skf/sdp, and IIA/?0A/spo0A is described in SI Materials and
Time-Lapse Microscopy. Cells were inoculated onto a thin semisolid agarose
see SI Materials and Methods.
Data Analysis. Custom analysis software (12) was supplemented with software
To correct tracking errors and include scoring for spore formation and cell lysis,
history and fate of individual B. subtilis cells growing in these microcolonies.
These datasets include information on the number of division cycles, its length,
width, fluorescence intensity, lineage information, and fate.
Phylogenetic Analyses. The parsimony reconstruction of ancestral states was
carried out with Mesquite version 2.0 (http://mesquiteproject.org). Every node
within the tree represents a cell division event. For Fig. 4B, lineage and cell fate
data contained information from 531 cells (including those that sporulated at a
lineage (Fig. 4A), fluorescence information of 356 cells after 669 min (?11h) of
growth was made binary (‘‘on’’ or ‘‘off’’) on the basis of the cutoff value of 40
arbitrary fluorescence units above background (see SI Fig. 19). When the fluo-
rescence data are randomized and the parsimony states are reconstructed, sig-
cells are more likely to show the same cell state (fluorescence) or fate (spore
formation or not), we used a test for phylogenetic signal (20). The phylogenetic
signal test was done on data for cell fate data (spore formers vs. non-spore
formers) after 25 h of growth. Although we only show the results of one
three independent microcolonies all showing significant ‘‘phylogenetic signal.’’
The phylogenetic signal test for fluorescence was done on data after 669 min of
growth (note that we used continuous fluorescence data for the test, unlike the
data used to generate Fig. 4A). The calculations were carried out as described in
ref. 20 with the use of the original Matlab code kindly provided by Ted Garland.
ACKNOWLEDGMENTS. We thank Richard Madden for the original program-
the anonymous referees for useful comments. J.-W.V. was supported by Grant
by a Ramsay Fellowship from the Royal Netherlands Academy of Arts and Sci-
ences, and by a grant from the Biotechnology and Biological Sciences Research
Biology Organization, Fondation pour la Recherche Me ´dicale, and Institut Na-
tional de la Sante ´ et de la Recherche Me ´dicale. F.T. was supported in part by a
Research Career Development Fellowship.
1. Brandman O, Ferrell JE, Jr, Li R, Meyer T (2005) Interlinked fast and slow positive-
feedback loops drive reliable cell decisions. Science 310:496–498.
2. Smits WK, Kuipers OP, Veening JW (2006) Phenotypic variation in bacteria: The role of
feedback regulation. Nat Rev Microbiol 4:259–271.
mammalian cells: Ultrasensitive and all-or-none responses. Curr Biol 13:315–320.
4. Lai K, Robertson MJ, Schaffer DV (2004) The sonic hedgehog signaling system as a
bistable genetic switch. Biophys J 86:2748–2757.
5. Kærn M, Elston TC, Blake WJ, Collins JJ (2005) Stochasticity in gene expression: From
theories to phenotypes. Nat Rev Genet 6:451–464.
6. Avery SV (2006) Microbial cell individuality and the underlying sources of heteroge-
neity. Nat Rev Microbiol 4:577–587.
7. Piggot PJ, Losick R (2002) Bacillus subtilis and Its Closest Relatives: From Genes to Cells,
8. Fujita M, Gonzalez-Pastor JE, Losick R (2005) High- and low-threshold genes in the
Spo0A regulon of Bacillus subtilis. J Bacteriol 187:1357–1368.
9. Veening JW, Hamoen LW, Kuipers OP (2005) Phosphatases modulate the bistable
sporulation gene expression pattern in Bacillus subtilis. Mol Microbiol 56:1481–1494.
10. Maughan H, Nicholson WL (2004) Stochastic processes influence stationary-phase
decisions in Bacillus subtilis. J Bacteriol 186:2212–2214.
11. Veening JW, Smits WK, Hamoen LW, Kuipers OP (2006) Single cell analysis of gene
expression patterns of competence development and initiation of sporulation in
Bacillus subtilis grown on chemically defined media. J Appl Microbiol 101:531–541.
12. Stewart EJ, Madden R, Paul G, Taddei F (2005) Aging and death in an organism that
reproduces by morphologically symmetric division. PLoS Biol 3:e45.
13. Silbersack J, et al. (2006) An acetoin-regulated expression system of Bacillus subtilis.
Appl Microbiol Biotechnol 73:895–903.
14. Gonzalez-Pastor JE, Hobbs EC, Losick R (2003) Cannibalism by sporulating bacteria.
15. Aguilaniu H, Gustafsson L, Rigoulet M, Nystrom T (2003) Asymmetric inheritance of
oxidatively damaged proteins during cytokinesis. Science 299:1751–1753.
17. Burkholder WF, Kurtser I, Grossman AD (2001) Replication initiation proteins regulate
a developmental checkpoint in Bacillus subtilis. Cell 104:269–279.
18. Dunn G, Mandelstam J (1977) Cell polarity in Bacillus subtilis: Effect of growth condi-
tions on spore positions in sister cells. J Gen Microbiol 103:201–205.
by a multicomponent phosphorelay. Cell 64:545–552.
20. Felsenstein J (2004) Inferring Phylogenies (Sinauer, Sunderland, MA).
21. Blomberg SP, Garland T, Jr, Ives AR (2003) Testing for phylogenetic signal in compar-
ative data: Behavioral traits are more labile. Evolution 57:717–745.
22. Predich M, Nair G, Smith I (1992) Bacillus subtilis early sporulation genes kinA, spo0F,
and spo0A are transcribed by the RNA polymerase containing sigma H. J Bacteriol
23. Ireton K, Rudner DZ, Siranosian KJ, Grossman AD (1993) Integration of multiple
developmental signals in Bacillus subtilis through the Spo0A transcription factor.
Genes Dev 7:283–294.
24. Fujita M, Losick R (2005) Evidence that entry into sporulation in Bacillus subtilis is
Genes Dev 19:2236–2244.
25. Perego M, Hoch JA (2002) in Bacillus subtilis and Its Closest Relatives: From Genes to
Cells, eds Sonenshein AL, Hoch JA, Losick R (Am Soc Microbiol, Washington, DC), pp
26. Koetje EJ, Hajdo-Milasinovic A, Kiewiet R, Bron S, Tjalsma H (2003) A plasmid-borne
Rap-Phr system of Bacillus subtilis can mediate cell density-controlled production of
extracellular proteases. Microbiology 149:19–28.
intermediate of the cytokinetic protein FtsZ. Cell 109:257–266.
28. Thattai M, van Oudenaarden A (2004) Stochastic gene expression in fluctuating
environments. Genetics 167:523–530.
fluctuating environments. Science 309:2075–2078.
30. Ackermann M, Stearns SC, Jenal U (2003) Senescence in a bacterium with asymmetric
division. Science 300:1920.
31. Dawes IW, Kay D, Mandelstam J (1971) Determining effect of growth medium on the
shape and position of daughter chromosomes and on sporulation in Bacillus subtilis.
32. Hauser PM, Errington J (1995) Characterization of cell cycle events during the onset of
sporulation in Bacillus subtilis. J Bacteriol 177:3923–3931.
33. Casadesus J, D’Ari R (2002) Memory in bacteria and phage. Bioessays 24:512–518.
34. Xiong W, Ferrell JE, Jr (2003) A positive-feedback-based bistable ‘‘memory module’’
that governs a cell fate decision. Nature 426:460–465.
tional feedback loops control white-opaque switching in Candida albicans. PLoS Biol
36. Kaufmann BB, Yang Q, Mettetal JT, van Oudenaarden (2007) A heritable stochastic
switching revealed by single-cell genealogy. PLoS Biol 5:e239.
38. Suel GM, Garcia-Ojalvo J, Liberman LM, Elowitz MB (2006) An excitable gene regula-
tory circuit induces transient cellular differentiation. Nature 440:545–550.
39. Branda SS, Gonzalez-Pastor JE, Ben-Yehuda S, Losick R, Kolter R (2001) Fruiting
body formation by Bacillus subtilis. Proc Natl Acad Sci USA 98:11621–
40. Veening JW, Kuipers OP, Brul S, Hellingwerf KJ, Kort R (2006) Effects of phosphorelay
perturbations on architecture, sporulation, and spore resistance in biofilms of Bacillus
subtilis. J Bacteriol 188:3099–3109.
www.pnas.org?cgi?doi?10.1073?pnas.0700463105Veening et al.