*For correspondence:
melowitz@caltech.edu
Competing interests:
The
authors declare that no
competing interests exist.
Funding:
See page 15
Received:
25 October 2017
Accepted:
21 May 2018
Published:
29 May 2018
Reviewing editor:
Martin
Ackermann, ETH Zurich,
Switzerland
Copyright Rosenthal et al. This
article is distributed under the
terms of the
Creative Commons
Attribution License,
which
permits unrestricted use and
redistribution provided that the
original author and source are
credited.
Metabolic interactions between dynamic
bacterial subpopulations
Adam Z Rosenthal
1,2
, Yutao Qi
1,2
, Sahand Hormoz
1,2
, Jin Park
1,2
,
Sophia Hsin-Jung Li
3
, Michael B Elowitz
1,2,4
*
1
Division of Biology and Biological Engineering, California Institute of Technology,
Pasadena, United States;
2
Department of Applied Physics, California Institute of
Technology, Pasadena, United States;
3
Department of Molecular Biology, Princeton
University, Princeton, United States;
4
Howard Hughes Medical Institute, Pasadena,
United States
Abstract
Individual microbial species are known to occupy distinct metabolic niches within multi-
species communities. However, it has remained largely unclear whether metabolic specialization
can similarly occur within a clonal bacterial population. More specifically, it is not clear what
functions such specialization could provide and how specialization could be coordinated
dynamically. Here, we show that exponentially growing
Bacillus subtilis
cultures divide into distinct
interacting metabolic subpopulations, including one population that produces acetate, and another
population that differentially expresses metabolic genes for the production of acetoin, a pH-neutral
storage molecule. These subpopulations exhibit distinct growth rates and dynamic interconversion
between states. Furthermore, acetate concentration influences the relative sizes of the different
subpopulations. These results show that clonal populations can use metabolic specialization to
control the environment through a process of dynamic, environmentally-sensitive state-switching.
DOI: https://doi.org/10.7554/eLife.33099.001
Introduction
Co-utilization of carbon sources was described alongside diauxie by Jacques Monod in his PhD the-
sis (
Monod, 1958
), and is common in many organisms (
Peyraud et al., 2012
). In the Gram-positive
bacterium
Bacillus subtilis,
two preferred carbon sources are co-utilized: glucose and malate
(
Kleijn et al., 2010
). When both of these carbon sources are available they are consumed simulta-
neously, generating growth rates that surpass those achieved with either substrate alone
(
Kleijn et al., 2010
). Under conditions of rapid growth, co-consumption of glucose and malate leads
to the accumulation of high levels of acetate (Kleijn et al., 2010). As a weak organic acid, acetate can
be harmful to cells even in buffered medium (
Rosenthal et al., 2008
). Acetate and related short-
chain fatty acids enter the cell passively in the neutral form and then dissociate intracellularly, releas-
ing a proton and transiently acidifying the cytoplasm (
Russell and Diez-Gonzalez, 1997
;
Roe et al.,
1998
). The intracellular dissociation of acetate also disrupts the cellular anion balance, with negative
effects on metabolism (
Roe et al., 1998
;
Roe et al., 2002
) and transcription (
Rosenthal et al.,
2008
). When extracellular acetate levels rise to toxic levels the growing
Bacillus subtilis
culture con-
sumes the acetate and produces acetoin, a non-toxic pH-neutral overflow metabolite that can be
used as a carbon source in later growth stages (
Speck and Freese, 1973
) (
Figure 1A
).
A biphasic growth strategy, in which acetate is produced to a toxic level and then reabsorbed
and replaced by a non-toxic metabolite (
Wolfe, 2005
), is common to many bacterial species and is
important both for understanding the basic biology of bacterial growth in culture, and for applica-
tions in metabolic engineering (
Papagianni, 2012
). However, it has generally been studied only at
the population level, implicitly assuming a homogeneous progression of the entire culture from
Rosenthal
et al
. eLife 2018;7:e33099.
DOI: https://doi.org/10.7554/eLife.33099
1 of 18
RESEARCH ARTICLE
acetate producing to acetate detoxifying states. By contrast, single cell approaches suggest that
bacterial populations can exhibit enormous heterogeneity in functional and gene expression states
(
Eldar et al., 2009
;
Locke et al., 2011
;
Su
̈
el et al., 2006
;
Levine et al., 2012
;
Davidson and Sure-
tte, 2008
;
Dubnau and Losick, 2006
;
Gefen and Balaban, 2009
). This prompts the questions of
whether microbial cells differentiate into metabolically distinct subpopulations, and more specifically,
whether acetate production and detoxification might occur in distinct cells specializing in acetate
production or detoxification, respectively.
Results
To address these questions we constructed a library of strains with reporters for key genes involved
in central carbon metabolism, acetate production, and organic acid detoxification (
Figure 1A
). We
introduced a fluorescent protein (YFP) under the control of promoters for 13 different metabolic
genes and stably incorporated them into the commonly used
sacA
site within the genome
(
Supplementary file 1
), (
Eldar et al., 2009
;
Locke et al., 2011
). We chose to use fluorescent pro-
moter reporters because they allow acquisition of dynamic measurements from individual living cells,
are easy to construct and integrate into the
B. subtilis
genome, allow for analysis of multiple genes
within the same cell, and can be used in fluorescence cell sorting for RNAseq experiments. Using
quantitative single-cell fluorescence microscopy, we analyzed the distribution of expression levels of
these 13 metabolic genes in individual cells at different times along the growth curve in buffered cul-
ture medium containing 22 mM glucose and 50 mM malate. To eliminate oxygen gradients, 10 mL
cultures were grown in 250 mL flasks with rapid shaking (250 RPM).
Four genes had expression levels that were at or near background and were not considered fur-
ther (
acoA
,
gntZ
,
pycA
,
sdhC
). Most of the genes showed unimodal distributions (
Figure 1—figure
eLife digest
The chemical reactions that occur within a living organism are collectively referred
to as its metabolism. Many metabolic reactions produce byproducts that will poison the cells if they
are not dealt with: fermenting bacteria, for example, release harmful organic acids and alcohols.
How the bacteria respond to these toxins has been most studied at the level of entire microbial
populations, meaning the activities of individual cells are effectively “averaged” together. Yet, even
two bacteria with the same genes and living in the same environment can behave in different ways.
This raises the question: do bacterial populations specialize into distinct subpopulations that play
distinct roles when dealing with metabolic products, or do all cells in the community act in unison?
Rosenthal et al. set out to answer this question for a community of
Bacillus subtilis
, a bacterium
that is commonly studied in the laboratory and used for the industrial production of enzymes. The
analysis focused on genes involved in fundamental metabolic processes, known as the TCA cycle,
which the bacteria use to generate energy and build biomass. The experiments revealed that, even
when all the cells are genetically identical, different
Bacillus subtilis
cells do indeed specialize into
metabolic subpopulations with distinct growth rates.
Time-lapse movies of bacteria that made fluorescent markers of different colors whenever certain
metabolic genes became active showed cells switching different colors on and off, indicating that
they switch between metabolic subpopulations. Further biochemical studies and measures of gene
activity revealed that the different subpopulations produce and release distinct metabolic products,
including toxic byproducts. Notably, the release of these metabolites by one subpopulation
appeared to activate other subpopulations within the community.
This example of cells specializing into unique interacting metabolic subpopulations provides
insight into several fundamental issues in microbiology and beyond. It is relevant to evolutionary
biologists, since the fact that fractions of the population can switch in and out of a metabolic state,
instead of evolving into several inflexible specialists, may provide an evolutionary advantage in
fluctuating natural environments by reducing the risk of extinction. It also has implications for
industrial fermentation processes and metabolic engineering, and may help biotechnologists design
more efficient ways to harness bacterial metabolism to produce useful products.
DOI: https://doi.org/10.7554/eLife.33099.002
Rosenthal
et al
. eLife 2018;7:e33099.
DOI: https://doi.org/10.7554/eLife.33099
2 of 18
Research article
Computational and Systems Biology
Microbiology and Infectious Disease
Figure 1.
Two genes in central carbon metabolism are heterogeneously expressed in a clonal population of
B. subtilis
. (
A
)
B. subtilis
uses glucose and
malate as preferred carbon sources, and under aerobic culture conditions produces acetate and acetoin as major overflow metabolites. Promoter
reporter strains were made for genes participating in the reactions marked with a yellow dot (
B
) Histograms depict the heterogeneous expression of
the central metabolism genes
sucC
(top panel) and
alsS
(bottom panel). Insets using merged phase and fluorescence images show typical fields of
cells, including cells in the high expressing tail of the distributions. (
C
) The heterogeneous expression of
sucC
(red line) and
alsS
(green line) is maximal
at different timepoints along the growth curve (black line). Black arrows denote the sampling timepoints shown in
Figure 1B
. (
D
) A line graph depicting
the accumulation of extracellular acetate and acetoin in the growth media during exponential and early stationary growth (OD
600
, black line). Acetate
(red line) is released around mid-exponential phase, and is reabsorbed at a later time during which acetoin is produced (green line).
DOI: https://doi.org/10.7554/eLife.33099.003
Figure 1 continued on next page
Rosenthal
et al
. eLife 2018;7:e33099.
DOI: https://doi.org/10.7554/eLife.33099
3 of 18
Research article
Computational and Systems Biology
Microbiology and Infectious Disease
supplement 1
), with relatively little skew (less than
±
0.7). Two genes,
sucC
and
alsS
, encoding succi-
nate co-A ligase and acetolactate synthase, respectively, were more heterogeneous (
Figure 1B
). We
observed skew values greater than 1 (4.72 and 1.14, respectively) with 3.8% of P
sucC
-YFP and 8.1%
of P
alsS
-YFP cells exhibiting high expression levels (
2 standard deviations above the mean) at
OD
600
~
0.8 (
sucC
) and OD
600
~
2 (
alsS
). In addition, for both genes, we observed cells whose expres-
sion exceeded the mean by >3 fold. While gene expression distribution can be broader immediately
after gene activation than at steady-state (
Shahrezaei and Swain, 2008
), both
sucC
and
alsS
main-
tained heterogeneous expression for several hours after the onset of expression. For these reasons,
we decided to focus on these two genes for further study.
To better understand when this heterogeneity emerges in batch culture, we performed a time
course analysis of the fraction of
sucC
and
alsS
positive cells (cells
2 standard deviations above the
mean were denoted
sucC+
and
alsS+
,
Figure 1C
). We observed that the subpopulation of
sucC+
cells only existed transiently, in mid- to late-exponential phase (
Figure 1C
), coinciding with the time
and culture optical density at which acetate production was observed (when the time derivative in
acetate, that is, the rate of change in acetate concentration, is positive
~
150–300 min,
Figure 1D
).
This observation suggested that
sucC
expression could be involved in acetate production. A parallel
analysis of
alsS
expression revealed the opposite behavior, with
alsS
expression dynamics coinciding
with a decrease in acetate and a concomitant increase in acetoin levels (
Figure 1C,D
). This behavior
is generally consistent with the known role of
alsS
in acetoin production in response to acetate toxic-
ity (
Speck and Freese, 1973
). Together, these results show that a dynamic change in acetate and
acetoin levels in the culture overlaps with changes in the population fraction of
sucC
and
alsS
expressing cells.
A role for sucC in acetate production has not been studied previously. To understand the rela-
tionship between the subpopulation marked
sucC+
and acetate production, we used fluorescence
activated cell sorting (FACS) of the P
sucC
YFP reporter strain to sort cells expressing YFP from a
SucC
promoter at the time of peak acetate levels, and performed RNAseq to compare gene expression
profiles (
Figure 2A
,
Figure 2—figure supplement 1
). As expected,
sucC
expression was elevated 2-
fold in the
sucC+
sorted subpopulation (blue dot,
Figure 2A
). This is particularly meaningful consid-
ering the fact that the fluorescent marker used for sorting is a stable reporter, making it likely that
some sorted cells may have high level of fluorescent signal even after exiting the transcriptionally
active state. For most genes, we observed a broad correlation in gene expression between the two
populations. However, RNAseq analysis with cuffdiff (
Trapnell et al., 2010
) and gene set enrichment
analysis with GSEA (
Subramanian et al., 2005
) showed that genetic competence genes
(
Berka et al., 2002
) were significantly enriched in the
~
300 upregulated genes in the
sucC+
subpop-
pulation (red dots and inset,