Heterogeneous Response to a Quorum-Sensing Signal in the Luminescence of Individual Vibrio fischeri
Full Citation
Permanent Link: http://ufdc.ufl.edu/IR00000298/00001
 Material Information
Title: Heterogeneous Response to a Quorum-Sensing Signal in the Luminescence of Individual Vibrio fischeri
Physical Description: Archival
Creator: Hagen, Stephen J.
Perez, Paul Delfino
Publisher: PLoS ONE
Publication Date: 2010
Spatial Coverage:
Funding: Publication of this article was funded in part by the University of Florida Open-Access publishing Fund.
 Record Information
Source Institution: University of Florida
Holding Location: University of Florida Institutional Repository
Rights Management: All rights reserved by the source institution and holding location.
System ID: IR00000298:00001

Full Text
Heterogeneous Response to a Quorum-Sensing Signal in the Luminescence of
Individual Vibrio fischeri
Pablo Delfino Perez and Stephen J. Hagen*

Physics Department, University of Florida, PO Box 118440
Gainesville FL 32611-8440 USA

The marine bacterium Vibrio fischeri regulates its bioluminescence through a
quorum sensing mechanism: the bacterium releases diffusible small molecules
(autoinducers) that accumulate in the environment as the population density
increases. This accumulation of autoinducer (Al) eventually activates
transcriptional regulators for bioluminescence as well as host colonization
behaviors. Although V.fischeri quorum sensing has been extensively
characterized in bulk populations, far less is known about how it performs at the
level of the individual cell, where biochemical noise is likely to limit the precision
of luminescence regulation. We have measured the time-dependence and Al-
dependence of light production by individual V.fischeri cells that are immobilized
in a perfusion chamber and supplied with a defined concentration of exogenous
Al. We use low-light level microscopy to record and quantify the photon
emission from the cells over periods of several hours as they respond to the
introduction of Al. We observe an extremely heterogeneous response to the Al
signal. Individual cells differ widely in the onset time for their luminescence and
in their resulting brightness, even in the presence of high Al concentrations that
saturate the light output from a bulk population. The observed heterogeneity
shows that although a given concentration of quorum signal may determine the
average light output from a population of cells, it provides far weaker control
over the luminescence output of each individual cell.
^Corresponding author. Email: sjhagen@ufl.edu

Numerous bacterial species use a form of chemical communication known as
quorum sensing (QS) to regulate gene expression [1]. The bacteria synthesize
and release small diffusible molecules known as autoinducers, which accumulate
as the bacterial population density grows. As their concentration rises, the
5 autoinducers activate transcriptional regulators that trigger important
phenotypic changes in the cells. QS therefore allows a population-sensitive
switch between different phenotypic states [1]. However, although QS is most
easily interpreted as a population-counting behavior, QS pathways are typically
complex, often employing multiple autoinducer signals and receptors. They may
10 also interact with other physical and biological parameters of the organism's
environment in addition to the population density [2-5].

The complexity of these pathways raises questions about how bacteria use QS to
probe their environment and exactly what types of information they may gather
15 through this mechanism. Understanding the capabilities and fundamental
limitations of QS requires detailed experimental and theoretical studies of QS
systems at the level of individual cells. The goal of this study is to characterize
the overall performance of QS at the single-cell level in one important model
organism. We aim to measure the precision with which an individual Vibrio
20 fischeri cell converts a well-defined QS signal input to a bioluminescence output.

V.fischeri is a Gram-negative marine bacterium that regulates its own
bioluminescence through QS [6]. The luminescence is produced by a bacterial
luciferase that utilizes FMNH2, O2, and a long-chain aldehyde as substrates. At
25 low cell densities, as in open seawater, the lux genes that synthesize the
luciferase and substrates are switched off and the bacterial cells are dark.
However, the bacterium also colonizes the light organs of fish and squid species,
where it attains high cell densities and the lux genes become strongly induced.
In the light organ of its symbiotic host squid Euprymna scolopes, V.fischeri may
30 attain 10^ 10^ cells/cm^ and a single cell may emit -10^ photons/s [7].

Studies of bulk populations of V.fischeri have revealed an intricate molecular
mechanism for this population-sensitive switch [6, 8]. The QS pathway employs
three autoinducer synthases, three corresponding autoinducers, and three
35 cognate receptors [8]. The full pathway integrates the separate autoinducer
signals to regulate not only the luminescence behavior but also other phenotypes
related to colonization of the symbiotic host [9]. Of the three signal channels, the
LuxI/LuxR pathway shown in Figure lA has been the subject of the most

extensive study. It consists of an autoinducer synthase Luxl, an autoinducer {N-
40 3-oxohexanoyl-L-homoserine lactone, 30C6HSL), and the transcriptional
activator LuxR, as well as the luminescence genes luxCDABEG. When the
concentration of 30C6HSL is sufficiently high, it forms a complex with LuxR that
activates transcription of the lux operon, leading to luciferase synthesis and
bioluminescence. The other two QS pathways (not shown in Figure lA) detect a
45 second homoserine lactone autoinducer (N-octanoyl-L-homoserine lactone,
C8HSL) that is produced by a synthase AinS and a third autoinducer Al-2 (as in
V. harveyi [8]) that is produced by LuxS.

Because it was the first known example of a Gram-negative QS system and
50 remains one of the best understood, LuxI/LuxR has been a model system for
theoretical and computational studies of the dynamics of quorum regulation.
Several authors have modelled its deterministic dynamics [10-14] as well as the
stochasticity [15-17] arising from the biochemical noise in gene expression [18].
The deterministic models characterize the stability of the "on" and "off" states of
55 LuxI/LuxR luminescence as well as the dynamics of switching and hysteresis.
Experiments on bulk cultures can provide a suitable test of such models [14].
However, bulk studies measure only average properties of the population. They
do not address stochasticity and they do not reveal exactly what information the
individual cell gathers in probing its environment with a QS mechanism. In
60 particular, the accuracy of the QS pathway as a sensor of the individual cell's
environment and as a regulator of phenotype, and the impact of stochasticity on
QS, can only be tested by experimental measurements on individual cells [14,19-
21]. Here we ask how accurately the autoinducer signal input to a single cell
defines or predicts the bioluminescence output from that cell.
A single-cell study of V.fischeri presents technical challenges, as the
bioluminescence emission from individual bacterial cells is exceedingly weak
and has rarely been measured quantitatively [22-25]. The light output from one
V. fischeri cell is estimated to lie in the range from 10"^ to lO'' photons/s,
70 depending on the strain, the environment, and whether the culture is fully
induced by its multi-input QS system [7, 26]. Only a fraction of this photon flux
can actually be collected, and therefore the measurable flux from one cell is
typically weaker than the signal that can be collected from even a single molecule
of a fluorescent reporter like EGFP [27, 28]. Under stable conditions and with
75 sufficiently long integration times, however, the luminescence from one cell can
be measured with a photomultiplier [24] or with an intensified or cryogenically-
cooled CCD camera [22, 23, 29]. We used an intensified camera and long image
exposures (10-15 minutes) to track the bioluminescent emission from individual

cells of V.fischeri strain MJll. The cells were immobilized on the window of an
80 observation chamber that was continuously perfused with medium containing
exogenous 30C6HSL autoinducer (Al), so that each cell was subject to a precisely
defined local Al concentration. Tracking individual cells over periods of several
hours, we found that cells differ widely in the time scale of their bioluminescence
response and in the overall intensity of that response. Hence, while QS can
85 coordinate and synchronize the average luminescence output of the bacterial
population, it has relatively imprecise control over the response of an individual

Individual bacterial cells emit very weak bioluminescence and the corresponding
signal levels are far weaker than (e.g.) the fluorescence that is typically collected
from a cell expressing GFP. Therefore, as explained in the Materials and Methods
and Text SI, we used several procedures to ensure that the microscopy imaging
95 and alignment were stable over the 3-4 hr period of luminescence observations
and that any observed heterogeneity in the light output from individual cells was
not a detection artifact. We verified that the cells remained stationary and in
focus during imaging (Figure SI), that the observed variations in luminescent
emission were larger than our measurement uncertainties (Figure S2), and that
100 the camera, microscope, and images were physically stable over periods of 4
hours or longer.

In the absence of exogenous autoinducer the V.fischeri cells in the perfusion
chamber produced no detectable luminescence. However, when at least -50 nM
105 of autoinducer {Al, 30C6HSL) was provided in the flowing medium the
luminescence of individual cells was clearly resolved. Figure 2 compares dark-
field {i.e. externally illuminated) and bioluminescence {i.e. luminescence emission
without external illumination) images of individual V. fischeri cells adhering to
the glass window in the presence of 500 nM Al. Qualitatively the image already
110 suggests that different cells emit with different intensities, even at a high Al
concentration that saturates the output of the bulk population (Figure IB). A
quantitative analysis of all data confirmed that the brightness of the cells was
heterogeneous at all autoinducer concentrations studied (0-1000 nM Al). At 1000
nM Al we found many individual cells emitting little light during a ten minute
115 exposure, even though we observed these same cells growing and dividing
during the ~4 hr duration of observation.

Studies of bulk cultures under our growth conditions established that the shape
of the luminescence versus Al response curve was established within 2-3 hrs
120 following introduction of Al (Figure IB). Therefore, an observation period of -3-4
hrs in a perfusion chamber should be sufficient to observe the response of
individual cells to introduction of Al. Figure 3 shows the time course of the
luminescence collected from an ensemble of individual cells. The luminescence
of each cell is tracked over time through a series of 10-minute camera exposures
125 (see Materials and Methods), following the introduction of exogenous Al at t = 0.
The initial response of the cells (0 < f < 100-150 minutes) is a transient increase or
decrease in average luminescence, as the Al concentration in the perfusion
chamber may be greater or less than in the starting culture. On a longer time
scale {t = 150-250 minutes) the cells attain average emission levels that are
130 consistent with the supplied concentration of exogenous Al. However a large
degree of cell-to-cell variability is apparent. The brightness of the different cells
diverges over time, with many cells luminescing at very modest levels while a
small fraction of cells emit much more brightly.

135 Heterogeneity is also apparent in the time scale of response. Figure 3 shows that,
when a high Al concentration (1000 nM) is introduced at f = 0, some cells begin to
respond quickly, with 250-350 photons/minute detected after 250 minutes. Other
cells however are only beginning to respond after -150 min. Figure 4 shows the
progression of the brightness distribution as a group of cells responds to the
140 introduction of 1000 nM Al. The variability in the time scale of response (the
kinetic heterogeneity) can be summarized by the distribution of the onset time
ti/2, which we define as the time at which the luminescence of a particular cell is
halfway between its initial {t = 0) and final {t ~ 250 minutes) values. Figure 5B
shows that ti/z has a very broad and flat distribution at 200 nM, and this
145 distribution remains broad even at a saturating Al concentration of 1000 nM.

Hence we observe several types of heterogeneity in the response of V.fischeri to a
defined Al concentration. Cells in the same environment respond on widely
differing time scales when Al is introduced, and they also differ in the overall
150 amplitude of that response. Furthermore the individual trajectories of Figure 3
suggest that the luminescence of at least some cells occasionally fluctuates by
-20-40% on time scales of -30 minutes.

As shown in Text SI and Figure S3, our experimental configuration also allows
155 us to observe other kinetic and steady state phenomena in single-cell V.fischeri
luminescence, such as the "rich medium effect" [30-32]. However we focus
here on the heterogeneity of the QS response.

The luminescence of V.fischeri is activated through a quorum sensing {QS)
mechanism in which the cells remain dark until their population reaches the high
densities that signify colonization of the light organ of the symbiotic host. Here
we ask how tightly this QS system regulates the luminescence output of an
165 individual cell in response to a defined chemical signal {i.e. the 30C6HSL
autoinducer concentration). We find that an ensemble of cells produces a
distinctly heterogeneous response to the Al input, with significant cell-to-cell
variability in the overall level of emission and in the onset time for this response,
as well as indications of short term fluctuations in brightness.
In the absence of exogenous Al the light emission from the cells was below
measureable levels. However, after -150-250 minutes in exogenous Al the
individual cells were significantly brighter on average, as in a bulk culture. The
addition of Al not only increases the average brightness, but also increases the
175 (absolute) differences in the brightness of individual cells; hence the individual
brightness levels eventually span an order of magnitude, as shown in Figure 5A.
Similarly the luminescence onset time tm shows a broad distribution at both 200
nM and 1000 nM Al (where the response of the bulk population in Figure IB is
seen to saturate). As the distributions for both the individual cell intensities and
180 the onset times in Figure 5B are not at all clustered about the mean values they
are clearly not Gaussian (normal) distributions.

Nevertheless these single-cell data are still consistent with the behavior of a bulk
culture, as can be seen by comparing the Al response curves of single cells and a
185 bulk culture under the same growth conditions. Figure 6B shows that a
nonlinear least squares fit of a cooperative binding model to the single-cell data
gives an equilibrium constant Kei, ~ 120 20 nM and a Hill coefficient n ~ 2.7 0.8.
By comparison, the average luminescence of a bulk culture of the same strain
(Figure IB) gives Kei, ~ 200 10 nM and n ~ 2.6 0.4. The smooth AJ-induced
190 luminescence response of the bulk population is a result of averaging over large
numbers of cells; it conceals a very heterogeneous character in the response of
individual cells in that population.

We find it remarkable that such large variations in emission persist in a
195 homogeneous Al environment, even several hours after introduction of the
exogenous signal. Even though we anticipate that stochasticity will generate
cell-to-cell variability, the coefficient of variation {cv = standard deviation / mean

1) in Figure 6A appears much greater than is expected from stochastic
simulations of the LuxI/LuxR system. For example. Cox et al. estimated the
200 kinetic parameters for a chemical model of the LuxI/LuxR network [15]. Their
stochastic simulations predicted relatively modest variability in the activation of
luxl as a function of Al concentration. Although the Luxl concentration was
variable at low (< 50 nM) Al concentrations, the simulations predicted minimal
fluctuation, with a standard deviation less than -10% of the Luxl concentration
205 once the Al concentration reached the induction threshold. By contrast we find a
large variation in light output persisting across the Al induction curve. The cv of
the luminescence is near unity even at 1000 nM Al. This variability is
presumably not attributable to heterogeneity in intracellular Al concentrations,
as the Al diffuses rapidly across the cell membrane [32] and the exogenous Al
210 level is well-controlled by the flow of medium.

Our emission versus time trajectories also show some evidence for short-term
fluctuations in the single-cell luminescence. The time series data of Figure 3
suggest that the light output from some cells occasionally fluctuates by -20-40%.
215 Furthermore, while the brightness of each cell is reasonably stable on short time
scales, the brightness of one cell is poorly correlated with its brightness 30-60
minutes later (Text SI and Figure S4). An early study of the time dependence of
V. fischeri luminescence found no significant oscillation or pulsing in the
luminescence output at frequencies 0.01 10 Hz, although it did not investigate
220 the low frequency behavior (-10"^ Hz) studied here [24]. Whether the noise in lux
gene expression does in fact have a bursting or intermittent character under
stable environmental conditions is an intriguing question that requires further
study. However the short time scale of these fluctuations suggests that they
originate in intrinsic {i.e. purely biochemical stochastic) noise [33]. By contrast
225 the slower intercellular variability in the onset times for Al response and in the
overall luminescence output is more suggestive of extrinsic noise originating in
the variable concentrations of cellular components such as ribosomes,
polymerases, or in different stages in the growth cycle, etc. [34].

230 A recent study of the QS bioluminescent emission of individual V. harveyi also
found very substantial cell-to-cell variability [19]. Anetzberger et al. allowed V.
harveyi cells to accumulate their own autoinducer for intervals up to 8 hours and
reach quorum conditions. This produced an approximately bimodal response,
with many cells luminescing brightly while roughly 25% of live cells remained
235 relatively dark, or roughly one-tenth as bright as the more luminescent cells.
Although the LuxI/LuxR pathway probed here has a different structure from the
lux regulatory system of V. harveyi {i.e. LuxI/LuxR does not directly include the

type of phosphorelay switch and sRNA regulation found in V.harveyi), these
findings are similar to ours: after several hours in Al, roughly 25% of V. fischeri
240 cells were emitting luminescence at or below our detection limits (Figure 4B).
Our results show that this heterogeneity occurs across a range of Al
concentrations and also extends to the kinetics of the onset of luminescence.

However another recent single-cell study of the V. harveyi QS pathway found a
245 more homogeneous response to autoinducer [20]. Long et al. constructed a qrr4:-
gfp transcriptional fusion that allowed them to use GFP fluorescence rather
than the native luminescence to monitor the effect of two autoinducer signals
on the activation of the quorum-regulatory RNAs that are controlled by the
phosphorylation of LuxO. LuxO phosphorylation is in turn regulated by the
250 three autoinducer receptors in V.harveyi. Long et al. found much less variance in
the response of different cells at the same autoinducer concentrations than
Anetzberger et al. observed in the bioluminescence response, and much less than
we report here in V.fischeri luminescence. For the two different Al receptor
mutants (each responsive to a single autoinducer) that they studied, they
255 observed a coefficient of variation cv 0.2-0.4 in the gfp expression, significantly
smaller than the cv ~1 that we observe here in V.fischeri luminescence.

To explain the observation of heterogeneity in the luminescence (but not in the
gfp reporter strains) of V.harveyi, TVnetzberger et al. suggested a possible role for
260 positive feedback in the V.harveyi master regulator LuxR (not homologous to
V.fischeri LuxR), which is regulated by the sRNAs and controls expression of the
lux genes for luminescence. They proposed that the absence of autoinducer
synthases in the GFP reporter strains eliminated possible feedback loops
involving Al synthesis and detection, leading to a more homogeneous behavior
265 in those strains. The fact that our system defines the Al concentration
exogenously also eliminating Al feedback yet still exhibits heterogeneity
argues against this interpretation. However a role for feedback in the observed
noise is nevertheless plausible in LuxI/LuxR. Williams et al. recently studied the
dynamics of Al sensing by an E.coli model strain luxOl, in which LuxR is
270 activated by 30C6HSL to control the expression of gfp while the autoinducer
synthase Luxl is absent [14]. Cell cytometry studies found a bimodal response of
gfp expression to the Al signal level, with the more responsive cells exhibiting a
roughly log-normal distribution in GFP fluorescence. They argue that the
external Al concentrations feed into an autoregulatory feedback loop for LuxR
275 expression, and that this generates hysteresis in the LuxI/LuxR system's response
to Al. That is, its activation at any particular Al concentration depends in part on
its prior history and initial LuxR levels. This LuxR mechanism would help to

explain some of the cell-to-cell variability that is observed in the luminescence
onset time in Figure 5B, as natural stochastic variations in initial LuxR levels
280 would be amplified by feedback to give large changes in activation of the
luminescence genes.

Alternatively it is possible that the heterogeneity in light output results from
some differences in the energy resources of different cells, with some cells in
285 bright (energy-intensive) states and others in dark (recovering) states. However
our data suggest that the overall luminosity state (brighter or darker) of a cell
tends to persist over relatively long periods of hours, comparable to the doubling
time. Cycles of energy depletion and recovery would presumably play out over
shorter time scales. We also found that the output variability was not due to a
290 shortage of the C14 long chain substrate needed for the luciferase reaction (see
Text SI). Furthermore, heterogeneity was not exclusive to a luminescence
reporter of the LuxI/LuxR system: under full induction of LuxI/LuxR, the
expression of a gfp reporter by V.fischeri mutant JBIO showed heterogeneity {cv =
0.8) similar to that of the bioluminescence (Figure 7). These points suggest that
295 cell-to-cell variability in luminescence response is not primarily due to a
deficiency of the luminescence substrate or energy resources.

Our findings raise some interesting questions about the performance of V.fischeri
QS at the single cell level. For example, the broad heterogeneity in the light
300 output from the cells which always remained short of the estimated maximum
output of -1000 photons/s/cell [7] raises the question of whether the observed
heterogeneity is still present in cultures emitting at maximum brightness {e.g.
within the symbiotic host). It would also be interesting to determine whether the
two other signal inputs in V.fischeri, i.e. the C8HSL and Al-2 autoinducers, drive
305 a similarly noisy response or whether they improve the noise performance of the
overall system. Mehta et al. recently argued that the processing of information
by the QS system of V. harveyi is limited primarily by interference between the
three input signal channels of the QS pathway, and secondarily by noise
originating within each pathway [35]. Because noise in any one signal input
310 channel ultimately feeds forward into the regulated output, a well-defined input
concentration for one of the three autoinducer species will not ensure a
predictable output. In the present V. fischeri study we have defined the 30C6HSL
level externally and also set the other two autoinducer concentrations (C8HSL
and Al-2) virtually to zero by advection; hence it appears unlikely that these
315 additional receptors contribute significant noise to the luminescence output.

The heterogeneity observed here may also argue against an interpretation of the
LuxI/LuxR system or at least its regulation of the bioluminescence genes as
allowing an individual cell to acquire much useful information about its local
320 microenvironment [4, 5, 36]. The individual luminescence response seems to
contain little information about {i.e. it is a poor indicator of) the local Al level, just
as the Al concentration is a weak predictor of the luminescence response. If a
group of cells in a well-defined and homogeneous environment exhibit widely
divergent responses, one cannot consider the QS system to be a reliable sensor of
325 local diffusion constants, for example. In a more heterogeneous natural
microenvironment one expects that the cell-to-cell variability in response would
only increase.

There are scenarios in which phenotypic variations arising from noisy gene
330 expression can provide a tangible benefit to the cell [37, 38]. Therefore it is
intriguing to consider whether noise in V.fischeri luminescence benefits the
bacterium or influences its symbiosis with a host animal. In the symbiotic
relationship V.fischeri is subject to a strong selective pressure to maintain bright
luminescence. For example, the squid E.scolopes does not tolerate colonization by
335 dark mutants of V.fischeri [39, 40]. However, although the host can select a strain
for its average luminescence output, the squid presumably cannot detect
temporal or other types of heterogeneity at the single-cell level. It may detect the
mean but not the variance of the cell brightness. Therefore the individual cell
is not likely to endure host pressure to minimize its brightness fluctuations.
340 Thus one possible interpretation of our results is that the signal response is
poorly coordinated across the population because the host cannot apply
feedback to enforce tight coordination.

Of course, this interpretation only raises the question of whether the
345 uncoordinated response brings any benefit to the bacteria. It would be
interesting to determine whether cells that emit a weak luminescence are
directing more energy into other QS-regulated behaviors, as if to divide
colonization tasks across the population. Alternatively, since bright emission is
energy intensive, one may speculate that a form of QS cheating occurs, with the
350 less luminous cells enjoying a growth advantage. In a fully induced cell the
luminescence may require more than -lO'' ATP molecules per second and
account for up to -20% of the oxygen consumption [7]. Such cheating does
appear to provide a benefit to individual bacteria, although it is expected to be
less pervasive in clonal populations where kin selection favors cooperation [41].
355 (Cultures grown from a single colony of MJll were as heterogeneous in light
output as cultures grown from multiple colonies, as described in Text SI.)

Finally, a variable luminescence output could be an optimal strategy in
fluctuating environments or when some of the autoinducer signals are weak or
absent, so that the cell's obligation to luminesce is uncertain. Noisy output
360 would be less advantageous in the rich, supportive environment of the host light

In summary we have observed that the luminescence response of individual,
wild-type V.fischeri cells is very imprecisely regulated by the local quorum signal
365 level. As QS regulation plays an important role not just in the bioluminescence
of V. fischeri but also in colonization of the symbiotic host [9] it will be interesting
to conduct mutational studies to investigate whether the noisy behavior
observed in this particular output also extends to other targets of QS regulation
in this organism, and how this influences the organism's ability to colonize the
370 heterogeneous microscopic environment within the host light organ.

Materials and Methods:

Vibrio fischeri strain MJll (NCBI Taxonomy ID: 388396), a strain derived from the
375 host fish Monocentris japonicus [42], was provided by Prof. M. Mandel and Prof.
E. Ruby. Cells were prepared in exponential phase at 24 C in defined artificial
seawater medium [43] containing glycerol as carbon source. Approximately 15
|il of culture in exponential phase was deposited at the center of the lower
window of a perfusion chamber. This chamber consisted of a cylinder (volume
380 approximately 1.5 cm^) constructed from two parallel, circular cover slips (25 mm
diameter) spaced 5 mm apart. The lower window was coated with poly-L-lysine
to promote adherence of the cells. The chamber was then closed and the cells
were allowed to settle and adhere to the window. After -15 minutes the
chamber was then washed with approximately two chamber volumes of defined
385 medium from a programmable syringe pump. This wash diluted away any
autoinducer that was present in the starting culture and removed any non-
adhering cells. The chamber was then placed on the stage of an inverted
microscope and the pump flow rate was reduced to 0.2 ml/hr in order not to
disturb the adhered cells during observation. The cells in the chamber were
390 primarily located within a small area (few mm^) of the window, directly above
the microscope objective, which was an infinite-conjugate lOOx plan oil
immersion objective, NA 1.25. The blue/green (near 490 nm) bioluminescence
from the cells on the lower window was collected by the objective and focused
onto an intensified CCD camera (512x512 pixel, I-MAX-512-T operating at
395 -35 C, Princeton Instruments, Princeton NJ) via an achromatic doublet lens, to
give a final image scale of 0.278 |im per pixel.

The concentration of 30C6HSL autoinducer was selected by adding exogenous
autoinducer {Al, N-(3-Oxohexanoyl)-L-homoserine lactone, CAS 143537-62-6, No.
400 K3007 from Sigma Aldrich, St. Louis) to the medium flowing in the chamber.
The continuous flow of medium removed unattached (freely swimming) cells
from the chamber and maintained the Al concentration at the selected level. Al
released by the few cells adhered on the glass was efficiently removed by
diffusion into the passing flow. This was verified in two ways. First, numerical
405 integration of the diffusion/advection equation for our experimental
configuration gives an Al accumulation of less than 50 pM at the window (for Al
synthesis at 10"^^ g/s/cell and diffusion at 100 jimVs [44]). This concentration is
insufficient to induce detectable luminescence. Second, when cells were
perfused with medium that contained no added autoinducer, we observed that
410 any luminescent emission from the immobilized cells soon diminished to
undetectable levels.

The doubling time for the growth of the cells in the chamber was approximately
2-3 hr, operating at 24 C. This growth rate set a practical limit of roughly 4 hrs
415 to our observations of individual cell luminescence in the perfusion chamber.
Once the cells on the window had divided more than once or twice, the cells
appeared as clusters and it became difficult to distinguish the luminescence of
neighboring cells in the camera images. We studied the luminescence of wild
type strains only. Preliminary studies of V.fischeri strain ATCC 7744 gave results
420 similar to those presented here for strain MJll. A fluorescence study of gfp-
reporter strain JBIO is described below.

In most of our studies, the programmable syringe pump supplied a flow of
defined medium containing 0-1000 nM added autoinducer to the chamber.
425 During the "rich medium" study (see Text SI) the syringe delivered commercial
photobacterium medium (No. 786230, Carolina Biological, Burlington NC) mixed
with defined medium and Al as indicated. For the tetradecanoic acid study (see
Text SI), we prepared a 1 mM stock solution of tetradecanoic acid (myristic acid,
M3128 from Sigma Aldrich) in ethanol and diluted this lOOOx into the defined
430 medium, to give a final concentration of 1 |lM.

After placing the perfusion chamber on the microscope stage and starting the
flow of medium + Al, we used dark field images {i.e. externally illuminated
images with brief exposure times) to locate and focus on individual cells. We
435 then disconnected the illumination source and collected a bioluminescence image
{i.e. collecting only bioluminescent emission) with an exposure time of (typically)

ten minutes, and then collected another dark field image for comparison. Figure
2 and Figure SI show sample images. We repeated this process over a period of
-4 hrs for each group of cells (at a fixed Al concentration), collecting alternately
440 both dark field and bioluminescence images at regular intervals. Comparisons
of successive dark-field images provided a running check of the physical and
optical stability of the cells and the scene being imaged.

To quantify the emission levels of individual cells in the bioluminescence images,
445 we first used the dark field images to obtain the pixel coordinates of individual
cells that had remained immobile during the experiment. We then defined a
small rectangular region surrounding each cell. We binned (2x2, to improve
SNR) the pixels of the corresponding region within the dark-subtracted
luminescence image, generated the brightness histogram of the pixels in that
450 region, and fit the lower portion (only) of that histogram to a Gaussian
distribution. This distribution accurately models the background intensity
distribution in cell-free regions of the image. We then subtracted the fit Gaussian
from the actual histogram and summed the residual. This provided a
satisfactorily robust count of the luminescence emission of a single cell, typically
455 10-100 photons/minute/cell. We confirmed that the luminescence emission count
from a single cell was insensitive to the precise size of the rectangular image
region used to estimate that count. Thermally generated background {e.g. dark
noise) in the CCD image contributes some uncertainty to this emission count. By
applying the above data analysis to several image regions that contained no cells,
460 we estimated the magnitude of this uncertainty as -20 photons/minute peak-to-
peak per cell per image frame. This defines a baseline noise level, prior to
Gaussian filtering of the emission level versus time record ("trajectory") of a cell.
The image intensifier itself also contributes some noise, which is best
characterized by imaging a stable light source, as discussed below. Camera
465 readout noise and photon shot noise were smaller than either of the above noise

We typically detected -10-100 photons/minute/cell from V.fischeri strain MJll in
our flow chamber, even in the presence of an Al concentration (1000 nM) that
470 would saturate the output of a bulk culture. Therefore, our single-cell
luminescence measurements involved signal levels that were drastically lower
than are commonly obtained in gene regulation studies using fluorescent
proteins like GFP. For this reason it was important to verify that the detected
signals and their variations were not due to experimental or analysis artifacts.
475 Text SI provides further detail on measures that we took to ensure the stability of
the optical configuration, with minimal drift in the focus and minimal movement

in the cells adhered to the glass. These included collecting and comparing a
series of dark-field images {i.e. one externally-illuminated dark-field image
between each pair of luminescence images) to check that cells under observation
480 remained in focus and had not physically moved.

Text SI also describes control experiments to verify the stability and sensitivity of
our detection system. That is, we verified that the observed variations in the
light output from individual V.fischeri cells were representative of cellular
485 emission, and were not generated within the image intensifier or due to
uncertainty in our detection or analysis. A suitable control must be a micron-
sized light source that is comparable in size to the V.fischeri cells, feebly
luminescent (no brighter than the weak luminescence of a single V.fischeri cell),
and absolutely stable in its output. For this purpose we used micron-sized green
490 fluorescent latex spheres (FluoSpheres, Invitrogen Inc.) dispersed at low density
onto the lower window of the perfusion chamber and illuminated with a heavily
attenuated blue LED excitation source. Under exceedingly faint excitation the
fluorescence from these spheres in a ten minute camera exposure was
comparable in magnitude to the emission detected from individual V.fischeri {i.e.
495 -100 photons/minute/particle) and it remained stable for extended periods. We
imaged these spheres with exactly the same instrumentation parameters (camera
gain and temperature, exposure time, magnification, etc.) as used for the
V.fischeri cells. Performing the same image analysis as used for the live cell
images, we obtained a highly stable and consistent photon count from the
500 spheres. Figure 3 shows that the emission detected from the control spheres
remained stable through more than four hours of observation, without any
manual adjustment of microscope focus. After Gaussian filtering (width o = 10
minutes) of all emission versus time trajectories, the noise level (standard
deviation) for the emission from the individual particles was 10-12
505 photons/minute. Figure S2 shows that the emission from different spheres in the
same image was closely similar as expected (standard deviation / mean 0.12).
These results show that the microscopy system and the data analysis were
sufficiently sensitive and stable for resolving heterogeneity in the luminescent
emission from different V.fischeri cells.
We also used fluorescence microscopy to measure GFP levels in individual cells
of V.fischeri strain JBIO, which was provided by Prof. E. Stabb. In the JBIO
mutant a chromosomal gfp reporter is placed under the control of the LuxI/LuxR
system by insertion between luxl and luxC, i.e. luxl-gfp-luxCDABEG, so as to
515 express GFP when the LuxI/LuxR system is activated by 30C6HSL [45]. Cells
were grown overnight in the same defined medium used for the luminescence

experiments and then transferred to fresh medium containing 1000 nM
exogenous Al. After incubating the cells with shaking for -2 hrs we dispersed
the cells on a coverslip and measured the fluorescence of 127 individual cells,
520 using an inverted microscope with a 60x oil immersion objective and a cooled
CCD camera (Micromax, Princeton Instruments).


525 We thank Prof. Mark J. Mandel, Prof. Edward G. Ruby, and Prof. Eric V. Stabb
for providing V.fischeri strains MJll (M.J.M., E.G.R.) and JBIO (E.V.S.) used in this
study, and for their helpful advice and suggestions. We also thank Leslie Pelakh
for assistance with data collection. Funding support was provided by the
National Science Foundation under award MCB 0347124. Publication of this
530 article was funded in part by the University of Florida Open-Access publishing

Figure Legends:
Figure 1. Schematic of LuxI/LuxR regulation of V.fischeri bioluminescence,
540 and bulk response. {A) Luxl synthesizes the autoinducer Al (N-3-oxohexanoyl-L-
homoserine lactone) which binds to LuxR, the transcriptional activator for the
luminescence genes luxCDABE [6]; (B) Luminescence response of a bulk culture
of V.fischeri strain MJll growing in defined medium at room temperature. The
points show the response of a (bulk) population of exponential phase cells in a
545 48-well plate, following addition of exogenous autoinducer {Al) at time t = 0.
After 70 minutes an AJ-dependent response is developing. After 130 minutes the
response has reached a steady state. Data for t > 130 minutes are fit to a
cooperative binding model (black dotted curve) to give an equilibrium constant
Kei, = 200 10 nM and Hill coefficient n = 2.6 0.4. Luminescence data are
550 normalized to the optical density at 600 nm to give the luminescence per cell, in
arbitrary units.

Figure 2. Individual V.fischeri imaged in dark field and bioluminescence. {A)
Dark field (externally illuminated) and (B) bioluminescence (light emission)
555 images of V.fischeri cells adhered to the glass window of the perfusion chamber at
24 C in the presence of 500 nM exogenous Al. The cells appear as rods (-3-5 |im
long) in the dark field image. The bioluminescence image shows in false color
the luminescent emission detected in a 16 minute exposure. Images were
collected in an inverted microscopy configuration with an intensified CCD
560 camera and a lOOx oil immersion objective.

Figure 3. Luminescence of individual V.fischeri cells following addition of
autoinducer, and detection stability test. At each autoinducer concentration,
roughly 25-40 MJll cells were imaged repeatedly over a period of -4 hrs
565 following introduction (at f = 0) of exogenous autoinducer Al at the indicated
concentration. The light emission from each cell was quantified through analysis
of a series of 10-minute camera exposures (see Materials & Methods). The state of
induction of the initial cell culture determines the luminescence of the cells att =
0. However, once adhered in the flow chamber and exposed to the flow of
570 medium (containing exogenous Al), the cells respond by adjusting their
luminescent output. This leads to a transient increase or decrease in the emission
over the next -1-2 hrs. After -3 hrs the cells have adapted to the applied Al level.
The control shows an experimental verification of the stability and sensitivity of
microscopy and data analysis. For this measurement, green fluorescent latex

575 spheres were illuminated with a severely attenuated blue light source and then
imaged with the same camera settings, magnification, 10-minute exposure time,
and data analysis, as used for the V.fischeri measurements. Image focus and
excitation intensity were not adjusted during the 4 hr measurement. Twelve
representative trajectories are shown. See Materials and Methods and Text SI. The
580 time-dependence of all emission versus time "trajectories" in this figure has been
smoothed by a Gaussian filter with width o = 10 minutes.

Figure 4. Spreading of the luminescence histogram over time. {A) Cell
brightness histograms for MJll cells at the indicated times, following
585 introduction of 1000 nM Al at t = 0. (B) Median (red curve) cell brightness and
the 25% and 75% percentiles of brightness (blue curves). The distribution of
intensities broadens as the cells response to the exogenous Al signal. A
substantial fraction of the cells emit near the detection threshold (-10-20
photons/minute/cell) even at f = 4 hrs.
Figure 5. Histograms of luminescence levels and onset times at high
autoinducer concentration. (A) Distribution of luminescence levels detected for
individual V.fischeri cells, at time t = 240 minutes after autoinducer {Al,
30C6HSL) was introduced at concentrations indicated. Cells emitting -10-20
595 photons/minute are at the measurement uncertainty, i.e. are consistent with no
emission. (B) Distribution of luminescence onset times tm in the presence of 200
nM and 1000 nM Al. The onset time tm is the time at which the luminescence
output l{t) of a particular cell is halfway between its initial value l{t = 0) and its
final value l{t ~ 250 minutes), when Al was introduced at f = 0.
Figure 6. Variation and mean of luminescence levels versus autoinducer
concentration. {A) Coefficient of variation {cv = standard deviation / mean) in the
luminescence of different cells. Variation is calculated from emission levels
recorded t > 100 minutes after introduction of exogenous Al; (B) Luminescence
605 emission detected from 188 individual cells (blue circles) after 150-250 minutes
exposure to Al. Data for each Al concentration represents a different group of
cells. Solid curve (blue) is a fit to a cooperative binding model, giving Kei, ~ 120
20 nM and Hill coefficient n ~ 2.7 0.8. For comparison with the expected
average behavior, the dashed curve (red) shows the Al response that is obtained
610 from a bulk population after 150-250 minutes in autoinducer (Figure 1).

Figure 7. Heterogeneity of native luminescence versus fluorescence reporter
for V.fischeri quorum system. (A) Histogram of bioluminescence emission
levels from 47 individual V.fischeri cells of wildtype strain MJll, following

615 induction by 1000 nM Al. The luminescence levels are normalized to the median
value. (B) Histogram of fluorescence levels for 127 individual V.fischeri cells of
mutant JBIO, following induction by 1000 nM Al. The JBIO mutant contains a
chromosomal gfp insertion between luxl and luxC in the LuxI/LuxR system.
Fluorescence values are normalized to the median value. Both luminescence and
620 fluorescence reporters for the QS system show a broadly heterogeneous response
at full induction, although the fluorescence shows slightly less variability {cv ~
0.8) than the luminescence {cv ~ 1.0).

625 Figure SI: Sequential dark field and luminescence images for one V.fischeri
cell. {A) Dark field and (B) bioluminescence images of an individual cell adhered
to the window of the perfusion chamber, and (C) luminescence levels extracted
from these images. (The luminescence trajectory has not been Gaussian filtered.)
Images were collected at the numbered time points indicated in (C).
Figure S2: Variability in signal levels for V.fischeri cells and for reference
particles. Histograms comparing the luminescent emission from individual
V.fischeri cells {A) to the fluorescent emission under weak excitation of a control
sample of individual micron-sized latex spheres (B). Each histogram shows the
635 number of individual emission measurements falling into the indicated
brightness bin, over a -30 minute period comprising three 10-minute camera
exposures. {A) and (B) have the same horizontal scales: All images for both cells
and fluorospheres were collected in ten minute exposures using identical camera
and microscope settings and image analysis. (For the fluorospheres, we used a
640 highly-attenuated blue LED as excitation source and inserted a Schott longpass
filter GG485 into the detection path.) The coefficient of variation for the
fluorospheres is 0.12, while the coefficient of variation for the V.fischeri cells is 1.3
(200 nM Al) and 1.0 (1000 nM Al).

645 Figure S3: Inhibition of V.fischeri bioluminescence by complete ("rich")
medium. Light emission from individual cells in the perfusion chamber was
tracked over time as the flowing medium was switched from an initial (100%
defined medium) to a final (70% defined medium, 30% complete medium)
composition. Al concentration remained 1000 nM at all times. Image times
650 represent the starting time of a 16-minute bioluminescence exposure. The
histograms, showing the fraction of observed cells emitting at the indicated level,
collapse rapidly as complete medium is introduced.

Figure S4: Temporal autocorrelation of individual cell luminescence. The
655 emission level l{t) of a cell at time t is compared to its emission at a later time
l{t+x). Data represent individual cell emission levels measured at least 100
minutes after introduction of 1000 nM Al: {A) (D) For small values of x, the
data are close to the (best fit) line, indicating that a cell's intensity at time f is a
reasonably good predictor of its intensity at time t + x. However as x approaches
660 40-60 minutes, the scatter around the average line increases, indicating that the
brightness of the cell at later times (relative to the average or best fit trend) is
poorly predicted by its earlier brightness or by the average behavior of the other
cells. The vertical distance d of each point from the trend line becomes larger at
large x. Panel (£) shows Od, (the standard deviation of d) as a function of x. At
665 high Al concentrations the standard deviation continues to grow for many
minutes, indicating that the brightness of the cells continues to diverge both from
its initial value and from the average growth trend. The Od of the control
(fluorescence spheres) is essentially flat as expected, except for a dip near x = 10
minutes (due to Gaussian filtering of the trajectories).


1. Waters CM, Bassler BL. (2005) Quorum sensing: Cell-to-cell communication
in bacteria. Annu Rev Cell Dev Biol 21: 319-346.

675 2. Boyer M, Wisniewski-Dye F. (2009) Cell-cell signalling in bacteria: Not
simply a matter of quorum. FEMS Microbiol Ecol 70(1): 1-19.

3. Dunn AK, Stabb EV. (2007) Beyond quorum sensing: The complexities of
prokaryotic parliamentary procedures. TVnal Bioanal Chem 387(2): 391-398.

4. Hense BA, Kuttler C, Mueller J, Rothballer M, Hartmann A, et al. (2007)
680 Opinion does efficiency sensing unify diffusion and quorum sensing?
Nature Reviews Microbiology 5(3): 230-239.

5. Redfield RJ. (2002) Is quorum sensing a side effect of diffusion sensing?
Trends Microbiol 10(8): 365-370.

6. Dunlap PV. (1999) Quorum regulation of luminescence in vibrio fischeri. J
685 Molec Microbiol Biotechnol 1(1): 5-12.

7. Dunlap PV, Greenberg EP. (1991) Role of intercellular chemical
communication in the vibrio fischeri-monocentrid fish symbiosis. In:
Dworkin M, editor. Microbial Cell-Cell Interactions. Washington, DC:
American Society for Microbiology, pp. 219-253.

690 8. Visick KL. (2005) Layers of signaling in a bacterium-host association. J
Bacteriol 187(11): 3603-3606.

9. Lupp C, Ruby EG. (2005) Vibrio fischeri uses two quorum-sensing systems
for the regulation of early and late colonization factors. J Bacteriol 187(11):

695 10. James S, Nilsson P, James G, Kjelleberg S, Fagerstrom T. (2000)
Luminescence control in the marine bacterium vibrio fischeri: An analysis of
the dynamics of lux regulation. J Mol Biol 296(4): 1127-1137.

11. Kuttler C, Hense BA. (2008) Interplay of two quorum sensing regulation
systems of vibrio fischeri. J Theor Biol 251:167-180.

700 12. Mueller J, Kuttler C, Hense BA, Rothballer M, Hartmann A. (2006) Cell-cell
communication by quorum sensing and dimension-reduction. J Math Biol
53(4): 672-702.

13. Goryachev AB, Toh DJ, Lee T. (2006) Systems analysis of a quorum sensing
network: Design constraints imposed by the functional requirements,
705 network topology and kinetic constants. BioSystems 83(2-3): 178-187.

14. Williams JW, Cui X, Levchenko A, Stevens AM. (2008) Robust and sensitive
control of a quorum-sensing circuit by two interlocked feedback loops.
Molec Syst Biol 4: 234.

15. Cox CD, Peterson GD, Allen MS, Lancaster JM, McCoUum JM, et al. (2003)
710 TVnalysis of noise in quorum sensing. OMICS: A Journal of Integrative
Biology 7(3): 317-334.

16. Zhou TS, Chen LN, Aihara K. (2005) Molecular communication through
stochastic synchronization induced by extracellular fluctuations. Phys Rev
Lett 95(17): 178103.

715 17. Tanouchi Y, Tu D, Kim J, You L. (2008) Noise reduction by diffusional
dissipation in a minimal quorum sensing motif. PLoS Comput Biol 4(8):

18. Kaern M, Elston TC, Blake WJ, Collins JJ. (2005) Stochasticity in gene
expression: From theories to phenotypes. Nat Rev Gen 6(6): 451-464.

720 19. TVnetzberger C, Pirch T, Jung K. (2009) Heterogeneity in quorum sensing-
regulated bioluminescence of vibrio harveyi. Mol Microbiol 73(2): 267-277.

20. Long T, Tu KC, Wang Y, Mehta P, Ong NP, et al. (2009) Quantifying the
integration of quorum-sensing signals with single-cell resolution. PLoS Biol
7(3): 640-649.

725 21. Teng S, Wang Y, Tu KC, Long T, Mehta P, et al. (2010) Measurement of the
copy number of the master quorum-sensing regulator of a bacterial cell.
Biophys J 98(9): 2024-2031.

22. Phiefer CB, Palmer RJ, White DC. (1999) Comparison of relative photon
flux from single cells of the bioluminescent marine bacteria vibrio fischeri
730 and vibrio harveyi using photon-counting microscopy. Luminescence 14(3):

23. Greer LF, Szalay AA. (2002) Imaging of light emission from the expression
of luciferases in living cells and organisms: A review. Luminescence 17(1):

735 24. Haas E. (1980) Bioluminescence from single bacterial-cells exhibits no
oscillation. Biophys J 31(3): 301-312.

25. Mihalcescu I, Hsing W, Leibler S. (2004) Resilient circadian oscillator
revealed in individual cyanobacteria. Nature 430(6995): 81-85.

26. Bose J, Rosenberg C, Stabb E. (2008) Effects of luxCDABEG induction in
740 vibrio fischeri: Enhancement of symbiotic colonization and conditional
attenuation of growth in culture. Arch Microbiol 190(2): 169-183.

27. Schwille P, Haupts U, Maiti S, Webb WW. (1999) Molecular dynamics in
living cells observed by fluorescence correlation spectroscopy with one- and
two-photon excitation. Biophys J 77(4): 2251-2265.

745 28. Shaner NC, Steinbach PA, Tsien RY. (2005) A guide to choosing fluorescent
proteins. Nat Meth 2(12): 905-909.

29. Sternberg C, Eberl L, Poulsen LK, Molin S. (1997) Detection of
bioluminescence from individual bacterial cells: A comparison of two
different low-light imaging systems. J Biolumin Chemilumin 12(1): 7-13.

750 30. Eberhard A. (1972) Inhibition and activation of bacterial luciferase
synthesis. J Bacteriol 109(3): 1101-1105.

31. Yarwood JM, Volper EM, Greenberg EP. (2005) Delays in pseudomonas
aeruginosa quorum-controlled gene expression are conditional. Proc Natl
Acad Sci U S A 102(25): 9008-9013.

755 32. Kaplan HB, Greenberg EP. (1985) Diffusion of autoinducer is involved in
regulation of the vibrio fischeri luminescence system. J Bacteriol 163(3): 1210-

33. Rosenfeld N, Young JW, Alon U, Swain PS, Elowitz MB. (2005) Gene
regulation at the single-cell level. Science 307(5717): 1962-1965.

760 34. Swain PS, Elowitz MB, Siggia ED. (2002) Intrinsic and extrinsic
contributions to stochasticity in gene expression. Proc Natl Acad Sci USA
99(20): 12795-12800.

35. Mehta P, Goyal S, Long T, Bassler BL, Wingreen NS. (2009) Information
processing and signal integration in bacterial quorum sensing. Molecular
765 Systems Biology 5: 325.

36. Hagen S, Son M, Weiss J, Young J. (2010) Bacterium in a box: Sensing of
quorum and environment by the LuxI/LuxR gene regulatory circuit. Journal
of Biological Physics 36(3): 317-327.

37. Balaban NQ, Merrin J, Chait R, Kowalik L, Leibler S. (2004) Bacterial
770 persistence as a phenotypic switch. Science 305(5690): 1622-1625.

38. Maamar H, Raj A, Dubnau D. (2007) Noise in gene expression determines
cell fate in bacillus subtilis. Science 317(5837): 526-529.

39. Tong D, Rozas NS, Oakley TH, Mitchell J, CoUey NJ, et al. (2009) Evidence
for light perception in a bioluminescent organ. Proceedings of the National
775 Academy of Sciences 106(24): 9836-9841.

40. Visick KL, Foster J, Doino J, McFall-Ngai M, Ruby EG. (2000) Vibrio fischeri
lux genes play an important role in colonization and development of the host
light organ. J Bacteriol 182(16): 4578-4586.

41. Diggle SP, Griffin AS, Campbell GS, West SA. (2007) Cooperation and
780 conflict in quorum-sensing bacterial populations. Nature 450(7168): 411-414.

42. Mandel MJ, WoUenberg MS, Stabb EV, Visick KL, Ruby EG. (2009) A single
regulatory gene is sufficient to alter bacterial host range. Nature 458(7235):

43. Ruby EG, Nealson KH. (1976) Symbiotic association of photobacterium
785 fischeri with the marine luminous fish monocentris japonica: A model of
symbiosis based on bacterial studies. Biol Bull 151(3): 574-586.

44. Parent ME, Snyder CE, Kopp ND, Velegol D. (2008) Localized quorum
sensing in vibrio fischeri. Colloids and Surfaces B-Biointerfaces 62(2): 180-

790 45. Bose JL, Kim U, Bartkowski W, Gunsalus RP, Overley AM, et al. (2007)
Bioluminescence in vibrio fischeri is controlled by the redox-responsive
regulator ArcA. Mol Microbiol 65(2): 538-553.

^>- o- r\l
o o
Cell exterior

Cell interior
^ 4


c 2
luxR]---\luxl\ CDABE>

* (LuxR)
1 1 1 1 IIIM O 70 minutes 130 160 190 ttMf 0" f "a "b
- n4 /^
j."## M* 1 II Jil_ . .
10 100 1000

Al concentration (nIVI)
Figure 1

Figure 2



^ 200








Al= 50 nM
added at t
= 0
100 nM
200 nM
1000 nM
0 50 100 150 200
time (minutes)
Figure 3

\ U.b 0.3 0.6 0.3 0.6 0.3 0.6 Lu f = 27min 1
J- 71 min 2
en o o J.. 100 min 3
5 0.3 0.6 J._ 131 min
0.3 0.6 ^ 159 min
0.3 186 min
0.6 IIILjbbl.
0.3 ^ 214 min
0 200 400
Detected emission (phot/min/cell)
B 140 ***
o 120 /75%
E CD ^ \ 1
o) -y 80 o .^ / ^
BE 60 1140 qS 20 J- m^ 0/ cfio/
*=*C.-- 25%
% 50 100 150 200
time (minutes)
Figure 4

A 0.2
n jllriirirhnn
nfTn nnnHi
Al= 1000 nM
rirnn nlTinrJin^nrTln,
12 5 10 20 50 100 200
Detected emission (pliotons/minute/cell)

? 0
Al = 200 nM
11 In I 11 Infill
Al =-\000nM
0 20 40 60 80 100 120140 160 180 200 220 240 260
t,,2 (minutes)
Figure 5

S 1 51

s 1

"S 0.5

; E

! c

> CL 1
1 -

:120 + 20nM I

) = 2.7 I 0.8 I

, = 200:::1DnM ,

1=2.6 : 0.4
400 600 800

Al (nIVI)
Figure 6

p mean nHnn. nnnnnnHn Hn n U
Luminescence (normalized to median)
w 0.2 a......]
o ni 1 +r
o t n D_ n 1nnn.i._._. .
0 12 3 4 5 6 7
Fluorescence (normalized to median)
Figure 7


Text SI

Heterogeneous response to a quorum-sensing signal in the luminescence of individual Vibrio fischeri

Pablo Delfino Perez and Stephen J. Hagen
The bioluminescence of an individual bacterium generates an exceedingly weak optical signal.
Measuring this signal quantitatively and over a period of several hours requires considerable care to
ensure the stability of the optical configuration i.e. to verify that there is minimal drift in the focus or
movement in the cells adhered to the glass and to verify that observed differences in light output from
individual cells are statistically significant and not artifacts of the detection or image analysis. This
supporting document provides further detail on these aspects. It also includes some additional
experimental results and analysis, including brief summaries of additional control experiments that we

1. Physical stability

We ensured the physical stability of the focus and the image scene (the cells in view) by collecting one
dark field image {i.e. externally illuminated) between each pair of bioluminescence images for the
duration of the observations i.e. at ten minute intervals over the 4 hr duration of the observations. These
dark field images provided a running check on the stability of individual cell position and focus. Owing
to the thermal and mechanical stability of the microscope stage and optics (which were enclosed in a
lightproof box on an optical table) small adjustments to the focus were only occasionally and not always
- needed after a few hours of observation.

Figure SI shows an excerpt from the image series collected for one cell. The dark field (gray scale)
images show that the orientation and position of the cell are stable and are easily monitored. The image
focus remains stable over time. However the emission level extracted from the luminescence (color
scaled) images varies over time.

Cells did occasionally detach from the glass during measurement and were washed away by the flow.
This was easily detected in the dark field images. Our data analysis (see Materials and Methods)
addressed only those cells that remained in place during the entire observation period. In some
experiments the microscope stage translated very slightly (few |am) during observations; this was frivially
corrected in the image analysis and did not affect emission counts.

2. Orientation of cells on glass

Although the cells remain stationary during measurement, we cannot ensure that all cells lie completely
flat on the glass coverslip. The orientation of the cell relative to the glass window cannot affect the
direction of its light emission. It can however affect the image focus and therefore one needs to consider
how this may affect the light collection efficiency. Cells that adhere to the glass window by standing
upright (partially or completely) will give a less focused image than cells lying flat on the glass, and this
will reduce the photon count per camera pixel. However we do not simply count the photons per camera
pixel; we find the total count (above detector background) in a larger image region that surrounds a cell.
{Sec Materials and Methods.) A partial separation of the cell body from the glass will slightly defocus
the cell image, but it will not greatly affect the photon count in the larger image region. This should tend
to reduce artifactual variations in cell brightness that owe to orientation of a cell relative to the glass
surface. Moreover, for our data analysis we selected cells that appeared clearly focused, immobile, and
well-demarcated through the entire series of dark field images. We did not attempt to track the emission
from cells whose dark field images were inconstant or poorly resolved. It is not surprising that the

majority of such stable, clearly-delineated cells were horizontal on the glass (see Figure 2), since the
horizontal cells make maximum contact with the surface and are therefore in best focus and should adhere
most sfrongly. The color scale of the bioluminescence image of Figure 2 shows that, even among cells
with similar size, shape, and orientation, the overall brightness varies noticeably. Therefore the specific
orientation of the cells conttibutes very little to the observed variability in the detected cell luminescence.

3. Noise levels

It is important to verify that the observed cell-to-cell variations in emission are inttinsic to the
luminescence and are not atfributable to uncertainty in the detection. We established that the observed
differences in the brightness of individual cells significantly exceed the uncertainty in detection.

The magnitude of the detection noise is apparent in the light levels measured from the fluorescent
particles that we used as a control sample. See Figure 3 and Figure S2. These particles generated a
signal level (~10^ photons/minute/particle) comparable to V.fischeri but with a much smaller noise level
of roughly 10-12 photons/minute/particle (rms) in 10 minute exposures collected over four hours. The
brightness of the different control particles also showed a modest coefficient of variation (cv = standard
deviation / mean) of 0.12. This is significantly narrower than the cv ~\ seen in the brightness of
individual V.fischeri.

The noise level can also be estimated from the magnitude of the parameter Oi{x) which is shown in
Figure S4. Oi{x) describes the variance in the difference between the intensity I{t) of a cell and its
intensity at a later time /(f-K). (See Autocorrelation of light output from individual cells, below.) In the
limit T ^ 0, Od extrapolates to roughly Od ~ 5-15 photons/minute/cell (Figure S4). This is consistent with
the noise estimate from the control spheres.

Therefore the estimated noise level in the Gaussian-filtered (width a = 10 minutes) emission vs time
"frajectories" is of order 10 photons/minute. This is dominated by noise in the image intensifier as well
as uncertainty in image background subtraction (see Materials and Methods). However it is much smaller
than the observed variations (often -100 photons/minute/cell) in the light output of different cells.
Furthermore those large cell-to-cell variations are sustained over periods of hours in some cases and are
qualitatively very dissimilar to the quiet signal obtained from the control spheres.

4. Autocorrelation of light output from individual cells

It is of interest to characterize the statistical properties of the luminescence behavior through its
autocorrelation. However, the trend and divergence in the emission levels make this difficult: At high Al
concenttation, the cells not only become brighter over time but the brightness of each cell diverges further
from the average brightness. Figure S4 characterizes the autocorrelation behavior. The figure compares
the brightness of each individual cell at time t to its brightness at a later time f + x, as determined from all
image frames collected at least 100 minutes after inttoduction of 1000 nM AL The dashed line is the best
fit line; it represents the average behavior of all cells under observation. The individual points are
clustered near the best fit line (quite obviously in the plot for x = 10 minutes). This is expected since the
overall brightness of one cell rarely changes by large amounts over short time intervals; bright cells tend
to remain bright and dark cells remain dark. However, for larger intervals T the points become more
widely scattered around the best fit line, indicating that as x grows larger the brightness of a cell at time t
becomes a progressively worse predictor of its brightness at f + x. Although the average distance d of the
points from the best fit line is zero, the standard deviation Od can be plotted as a function of x. The
bottom panel of the figure shows that Od saturates in -30-40 minutes at a low Al concenttation of 100
nM. (The cells remain fairly dark and therefore cannot diverge far from the average brightness.)
However, at higher Al concentrations Od is still increasing at X 60 minutes. That is, upon the addition of

200-1000 nM Al, the brightness of each cell not only increases on average, but also diverges away from
the population-averaged ttend. This is a signature of the broad disttibution of time scales that is present
in the luminescent response of individual cells. (As expected, the conttol spheres show a constant Od,
except at X < 20 minutes where the Gaussian filter smooths the noise.)

5. Substrate deficiency

We tested whether the variability in single-cell luminescence could be atttibuted to a shortage of the long
chain aldehyde substtate that is needed for the bacterial luciferase reaction. The luciferase reaction
involves the oxidation of tetradecanal (a CI4 aldehyde) to produce the fatty acid (tetradecanoic acid),
which is then recycled back to aldehyde [1,2]. In cells that are deficient in the production of aldehyde,
the addition of exogenous tetradecanal or tetradecanoic acid restores the CI4 supply to the reaction cycle
and stimulates a dramatic enhancement in luminescence output. We repeated our 1000 nM Al study in
the presence of I |iM tettadecanoic acid and found no difference in our results: over a four hour
observation period, the average intensity of the cells, the distribution in onset times t^, and the variability
in the light output were all similar to results observed in the absence of exogenous CI4. The variability in
luminescence output cannot be attributed simply to a deficiency of CI4 substtate for the luminescence

6. QS heterogeneity in a single colony

Although we did not generally start each exponential culture from a single colony of MJI1 we did verify
that the observed heterogeneity of bioluminescence emission was not due to a mixture of mutants. We
grew an exponential (liquid) culture from an isolated single colony (selected from a photobacterium agar
plate) and confirmed that in this culture both bright and dark cells were present at high Al concenttations
(1000 nM Al). Cells originating from a single colony luminesced with different brightness levels under
the same Al condition, even when nearby cells of equal size lying just a few microns apart were
compared. This finding indicates that the heterogeneity cannot be easily atttibuted to appearance of a
"dark mutant" in the strain. Anetzberger et al report a similar result for V.harveyi luminescence [3].

7. Rich medium effect

Our observations confirm that individual cells exposed to new rich growth medium rapidly lose
brightness, even with 1000 nM^/present. Luminescent cells ttansferred into fresh complete medium
(even with exogenous^ cease luminescence until they have grown for a period of time and
(presumably) consumed or inactivated an inhibitor that is present in new medium [4, 5]. This "rich
medium effect" was described in earlier literature on bulk cultures of V.fischeri although its cause has not
been identified. Similar inhibition of homoserine-lactone dependent gene expression has also been
observed in Pseudomonas aeruginosa QS [6] and has been attributed to a small molecule present in
complete medium. In Figure S3 luminescent MJll V.fischeri in defined medium and 1000 nM AZ^were
switched at f 80 minutes to a mixture of defined medium (70%) and new complete medium (30%).
Although the Al concenttation was 1000 nM in both environments, the addition of complete medium
causes the cell brightness histogram to collapse quickly to a lower average brightness. This response is
much more rapid than the slower (- hrs) timescale of response to added ^/. For the luminescence studies
described in the main text we avoided rich medium and used only defined medium.


I. Ulitzur S, Hastings JW. (1978) Myristic acid stimulation of bacterial bioluminescence in
aldehyde mutants. Proc Natl Acad Sci U S A 75(1): 266-269.

2. Ulitzur S, Hastings JW. (1979) Evidence for tetradecanal as the natural aldehyde in
bacterial bioluminescence. Proc Natl Acad Sci U S A 76(1): 265-267.

3. Anetzberger C, Pirch T, Jung K. (2009) Heterogeneity in quorum sensing-regulated
bioluminescence oi vibrio harveyi. Mol Microbiol 73(2): 261-211.

4. Kaplan HB, Greenberg EP. (1985) Diffusion of autoinducer is involved in regulation of the
vibrio fischeri luminescence system. J Bacteriol 163(3): 1210-1214.

5. Eberhard A. (1972) Inhibition and activation of bacterial luciferase synthesis. J Bacteriol
109(3): II0I-II05.

6. Yarwood JM, Volper EM, Greenberg EP. (2005) Delays in pseudomonas aeruginosa
quorum-controlled gene expression are conditional. Proc Natl Acad Sci U S A 102(25):


3 -I 5 8
12 3 4 5 6
C 150
O o
4 5
50 100 150 200
time (minutes)
Figure SI

0 100 200 300 400
] 200 nM Al
11000 nM
In n n
0 100 200 300 400
Detected photons/minute
Figure S2

-H___ f = 23min .jiai..i..l _
43 min
63 min ......il...HH .n 1
iJlL. 93 min
JJil.ii 113 min
jilL. 137 min
Ill 159 min
iili 183 min
iJl L .. 202 min
i.j| 1.... 243 min m
0 200 400 600
Detected emission
Figure S3

A 400
B 400
i 200
C 400
i 200

D 400

/\/=1000nM X = 10 min

X = 20 min

lie--------- X = 40 min

X = 60 min
0 100 200 300 400
l(t) (phot/min/cell)

= 50
1 40
I 30


.- 10
200 nM
100 nM
50 nM
0 20 40 60
X (minutes)
Figure S4

University of Florida Home Page
© 2004 - 2011 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Powered by SobekCM