Engineered bidirectional communication mediates
a consensus in a microbial biofilm consortium
Katie Brenner*, David K. Karig†, Ron Weiss†‡§, and Frances H. Arnold*¶
*Division of Chemistry and Chemical Engineering, California Institute of Technology, MC 210-41, Pasadena, CA 91125; and Departments of†Electrical
Engineering and‡Molecular Biology, Princeton University, Princeton, NJ 08544
Edited by Charles R. Cantor, Sequenom, Inc., San Diego, CA, and approved September 10, 2007 (received for review May 7, 2007)
Microbial consortia form when multiple species colocalize and
communally generate a function that none is capable of alone.
Consortia abound in nature, and their cooperative metabolic ac-
tivities influence everything from biodiversity in the global food
chain to human weight gain. Here, we present an engineered
consortium in which the microbial members communicate with
each other and exhibit a ‘‘consensus’’ gene expression response.
Two colocalized populations of Escherichia coli converse bidirec-
tionally by exchanging acyl-homoserine lactone signals. The con-
populations are present at sufficient cell densities. Because neither
population can respond without the other’s signal, this consensus
function can be considered a logical AND gate in which the inputs
are cell populations. The microbial consensus consortium operates
in diverse growth modes, including in a biofilm, where it sustains
its response for several days.
biological engineering ? cellular circuits ? synthetic biology
coated in these living films. In many cases, the microorganisms
composing this ubiquitous coating form complex, interactive
communities (1–5). Despite their abundance, these microbial
communities are poorly understood. Reflecting this relative
ignorance of how bacteria behave in biofilms, efforts to program
biofilm functions are still in their infancy. The ability to manip-
ulate these films, however, would enable controlled studies of
microbial ecosystem dynamics and microscale environmental
manipulation. To begin to explore these possibilities, we have
engineered de novo cellular circuits that control Escherichia coli
behavior in a stable, robust mixed-population biofilm commu-
nity. The populations communicate, come to a consensus, and
respond to each other’s presence with a flexible, combinatory
Engineered circuits have been used to control the behavior of
and space. Cell–cell communication is a prerequisite for coor-
dination of cellular circuit dynamics on the population level.
Engineered communication, via broadcasting and receiving
small-molecule signals, can enable the programming of robust
and predictable population dynamics (13). One-way engineered
cell–cell communication has been used to coordinate biofilm
formation in a single population at a predictable cell density (8)
Here, we demonstrate an engineered bidirectional cell–cell
communication network that can coordinate gene expression
from a mixed population. We have characterized the spatial and
temporal behavior of this communication network in liquid,
agar, and biofilm growth systems.
ost bacteria live in heterogeneous surface-bound congre-
gations called biofilms, and vast reaches of the earth are
Results and Discussion
Microbial Consensus Consortium (MCC) Design and Implementation.
The MCC signaling network was constructed in E. coli from
components of the LasI/LasR and RhlI/RhlR quorum sensing
systems (16) found in Pseudomonas aeruginosa, an opportunistic
pathogen that forms a biofilm in the lungs of cystic fibrosis
patients (Fig. 1). These two systems enable P. aeruginosa cells to
sense their environment and population density and correspond-
ingly regulate hundreds of genes (17–19). LasI in Consensus
Circuit A and RhlI in Consensus Circuit B catalyze the synthesis
of the small acyl-homoserine lactone (acyl-HSL) signaling mol-
ecules 3-oxododecanoyl-HSL (3OC12HSL) and butanoyl-HSL
(C4HSL). The LasR transcriptional regulator in Circuit B is
activated by the 3OC12HSL signal emitted by Circuit A, whereas
RhlR in Circuit A is activated by the C4HSL signal emitted by
Circuit B. The acyl-HSL concentrations are low at low cell
densities but rise as the densities of Circuit A and Circuit B cells
increase. When the signal molecules accumulate at high enough
concentrations to activate the R proteins, the active R proteins
can up-regulate target gene expression. Thus, both Circuit A and
Circuit B cells must be present and at sufficient density before
generating a ‘‘consortium’’ response, in this case red and green
fluorescence. The MCC signaling network implements a logical
AND gate in which the two inputs are population levels of cells
containing Circuit A and cells containing Circuit B, and the
output is target gene expression by the two populations (Fig. 1,
lower left corner).
Proper function of the MCC is defined by minimal target gene
expression when the cells grow in isolation (neither can generate
a response without a signal from the other) and high target gene
expression when they are colocalized in sufficient densities to
activate the R proteins. Preventing a single MCC member
population from self-activating in isolation requires minimal
‘‘crosstalk’’ interactions between the Rhl and Las signaling
systems. This constraint means that the Rhl promoter p(rhl)
must respond specifically to C4HSL, the primary RhlI product
(20), and the Las promoter p(las) must respond specifically to
3OC12HSL, the primary LasI product (18, 21, 22). However,
initial experiments revealed minor crosstalk between these
promoter–activator pairs; particularly, p(rhl) responded to high
levels of 3OC12HSL [supporting information (SI) Fig. 6]. Thus,
engineering of the MCC began with modeling to investigate the
effects of this crosstalk and how these effects might be mitigated.
The model was used to choose between circuit designs based on
their ability to minimize the population densities required for
Author contributions: K.B. and D.K.K. contributed equally to this work; K.B., D.K.K., R.W.,
and F.H.A. designed research; K.B. and D.K.K. performed research; K.B. and D.K.K. con-
and F.H.A. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
3-oxododecanoyl-HSL; C4HSL, butanoyl-HSL.
§To whom correspondence may be addressed at: B-312 E-Quad, Princeton, NJ 08544-5263.
¶To whom correspondence may be addressed at: MC 210-41, 1200 East California Boule-
vard, Pasadena, CA 91125. E-mail: firstname.lastname@example.org.
This article contains supporting information online at www.pnas.org/cgi/content/full/
© 2007 by The National Academy of Sciences of the USA
October 30, 2007 ?
vol. 104 ?
activation when Circuit A cells and Circuit B cells are grown
together (activation by consensus), while maximizing the popu-
lation densities required for self-activation when Circuit A cells
and Circuit B cells are grown in isolation (isolation activation).
The model suggests that the MCC should be designed with
positive feedback on the I proteins, as illustrated in SI Figs. 7 and
8. The presence of the cognate signal, C4HSL, in cells containing
Circuit A should be a prerequisite for expression of LasI and
production of the signal 3OC12HSL; in this way, the crosstalk-
signal concentration is minimized in the absence of Circuit B.
Likewise, 3OC12HSL should up-regulate expression of RhlI in
Circuit B, limiting the concentration of the crosstalk signal,
C4HSL, in the absence of Circuit A. Modeling results illustrating
target gene expression profiles in the presence of positive
feedback are shown in Fig. 2A. The construction of Circuits A
and B therefore proceeded with lasI under control of p(rhl) in
Circuit A and rhlI under control of p(las) in Circuit B (Fig. 1 and
SI Fig. 9).
MCC Validation in Liquid Culture. We confirmed these design
choices by initial characterization of the MCC system in liquid
culture. To eliminate behavioral differences arising from varia-
tions in fluorophore maturation time and toxicity between
Circuits A and B, we used GFP as the target gene in both circuits
(GFP replaced Ds-Red; SI Fig. 9). Cells containing each circuit
were grown in isolation. Single-cell fluorescence measured as a
function of time demonstrated that isolated circuits are unable
to produce a significant response (Fig. 2B and SI Fig. 10). Cells
that allowed passage of small molecules between the two pop-
ulations through a 0.2-?m membrane. When the two circuits
were allowed to communicate with one another and grow to
sufficient density, responses from both were ?100-fold greater
than the responses of the circuits in isolation (Fig. 2B and SI Fig.
10). These results confirm our model-based design and verify
that the response is specific and combinatorial: MCC compo-
nents are distributed among different cell populations, providing
response control based on presence or absence of one of the cell
populations from the mixture.
MCC Behavior Requires Colocalization. To explore the need for
colocalization in preparation for biofilm experiments, we tested
MCC function in solid-phase cultures. Circuit A cells were
embedded in solid medium and placed in physical contact with
containing Circuits A and B. (A) A gradient of fluorescence emerges from the
interface between an agar slice with embedded Circuit A cells and another
slice with embedded Circuit B cells. (B) Image analysis of the experiment in A,
depicting the log of fluorescence. The pixels immediately surrounding the
interface between agar slices were not quantified and were replaced with a
black strip, because fluorescence in the boundary region may not accurately
represent target-gene expression. Details regarding image processing are
available in SI Text, Solid-Phase Imaging Equipment and Settings.
The MCC response is achieved by spatial colocalization of cells
aeruginosa quorum sensing components to achieve a consensus response. In
promoter. Similarly, RhlI catalyzes production of C4HSL in Circuit B, which
diffuses into Circuit A, forms a complex with RhlR, and activates the Rhl
promoter. Expression of both Targets A and B constitutes the MCC response
and can be regarded as implementing a logical AND gate operation (lower
is the output. Detailed plasmid maps for these circuits are shown in SI Fig. 9.
The MCC. Two E. coli cell populations communicate by using P.
high levels only when both populations are present at adequate population densities. To optimize performance of the AND gate, it is necessary to maximize
circuit in the presence of the other (activation by consensus). A more formal analysis is included in SI Text, Rate-Equation Based Model. (B) Liquid phase
characterization of the MCC confirms the modeling results in A. Median single-cell fluorescence is depicted for each circuit as a function of the OD of cells
than when they are grown in isolation. Fluorescence with respect to time is illustrated in SI Fig. 10. Circuit A cells grow more slowly than Circuit B cells in liquid
phase, possibly because a higher metabolic cost is associated with production of 3OC12HSL (from LasI in Circuit A) than production of C4HSL (from RhlI in Circuit
B) or because high intracellular concentrations of 3OC12HSL may have toxic effects. However, both populations reach stationary phase within 20 h of growth
in liquid culture (Inset).
Initial characterization of the MCC. (A) Modeling results depicting AND gate activity of the Circuit A and B populations. Target genes are expressed at
Brenner et al.
October 30, 2007 ?
vol. 104 ?
no. 44 ?
solid medium containing the same density of Circuit B cells (Fig.
3A). Function of both circuits was again indicated by green
fluorescence to enable quantitative comparison, and images of
green fluorescence were captured every 30 min. Image analysis
(Fig. 3B) revealed that, in the Circuit B cells closest to Circuit A,
A cells as a whole is lower than that of Circuit B cells, likely
because of the slower growth of Circuit A cells (Fig. 2B Inset).
Both populations reach maximal reporter-gene expression
within 20 h of incubation in spatial proximity and maintain these
fluorescence levels through at least 24 h (Fig. 3B). The level of
fluorescence decreases with distance from the interface, reflect-
ing the signal gradient. This finding illustrates the requirement
that cells containing Circuit A and cells containing Circuit B
must grow to adequate cell densities in spatial proximity to one
another to generate the consensus response.
MCC Function in Biofilms. Biofilms enable spatially proximate,
sheltered bacterial growth and provide for development of
predictable environmental niches in otherwise changeable mac-
roenvironments (23). The ability to engineer living films may
enable unprecedented stand-alone sensor design and environ-
mental manipulation opportunities. To explore these possibili-
ties, we studied the behavior of MCC circuit-containing cells
growing in biofilms (SI Fig. 11). First, thin conformal biofilms
were imaged by a confocal laser scanning microscope to deter-
mine whether Circuit A and B cells would respond to increasing
concentrations of acyl-HSL with increasing levels of fluores-
cultures. No biofilms used in this analysis were allowed to grow
deeper than 10 ?m, removing acyl-HSL diffusion through the
biofilm as a variable. All cells expressed a constitutive cyan
fluorescent protein, enhanced cyan fluorescent protein (eCFP),
to enable total cell counts (24), and both circuit responses were
indicated by green fluorescence. Results revealed that Circuit A
and Circuit B cells are individually able to initiate and maintain
healthy monoculture biofilms for periods of up to 2 weeks.
Consistent with the liquid phase results, both populations re-
spond strongly to their cognate acyl-HSL, and Circuit B cells
exhibit greater sensitivity to exogenous acyl-HSL than Circuit A
cells (Fig. 4).
Mixed-culture MCC biofilms were then monitored by a con-
focal laser scanning microscope. In contrast to the thin biofilm
dosage experiments detailed above, in which acyl-HSL was
provided exogenously and concentration was uniform through-
out, here, the medium served as a sink for endogenously
culture MCC analysis were allowed to grow deeper than the
monoculture biofilms so that signal molecules from Circuits A
and B could accumulate. These biofilms grew no deeper than 80
?m, a depth at which oxygen diffusion is not a variable in
fluorophore expression (25). Circuit A function was identified by
green fluorescence and Circuit B function was identified by red
As demonstrated in Fig. 5A, Circuit A and B cells grow
together and display MCC function in the mixed culture biofilm.
Images of MCC biofilms taken between 24 and 120 h after
biofilm inoculation reveal that Circuits A and B grow in intimate
contact within the biofilms. Cells containing Circuit A grow
more slowly than Circuit B cells in liquid culture (Fig. 2B Inset);
Circuit A cells grow more slowly in the biofilm, as well. Consis-
tent with the liquid- and solid-phase results, fluorescence
emerges in both strains within 24–36 h of inoculation (Fig. 5B).
Steady MCC behavior, similar to that illustrated in Fig. 5A, is
observed for at least 6 days after inoculation, after which time
biofilm depth generally exceeds the 80-?m experimental limit
(Fig. 5B). Neither circuit exhibits significant fluorescence when
grown separately in a similarly thick monoculture biofilm (Fig.
5 C and D). These results demonstrate sustained and specific
consensus consortium behavior in an engineered biofilm.
E. coli can be engineered to detect and respond to highly
varied stimuli including temperature, pH, gas concentrations,
and liquid concentrations (8, 14, 26, 27). The MCC’s population-
level AND gate enables a convenient and efficient integration of
the function of multiple engineered cells that have each been
specialized to sense and respond to particular conditions. The
MCC might also be engineered into existing industrial strains,
for example, to guarantee that in mixed-culture batch reactors,
optimal population densities are reached before onset of mul-
tispecies enzymatic activity. We have demonstrated that E. coli
growing in biofilms can be engineered like their planktonic
counterparts. Communication among the cell populations in the
MCC biofilm is essential, and it is noteworthy that some bacteria
naturally depend on quorum sensing to coordinate biofilm
formation (28, 29), whereas others are known to disrupt their
competitors’ biofilms by intercepting these signals (30). An
engineered living film could comprise such natural systems and
be tuned to interact with them to engineer its environment. For
(Inset) Mean fluorescence at each C4HSL concentration. (B) Circuit B fluoresces minimally in response to 0.001 ?M (purple) C12HSL, but fluorescence increases
concentration. Details regarding image processing are available in SI Text, Biofilm Imaging Equipment and Settings.
Monoculture biofilms respond to higher concentrations of acyl-HSL with higher levels of GFP expression. (A) Circuit A fluoresces minimally when 0.1
www.pnas.org?cgi?doi?10.1073?pnas.0704256104 Brenner et al.
example, such an engineered biofilm might be used to better
understand the interactions, or to interrupt the normal pro-
cesses, of a quorum-sensing dependent pathogenic biofilm.
In establishing the MCC, we have taken a step toward such an
engineered living film. The consensus response might be com-
posed of an enzyme and prodrug pair, or two inactive fragments
of a toxin. Leakage from either circuit in the absence of its
partner or without adequate population density would be inert,
but a highly targeted therapeutic or destructive response would
occur where and when the MCC becomes active. This type of
multisignal engineered living film could also be expanded to
include many conversation partners and to incorporate commu-
nication mechanisms other than quorum sensing. Potential
applications of such multisignal, synthetic multicellular systems
include synthesis of materials (31) in response to integrated
stimuli or surveillance and early detection of environmental
changes related to epidemiology or material degradation. As a
medical technology, an engineered biofilm consortium might
eliminate unwanted infection or even destroy harmful cells in the
body (32). In such applications, the engineered bacterial biofilm
consortium would carry out its function over long periods of
time, under a variety of conditions, with minimal human aware-
ness or intervention.
Plasmids. The Circuit A plasmid pFNK-601 encodes lasI and
gfp(LVA) under control of the Pseudomonas aeruginosa rhlAB
promoter (qsc119), as well as constitutive RhlR production from
amplified fragment of P. aeruginosa PAO-1 containing the lasI
gene and from the Receiver A plasmid pFNK-202-qsc119 (33),
shown in SI Fig. 6a. The Circuit B plasmid pFNK-602 expresses
rhlI and gfp(LVA) from the P. aeruginosa p(las) promoter p(rsal).
This plasmid also encodes constitutive lasR from the p(lacIq)
promoter. Plasmid pFNK-602 was constructed by inserting rhlI
into parent plasmid pFNK-503-qscrsaL. Plasmid pFNK-503-
qscrsaL is the Receiver B plasmid (SI Fig. 6a) and expresses lasR
from p(lacIq) and gfp(LVA) from the P. aeruginosa p(rsal)
promoter. Plasmid pFNK-602-red was constructed by replacing
gfp(LVA) in pFNK-602 with dsred-exp from Clontech plasmid
pDsRed-Exp. These plasmids are illustrated in SI Fig. 9.
Model and Simulations. Continuous differential equations were
used to model promoter activation by R proteins, I protein
production and degradation (which is proportional to target
protein expression at steady state), acyl-HSL synthesis and
degradation, and saturation of acyl-HSL synthesis. The model is
described in detail in SI Text, Rate-Equation-Based Model.
Liquid-Phase Data Acquisition and Analysis. To study the MCC
response in liquid phase, starter cultures of E. coli JM2.300 cells
[F?lacI22 ??e14?rpsL135(StrR) thi-1] harboring either Circuit
A or Circuit B plasmids were grown to OD ?0.3 in M9 medium
(2 mM MgSO4/0.2% casamino acids/0.5% glycerol/300 ?M
thiamine) with 50 ?g ml?1kanamycin at 37°C in a shaking
incubator. The cells were then washed and diluted to an OD of
0.02 in M9 medium supplemented with 50 ?g ml?1kanamycin.
Holes were bored into the sides of two 50-ml Corning (Corning,
NY) centrifuge tubes, and 20 ml of the Circuit A dilution was
placed in one tube, and 20 ml of the Circuit B dilution was placed
in the other. A Millipore (Billerica, MA) Steriflip vacuum
filtration unit was used to provide a 0.22 ?M filter interface
between the two cultures for the consensus experiments. Both
50-ml tubes were affixed horizontally to the platform of a shaker
replaced with 1 ml of fresh M9 medium. Sample OD was
measured by using a Beckman Coulter (Fullerton, CA) DU 800
spectrophotometer, and fluorescence measurements were taken
both Circuit A and B cells are present and fluorescing. Circuit A cells constitutively express enhanced yellow fluorescent protein (eYFP; shown in yellow) and
express GFP when the circuit is ‘‘on.’’ Circuit B cells constitutively express enhanced cyan fluorescent protein (eCFP; shown in cyan) and express ds-Red when the
circuit is ‘‘on.’’ Circuit A cells are a minority, possibly because of their slower growth. (B) Mean intensities for Circuit A and Circuit B cells remain significant for
at least 120 h after inoculation. (C) Monoculture biofilms of Circuit A cells fluoresce minimally. (D) Monoculture biofilms of Circuit B cells fluoresce minimally.
All gridlines are 20 ?m apart. Details regarding image processing are available in SI Text, Biofilm Imaging Equipment and Settings.
The MCC functions for at least 6 days when grown in a biofilm. (A) Three-dimensional rendering of an MCC biofilm, 24 h after inoculation, shows that
Brenner et al.
October 30, 2007 ?
vol. 104 ?
no. 44 ?
on a Beckman Coulter Altra flow cytometer equipped with a Download full-text
488-nm argon excitation laser and a 515–545 nm emission filter.
Median fluorescence values were converted to equivalent fluo-
rescein molecule counts by using SPHERO Rainbow Calibration
Particles (Spherotech RCP-30-5A; Spherotech, Lake Forest, IL)
that were measured during each session.
Solid-Phase Experiments. Two starter cultures of E. coli JM2.300,
one harboring the Circuit A plasmid and one harboring the
Circuit B plasmid, were grown to OD ?0.3 in M9 medium with
50 ?g ml?1kanamycin as described above. Cells from each
culture were aliquotted into 6 ml of 37°C molten 1.5% low-melt
agarose (SeaPlaque; Lonza, Rockland, ME) containing M9 and
poured into 60 ? 15 mm Petri dishes (Falcon, Oxnard, CA), and
rectangular segments containing either Circuit A or Circuit B
were excised from the solidified products. A Circuit A segment
was placed end to end with a circuit B segment in a sterile WillCo
glass-bottom dish. The plate was then incubated at 37°C, and
images were taken every 30 min by using a Zeiss (Thornwood,
CCD camera. Images were captured with a ?2.5 brightfield
objective and a GFP filter with 470/40 excitation and 525/50
emission. Additional information is available in SI Text, Solid-
Phase Imaging Equipment and Settings.
Biofilm Experiments. Starter cultures of E. coli JM2.300 harboring
plasmid pMP4641 (24) and either the Circuit A or Circuit B
plasmid were grown to saturation at 37°C in M9 biofilm medium
(2 mM MgSO4/0.1% casamino acids/0.4% glucose/0.01% thy-
mine/100 ?M CaCl2) containing 50 ?g ml?1kanamycin and 20
?g ml?1tetracycline. Starter cultures were then diluted to OD
0.2 in fresh M9 biofilm medium with 50 ?g ml?1kanamycin and
40 mm flow chambers (Stovall Life Science, Greensboro, NC)
with glass microscope coverslips. Monoculture biofilms were
inoculated with 1 ml of the dilution of cells of the appropriate
circuit, and MCC biofilms were inoculated with a mixture of 500
?l of each. After inoculation, flow chambers were incubated for
1 h without flow and then perfused at a low flow rate with M9
biofilm medium containing 50 ?g ml?1kanamycin and 20 ?g
ml?1tetracycline. The flow chambers were incubated at 30°C.
Images of the biofilms were captured at 24-h intervals with a
Zeiss 510 upright confocal laser scanning microscope, controlled
by Zeiss AIM. A Zeiss Achroplan ?40/0.8 W objective was used
to capture all images. Images were captured with 512 ? 512 pixel
resolution, and all images used in quantitative comparisons were
captured with identical pinhole and gain settings. Enhanced cyan
fluorescent protein (eCFP) excitation: 458 nm with an Argon
laser; emission filter: BP 480–520 nm. GFP excitation: 488 nm
with an Argon laser; emission filter: BP 500–530 nm. ds-Red
excitation: 543 nm with a Helium-neon laser; emission filter: LP
560 nm. Images were processed for quantitative comparison with
custom-written Matlab-based tools. Three-dimensional render-
ing was performed in Imaris 4.5.2. Biofilms prepared solely for
three-dimensional rendering incorporated a fourth fluorophore,
enhanced yellow fluorescent protein (eYFP), on plasmid
pMP4658 (24); excitation: 514 nm with an Argon laser; emission
filter: LP 530 nm. More information regarding procedures,
equipment settings, and processing can be found in SI Text,
Biofilm Experimental Setup and Biofilm Imaging Equipment and
We thank Jared Leadbetter and Ernesto Andrianantoandro for discus-
sions or comments on the manuscript; and Chris Waters, Tracy Teal, and
the Caltech Biological Imaging Center for assistance with biofilm
imaging. This material is based on work supported by 2005 National
Science Foundation Emerging Models and Technologies for Computa-
tion Grant CCF-0523195 and 2006 National Institutes of Health Grants
R01 GM074712-01A1 and 5R01CA118486-2.
1. Hooper LV, Midtvedt T, Gordon JI (2002) Annu Rev Nutr 22:283–307.
2. Kato S, Haruta S, Cui ZJ, Ishii M, Igarashi Y (2005) Appl Environ Microbiol
3. Kleessen B, Blaut M (2005) Br J Nutr 93(Suppl 1):S35–S40.
4. Macfarlane S, Woodmansey EJ, Macfarlane GT (2005) Appl Environ Microbiol
5. Mishra S, Jyot J, Kuhad RC, Lal B (2001) Curr Microbiol 43:328–335.
6. Becskei A, Serrano L (2000) Nature 405:590–593.
7. Becskei A, Seraphin B, Serrano L (2001) EMBO J 20:2528–2535.
8. Kobayashi H, Kaern M, Araki M, Chung K, Gardner TS, Cantor CR, Collins
JJ (2004) Proc Natl Acad Sci USA 101:8414–8419.
9. Kramer BP, Viretta AU, Daoud-El-Baba M, Aubel D, Weber W, Fussenegger
M (2004) Nat Biotechnol 22:867–870.
10. Elowitz MB, Leibler S (2000) Nature 403:335–358.
11. Bulter T, Lee SG, Wong WW, Fung E, Connor MR, Liao JC (2004) Proc Natl
Acad Sci USA 101:2299–2304.
12. Atkinson MR, Savageau MA, Myers JT, Ninfa AJ (2003) Cell 113:597–607.
13. You L, Cox RS, Weiss R, Arnold FH (2004) Nature 428:868–871.
14. Basu S, Gerchman Y, Collins CH, Arnold FH, Weiss R (2005) Nature
15. Basu S, Mehreja R, Thiberge S, Chen MT, Weiss R (2004) Proc Natl Acad Sci
16. Pesci EC, Iglewski BH (1997) Trends Microbiol 5:132–134.
17. De Kievit TR, Gillis R, Marx S, Brown C, Iglewski BH (2001) Appl Environ
18. Pearson JP, Gray KM, Passador L, Tucker KD, Eberhard A, Iglewski BH,
Greenberg EP (1994) Proc Natl Acad Sci USA 91:197–201.
19. Pesci EC, Pearson JP, Seed PC, Iglewski BH (1997) J Bacteriol 179:3127–3132.
20. Winson MK, Camara M, Latifi A, Foglino M, Chhabra SR, Daykin M, Bally
M, Chapon V, Salmond GP, Bycroft BW, et al. (1995) Proc Natl Acad Sci USA
21. Passador L, Tucker KD, Guertin KR, Journet MP, Kende AS, Iglewski BH
(1996) J Bacteriol 178:5995–6000.
22. Gould TA, Herman J, Krank J, Murphy RC, Churchill ME (2006) J Bacteriol
23. Christensen BB, Haagensen JA, Heydorn A, Molin S (2002) Appl Environ
24. Bloemberg GV, Wijfjes AH, Lamers GE, Stuurman N, Lugtenberg BJ (2000)
Mol Plant–Microbe Interact 13:1170–1176.
25. Sternberg C, Christensen BB, Johansen T, Toftgaard Nielsen A, Andersen JB,
Givskov M, Molin S (1999) Appl Environ Microbiol 65:4108–4117.
26. Anderson JC, Clarke EJ, Arkin AP, Voigt CA (2006) J Mol Biol 355:
27. Levskaya A, Chevalier AA, Tabor JJ, Simpson ZB, Lavery LA, Levy M,
28. Eberl L (1999) Syst Appl Microbiol 22:493–506.
29. Parsek MR, Greenberg EP (2005) Trends Microbiol 13:27–33.
30. Houdt R, Aertsen A, Moons P, Vanoirbeek K, Michiels CW (2006) FEMS
Microbiol Lett 256:83–89.
31. Klaus T, Joerger R, Olsson E, Granqvist CG (1999) Proc Natl Acad Sci USA
32. Swartz JR (2001) Curr Opin Biotechnol 12:195–201.
33. Karig D, Weiss R (2005) Biotechnol Bioeng 89:709–718.
www.pnas.org?cgi?doi?10.1073?pnas.0704256104 Brenner et al.