Blackwell Science, LtdOxford, UKMMIMolecular Microbiology0950-382X© 2005 The Authors; Journal compilation © 2005 Blackwell Publishing Ltd?

200558411861202Original ArticleRegulation of quorum sensing in Vibrio choleraeD. H. Lenz

et al.

Molecular Microbiology (2005) 58(4), 1186–1202 doi:10.1111/j.1365-2958.2005.04902.x

First published online 14 October 2005

CsrA and three redundant small RNAs regulate quorum

sensing in Vibrio cholerae
Derrick H. Lenz,1 Melissa B. Miller,1† Jun Zhu,2 synchronize the gene expression of the community (Miller
Rahul V. Kulkarni3 and Bonnie L. Bassler1* and Bassler, 2001). Quorum sensing involves production,
1
Department of Molecular Biology, Howard Hughes release and detection of signalling molecules called auto-
Medical Institute, Princeton University, Princeton, NJ inducers. Quorum-sensing-controlled behaviours are typ-
08544-1014, USA. ically ones that are only productive when carried out in
2
Department of Microbiology, University of Pennsylvania unison by a group of cells. Examples include DNA
School of Medicine, Philadelphia, PA 19104, USA. exchange, sporulation, biofilm formation, virulence factor
3
Department of Physics, Virginia Polytechnic Institute and production and bioluminescence (Miller and Bassler,
State University, Blacksburg, VA 24061, USA. 2001; Zhu et al., 2002; Hammer and Bassler, 2003; Zhu
and Mekalanos, 2003; Yildiz et al., 2004).
The human pathogen Vibrio cholerae has multiple
Summary
quorum-sensing systems that act in parallel (Fig. 1) (Miller
Bacteria communicate using a process called quorum et al., 2002). System 1 is composed of the CAI-1 autoin-
sensing which involves production, secretion and ducer of unknown structure and CqsS, its two-component
detection of signalling molecules called auto- sensor. System 2 is made up of AI-2, a furanosyl borate
inducers. Quorum sensing allows populations of diester, the periplasmic binding protein, LuxP, and the two-
bacteria to simultaneously regulate gene expression component sensor, LuxQ (Bassler et al., 1994; Chen
in response to changes in cell density. The human et al., 2002). At low-cell density, when concentrations of
pathogen, Vibrio cholerae, uses a quorum-sensing autoinducers are low, the sensors are kinases that auto-
circuit composed of parallel systems that transduce phosphorylate on conserved histidines (H1) and transfer
information through four redundant regulatory small phosphate to conserved aspartates (D1) on their
RNAs (sRNAs) called quorum regulatory RNAs (Qrr) response regulator modules (Freeman et al., 2000). Phos-
to control the expression of numerous genes, most phate is passed to a histidine (H2) on the phospho-
notably those required for virulence. We show that the transfer protein LuxU and next to an aspartate (D2) on
VarS/VarA two-component sensory system comprises the response regulator LuxO (Freeman and Bassler,
an additional regulatory input controlling quorum- 1999; 2000). Phospho-LuxO, with the alternative sigma
sensing-dependent gene expression in V. cholerae. factor σ54, activates expression of four genes encoding
VarS/VarA controls transcription of three previously small RNAs (sRNAs) (Lilley and Bassler, 2000; Lenz
unidentified small regulatory RNAs (sRNAs) that are et al., 2004). The sRNAs (called Qrr sRNAs for quorum
similar to the sRNAs CsrB and CsrC of Escherichia regulatory RNA) with the sRNA chaperone Hfq, bind to
coli. The three V. cholerae sRNAs, which we name the hapR mRNA encoding the master regulator of the
CsrB, CsrC and CsrD, act redundantly to control quorum-sensing cascade, HapR, and destabilize it. Thus,
the activity of the global regulatory protein, CsrA. little HapR protein is made at low-cell density. At high cell
The VarS/VarA-CsrA/BCD system converges with the density, in response to autoinducers, the sensors switch
V. cholerae quorum-sensing systems to regulate the from kinases to phosphatases. Phosphate flows backward
expression of the Qrr sRNAs, and thus, the entire through the circuit rendering LuxO inactive. The qrr genes
quorum-sensing regulon. are not transcribed, hapR mRNA is stabilized, HapR is
translated, and HapR activates or represses a variety of
genes. For example, when HapR is produced, it activates
Introduction
protease production and represses virulence factor pro-
Quorum sensing is a process of chemical communication duction and biofilm formation (Fig. 1) (Jobling and
that bacteria use to monitor cell population density and Holmes, 1997; Zhu et al., 2002; Hammer and Bassler,
2003; Zhu and Mekalanos, 2003).
Accepted 29 August, 2005. *For correspondence. E-mail A previous study predicted that a third sensory system
bbassler@molbio.princeton.edu; Tel. (+1) 609 258 2857; Fax exists in V. cholerae and functions through LuxO but inde-
(+1) 609 258 2957. †Present address: University of North Carolina
School of Medicine, Department of Pathology and Laboratory Medi- pendently of LuxU (Miller et al., 2002). Here we report
cine, Chapel Hill, NC 27599, USA. experiments that identify a sensor kinase-response regu-
© 2005 The Authors
Journal compilation © 2005 Blackwell Publishing Ltd
 Regulation of quorum sensing in Vibrio cholerae 1187

OM

VarS/VarA-CsrA/BCD CAI-1-CqsS AI-2-LuxPQ

System System System

CAI-1 LuxP AI-2

IM
H1 H1 H1
VarS CqsS LuxQ
D1 D1
D1 P

H2

LuxU H2

VarA D2
H2
54
LuxO D2 +
CsrB, C, D
X sRNAs+ Hfq
sRNAs

CsrA HapR
virulence lux
biofilms protease
other genes other genes

Fig. 1. Model of the V. cholerae quorum-sensing circuit. The VarS/VarA-CsrA/BCD system functions with the CAI-1-CqsS and AI-2-LuxPQ
systems. Pentagons and triangles denote CAI-1 and AI-2 respectively. The flow of phosphate indicated by the arrows depicts the low-cell density
state. Flow is reversed at high-cell density. Dotted lines denote hypothetical interactions.

lator pair that we call VarS/VarA as part of this third sys- roles in the inhibition of V. cholerae CsrA. We further show
tem. The response regulator, VarA had previously been that VarS/VarA-CsrA/BCD act through the quorum-sens-
identified as a regulator of virulence in V. cholerae (Wong ing network to regulate the expression of hapR. Our find-
et al., 1998). VarS/VarA homologues exist in a variety of ings suggest that the VarS/VarA-CsrA/BCD system could
Gram-negative bacteria, including Escherichia coli (BarA/ function as part of the third sensory system in V. cholerae.
UvrY), Salmonella typhimurium (BarA/SirA), members of
the genus Pseudomonas (GacS/GacA) and Legionella
pneumophila (LetS/LetA) (Heeb and Haas, 2001; Molof- Results
sky and Swanson, 2004). While the signal(s) for these
A screen for V. cholerae quorum-sensing regulators
systems are not known, it is established that VarS-type
reveals VarS and VarA
proteins act as sensor kinases for VarA-type response
regulators (Rich et al., 1994; Pernestig et al., 2001). In E. The V. cholerae quorum-sensing network is similar to that
coli, the VarS/VarA homologues BarA/UvrY activate tran- of the bioluminescent marine bacterium Vibrio harveyi
scription of genes encoding the sRNAs CsrB and CsrC (Miller et al., 2002). In V. harveyi, quorum sensing con-
which bind to and inhibit the activity of the CsrA protein trols the expression of the luciferase operon (luxCDABE)
(Liu et al., 1997; Weilbacher et al., 2003). CsrA is a post- (Nealson and Hastings, 1979; Bassler et al., 1993; 1994).
transcriptional regulator of diverse processes including We showed that luxCDABE from V. harveyi is a
carbon metabolism and biofilm formation (Romeo et al., faithful reporter of quorum-sensing gene expression
1993; Romeo, 1998). In Pseudomonas species, the VarS/ in V. cholerae. Using luxCDABE as the reporter, Tn5
VarA homologues, called GacS/GacA, activate transcrip- mutagenesis of V. cholerae was performed, and followed
tion of rsm genes encoding sRNAs that inhibit the CsrA by a screen for decreased light production. This led to
homologue RsmA (Aarons et al., 2000; Heeb et al., 2002; identification of the CAI-1-CqsS quorum-sensing system
Valverde et al., 2003; Burrowes et al., 2005). Here we and we showed it functions in parallel with the AI-2-LuxPQ
show that, in V. cholerae, VarS/VarA activate expression system (Miller et al., 2002).
of three previously unidentified genes encoding sRNAs The above screen yielded mutants with dim/dark phe-
similar to CsrB of E. coli. We name these sRNAs CsrB, notypes that did not have insertions in genes encoding
CsrC and CsrD, and demonstrate that they play redundant known quorum-sensing regulators. We have sequenced
© 2005 The Authors
Journal compilation © 2005 Blackwell Publishing Ltd, Molecular Microbiology, 58, 1186–1202
1188 D. H. Lenz et al.
these insertions, and several are located in two unlinked mutants produce low levels of light (closed circles and
genes, annotated VC2453 and VC1213. VC2453 and closed triangles respectively). The varS, luxO and varA,
VC1213 encode proteins homologous to GacS and GacA, luxO double mutants display constitutive bright lux pheno-
respectively, which were first identified as global regula- types (open circles and open triangles respectively). Thus,
tors of secondary metabolite production in Pseudomonas VarS and VarA require LuxO to transduce information
syringae and Pseudomonas fluorescens (Hrabak and through the circuit, so we place them upstream of LuxO
Willis, 1992; Laville et al., 1992). VC1213 has previously in the regulatory hierarchy (Fig. 1). We note that the luxO
been identified in V. cholerae as the response regulator mutant is 10-fold brighter than the varS, luxO and varA,
VarA (Virulence Associated Regulator) and a varA mutant luxO double mutants, however, the double mutants have
has reduced virulence due to decreased cholerae toxin severe growth defects, and we suspect that this contrib-
(CT) and toxin-coregulated pilus (TCP) production (Wong utes to their inability to produce maximal levels of light.
et al., 1998). Our varA mutant has a virulence defect iden- We have observed similar effects in other mutants that
tical to that previously reported (not shown). VC2453 is show impaired growth (Lenz et al., 2004).
predicted to encode a two-component sensor kinase with We also examined VarS/VarA function relative to LuxU.
three conserved domains (H1, D1 and H2). This domain Unlike a luxO mutant, a luxU mutant retains density-
organization is identical to other GacS homologues, and dependent expression of lux (Fig. 2C, open diamonds).
VC2453 is 34% identical to GacS of Pseudomonas spp., We have argued previously that this phenotype is due to
and 52% and 51% identical to BarA of E. coli and S. an unidentified System 3 channelling information into the
typhimurium respectively. No cognate sensor kinase has circuit through LuxO but independently of LuxU. The varS,
been identified for VarA; we suggest VC2453 fulfils this luxU and varA, luxU double mutants do not give a clear
function. Therefore, we name the VC2453 gene product epistasis result, but instead produce levels of light inter-
VarS. mediate between the single varS and varA mutants and
To investigate the roles of VarS and VarA in quorum the single luxU mutant (Fig. 2C open circles and open
sensing, we measured cell density-dependent light pro- triangles respectively). Mutation of luxU has the opposite
duction from luxCDABE in various V. cholerae strains. All effect on LuxO-phosphate, and in turn on biolumines-
strains were grown to high-cell density and diluted 1000- cence, than does mutation of varS or varA. In the absence
fold. The characteristic quorum-sensing behaviour is of LuxU, all input from the CAI-1-CqsS and AI-2-LuxPQ
displayed by the wild type (Fig. 2A, open squares). systems is eliminated, reducing the level of phospho-
After dilution, light production declines precipitously in LuxO, so a luxU strain is brighter than wild type at low-
response to dilution of the autoinducers to a concentration cell density. In contrast, varS and varA mutants are darker
below the level required for detection. As the culture than wild type suggesting that increased phosphate is
grows, autoinducers are produced and accumulate. The shuttled to LuxO in these mutants. When we combine the
cells respond by commencing light production. varS and varS or varA mutation with the luxU mutation, we believe
varA mutants are impaired in production of light. At high- we observe an intermediate phenotype because of the
cell density, the varS and varA mutants produce 100-fold competing effects each mutation has on the luminescence
and 10 000-fold less light than the wild type respectively. readout. Additional evidence follows to support this
Mutation of varA is more severe than mutation of varS. A hypothesis.
likely interpretation is that, similar to E. coli, mutation of We wondered if the decreased Lux production of varS
the barA homologue, varS, partially disrupts signalling and varA mutants depended on the kinase input from the
through VarA, whereas mutation of the downstream uvrY CqsS and LuxQ sensors. To investigate this, we con-
homologue, varA, produces a null phenotype (Suzuki structed cqsS, luxQ, varS and cqsS, luxQ, varA triple
et al., 2002). mutants. Mutation of cqsS and luxQ does not alter the
varS or varA phenotype (Fig. 2D; compare closed and
open circles and closed and open triangles, respectively),
VarS/VarA function in parallel to CAI-1-CqsS and
indicating that VarS/VarA operate independently of CqsS
AI-2-LuxPQ
and LuxQ. We suggest that the VarS/VarA system acts in
To determine where VarS/VarA channel information into parallel to the CAI-1-CqsS and AI-2-LuxPQ systems and
the quorum-sensing pathway, we performed genetic channels information into the circuit at the level of LuxO
epistasis tests. First we analysed whether LuxO is (Fig. 1).
required for the VarS/VarA effect on bioluminescence
(Fig. 2B). V. cholerae luxO mutants carrying luxCDABE
VarS/VarA require CsrA to control quorum sensing
are constitutively bright because, in the absence of LuxO,
HapR is always present and constitutively activates lux Based on homologous systems, we predicted that
expression (Fig. 2B; open diamonds). varS and varA an additional regulatory component(s) must exist

© 2005 The Authors

Journal compilation © 2005 Blackwell Publishing Ltd, Molecular Microbiology, 58, 1186–1202
 Regulation of quorum sensing in Vibrio cholerae 1189
Fig. 2. The VarS/VarA two-component system
A controls quorum sensing in V. cholerae.
Density-dependent lux assays for V. cholerae
1011
strains: A. MM227 (WT, open squares), MM609
1010 (∆varS, closed circles), MM560 (∆varA, closed
WT triangles). Relative light units (indicated as
109 RLU) are counts min−1 ml−1/OD600.
B. MM227 (WT, open squares), MM349 (∆luxO,
RLU

108 open diamonds), MM609 (∆varS, closed cir-

varS
cles), MM560 (∆varA, closed triangles),
107 DL3698 (∆luxO, ∆varS, open circles), DL3696
(∆luxO, ∆varA, open triangles).
106
C. MM227 (WT, open squares), MM359 (∆luxU,
varA
105 open diamonds), MM609 (∆varS, closed cir-
cles), MM560 (∆varA, closed triangles), MM634
104 (∆luxU, ∆varS, open circles), MBM32 (∆luxU,
0.001 0.01 0.1 1 10 ∆varA, open triangles).
B OD600
D. MM227 (WT, open squares), DL2820
(∆cqsS, luxQ, open diamonds), MM609 (∆varS,
1011 closed circles), MM560 (∆varA, closed trian-
gles), DL3282 (∆cqsS, ∆luxQ, ∆varS, open cir-
1010 WT
cles), DL3268 (∆cqsS, ∆luxQ, ∆varA, open
luxO triangles).
10 9
varS
RLU

10 8
varA
10 7
luxO, varS
10 6
luxO, varA
10 5

10 4
0.0001 0.001 0.01 0.1 1 10

OD600
C
10 11
WT
10 10

10 9 luxU

10 8 varS
RLU

10 7 varA
10 6
luxU, varS
10 5
luxU, varA
10 4
0.0001 0.001 0.01 0.1 1 10

OD600

D 10 11
WT
10 10
cqsS, luxQ
10 9
varS
10 8
RLU

10 7 varA

10 6 cqsS, luxQ, varS

10 5
cqsS, luxQ, varA
10 4
0.001 0.01 0.1 1 10

OD600

© 2005 The Authors

Journal compilation © 2005 Blackwell Publishing Ltd, Molecular Microbiology, 58, 1186–1202
1190 D. H. Lenz et al.
downstream of VarS/VarA and link it to the quorum- notype (data not shown). These results indicate that VarS/
sensing system. To identify this component(s), we per- VarA control quorum sensing through CsrA.
formed a varA bypass suppressor screen by subjecting We defined the epistatic relationships between csrA
the dim V. cholerae varA mutant carrying luxCDABE to and the genes encoding the quorum-sensing regulators.
Tn5 mutagenesis and screening for mutants restored to Analogous to the results with the varS and varA mutants,
a bright phenotype. Tn5 insertions were located in luxO, the phenotype of the csrA::Tn5 mutant is converted to the
rpoN (encoding σ54) and VC0548. Mutations in luxO constitutive bright phenotype of the luxO mutant in the
and rpoN were anticipated because their inactivation double csrA::Tn5, luxO mutant showing that input from
results in constitutive light production. VC0548 encodes CsrA depends on LuxO (data not shown). Similar to
a protein that is 89% identical to CsrA, a sRNA-binding, above, the csrA::Tn5 mutation does not alter the pheno-
post-transcriptional regulator in E. coli (Romeo et al., type of the double cqsS, luxQ mutant. Thus, input from
1993). CsrA does not depend on the known sensors (data not
The phenotypes of the varS, varA and csrA::Tn5 single shown). Combining the csrA::Tn5 mutation with the luxU
mutants and the varS, csrA::Tn5 and varA, csrA::Tn5 dou- null mutation increases light production from the luxU
ble mutants are shown in Fig. 3A. Mutation of csrA in the mutant 50-fold at low-cell density (Fig. 3B; compare open
wild type causes no significant change in quorum-sensing squares and open diamonds). Mutation of luxU eliminates
phenotype (compare open squares to closed diamonds). significant phospho-flow to LuxO, making the strain
However, in the varS or varA dim mutants, Tn5 insertion brighter than wild type at low-cell density. Combining the
in csrA restores the wild type profile (open circles and luxU mutation with the csrA::Tn5 mutation results in a
open triangles respectively). Introduction of the wild-type strain with an even brighter phenotype at low-cell density,
V. cholerae csrA gene into the varA, csrA::Tn5 mutant indicating that phosphate flow to LuxO has been further
complements the defect and restores the dark varA phe- reduced by mutation of csrA. These results are in contrast

Fig. 3. CsrA operates downstream of VarS/

A VarA. Density-dependent lux assays for: A
MM227 (WT, open squares), MM609 (∆varS,
1011
closed circles), MM560 (∆varA, closed trian-
WT gles), DL2493 (csrA::Tn5, closed diamonds),
1010
DL3404 (∆varS, csrA::Tn5, open circles),
varS DL2499 (∆varA, csrA::Tn5, open triangles) and
109 B MM359 (∆luxU, open squares), DL2493
(csrA::Tn5, closed diamonds), DL2489 (∆luxU,
108 varA
csrA::Tn5, open diamonds).
RLU

107 csrA::Tn5

106
varS, csrA::Tn5
105
varA, csrA::Tn5
104
0.001 0.01 0.1 1 10
OD600
B
1011

1010 luxU

10 9

10 8
RLU

10 7 csrA::Tn5

10 6

10 5
csrA::Tn5, luxU
10 4
0.0001 0.001 0.01 0.1 1 10

OD600

© 2005 The Authors

Journal compilation © 2005 Blackwell Publishing Ltd, Molecular Microbiology, 58, 1186–1202
 Regulation of quorum sensing in Vibrio cholerae 1191
to the epistasis results with varS and varA and luxU 17 AGGA and two AGGGA motifs, and CsrD has 13
(Fig. 2B). Our interpretation is that, in the case of luxU AGGA and two AGGGA motifs. These results suggest
and varS/varA, the effects of the mutations oppose each that CsrBCD of V. cholerae likely act in a manner similar
other so combining the mutations results in an intermedi- to their homologues in E. coli, and control CsrA by bind-
ate phenotype. In the case of luxU and csrA, the effects ing multiple copies of the protein at the AGGA loops. To
of the two mutations reinforce one another, and a syner- test if these components are equivalent to the analogous
gistic effect is observed when the mutations are com- E. coli components, we examined their roles in glycogen
bined. All of the results are consistent with VarS/ production because CsrA is a repressor of glycogen pro-
VarA-CsrA being components of a third sensory system duction in E. coli. Introduction of V. cholerae csrA into an
functioning as depicted in Fig. 1. E. coli csrA mutant reduced glycogen production as
judged by iodine staining of glycogen on M63-minimal
glucose plates (Romeo et al., 1993). Consistent with this,
Identification and analysis of V. cholerae CsrB, CsrC and
overexpression of the V. cholerae csr sRNA genes in
CsrD
wild-type E. coli significantly increased glycogen produc-
In the E. coli BarA/UvrY-CsrA system, the sRNAs CsrB tion. These results show that the V. cholerae CsrA and
and CsrC control the activity of CsrA (Liu et al., 1997; CsrBCD components are equivalent to their counterparts
Weilbacher et al., 2003). Likewise, in Pseudomonas in E. coli and justify the naming convention we have
GacS/GacA control RsmA activity through the Rsm chosen.
sRNAs (Aarons et al., 2000; Heeb et al., 2002; Valverde To test whether CsrB, CsrC and CsrD regulate
et al., 2003; Burrowes et al., 2005). In these cases, the V. cholerae quorum sensing we measured biolumines-
sRNAs bind and titrate the activity of CsrA and RsmA. cence in csrB, csrC and csrD single, double and triple
Thus, the sRNAs have a negative effect on their sRNA- mutants. Figure 4C shows that each single mutant and all
binding proteins. the combinations of double sRNA mutants are wild type
To identify putative sRNAs that could act in conjunction with respect to density-dependent bioluminescence. How-
with CsrA in V. cholerae, we performed a BLAST search of ever, the triple csrBCD mutant is dark (closed black cir-
the V. cholerae genomic contigs using the E. coli csrB cles), and thus quorum sensing is eliminated in this
sequence. This analysis identified one candidate csrB mutant. Importantly, the dark csrBCD phenotype differs
gene. We suspected multiple sRNAs act downstream of from the density-dependent bright Lux phenotype of the
VarA, so we used the V. cholerae csrB sequence we had csrA mutant (Fig. 3A) suggesting that CsrBCD act oppo-
identified to perform a second BLAST search against sitely to CsrA with respect to quorum sensing. An epista-
the V. cholerae genome. Two additional sRNAs were sis test demonstrated that the csrA, csrBCD quadruple
revealed. We name the three genes csrB (located mutant phenotype is identical to that of the single csrA
between VC0190/VC0191), csrC (located between mutant (not shown) showing that CsrBCD act upstream
VC0882/VC0883) and csrD (located between VCA0839/ of CsrA (Fig. 1). As in other systems, the CsrBCD sRNAs
VCA0840). While the genome contexts of the E. coli csrB most likely work by inhibiting the action of the CsrA bind-
and csrC genes are conserved at least partially in closely ing protein.
related species, none of the genes flanking the csr sRNAs To verify that CsrBCD function in the VarS/VarA path-
genes in V. cholerae are orthologues of the genes flanking way, we overexpressed each in the varA mutant and mea-
csrB and csrC in E. coli. The three V. cholerae csr sRNA sured light production (Fig. 4D). Expression of any one of
genes most likely arose from gene duplications because the sRNAs causes over a 1000-fold increase in light pro-
they are more homologous to each other than to the csr duction. Thus, each sRNA can bypass the varA mutation
sRNA genes from other organisms. suggesting that these sRNAs act downstream of VarA
An alignment of the V. cholerae csrB, csrC and csrD (Fig. 1). Because deletion of all three csr sRNA genes is
genes is shown in Fig. 4A and the predicted secondary required to obtain the varA phenotype, we suggest that
structures of the corresponding sRNAs are shown in CsrB, CsrC and CsrD play redundant roles in quorum-
Fig. 4B. In E. coli, secondary structures for CsrB and sensing regulation. Deletion of all three is sufficient to
CsrC show 18 and nine repeats, respectively, containing simulate the varA phenotype, so we conclude that no
AGGA/AGGGA sequences located primarily in the loop additional sRNAs function at this step.
regions of stem-loops or in other single-stranded regions
which are the predicted sites of interaction with CsrA (Liu
VarS/VarA activate expression of csrB, csrC and csrD
et al., 1997; Weilbacher et al., 2003). The V. cholerae
Csr sRNA predictions are consistent with the proposed In other bacteria, VarS/VarA homologues are proposed to
binding motif for CsrA (Liu et al., 1997). In V. cholerae, activate expression of genes encoding the CsrB-type
CsrB has 15 AGGA and three AGGGA motifs, CsrC has sRNAs via conserved DNA regions speculated to bind the

© 2005 The Authors

Journal compilation © 2005 Blackwell Publishing Ltd, Molecular Microbiology, 58, 1186–1202
1192 D. H. Lenz et al.

CsrB CsrC CsrD

C D
10 11 1010
WT 10 9
10 10
luxO 10 8
10 9 varA 10 7
csrBCD 10 6
RLU

10 8
RLU

csrB
10 5
10 7 csrC
10 4
csrD
10 6 10 3
csrBC
10 2
10 5 csrBD
10 1
csrCD
10 4 10 0
varA (pKK177)

1
varA (pcsrB)

varA (pcsrC)

varA (pcsrD)

0.001 0.01 0.1 10

OD600

Strain

Fig. 4. Identification and activity of the CsrB, CsrC and CsrD sRNAs.
A. Alignment of the DNA sequences encoding the CsrB, CsrC and CsrD sRNAs. The sequences are shown from the predicted transcription start
to termination sites.
B. Lowest-energy secondary-structural predictions for V. cholerae CsrB, CsrC and CsrD sRNAs. AGGA and AGGGA motifs are in bold.
C. Density-dependent lux assays for: MM227 (WT, open squares), MM349 (∆luxO, open diamonds), MM560 (∆varA, closed black diamonds),
DL3232 (∆csrBCD, closed black circles), DL2479 (∆csrB, dark grey squares), DL3227 (∆csrC, dark grey diamonds), DL3228 (∆csrD, dark grey
circles), DL3230 (∆csrBC, light grey squares), DL3229 (∆csrBD, light grey diamonds), DL3231 (∆csrCD, light grey circles).
D. Single time point lux assays following overexpression of csrB, csrC and csrD in a V. cholerae ∆varA mutant: DL3610 (∆varA/pKK177-3RI),
DL3611 (∆varA/pKK177-3RI-csrB), DL3612 (∆varA/pKK177-3RI-csrC), DL3613 (∆varA/pKK177-3RI-csrD).

© 2005 The Authors

Journal compilation © 2005 Blackwell Publishing Ltd, Molecular Microbiology, 58, 1186–1202
 Regulation of quorum sensing in Vibrio cholerae 1193
VarA-type proteins (Valverde et al., 2003). Alignment of We wondered if the difference in the levels of VarS
the regions upstream of the three sRNA genes in and VarA regulation of csrB and csrD accounted for the
V. cholerae shows a conserved 18 bp sequence similar to distinct quorum-sensing phenotypes of the varS and
the putative GacA upstream activating sequence identi- varA mutants (Fig. 2A). If so, we predict that, in a varS
fied in the promoter region of rsmY in P. fluorescens background, deletion of csrB and csrD would confer the
(Valverde et al., 2003). In V. cholerae, the consensus varA phenotype. Only this specific deletion combination
binding site is TGTGCGAGATCTCTTACA. The binding and none of the other single or combinations of csrB,
site upstream of V. cholerae csrB is identical to the con- csrC, or csrD double mutations in conjunction with varS
sensus binding site, whereas the sites for csrC and csrD should produce the phenotype. Figure 5C shows this is
contain two and one mismatches from the consensus indeed the case. The varS single mutant is dim and the
respectively. Moreover, the above binding site is nearly csrBD double mutant displays density-dependent lux
identical (1 mismatch) to the sequence (TGTGAGAGAT expression (closed squares and closed triangles
CTCTTACA) upstream of csrB in E. coli. respectively). In contrast, the varA single mutant and
To test whether VarS/VarA regulate csrB, csrC and the triple varS, csrBD mutant have dark phenotypes
csrD expression, we constructed transcriptional reporter (closed diamonds and stars). Importantly, Fig. 5C also
fusions to each. Figure 5A shows that transcription of shows that deletion of any single csr sRNA gene, or
csrB, csrC and csrD is 200-fold higher in the wild type double deletion of csrBC or csrCD in the varS back-
than in the varS mutant. Transcription of csrC is 200-fold ground does not alter the phenotype. We propose that
higher, and transcription of csrB and csrD is 2000-fold without VarS, the remaining VarA-mediated regulation of
higher in the wild type than in the varA mutant (Fig. 5A). csrB and csrD leads to residual density-dependent lux
Thus, VarS/VarA activate transcription of csrB, csrC and expression. Taken together, the results in Fig. 5 suggest
csrD, and we hypothesize this regulation is direct. that the original quorum-sensing defect observed in the
VarS and VarA regulate csrC identically but have differ- varA mutant is due to misregulation of csrB, csrC and
ent effects on csrB and csrD. The former result signifies csrD.
that all VarA regulation of csrC is due to input from the
sensor VarS. However, in the case of csrB and csrD, VarA
VarS/VarA-CsrA/BCD controls qrr expression and
has a 10-fold stronger regulatory effect that does VarS
HapR production
indicating that a portion of the VarA regulation of csrB and
csrD is independent of VarS. This latter result suggests The present work shows that the VarS/VarA-CsrA/BCD
that VarA receives input in the form of phosphate from system operates through LuxO which controls expression
cellular component(s) in addition to VarS, and further, that of genes encoding the four Qrr sRNAs. The sRNAs control
the effect of this additional input is manifested in the the stability of hapR mRNA (Lenz et al., 2004). It follows
regulation of csrB and csrD. One explanation is that the therefore, that qrr gene expression and HapR protein pro-
csrB and csrD promoters have a higher affinity for phos- duction should be controlled by VarS/VarA-CsrA/BCD. To
pho-VarA than does the csrC promoter. This predicted verify this, we measured levels of Qrr1, Qrr2, Qrr3, Qrr4
hierarchy of promoter affinity (csrB ∼ csrD > csrC) follows and HapR protein in selected mutants (Fig. 6A). For ref-
the conservation of the respective binding sites upstream erence, we have included the wild type, luxO null strain
of these genes (csrB; no mismatches, csrD one mis- and luxO D47E strain. In wild-type cells, at high-cell den-
match, csrC two mismatches). Thus, we propose that the sity, LuxO is dephosphorylated, the qrr genes are not
phospho-VarA present in the absence of VarS acts at transcribed and hapR mRNA is stabilized (Fig. 1). Thus,
csrB and csrD, but because there is a limited supply, not under this condition very little of the Qrr sRNAs can be
at csrC. detected while high levels of HapR protein are present
To verify that the csrBCD genes indeed encode sRNAs, (Fig. 6A; WT). The luxO null strain is locked in a state
and to examine whether the differences in transcription mimicking high-cell density (Fig. 2B), and the luxO null
we observed with the reporter fusions are reflected in the profile is similar to the wild type with respect to Qrr sRNAs
levels of the Csr sRNAs, the Csr sRNAs were measured and HapR protein (Fig. 6A; luxO). The opposite pattern is
at high-cell density via steady-state Northern blots observed for the luxO D47E strain (Fig. 6A; luxO D47E).
(Fig. 5B). All three Csr RNAs are produced by the wild The LuxO D47E allele of LuxO locks LuxO into a form
type and significantly less of each is produced by the varS simulating the phosphorylated, low-cell density state.
and varA mutants. Consistent with the transcriptional V. cholerae strains carrying LuxO D47E constitutively
reporter data, the same level of CsrC RNA is produced transcribe the Qrr sRNAs, and thus hapR mRNA is always
by the varS and varA mutants, whereas the varA mutant destabilized (Lenz et al., 2004). Figure 6A shows that this
produces less CsrB and CsrD RNA than does the varS strain over-produces the Qrr sRNAs and no HapR protein
mutant. is detected.

© 2005 The Authors

Journal compilation © 2005 Blackwell Publishing Ltd, Molecular Microbiology, 58, 1186–1202
1194 D. H. Lenz et al.
Fig. 5. VarS/VarA positively regulate csrB,
A csrC and csrD expression.
10 10
A. Single time point lux assay for csrB, csrC and
10 9 csrD–lux transcriptional fusions in: WT, ∆varS
and ∆varA strains: DL3273 (WT + csrB-lux),
10 8
DL3274 (WT + csrC-lux), DL3275 (WT + csrD–
10 7 lux), DL3279 (∆varS + csrB-lux), DL3280
csrB-lux
(∆varS + csrC-lux), DL3281 (∆varS + csrD–
10 6 lux), DL3276 (∆varA + csrB-lux), DL3277
10 5 csrC-lux (∆varA + csrC-lux), DL3278 (∆varA + csrD–
RLU

lux).
10 4 B. Northern blots of RNA isolated from
csrD-lux
10 3 V. cholerae C6706str2 (WT), MM533 (∆varA)
and MM597 (∆varS) probed for the sRNAs
10 2 CsrB, CsrC and CsrD. Gels stained for riboso-
10 1 mal RNA are shown as loading controls.
C. Density-dependent lux assays for: MM609
10 0 (∆varS, closed squares), MM560 (∆varA,
WT varS varA closed diamonds), DL3229 (∆csrBD, closed tri-
Strain angles), DL3704 (∆varS, ∆csrB, open squares),
DL3705 (∆varS, ∆csrC, open diamonds),
DL3706 (∆varS, ∆csrD, open circles), DL3707
(∆varS, ∆csrBC, open triangles), DL3406
B (∆varS, ∆csrBD, stars), DL3708 (∆varS,
varA
varS

∆csrCD, plus symbols).

WT

CsrB

CsrC

CsrD

C
1011
varS
1010 varA

10 9 csrBD
RLU

varS, csrB
10 8
varS, csrC
10 7
varS, csrD
10 6
varS, csrBC
10 5 varS, csrBD

10 4 varS, csrCD
0.001 0.01 0.1 1 10

OD600

© 2005 The Authors

Journal compilation © 2005 Blackwell Publishing Ltd, Molecular Microbiology, 58, 1186–1202
 Regulation of quorum sensing in Vibrio cholerae 1195
Fig. 6. Qrr1–4 and HapR protein levels are
A regulated by VarS/VarA and CsrA/BCD.

luxO D47E

csrA::Tn5
A. RNA isolated from V. cholerae C6706str2
(WT), MM307 (∆luxO), BH38 (luxO D47E),

csrBCD
MM533 (∆varA), MM597 (∆varS), DL3183

varA
varS
luxO
(∆csrBCD) and DL2413 (csrA::Tn5) was

WT
probed for the sRNAs Qrr1, Qrr2, Qrr3 and
Qrr4. Gels stained for ribosomal RNA are
shown as loading controls. The bottom panel
shows a Western blot prepared for the same
Qrr1 strains probed with antibody to HapR.
B. Single time point lux assay following IPTG
induction of hapR in: DL3336 (WT/pJZ146),
DL3338 (∆hapR/pJZ146), DL3340 (luxO D47E/
Qrr2 pJZ146), DL3379 (∆varS/pJZ146), DL3342
(∆varA/pJZ146), DL3381 (∆csrBCD/pJZ146).

Qrr3

Qrr4

HapR

B
1010

10 9

10 8

10 7

10 6 - IPTG
RLU

10 5
+IPTG
10 4

10 3

10 2

10 1

10 0
WT hapR luxO D47E varS varA csrBCD
Strain

Figure 6A shows that, compared with the wild type, edly, Qrr4 levels appear minimally altered in all of the
increased levels of Qrr1, Qrr2 and Qrr3 are present in single mutants. qrr4 is the most highly expressed of the
varA and varS single and csrBCD triple mutants, consis- four qrr sRNA genes (Lenz et al., 2004). Most likely, LuxO-
tent with mutations in these genes promoting the low-cell P has the highest affinity for the qrr4 promoter and, at low-
density state. HapR protein is also not detected in the cell density, phospho-LuxO more fully occupies the qrr4
varA and varS mutants. A minor amount of HapR is promoter than those of qrr1–3. Thus, alterations in levels
present in the csrBCD mutant suggesting that mutation of of phospho-LuxO, either up or down, would have more
csrBCD does not completely mimic mutation of varA. In dramatic effects at the latter three promoters than at qrr4.
contrast, the csrA single mutant produces low (wild type) VarS/VarA-CsrA/BCD input into LuxO is transduced
levels of Qrr1, Qrr2 and Qrr3 and a high level of HapR into effects on qrr expression and HapR production. We
protein. This is expected given that CsrA acts inversely to therefore place the entire VarS/VarA-CsrA/BCD system
and downstream of VarS, VarA and CsrBCD. Unexpect- upstream of LuxO in the model in Fig. 1. Consistent with
© 2005 The Authors
Journal compilation © 2005 Blackwell Publishing Ltd, Molecular Microbiology, 58, 1186–1202
1196 D. H. Lenz et al.
this arrangement, Fig. 6B shows that overexpression of mentioned, causes a dark Lux phenotype. If active CsrA
hapR in the hapR, luxO D47E, varS, varA and csrBCD increases the level of LuxO, then inactivation of csrA in
mutants restores light production, demonstrating that the luxO D47E background should result in an increase
HapR is the most downstream component in the circuit. in bioluminescence because reduced LuxO would be
present. Alternatively, if active CsrA increases phospho-
LuxO activity, in the absence of CsrA, no change in lumi-
VarS/VarA-CsrA/BCD functions through LuxO
nescence output should occur in the luxO D47E mutant.
When VarS/VarA is functioning, CsrBCD are expressed, The figure shows that the luxO D47E, csrA::Tn5 double
and inhibit CsrA. Under this condition, the Lux quorum- mutant has a dark phenotype indistinguishable from the
sensing system functions normally. When VarS/VarA is luxO D47E single mutant. While we do not know the
inactivated, CsrBCD are not made, and CsrA is active. In molecular mechanism linking active CsrA to LuxO activity,
this mode, the cells are dark, implying that there is this result provides preliminary evidence supporting a
increased active LuxO. Two ways CsrA could accomplish model in which the VarS/VarA-CsrA/BCD system influ-
this are by increasing LuxO levels or increasing LuxO ences the activity of LuxO but not the absolute levels of
activity. In both cases, the effect would be to inhibit the the LuxO protein in the cell. Reporter fusions to luxO and
transition of the cells from the low to the high-cell density real-time polymerase chain reaction (PCR) were used to
state. determine the levels of luxO mRNA in the wild type varA,
In an initial attempt to distinguish between whether varS and csrBCD mutants. No significant differences in
CsrA influences LuxO levels or phospho-LuxO activity, we luxO transcript levels were observed supporting the above
examined the lux phenotype of the luxO D47E, csrA::Tn5 model (data not shown). In addition to functioning
double mutant (Fig. 7A). The luxO D47E mutation, as upstream of LuxO, CsrA could also control hapR expres-

Fig. 7. CsrA affects the activity of LuxO.

A A. Density-dependent lux assays for: MM227
(WT, open squares), BH48 (luxO D47E, open
circles), DL2493 (csrA::Tn5, open diamonds)
10 11
and DL3265 (luxO D47E, csrA::Tn5, closed
WT diamonds).
10 10
B. Western blot of HapR in whole cell lysates
10 9 prepared from high-cell density cultures of:
MM307 (∆luxO), DL2953 (∆qrr1, 2, 3, 4),
10 8 luxO D47E
RLU

MM533 (∆varA), DL2413 (csrA::Tn5), MM851

(∆luxO, ∆varA), DL2417 (∆luxO, csrA::Tn5),
10 7 DL3743 (∆qrr1, 2, 3, 4, ∆varA), DL3747 (∆qrr1,
2, 3, 4, csrA::Tn5).
10 6 csrA::Tn5
10 5

10 4
luxO D47E, csrA::Tn5
10 3
0.001 0.01 0.1 1 10

OD600

B
qrr1, 2, 3, 4, csrA::Tn5
qrr1, 2, 3, 4, varA
luxO, csrA::Tn5
qrr1, 2, 3, 4

luxO, varA
csrA::Tn5
luxO

varA

HapR

© 2005 The Authors

Journal compilation © 2005 Blackwell Publishing Ltd, Molecular Microbiology, 58, 1186–1202
 Regulation of quorum sensing in Vibrio cholerae 1197
sion directly, or at least, independently of LuxO. To test these genes that occurred at low-cell density. To verify that
this possibility, we analysed CsrA effects on HapR protein VarS/VarA-CsrA/BCD control of vpsL and vpsT expres-
levels in strains in which quorum sensing had been elim- sion is mediated through LuxO and the downstream com-
inated by mutation of luxO or deletion of the four qrr ponents of the quorum-sensing cascade, we performed
genes. Specifically, we measured HapR protein levels in an epistasis experiment by combining the luxO D47E and
the varA mutant (i.e. highly active CsrA), and the csrA::Tn5 mutations and the luxO and varA mutations. In
csrA::Tn5 mutation (inactive CsrA) in a luxO mutant and both cases, the luxO mutations were epistatic, showing
a quadruple qrr mutant. In both of these backgrounds, that indeed VarS/VarA-CsrA/BCD control of exopolysac-
hapR is constitutively expressed. Our rationale was that if charide production functions through the quorum-sensing
CsrA regulates hapR expression independently of the circuit (data not shown). The csrA::Tn5 mutation strongly
quorum sensing system, there should be different HapR reduces vpsL and vpsT expression even though this muta-
protein levels in the high and low CsrA activity back- tion has no discernible effect on lux regulation. Thus, CsrA
grounds. No such differences were observed (Fig. 7B), could have additional effects on vpsL and vpsT expression
consistent with CsrA regulating hapR expression through that are independent of quorum sensing. Finally, the hapA
the V. cholerae quorum sensing cascade. gene, encoding the H/A protease, is activated by quorum
sensing at high-cell density (Jobling and Holmes, 1997).
Mutations in the VarS/VarA-CsrA/BCD pathway affected
VarS/VarA-CsrA/BCD control the expression of vpsL
hapA expression, and the effect of each mutation was
and vpsT
opposite to that on vpsT or vpsL (not shown). Taken
Many of the above experiments make use of the non- together, it is likely that VarS/VarA-CsrA/BCD control,
native Lux target as a readout of quorum-sensing. To through the circuit shown in Fig. 1, extends to all quorum-
investigate whether VarS/VarA-CsrA/BCD control endog- sensing-regulated genes in V. cholerae.
enous V. cholerae quorum-sensing targets, we conducted
preliminary microarray analyses. These indicated that
Discussion
VarA regulates genes specifying exopolysaccharide pro-
duction (data not shown). Exopolysaccharide biosynthe- The VarS/VarA phosphorelay system, originally identified
sis is involved in biofilm formation, and in V. cholerae, for its role in V. cholerae pathogenesis (Wong et al.,
quorum sensing activates exopolysaccharide production 1998), acts in parallel to the CAI-1-CqsS and AI-2-LuxPQ
at low-cell density and represses this trait at high-cell quorum-sensing systems. VarS/VarA activate the expres-
density (Hammer and Bassler, 2003; Zhu and Mekalanos, sion of genes encoding three sRNAs, CsrB, CsrC and
2003; Yildiz et al., 2004). CsrD which, in turn, inhibit the activity of the sRNA-binding
To investigate the role of VarS/VarA-CsrA/BCD in protein CsrA. Together the VarS/VarA-CsrA/BCD regula-
exopolysaccharide production, transcriptional fusions to a tory components relay information, in a LuxO-dependent
luciferase reporter were constructed to two regulators of manner, to control the expression of the genes encoding
exopolysaccharide production, vspR (VC0665) and vpsT the four Qrr sRNAs. The Qrr sRNAs, with the sRNA chap-
(VCA0952) and to the first gene in one of the exopolysac- erone Hfq, control the stability of the mRNA encoding the
charide biosynthetic operons, vpsL (VC0934) (Yildiz and master regulator of the quorum-sensing cascade, HapR
Schoolnik, 1999; Yildiz et al., 2001; Casper-Lindley and (Fig. 1).
Yildiz, 2004). Figure 8A shows that vpsR expression is not Previously we suggested that an additional sensory
significantly affected by mutation of any lux, var, or csr system transduces information into the V. cholerae Lux
gene, whereas these signalling systems do affect vpsT circuit in a manner that bypasses LuxU but requires LuxO.
and vpsL expression (Fig. 8B and C). Unlike in our anal- This supposition was based on findings that V. cholerae
yses, an earlier experiment showed that vpsR was strains lacking the two known autoinducer detectors
modestly affected by quorum sensing (Yildiz et al., 2004). (CqsS and LuxPQ) or lacking LuxU retain density-
We find that expression of vpsT and vpsL is high in dependent gene expression while V. cholerae luxO null
V. cholerae mutants locked in low-cell density mode (luxO mutants are incapable of density-dependent gene expres-
D47E and hapR, varA, varS and the triple csrBCD sion (Miller et al., 2002). The VarS/VarA-CsrA/BCD could
mutants) and expression is significantly reduced in the be part of the unidentified sensory apparatus. Our evi-
luxO mutant, that mimics the high-cell density state, and dence shows that VarS/VarA-CsrA/BCD system acts inde-
the csrA::Tn5 mutant. This shows that VarS/VarA-CsrA/ pendently of the known autoinducer sensors (Fig. 2D),
BCD regulates expression of vpsT and vpsL. Nearly max- and signalling through VarS/VarA-CsrA/BCD requires
imal expression of the vpsT and vpsL occurs in the high- LuxO (Fig. 2B).
cell density, wild-type strain. This result is expected CsrA is an important post-transcriptional regulator in
because of the LuxO-phosphate-mediated induction of many bacteria. A model has been proposed for E. coli in

© 2005 The Authors

Journal compilation © 2005 Blackwell Publishing Ltd, Molecular Microbiology, 58, 1186–1202
1198 D. H. Lenz et al.
Fig. 8. vpsT and vpsL are regulated by the
A 109
VarSA-CsrA/BCD System.
108 Single time point lux assay for vpsR–lux tran-
scriptional fusions in: A. DL3572 (WT), DL3573
107 (∆luxO), DL3574 (luxO D47E), DL3582
(∆hapR), DL3575 (∆varA), DL3576 (∆varS),
106 DL3577 (∆csrBCD), DL3578 (csrA::Tn5).
B. vpsT–lux transcriptional fusions in: DL3479
105 (WT), DL3480 (∆luxO), DL3481 (luxO D47E),
RLU

vpsR-lux DL3482 (∆hapR), DL3483 (∆varA), DL3484

104 (∆varS), DL3485 (∆csrBCD), DL3498
(csrA::Tn5).
103
C. vpsL–lux transcriptional fusions in: DL3438
(WT), DL3543 (∆luxO), BH931 (luxO D47E),
102
BH633 (∆hapR), DL3440 (∆varA), DL3442
101
(∆varS), DL3443 (∆csrBCD), DL3491
(csrA::Tn5).
100
B 109

108

107

106

105
RLU

vpsT-lux
104

103

102

101

100
C 109

108

107

106

105
RLU

vpsL-lux
104

103

102

101

100
hapR
luxO

luxO D47E
WT

csrA::Tn5
varS

csrBCD
varA

Strain

which CsrA promotes exponential phase and prevents (Suzuki et al., 2002; Weilbacher et al., 2003). These two
stationary phase gene expression (Romeo, 1998). CsrA sRNAs have numerous AGGA-containing loops to which
functions by binding to specific sequences in the 5′ CsrA binds, which titrates CsrA away from target mRNAs.
untranslated regions of target mRNAs and altering their We suggest a role for CsrA in controlling V. cholerae
stability and/or translation. Inhibition of CsrA requires the quorum sensing, and our results show that CsrA acts
VarS/VarA homologues BarA/UvrY to activate the expres- upstream of LuxO. There is no CsrA consensus binding
sion of the genes encoding the sRNAs CsrB and CsrC site upstream of the luxO gene and mutation of csrA did
© 2005 The Authors
Journal compilation © 2005 Blackwell Publishing Ltd, Molecular Microbiology, 58, 1186–1202
 Regulation of quorum sensing in Vibrio cholerae 1199
not alter the levels of luxO transcript, suggesting that CsrA Although the VarS/A-CsrA/BCD system operates in
does not act directly on luxO expression. Rather, we pre- parallel to the two V. cholerae quorum-sensing systems,
dict that an additional component (denoted X in Fig. 1) we do not know if VarS responds to an extracellular mol-
exists and couples signalling from VarS/VarA-CsrA/BCD ecule. While several reports connect VarS/VarA-CsrA-
to LuxO. Because we do not know the molecular mecha- sRNA homologues to quorum sensing (Reimmann et al.,
nism underlying this event, we have denoted this interac- 1997; Whitehead et al., 2002), no signal has been identi-
tion with a dotted line. Under conditions where VarS/VarA fied that initiates the phosphorelay. Recently, organic
do not induce expression of csrBCD, CsrA is active and extracts of conditioned medium prepared from stationary
promotes enhanced LuxO functioning. This could be phase cultures of Pseudomonas fluorescens were shown
accomplished through CsrA-mediated regulation of some to contain an activity capable of modest activation of
factor, X, whose function is to increase the concentration GacS/GacA-regulated targets (Valverde et al., 2003;
or activity of phospho-LuxO. The component X could be Zuber et al., 2003). This finding suggests that GacS could
a kinase, a small molecule phospho-donor, or a regulatory respond to an extracellular molecule that is produced or
component that controls the expression or activity of a accumulates in stationary phase. If so, and if this holds
phospho-donor. An obvious candidate for the phospho- true for VarS, it would be especially exciting because the
donor is acetyl-phosphate. However, inactivation of the remainder of the quorum-sensing regulators in vibrios
genes responsible for acetyl-phosphate production in and in pseudomonads share no similarity. We found that
V. cholerae (pta, ack1 and ack2) in the varA background csrB, csrC and csrD transcription is induced approxi-
did not have any effect on light production (data not mately 50-fold from low- to high-cell density. Thus, CsrB,
shown) suggesting that if some small molecule phospho- CsrC and CsrD levels are growth phase regulated. How-
donor is responsible, it is not acetyl-phosphate. Although ever, we were unable to induce their expression at low-
we have preliminary evidence indicating that VarS/VarA- cell density by the addition of cell-free culture fluids
CsrA/BCD primarily functions through LuxO to control collected from wild-type V. cholerae grown to high-cell
hapR (Fig. 7B), we have not fully eliminated the possibility density.
that, in addition, VarS/VarA-CsrA/BCD inputs information A final, but important point regarding the complexity of
into the quorum-sensing cascade at other points. the circuit shown in Fig. 1 comes from examination of the
Bioinformatics, overexpression and deletion analysis phenotypes of some of the mutants shown in Fig. 2C and
revealed that the link between VarS/VarA and CsrA in D. We observe that the phenotype of a luxU mutant is not
V. cholerae is three sRNAs we name CsrB, CsrC and identical to that of the double cqsS, luxQ mutant. We find
CsrD. Finding three CsrB-type sRNAs is surprising that luxU mutants are consistently 100-times brighter than
because this level of redundancy has not been observed cqsS, luxQ double mutants. This result implies that an
in other bacterial systems in which CsrA regulation has additional sensory input(s) is funnelled into LuxU. This
been investigated. We suspect that VarA binds upstream input cannot be attributed to VarS/A-CsrA/BCD because
of each of the csr sRNA genes based on transcriptional that system feeds information into the circuit at the level
fusions showing that VarS/VarA activate their expression of LuxO. We therefore conclude that, AI-1-CqsS is System
200- to 2000-fold combined with an analysis of their 1, AI-2-LuxPQ is System 2, VarS/A-CsrA/BCD is part of
upstream DNA sequences showing that a site very System 3, and there must also exist a System 4 that relays
similar to the putative upstream activating sequence in information through LuxU, and thus must influence the
Pseudomonas spp. (Valverde et al., 2003) exists activity of LuxO, the levels of the Qrrs and HapR, and, in
upstream of each. We believe that our analysis enabled turn, the expression of all the genes that lie downstream
identification of the entire suite of sRNAs involved in con- of HapR in the hierarchy.
trolling CsrA activity in V. cholerae because the triple csr- The molecular architecture controlling V. cholerae quo-
BCD mutant has a phenotype indistinguishable from the rum sensing requires at least three signalling channels,
varA mutant (Fig. 4C). We also know that csrBCD act and at a minimum, seven small regulatory RNAs (four
redundantly because the triple csrBCD mutant affects Qrr and three Csr sRNAs). The circuit is highly redundant
HapR protein production (Fig. 6A), whereas the single in terms of the extracellular signals and the signal trans-
mutants and all combinations of double mutants do not duction machinery. This degree of overlap probably exists
(data not shown). While redundancy in sRNA function has because high fidelity information flow is required, and the
been observed in other CsrA-type systems, unlike in mechanism V. cholerae has evolved for ensuring accu-
V. cholerae, none of the previously described systems is racy relies on extreme functional redundancy. If the sig-
known to have greater than two sRNAs that act on CsrA. nal detected by VarS/VarA accumulates in stationary
However, in some cases, the complete set of sRNAs have phase, this could be the link connecting growth phase to
probably not yet been identified, preventing the absolute community number and composition. Funnelling all the
definition of their regulatory roles. information into one integrated circuit could provide

© 2005 The Authors

Journal compilation © 2005 Blackwell Publishing Ltd, Molecular Microbiology, 58, 1186–1202
1200 D. H. Lenz et al.
V. cholerae a combinatorial mechanism for broadly prob- was introduced into V. cholerae strains by conjugation. Single
ing the environment for fluctuations and responding time point bioluminescence assays were performed at high-
accurately. cell density in triplicate.

Experimental procedures Western blot analysis

Western blots to monitor HapR protein were performed as in
Bacterial strains and media
Lenz et al. (2004). Strains were grown to high-cell density.
Vibrio cholerae strains are derived from the El Tor strain
C6706str2 (Thelin and Taylor, 1996). Details are provided in
figure legends. V. cholerae strains were grown with aeration Northern blot analysis
at 30°C in SOC broth (Sambrook et al., 1989) for lux assays
Steady-state Northern blots were as reported (Lenz et al.,
and in Luria–Bertani (LB) broth for other experiments.
2004) except that cultures were grown to OD600 of 2 prior to
Plasmids were maintained at 37°C in LB in E. coli JM109
RNA preparation and twice as much RNA was loaded. Gels
(Yanisch-Perron et al., 1985) or E. coli S17-1λpir (de Lorenzo
containing ribosomal RNA stained with ethidium bromide
and Timmis, 1994). Antibiotics concentrations used are:
prior to transfer are shown as loading controls.
ampicillin 100 µg ml−1; tetracycline 10 µg ml−1; kanamycin
100 µg ml−1; chloramphenicol 10 µg ml−1; streptomycin 1–
5 mg ml−1; and polymixin B 50 units ml−1. Genetic screen to identify varS/varA and csrA
The Tn5 mutagenesis and analysis yielding insertions in varS
DNA manipulations and varA has been described (Miller et al., 2002). Bypass
DNA manipulations were performed according to Sambrook suppressors were obtained by introduction of the pRL27c Tn5
et al. (1989). Deletions were constructed according to Sko- suicide vector (Larsen et al., 2002) into the varA in-frame
rupski and Taylor (1996). csrA::Tn5 strains were constructed deletion strain MM533. Kanamycin resistant colonies were
by PCR amplification of the csrA::Tn5 region, cloning into pooled, and pBB1 conjugated into them. Exconjugants were
pKAS32 vector (Skorupski and Taylor, 1996) and gene screened for a bright phenotype. The Tn5-chromosome junc-
replacement onto the chromosome with kanamycin selection. tions were characterized as described (Miller et al., 2002).
PCR reactions for cloning used PFU turbo polymerase (Strat-
agene); other PCR reactions used Taq polymerase (Roche).
dNTPs, restriction endonucleases and T4 ligase were pur-
Acknowledgements
chased from New England Biolabs. DNA purification was This work was supported by HHMI and NSF Grant MCB-
performed by QIAGEN methods. Primer sequences are avail- 0343821, NIH Grant 5RO1 G065859, ONR Grant N00014-
able by request. The csrB, csrC and csrD overexpression 03-0183 and an HHMI Predoctoral Fellowship (D.H.L.), The
plasmids were constructed in pKK177-3RI as in Lenz et al. Jeffress Foundation (R.V.K.). We thank the Bassler Lab, Dr
(2004), and were introduced into V. cholerae by electropora- T. Silhavy and Dr N. Wingreen for insightful discussions.
tion. The lux transcriptional fusion plasmids were constructed
as in Lenz et al. (2004) and introduced into V. cholerae by
conjugation. The hapR overexpression construct pJZ146 References
came from Zhu et al. (2002) and was induced with 0.5 mM
IPTG. Aarons, S., Abbas, A., Adams, C., Fenton, A., and O’Gara,
F. (2000) A regulatory RNA (PrrB RNA) modulates expres-
sion of secondary metabolite genes in Pseudomonas flu-
Bioinformatics orescens F113. J Bacteriol 182: 3913–3919.
Bassler, B.L., Wright, M., Showalter, R.E., and Silverman,
The locations of csrBCD in the V. cholerae genome were
M.R. (1993) Intercellular signalling in Vibrio harveyi:
identified using BLAST at The Institute for Genomic Research
sequence and function of genes regulating expression of
(TIGR, http://www.tigr.org). Multiple alignments were carried
luminescence. Mol Microbiol 9: 773–786.
out with TCOFFEE (Poirot et al., 2003). MFOLD was used
Bassler, B.L., Wright, M., and Silverman, M.R. (1994) Multi-
for secondary structure predictions (Jacobson and Zuker,
ple signalling systems controlling expression of lumines-
1993; Zuker, 2003). Default parameters were used for the
cence in Vibrio harveyi: sequence and function of genes
folding predictions, and the structure with the lowest free
encoding a second sensory pathway. Mol Microbiol 13:
energy is shown. To avoid spurious loop formation in the
273–286.
secondary structure, we set a cut-off of 40 for the maximum
Burrowes, E., Abbas, A., O’Neill, A., Adams, C., and O’Gara,
distance between paired bases.
F. (2005) Characterisation of the regulatory RNA RsmB
from Pseudomonas aeruginosa PAO1. Res Microbiol 156:
Bioluminescence assays 7–16.
Casper-Lindley, C., and Yildiz, F.H. (2004) VpsT is a tran-
Vibrio cholerae bioluminescence assays were performed as scriptional regulator required for expression of vps biosyn-
in Lenz et al. (2004). pBB1, containing V. harveyi luxCDABE, thesis genes and the development of rugose colonial

© 2005 The Authors

Journal compilation © 2005 Blackwell Publishing Ltd, Molecular Microbiology, 58, 1186–1202
 Regulation of quorum sensing in Vibrio cholerae 1201
morphology in Vibrio cholerae O1 El Tor. J Bacteriol 186: de Lorenzo, V., and Timmis, K.N. (1994) Analysis and con-
1574–1578. struction of stable phenotypes in Gram-negative bacteria
Chen, X., Schauder, S., Potier, N., Van Dorsselaer, A., Pel- with Tn5- and Tn10-derived minitransposons. Methods
czer, I., Bassler, B.L., and Hughson, F.M. (2002) Structural Enzymol 235: 386–405.
identification of a bacterial quorum-sensing signal contain- Miller, M.B., and Bassler, B.L. (2001) Quorum sensing in
ing boron. Nature 415: 545–549. bacteria. Annu Rev Microbiol 55: 165–199.
Freeman, J.A., and Bassler, B.L. (1999) A genetic analysis Miller, M.B., Skorupski, K., Lenz, D.H., Taylor, R.K., and
of the function of LuxO, a two-component response regu- Bassler, B.L. (2002) Parallel quorum sensing systems con-
lator involved in quorum sensing in Vibrio harveyi. Mol verge to regulate virulence in Vibrio cholerae. Cell 110:
Microbiol 31: 665–677. 303–314.
Freeman, J.A., and Bassler, B.L. (2000) Sequence and func- Molofsky, A.B., and Swanson, M.S. (2004) Differentiate to
tion of LuxU: a two-component phosphorelay protein that thrive: lessons from the Legionella pneumophila life cycle.
regulates quorum sensing in Vibrio harveyi. J Bacteriol Mol Microbiol 53: 29–40.
181: 899–906. Nealson, K.H., and Hastings, J.W. (1979) Bacterial biolumi-
Freeman, J.A., Lilley, B.N., and Bassler, B.L. (2000) A nescence: its control and ecological significance. Microbiol
genetic analysis of the functions of LuxN: a two-component Rev 43: 496–518.
hybrid sensor kinase that regulates quorum sensing in Pernestig, A.K., Melefors, O., and Georgellis, D. (2001) Iden-
Vibrio harveyi. Mol Microbiol 35: 139–149. tification of UvrY as the cognate response regulator for the
Hammer, B.K., and Bassler, B.L. (2003) Quorum sensing BarA sensor kinase in Escherichia coli. J Biol Chem 276:
controls biofilm formation in Vibrio cholerae. Mol Microbiol 225–231.
50: 101–104. Poirot, O., O’Toole, E., and Notredame, C. (2003)
Heeb, S., and Haas, D. (2001) Regulatory roles of the GacS/ Tcoffee@igs: a web server for computing, evaluating and
GacA two-component system in plant-associated and combining multiple sequence alignments. Nucleic Acids
other gram-negative bacteria. Mol Plant Microbe Interact Res 31: 3503–3506.
14: 1351–1363. Reimmann, C., Beyeler, M., Latifi, A., Winteler, H., Foglino,
Heeb, S., Blumer, C., and Haas, D. (2002) Regulatory RNA M., Lazdunski, A., and Haas, D. (1997) The global activator
as mediator in GacA/RsmA-dependent global control of GacA of Pseudomonas aeruginosa PAO positively controls
exoproduct formation in Pseudomonas fluorescens CHA0. the production of the autoinducer N-butyryl-homoserine
J Bacteriol 184: 1046–1056. lactone and the formation of the virulence factors pyocya-
Hrabak, E.M., and Willis, D.K. (1992) The lemA gene nin, cyanide, and lipase. Mol Microbiol 24: 309–319.
required for pathogenicity of Pseudomonas syringae Rich, J.J., Kinscherf, T.G., Kitten, T., and Willis, D.K. (1994)
pv. syringae on bean is a member of a family of two- Genetic evidence that the gacA gene encodes the cognate
component regulators. J Bacteriol 174: 3011–3020. response regulator for the lemA sensor in Pseudomonas
Jacobson, A.B., and Zuker, M. (1993) Structural analysis by syringae. J Bacteriol 176: 7468–7475.
energy dot plot of a large mRNA. J Mol Biol 233: 261–269. Romeo, T. (1998) Global regulation by the small RNA-binding
Jobling, M.G., and Holmes, R.K. (1997) Characterization of protein CsrA and the non-coding RNA molecule CsrB. Mol
hapR, a positive regulator of the Vibrio cholerae HA/ Microbiol 29: 1321–1330.
protease gene hap, and its identification as a functional Romeo, T., Gong, M., Liu, M.Y., and Brun-Zinkernagel,
homologue of the Vibrio harveyi luxR gene. Mol Microbiol A.M. (1993) Identification and molecular characterization
26: 1023–1034. of csrA, a pleiotropic gene from Escherichia coli that
Larsen, R.A., Wilson, M.M., Guss, A.M., and Metcalf, W.W. affects glycogen biosynthesis, gluconeogenesis, cell
(2002) Genetic analysis of pigment biosynthesis in Xantho- size, and surface properties. J Bacteriol 175: 4744–
bacter autotrophicus Py2 using a new, highly efficient 4755.
transposon mutagenesis system that is functional in a wide Sambrook, J., Fritsch, E.F., and Maniatis, T. (1989) Molecu-
variety of bacteria. Arch Microbiol 178: 193–201. lar Cloning: a Laboratory Manual. Cold Spring Harbor, NY:
Laville, J., Voisard, C., Keel, C., Maurhofer, M., Defago, G., Cold Spring Harbor Laboratory Press.
and Haas, D. (1992) Global control in Pseudomonas fluo- Skorupski, K., and Taylor, R.K. (1996) Positive selection vec-
rescens mediating antibiotic synthesis and suppression of tors for allelic exchange. Gene 169: 47–52.
black root rot of tobacco. Proc Natl Acad Sci USA 89: Suzuki, K., Wang, X., Weilbacher, T., Pernestig, A.K., Mele-
1562–1566. fors, O., Georgellis, D., et al. (2002) Regulatory circuitry of
Lenz, D.H., Mok, K.C., Lilley, B.N., Kulkarni, R.V., Wingreen, the CsrA/CsrB and BarA/UvrY systems of Escherichia coli.
N.S., and Bassler, B.L. (2004) The small RNA chaperone J Bacteriol 184: 5130–5140.
Hfq and multiple small RNAs control quorum sensing in Thelin, K.H., and Taylor, R.K. (1996) Toxin-coregulated pilus,
Vibrio harveyi and Vibrio cholerae. Cell 118: 69–82. but not mannose-sensitive hemagglutinin, is required for
Lilley, B.N., and Bassler, B.L. (2000) Regulation of quorum colonization by Vibrio cholerae O1 El Tor biotype and O139
sensing in Vibrio harveyi by LuxO and sigma-54. Mol strains. Infect Immun 64: 2853–2856.
Microbiol 36: 940–954. Valverde, C., Heeb, S., Keel, C., and Haas, D. (2003) RsmY,
Liu, M.Y., Gui, G., Wei, B., Preston, J.F., 3rd, Oakford, L., a small regulatory RNA, is required in concert with RsmZ
Yuksel, U., et al. (1997) The RNA molecule CsrB binds to for GacA-dependent expression of biocontrol traits in
the global regulatory protein CsrA and antagonizes its Pseudomonas fluorescens CHA0. Mol Microbiol 50: 1361–
activity in Escherichia coli. J Biol Chem 272: 17502–17510. 1379.

© 2005 The Authors

Journal compilation © 2005 Blackwell Publishing Ltd, Molecular Microbiology, 58, 1186–1202
1202 D. H. Lenz et al.
Weilbacher, T., Suzuki, K., Dubey, A.K., Wang, X., Gudapaty, VpsR, a Member of the Response Regulators of the Two-
S., Morozov, I., et al. (2003) A novel sRNA component of Component Regulatory Systems, Is Required for Expres-
the carbon storage regulatory system of Escherichia coli. sion of vps Biosynthesis Genes and EPS (ETr) -Associated
Mol Microbiol 48: 657–670. Phenotypes in Vibrio cholerae O1 El Tor. J Bacteriol 183:
Whitehead, N.A., Byers, J.T., Commander, P., Corbett, M.J., 1716–1726.
Coulthurst, S.J., Everson, L., et al. (2002) The regulation Yildiz, F.H., Liu, X.S., Heydorn, A., and Schoolnik, G.K.
of virulence in phytopathogenic Erwinia species: quorum (2004) Molecular analysis of rugosity in a Vibrio chol-
sensing, antibiotics and ecological considerations. Antonie erae O1 El Tor phase variant. Mol Microbiol 53: 497–
Van Leeuwenhoek 81: 223–231. 515.
Wong, S.M., Carroll, P.A., Rahme, L.G., Ausubel, F.M., and Zhu, J., and Mekalanos, J.J. (2003) Quorum sensing-
Calderwood, S.B. (1998) Modulation of expression of the dependent biofilms enhance colonization in Vibrio chol-
ToxR regulon in Vibrio cholerae by a member of the two- erae. Dev Cell 5: 647–656.
component family of response regulators. Infect Immun 66: Zhu, J., Miller, M.B., Vance, R.E., Dziejman, M., Bassler,
5854–5861. B.L., and Mekalanos, J.J. (2002) Quorum-sensing regula-
Yanisch-Perron, C., Vieira, J., and Messing, J. (1985) tors control virulence gene expression in Vibrio cholerae.
Improved M13 phage cloning vectors and host strains: Proc Natl Acad Sci USA 99: 3129–3134.
nucleotide sequences of the M13mp18 and pUC19 vec- Zuber, S., Carruthers, F., Keel, C., Mattart, A., Blumer, C.,
tors. Gene 33: 103–119. Pessi, G., et al. (2003) GacS sensor domains pertinent to
Yildiz, F.H., and Schoolnik, G.K. (1999) Vibrio cholerae O1 the regulation of exoproduct formation and to the biocontrol
El Tor: identification of a gene cluster required for the potential of Pseudomonas fluorescens CHA0. Mol Plant
rugose colony type, exopolysaccharide production, chlo- Microbe Interact 16: 634–644.
rine resistance, and biofilm formation. Proc Natl Acad Sci Zuker, M. (2003) Mfold web server for nucleic acid folding
USA 96: 4028–4033. and hybridization prediction. Nucleic Acids Res 31: 3406–
Yildiz, F.H., Dolganov, N.A., and Schoolnik, G.K. (2001) 3415.

© 2005 The Authors

Journal compilation © 2005 Blackwell Publishing Ltd, Molecular Microbiology, 58, 1186–1202