APPLIED AND ENVIRONMENTAL MICROBIOLOGY, Oct. 2011, p. 7227–7235
Copyright © 2011, American Society for Microbiology. All Rights Reserved.
Vol. 77, No. 20
Spontaneous Gac Mutants of Pseudomonas Biological Control
Strains: Cheaters or Mutualists??
William W. Driscoll,1John W. Pepper,1,2Leland S. Pierson III,3and Elizabeth A. Pierson4*
Department of Ecology and Evolutionary Biology, University of Arizona, Tucson, Arizona1; Santa Fe Institute,
Santa Fe, New Mexico2; and Departments of Plant Pathology and Microbiology3and Horticultural Sciences,4
Texas A&M University, College Station, Texas
Received 25 March 2011/Accepted 18 August 2011
Bacteria rely on a range of extracellular metabolites to suppress competitors, gain access to resources, and
exploit plant or animal hosts. The GacS/GacA two-component regulatory system positively controls the
expression of many of these beneficial external products in pseudomonad bacteria. Natural populations often
contain variants with defective Gac systems that do not produce most external products. These mutants benefit
from a decreased metabolic load but do not appear to displace the wild type in nature. How could natural
selection maintain the wild type in the presence of a mutant with enhanced growth? One hypothesis is that Gac
mutants are “cheaters” that do not contribute to the public good, favored within groups but selected against
between groups, as groups containing more mutants lose access to ecologically important external products. An
alternative hypothesis is that Gac mutants have a mutualistic interaction with the wild type, so that each
variant benefits by the presence of the other. In the biocontrol bacterium Pseudomonas chlororaphis strain
30-84, Gac mutants do not produce phenazines, which suppress competitor growth and are critical for biofilm
formation. Here, we test the predictions of these alternative hypotheses by quantifying interactions between the
wild type and the phenazine- and biofilm-deficient Gac mutant within growing biofilms. We find evidence that
the wild type and Gac mutants interact mutualistically in the biofilm context, whereas a phenazine-defective
structural mutant does not. Our results suggest that the persistence of alternative Gac phenotypes may be due
to the stabilizing role of local mutualistic interactions.
In recent years, it has become clear that microbes are pro-
foundly cooperative organisms (4, 29, 44, 45). In particular,
much recent work has focused on the production of beneficial
extracellular secondary metabolites, apparently for the “public
good.” Such external products are often crucial to the success
of microbial populations in natural, medical, agricultural, and
industrial settings. The terms “cheater” and “defector” have
been used to describe mutants that benefit from these external
products but do not produce or underproduce their share (11,
12, 19, 21, 36). Cheaters are evolutionarily significant because
they illustrate a classic paradox in social evolution, whereby
cheaters are favored over their cooperative counterparts within
groups, yet altruists persist in the population (30, 46, 47). This
suggests a conflict of selection at alternate levels, with groups
producing more external products outreproducing those that
do not, potentially countering the within-group (local) advan-
tage of cheating (29, 34). Furthermore, in the case of strains
that utilize external products in pathogenesis or biocontrol
functions, cheaters may reduce virulence or competitive ability
and thus have applied significance (6, 14, 19, 38).
The GacS/GacA two-component system positively regulates
the expression of many extracellular enzymes, secondary me-
tabolites, and some carbon storage compounds as well as oxi-
dative stress response and other functions that play roles in
biological control or pathogenicity (hence the name gac, for
global activator) (3, 7, 15). The GacS/GacA regulatory system
is highly conserved among Gram-negative bacteria, including
many beneficial biological control and plant pathogenic bac-
teria. The GacS/GacA global regulatory system is prevalent
among Pseudomonas species, and homologous systems have
been identified in other genera, including Escherichia, Vibrio,
Legionella, Erwinia, Azotobacter, and Salmonella (16). Mecha-
nistically, the GacS sensor kinase autophosphorylates and ac-
tivates the GacA response regulator by phosphorylation in
response to as yet unidentified signals or stimuli. Global reg-
ulation by GacS/GacA in many Pseudomonas species occurs via
transcriptional activation of one to several genes for small
noncoding RNAs (ncRNAs). These ncRNAs subsequently se-
quester RsmA and RsmE, small RNA-binding proteins, reliev-
ing translational repression by these regulatory proteins (re-
viewed in references 41 and 42). GacS/GacA also negatively
regulates a spectrum of traits by mechanisms largely unidenti-
fied. For example, Hassan et al. (15) showed that inactivation
of GacS/GacA resulted in the altered expression of 10% of the
genes in Pseudomonas fluorescens strain Pf-5. Phenotypic vari-
ants with mutations disrupting gacA or gacS have identical
phenotypes: loss of production of many positively regulated
external products (e.g., exoproteases, antibiotics, quorum-sens-
ing signals, and exotoxins ) as well as overproduction of
some negatively regulated metabolites (e.g., siderophores [2,
35]), although other cell-intrinsic traits (e.g., flagella) are also
Many Pseudomonas species identified for biological control
exhibit phenotypic variation resulting from spontaneous muta-
tion in gacS or gacA (reviewed in reference 41). These mutants
are hypothesized to have a reduced metabolic load compared
* Corresponding author. Mailing address: Department of Horticul-
tural Sciences, 202 Horticulture and Forestry Sciences Building, Texas
A&M University, College Station, TX 77843-2133. Phone: (979) 862-
1307. Fax: (979) 845-0627. E-mail: email@example.com.
?Published ahead of print on 26 August 2011.
with the wild type (39, 41), and consistent with this hypothesis,
Gac mutants generally have a growth advantage compared
with the wild type (41). In nutrient-rich fermentation cultures,
spontaneous Gac mutants often become the majority popula-
tion (6). This has been a cause for significant concern in the
development of biological control because the occurrence of
mutants in production cultures results in the loss of secondary
metabolites essential for their biological control activity (2, 3,
6, 15). Natural rhizosphere populations also frequently contain
variants with defective Gac systems (3, 6, 39). For these rea-
sons, Jousset et al. (19) recently hypothesized that Gac mutants
are cheaters, favored by within-group selection but selected
against at the group level.
An alternative hypothesis is that Gac mutants are mutualis-
tic or commensalistic in their interactions with the wild type in
natural settings (2, 39, 41). This view is contingent primarily on
observations that while Gac mutants are frequently observed in
nature, they do not appear to displace wild-type strains in
natural environments. Better performance of wild-type popu-
lations in a mixture with mutants suggests that Gac mutants
may interact mutualistically, rather than parasitically, with the
wild type in the rhizosphere (2, 39). Both parasitic and mutu-
alistic hypotheses predict that Gac mutants benefit from some
extracellular product(s) secreted by the wild type, but offer
contrasting predictions about the impact of the Gac mutant on
the wild type. The parasitic hypothesis suggests that between-
group selection favors the wild type, so that Gac mutants im-
pose fitness costs on neighbors. Specifically, groups containing
greater frequencies of the wild type are more productive
(higher fitness) than groups of their counterparts containing
fewer wild-type individuals, despite local (within-group) selec-
tion favoring Gac mutants. In contrast, if Gac mutants are
mutualists, mixed groups should have higher fitness than sin-
gle-strain groups. Exploitation of external products—particu-
larly when such products are central to survival—requires that
the cheater and mutualist co-occur in relatively close proxim-
ity, because diffusion-mediated transfer of extracellular prod-
ucts imposes a physical barrier to long-distance interactions
(10). Thus, diffusion of metabolites defines the neighborhood
within which parasitic or mutualistic interactions may occur.
For this study, we employed the biological control strain
Pseudomonas chlororaphis 30-84. Strain 30-84 is an effective
biological control of take-all disease of wheat caused by the
fungal pathogen Gaeumannomyces graminis var. tritici and has
become a useful system for studying biocontrol (32). GacS/
GacA controls the production of a number of secondary
metabolites shown to be important in biological activities,
including orange-pigmented phenazines, hydrogen cyanide,
gelatinase, lipase, and exoprotease. Phenazines are responsible
for the majority of strain 30-84’s ability to inhibit fungal patho-
gens (31, 40), but they are also required for biofilm formation
in this strain (22, 23). In laboratory flow cell assays, phenazine-
deficient mutants were unable to establish biofilms; however,
biofilm formation was restored by supplying mutants with
In order to differentiate between the parasitic and mutual-
istic hypotheses, we observed and quantified the interactions
between wild-type strain 30-84, 30-84GacA, a GacA mutant
derivative, and 30-84ZN, a phenazine biosynthetic mutant,
using differentially fluorescently labeled strains in growing
biofilms. Because biofilm formation is central to the fitness
of rhizosphere bacteria (27, 28) (whereas anticompetitor
functionality may vary in importance across diverse environ-
ments), we focus on biofilm formation as a proxy for fitness
MATERIALS AND METHODS
Bacterial strains and media. The bacterial strains and plasmids used in this
study are described in Table 1. A spontaneous rifampin-resistant derivative of P.
chlororaphis strain 30-84 was used in all studies (33), and all mutants were
derived from this parental strain. Strain 30-84ZN contains a phzB::lacZ genomic
fusion (phzB is one of the five genes in a single operon necessary for phenazine
biosynthesis) and produces no phenazines (Table 1). Strain 30-84GacA contains
a gacA::Kmrgenomic fusion (Table 1), completely inactivating gacA. Triparental
conjugations into strain 30-84 or its derivates were performed as described
previously (31). All strains of P. chlororaphis were grown at 28°C in LB medium
supplemented with 5 g of NaCl/liter or in AB minimal medium supplemented
with 2% Casamino Acids (AB ? CAA) (22). Escherichia coli strains were grown
at 37°C in LB. Antibiotics were used where appropriate at the following con-
centrations: for E. coli, ampicillin at 100 ?g/ml, kanamycin sulfate (Km) at 50
?g/ml, and chloramphenicol (Cm) at 30 ?g/ml; for P. chlororaphis, Km at 50
?g/ml, Cm at 300 ?g/ml, and rifampin at 100 ?g/ml (22).
Construction of fluorescent derivatives. To visualize P. chlororaphis strain
30-84 derivatives during biofilm development, we constructed plasmids pKI2CM-
EGFP and pKI2CM-DsRed carrying either an enhanced variant of the Aequorea
victoria green fluorescent protein (EGFP) or the Discosoma sp. red fluorescent
protein (DsRed-Express), respectively. The expression of both is under the
control of the lac promoter. The 974-bp PvuII-EcoRI fragment carrying the
promoter plus EGFP and the 930-bp PvuII-EcoRI fragment carrying the pro-
moter plus DsRed-Express were isolated from the commercial vectors pEGFP
and pDsRed-Express (Clontech), respectively, gel purified, treated with Klenow
fragment, and inserted into the BamHI site of pProbe-KI? (also Klenow frag-
ment treated). pProbe-KI? contains the pVS1 and p15a replicons (25) and has
been shown to be highly stable without selection for more than 30 generations in
strain 30-84 (22). A 4.3-kb EcoRI fragment carrying cat (Cmr) was introduced
into the EcoRI site of pProbe-KI? prior to the insertion of the fluorescent protein
gene (Table 1). Each plasmid was introduced into strain 30-84 or its derivatives
via triparental conjugation.
Growth curves in liquid culture. Overnight cultures of strains 30-84red, 30-
84GacAgreen, and 30-84ZNgreen were grown in 5 ml LB medium. Cultures were
adjusted to an optical density at 620 nm (OD620) of 0.4 using AB ? CAA, and
10 ?l of each dilute culture was added to 20 ml AB ? CAA. Each treatment was
replicated three times. Three OD620values were taken per biological replicate,
once every 1 to 3 h during the exponential phase of growth (between 10 h and
25 h). Additional measurements at 32 h and 35 h confirmed that the cultures had
entered stationary phase. A logistic curve was fitted to these OD620time series
data and then linearized via logistic transformation. Regression analysis was used
to compare the observed data to those predicted by a logistic model. In all cases,
the coefficient of determination (R2) was 0.95 or greater. The intrinsic rate of
growth and doubling time for each strain were estimated from the slope of the
Flow cell experiments. Single-pass flow cell assays were used to visually com-
pare biofilm formation by strain 30-84 and its derivatives as described previously
(22), with slight modifications. Briefly, inocula were prepared from exponential-
phase cultures grown in 5 ml LB. These cultures were centrifuged and washed
with AB ? CAA and then diluted (OD600of 0.4) with AB ? CAA. For each
treatment, a 300-?l aliquot of dilute culture was inoculated into individual flow
cell chambers (4 by 40 by 1 mm) of a three-chamber flow cell (Stovall Life
Sciences, Inc., Greensboro, NC). To ensure sterile conditions, kanamycin (50
?g/ml) was added to the medium. After inoculation, chambers were inverted and
maintained under static conditions (no flow) for 1 h to allow cell attachment.
Following this, the chamber was placed right side up, and a continuous flow of
fresh medium (160 ?l/min) was initiated using a 12-channel Ismatec pump
(Ismatec SA, Labortechnik-Analytik, Glattbrugg, Switzerland). Biofilms were
visualized using a Nikon E800 confocal microscope equipped with argon and
green HeNe lasers. Images were collected at 6, 12, and 24 h after inoculation and
every 24 h thereafter for 5 days. Images were collected using Simple PCI software
(Compix, Inc., Cranberry Township, PA). Each experiment was repeated at least
Image analysis and calculation of biofilm characteristics. Three-dimensional
images were assembled in ImageJ (1), and the ImageJ-based program
7228 DRISCOLL ET AL.APPL. ENVIRON. MICROBIOL.
COMSTAT 2 (18; M. Vorregaard, B. K. Ersbøll, L. Yang, J. A. J. Haagensen,
A. Heydorn, S. Molin, and C. Sternberg, personal communication) was used to
acquire biofilm data. Biovolume data were collected from between 8 and 10
independent stacks of images across two separate biofilms per sample in the
cases of the 30-84red-plus-30-84green and 30-84red-plus-30-84GacAgreen bio-
films presented here, while between 7 and 9 stacks were used in the case of
30-84red-plus-30-84ZNgreen biofilms. In treatments in which biofilms failed to
establish, two or three stacks were analyzed per time point. Structural charac-
teristics of biofilms were calculated from at least five independent stacks taken
after 96 h for each treatment. Images were also collected of reciprocally labeled
strains in mixed biofilms to confirm strain behavior.
We used the normalized (mean removed) cross-covariance between the red
and green channels in order to quantify the degree of clustering between oppo-
sitely labeled strains in wild-type-only and wild-type-plus-30-84GacA mixed bio-
films. We obtained two-dimensional maximum brightness vertical projections for
five independent stacks of images of the mature biofilm taken at h 96. We
measured the average covariance, c, between a focal pixel, x, in the red channel
and a pixel in the green channel within the same row at a distance, d, where 0 ?
d ? 100 (? 47 ?m). The covariance (Cov) is given by
Covdy?Gy, Ry? ?
n ? 1?
n??Gx?d,y? G?y??Rx,y? R?y??
where bars denote average values for an entire row y. This yields the mean
covariance taken across all points n in a row, y, for the set of lag distances, d, from
0 to 100. The vector of average covariances at each lag distance, cd, is
m ? 1?
This was repeated four times for each projection, with a 90° rotation in each case,
ensuring no spatial biases due to sampling with respect to flow direction. Four
direction-specific vectors of mean c values were obtained from all five stacks for
each treatment. Finally, the grand means and standard errors were calculated
across the four cardinal directions, yielding a single vector of cdvalues for each
Replacement series. We determined the fitness of wild-type strain 30-84,
30-84GacA, and 30-84ZN in mixed biofilms by introducing strains in a replace-
ment series containing specific proportions of wild-type 30-84red and either
strain 30-84green (control), 30-84GacAgreen, or 30-84ZNgreen (fractions of
1.0:0, 0.5:0.5, and 0:1.0, respectively, keeping the total inoculation density con-
stant). Briefly, a replacement series analysis is intended to highlight interactions
between co-occurring varieties by generating a baseline additive expectation of
performance based on each variety in isolation. Deviations from this expected
performance indicate an interaction between varieties. For example, a strain that
grows to an average biovolume of 1 ?m3/?m2in isolation is expected to grow to
0.5 ?m3/?m2when introduced in a 1:1 ratio with another strain; the observation
that it actually achieves a biovolume of only 0.1 ?m3/?m2may suggest a com-
petitive or parasitic interaction.
Expected values were calculated on the basis of the biofilm biovolume of each
strain inoculated alone and the proportion of the wild type and mutants in the
original inoculum for each treatment after 96 h. Comparisons of observed bio-
volume in a 50:50 mixture to expected biovolume were used to assess whether
individual strain fitness is enhanced or degraded in a mixture compared to their
predicted biofilm development given a 50% lower introduction density. Biovol-
ume values were calculated using at least five image stacks from a single biofilm
for 30-84red-plus-30-84green and 30-84red-plus-30-84GacAgreen populations
and three image stacks from a single biofilm for 30-84red-plus-30-84ZNgreen
When grown in shaken liquid culture, 30-84GacA mutants
entered the exponential growth phase earlier than both wild-
type strain 30-84 and the phz structural mutant 30-84ZN (Fig.
1), although doubling times were approximately 5 h for each
strain. When these data were fit to a logistic growth model,
30-84GacA had a significantly (P ? 0.05) higher intrinsic rate
TABLE 1. Bacterial strains and plasmids used in this study
Strain or plasmid Descriptiona
Source or reference
30-84Wild type, Rifr
W. Bockus, personal
phzB::lacZ genomic fusion, Rifr
Wild type containing pKI2CM-EGFP; gfp?
Wild type containing pKI2CM-DsRed; rfp?
30-84GacA containing pKI2CM-EGFP; gfp?
30-84GacA containing pKI2CM-DsRed; rfp?
30-84ZN containing pKI2CM-EGFP; gfp?
30-84ZN containing pKI2CM-DsRed; rfp?
F?recA1 endA1 hsdR17 supE44 thi-1 gyrA96 relA1 ?(argF-lacZYA)I169 ?80lacZ ?M15 ??
?) supE44 recA1 ara14 proA2 lacY1 galK2 rpsL20 xyl-5 mtl-5 ??
pVS1 replicon, p15a origin of replication, inaZ transcriptional fusion; Kmr
ColE1, contains an enhanced variant of the Aequorea victoria green fluorescent protein
ColE1, contains Discosoma sp. red fluorescent protein (DsRed-Express); AprKmrrfp?
Interposon, contains cat; Cmr
pPROBE-KI? containing the 4.3-kb EcoRI fragment carrying cat and a 974-bp PvuII-
EcoRI fragment carrying EGFP; KmrCmrgfp?
pPROBE-KI? containing the 4.3-kb EcoRI fragment carrying cat and a 930-bp PvuII-
EcoRI fragment carrying DsRed-Express; KmrCmrrfp?
aApr, Cmr, Kmr, and Rifrindicate resistance to ampicillin, chloramphenicol, kanamycin, and rifampin, respectively.
VOL. 77, 2011Gac REGULATORY MUTANTS ARE MUTUALISTS IN BIOFILMS 7229
of growth than 30-84 and 30-84ZN (0.413 ? 0.008, 0.377 ?
0.007, and 0.380 ? 0.007, respectively). In contrast, each strain
approached carrying capacities that were statistically indistin-
guishable (2.173 ? 0.076, 2.179 ? 0.078, and 2.152 ? 0.087,
respectively). Thus, the higher densities of 30-84GacA during
exponential growth are due to the mutant strain entering ex-
ponential phase growth earlier than the wild type.
Both strains 30-84ZN and 30-84GacA, which were deficient
in phenazine production, were unable to form single-strain
biofilms (Fig. 2B and D, respectively). However, both mutants
became established in biofilms when grown in a mixture with
strain 30-84 (Fig. 2B and D). The biovolume of the wild type
was significantly reduced when grown with the phenazine
structural mutant 30-84ZN (Fig. 2A). In contrast, strain 30-84
experienced a significant benefit when grown with 30-84GacA
(Fig. 2C). The replacement series analysis further supports the
observation that strain 30-84 grew to lower and higher biovol-
umes with 30-84ZN and 30-84GacA, respectively, than would
be expected based on initial frequencies in the inocula (Fig. 3A
and B, respectively). These data indicate that 30-84GacA pro-
vided a net growth benefit to strain 30-84, whereas 30-84ZN
imposed a net cost.
FIG. 1. Growth curves for strains 30-84 (E), 30-84ZN (‚), and
30-84GacA (?). Error bars representing standard errors are mostly
too small to be seen. 30-84GacA entered the exponential growth phase
significantly earlier than the other two strains.
FIG. 2. Time series of the growth of strains in single-strain (E) and mixed (?) biofilms. Error bars represent standard errors. (A) The
biovolume of strain 30-84 was greater in single-strain biofilms than in mixed biofilms with 30-84ZN. (B) Strain 30-84ZN was unable to establish
single-strain biofilms, but it showed biofilm formation in the presence of 30-84. (C) Strain 30-84 grew to greater volumes in chimeric biofilms with
30-84GacA than in single-strain biofilms. (D) The GacA mutant failed to establish single-strain biofilms, but it formed extensive chimeric biofilms
with the wild type.
FIG. 3. Replacement series analysis of mixed biofilms after 96 h.
(A) Observed biovolume of 30-84red (gray bars), 30-84ZNgreen (open
bar), and total biofilm (both bars). The dashed line indicates the
expected biovolume of strain 30-84 in mixed biofilms predicted from
the performance of 30-84 without 30-84ZN (given a 50% or 100%
reduction in inoculum). Wild-type strain 30-84 achieved a lower bio-
volume in the presence of 30-84ZN than would be predicted by its
performance without 30-84ZN, indicating that the wild type was neg-
atively impacted by the presence of 30-84ZN. 30-84ZN was unable to
form a biofilm without the wild type, but it formed a mixed biofilm with
the wild type. (B) Expected (dashed lines) and observed (gray bars)
biovolumes of 30-84red, 30-84GacAgreen (open bar), and total biofilm
(both bars). The wild type reached a higher biovolume in the mixture
with 30-84GacA than expected based on its performance without 30-
84GacA. 30-84GacA was unable to form a biofilm without the wild
type, but it formed an extensive chimeric biofilm with the wild type.
The total population of the mixed biofilm significantly exceeded the
population of the wild type alone. Error bars denote standard errors.
7230 DRISCOLL ET AL.APPL. ENVIRON. MICROBIOL.
Chimeric biofilms comprised of strains 30-84 and 30-
84GacA developed more extensively (Fig. 4C and 5B) than
30-84-only biofilms (Fig. 4A and 5A). After 96 h, mixed bio-
films containing 30-84 and 30-84GacA developed significantly
more total biovolume, were thicker, and had approximately
twice as many microcolonies as the wild-type biofilms (Table 2,
Fig. 4A and C, and Fig. 5). In contrast, mixed populations
containing 30-84 and the phz structural mutant 30-84ZNgreen
formed biofilms that lacked well-developed multicellular struc-
tures (Fig. 4B). This is reflected in the extremely low mean
diffusion distance of biofilms containing phz mutants (Table 2).
These biofilms also were characterized by a relatively low oc-
currence of 30-84red (Fig. 4B).
Wild-type strain 30-84 clustered significantly more with the
FIG. 4. Maximum-intensity z-axis projections of representative image stacks taken from biofilms comprised of strain 30-84red with 30-84green
(A), 30-84ZNgreen (B), and 30-84GacAgreen (C) after 6, 24, and 72 h. Initially, 30-84ZNgreen and 30-84GacAgreen were significantly reduced
in attachment compared to 30-84green. Thus, initial attachment of 30-84green (6 h) is enhanced relative to the phenazine-deficient mutants. By
72 h, biofilms containing phenazine structural mutant 30-84ZN had formed thin, undifferentiated biofilms without distinct microcolonies, whereas
biofilms containing 30-84GacA formed thicker, more extensive biofilms than the wild type alone.
FIG. 5. Orthogonal views of 30-84red with 30-84green (A) and 30-84red and 30-84GacAgreen mixed biofilms (B) after 96 h. The large image
is a horizontal cross-section through a biofilm. The smaller images (side and bottom panels) are vertical sections. The white arrowheads indicate
the placement of the vertical and horizontal cross-sections. Note the frequent occurrence of 30-84 at the attachment surface compared with
30-84GacA mutants in mixed biofilms, suggesting that 30-84red is important for attachment of 30-84GacAgreen. (The substratum is at the outer
edge of each panel.)
VOL. 77, 2011 Gac REGULATORY MUTANTS ARE MUTUALISTS IN BIOFILMS7231
GacA mutants than with itself (Fig. 6). The spatial cross-cova-
riance between red and green biovolume, c, and distance from
focal pixel, d, followed a negative exponential relationship of
the form cd? abd, where a and b are parameters (Fig. 6A). The
decay rate parameter, b, was not significantly different between
the two treatments, but the base parameter, a, was significantly
greater in biofilms containing 30-84 and 30-84GacA than those
containing only the wild type at the 99% confidence level
(0.163 ? 0.015 and 0.103 ? 0.015, respectively). These data
indicated that co-occurrence is higher over short distances
(0 to 5 ?m) and decays rapidly thereafter. Furthermore, the
average cdwithin 10 ?m was significantly higher in mixed
biofilms containing 30-84 and 30-84GacA than single-strain
biofilms containing 30-84 (Fig. 6B). These relationships are
reflected in representative comparisons between three-dimen-
sional images of wild-type-only biofilms (Fig. 6C) and biofilms
containing strains 30-84 and 30-84GacA (Fig. 6D). Overall,
these data show that local spatial clustering between the wild
type and 30-84GacA was significantly higher than that of the
wild type with itself.
Do Gac mutants behave as “cheaters” or social parasites in
mixed populations with the wild type, or do they behave as
beneficial mutualists, having a positive effect on the wild type?
As expected, both 30-84ZN and 30-84GacA were unable to
form single-strain biofilms in the absence of wild-type 30-84
(Fig. 2A and B). However, our results clearly showed that both
phenazine-deficient mutants benefited significantly from the
presence of the wild type, presumably due to the production of
phenazines produced by neighboring wild-type cells (Fig. 2B
and D). Previous work demonstrated that phenazines are re-
quired for biofilm formation by strain 30-84 in flow cell assays
TABLE 2. Biofilm characteristics for representative biofilms formed by wild-type strain 30-84 with itself and each mutant after 96 h
30-84 ? 30-84 30-84 ? 30-84GacA 30-84 ? 30-84ZN
Diffusion distance (?m)b
No. of microcoloniesc
Microcolony area (?m2)
0.474 ? 0.112 B
2.811 ? 0.453 B
0.007 ? 0.002 B
12.200 ? 4.716 B
50.879 ? 7.550 A
0.990 ? 0.124 A
6.334 ? 1.190 A
0.013 ? 0.001 A
31.000 ? 3.493 A
64.584 ? 8.430 A
0.299 ? 0.071 C
2.180 ? 0.643 B
0.001 ? 0.000 C
aThe numbers reported are averages ? standard errors, and the letters following the values indicate significant differences.
bDiffusion distance is defined as the minimum distance between a pixel of biovolume and a pixel of void space. Higher values thus indicate larger aggregates of
cMicrocolonies were defined as aggregates that occupy an area of at least 25 ?m2in a maximum-intensity vertical projection of image stacks.
dNA, microcolony counts and sizes are not reliable in the case of the relatively undifferentiated biofilms formed by 30-84 ? 30-84ZN.
FIG. 6. (A) Spatial covariance between 30-84red with strain 30-84GacAgreen (E) or 30-84green (F). Higher spatial covariance between the red
and green pixels at short distances in mixed biofilms than wild-type-only biofilms indicates that 30-84 clusters more with 30-84GacA than with itself.
Error bars denote standard errors between replicates in four cardinal directions. (B) Average covariances within 10 ?m are significantly higher in
mixed biofilms than wild-type-only biofilms. Error bars denote standard errors between average pixel-level covariances within 10 ?m. Three-
dimensional images of wild-type-only (C) and 30-84red-plus-30-84GacAgreen (D) biofilms after 96 h.
7232 DRISCOLL ET AL.APPL. ENVIRON. MICROBIOL.
(23). Furthermore, addition of exogenous phenazines restored
biofilm formation capabilities to 30-84ZN in flow cell assays
(24). The fact that both the wild type and mutants were able to
utilize phenazines and/or other secondary metabolites or cel-
lular components produced by wild-type strain 30-84 and be-
come established in mixed biofilms suggests that a diffusible
external product is being shared between wild-type and mutant
participants. However, this fact alone does not differentiate
between the competing views of the Gac mutant as a “cheater”
or a beneficial mutualist, because a positive effect of 30-84 on
GacA is a prediction of both hypotheses.
The data presented here indicated that wild-type strain
30-84 also benefited from the presence of 30-84GacA; notably
the biovolume of the similarly labeled wild type was higher in
mixed biofilms than in single-strain biofilms (Fig. 2D). The
replacement series further demonstrated that the biovolume of
the wild type in the mixed biofilm was higher than would be
predicted based on the inoculum density (Fig. 3A). Mixed
biofilms containing both Gac phenotypes also were signifi-
cantly larger than wild-type biofilms, suggesting a net benefit to
phenotypic variation for strain 30-84 (Fig. 3A). In contrast,
30-84ZN competitively displaced 30-84 in mixed biofilms, sug-
gesting a potential explanation for why Gac mutants are fre-
quently found in nature, whereas phenazine biosynthesis mu-
tants fail to be recovered. Our results confirm previous
observations of better survival of wild-type 30-84 populations
in mixtures with Gac mutants in the natural rhizosphere (3).
Although the biofilm experiments were not intended to deter-
mine the specific mechanism by which the wild type benefits
from the GacA mutant, these results suggest that enhanced
biofilm development by mixed populations may be one factor
contributing to the success of mixed populations in the rhizo-
The idea that phenotypic variation may be beneficial to
rhizosphere populations is somewhat surprising given that Gac
mutants have been traditionally viewed as having a negative
global effect on secondary metabolite production. However, it
has been recognized previously that Gac also positively regu-
lates certain phenotypes, including siderophore production,
motility, and exopolysaccharide production, providing 30-84
Gac mutants with their characteristic hyperfluorescent, large
colony appearance on solid media (3). Recently, transcriptome
analysis comparing the wild type and a gacA mutant of P.
fluorescens PF-5 using microarrays demonstrated that approx-
imately 10% of the genome is under positive regulation by
Gac, including specific genes involved in iron uptake (sidero-
phore), motility, and stress response (15). Preliminary experi-
ments designed to identify secondary metabolites or other po-
tentially beneficial substances using cell-free supernatants have
failed to identify a diffusible product produced by 30-84 GacA
mutants that contributes to enhanced biofilm formation by the
wild type (data not shown). This is not unexpected as cellular
traits that contribute to motility (flagella, pili, and exopolysac-
charides), nutrient uptake, or stress response may also contrib-
ute to cell attachment and biofilm development. Future work
will be directed toward the identification of genes and pheno-
types of GacS/GacA variants that contribute to improved spe-
Overall, our data do not support the role of GacA mutants
as cheaters in mixed biofilms. On the contrary, we show that
the presence of 30-84GacA stimulates 30-84 growth in mixed
biofilms (i.e., at intermediate frequencies). All of our exper-
iments were performed under sterile conditions in order to
isolate and test the most basic prediction of the cheater
hypothesis—freed from competitive or predatory (between-
group selective) pressures, GacA mutants will increase in fre-
quency at a net cost to the wild type. Such a cost was observed
only in the case of the phenazine structural mutant, 30-84ZN,
supporting the hypothesis that the presence of Gac mutants
provides some benefits to the wild-type strain. In nature, com-
petitive and predatory pressures are expected to place a cap on
Gac fitness (3, 19), but this alone does not qualify Gac mutants
as cheaters. It is likely that the interaction between Gac mu-
tants and the wild-type strain is dynamic and is thus expected
to be different across species and environments. However, we
note that the idea of Gac mutants as individually selected
cheaters is also problematic in light of the presence of mech-
anisms that apparently ensure the presence of Gac mutants,
even in clonal groups (e.g., programmed phase variation such
as slipped-strand mispairing, genomic rearrangements, and dif-
ferential methylation and unprogrammed phase variation via
mutation [see reference 42 for a full review of these mecha-
nisms]). This runs counter to the theoretical prediction that
cooperative traits should evolve novel mechanisms (e.g.,
pleiotropy) to ensure that efficient cheaters do not easily arise
through mutation (9, 43). Although much remains to be un-
derstood about the interactions between Gac mutants and the
wild type under various ecological conditions, we have dem-
onstrated that interactions between the wild type and a GacA
mutant in single-species flow cell biofilms are mutually bene-
Another interpretation of the role of alternative phenotypes
in biofilms is possible when loss-of-function modifications are
reversible. Recently, van den Broek et al. (41) proposed that
reversible switching between functional and nonfunctional Gac
phenotypes may be a conserved strategy among rhizosphere
pseudomonads. By encoding both phenotypes, a single geno-
type can capitalize on the strengths of each strategy. Phase
variation in the Gac system may simultaneously confer su-
perior competitive or host-exploitative abilities while allow-
ing for efficient exploitation of hard-won habitat by a fast-
growing variant. Mechanisms to generate frequent switching
between these phenotypes are crucial to this hypothesis. Al-
though such reversible modification has been shown only for
two strains within the genus Pseudomonas (13, 42), the possi-
bility of reversible modifications to gac in other pseudomonads
clearly warrants further study. However, in current and previ-
ous work with P. chlororaphis strain 30-84, spontaneous rever-
sions to the Gac?phenotype have not been observed (2).
Our results underscore the importance of environmental
context in governing bacterial ecology and evolution. The phys-
ical environment may greatly influence the spatial distribution
of beneficial metabolites, thus profoundly altering the selective
pressures within a population. For this reason, future investi-
gations of external products in bacteria should take into ac-
count the biofilm context in which these traits naturally occur
or risk overestimating the degree of sharing—and thus, the
potential for cheating (5). This point parallels recent develop-
ments in the study of quorum-sensing signals. In natural set-
tings, quorum-sensing signals may reach very high local con-
VOL. 77, 2011 Gac REGULATORY MUTANTS ARE MUTUALISTS IN BIOFILMS7233
centrations in even sparse populations when diffusion is
restricted (7, 10, 17, 37, 48). Studies of quorum sensing con-
ducted with liquid cultures tend to overestimate the relative
importance of population size, because local increases in signal
concentration are prevented. Indeed, as few as two cells have
been observed to constitute a “quorum” where diffusion is
limited (7). Similarly, we have found evidence for heightened
co-occurrence of wild-type and gac mutant phenotypes over
relatively small (?10 ?m) distances, suggesting that phenazines
may be restricted to the immediate neighborhood of produc-
ers. In natural rhizosphere settings, in which populations oc-
cupy microsites that may be more limited in diffusion than flow
cells, beneficial metabolites such as phenazines may accumu-
late to higher levels and thus increase the neighborhood size
over which interactions between alternative phenotypes occur.
Understanding the role of multilevel selection in maintain-
ing or undermining biofilm-associated traits in nature will help
in evaluating competing views of biofilms as collectives of self-
ish individuals (20) or incipient multicellular organisms (26,
44). An important first step is to properly appreciate how
experimental conditions affect parameters of central impor-
tance to social evolution models, such as neighborhood size
and population structure. This will, in turn, permit a more
accurate understanding of the selective forces shaping bacte-
rial traits involved in biocontrol or pathogenesis and how these
forces may be sensitive to environmental conditions.
Conclusions. The nature of the interaction between coexist-
ing phenotypic variants with alterations to the Gac and or-
thologous regulatory systems depends strongly on ecological
context. Although it has been observed that Gac mutants dis-
place the wild type in nutrient-rich liquid competition (6) and
incur group-level costs under some realistic natural conditions
(19), we have found evidence for a mutualism between these
alternative phenotypes in laboratory biofilms. Such an inter-
pretation appears more consistent with the high rates of alter-
ation to the GacS/GacA system and orthologous systems ob-
served in nature (41, 42), as evolution at higher levels (e.g., the
clone level) is expected to disrupt the efficient emergence of
cheaters (9, 43).
This work was supported by NSF IOS-1010669 to W.W.D. and
NIFA 2008-35319-21879 to L.S.P. and E.A.P.
We thank O. Eldakar, G. Holt, R. Ferriere, N. Young, and C. Seeve
for helpful comments and discussions.
1. Abramoff, M. D., P. J. Magelhaes, and S. J. Ram. 2004. Image processing
with ImageJ. Biophotonics Int. 11:36–42.
2. Chancey, S. T., D. W. Wood, E. A. Pierson, and L. S. Pierson. 2002. Survival
of GacS/GacA mutants of the biological control bacterium Pseudomonas
aureofaciens 30-84 in the wheat rhizosphere. Appl. Environ. Microbiol. 68:
3. Chancey, S. T., D. W. Wood, and L. S. Pierson. 1999. Two-component
transcriptional regulation of N-acyl-homoserine lactone production in Pseu-
domonas aureofaciens. Appl. Environ. Microbiol. 65:2294–2299.
4. Crespi, B. J. 2001. The evolution of social behavior in microorganisms.
Trends Ecol. Evol. 16:178–183.
5. Driscoll, W. W., and J. W. Pepper. 2010. Theory for the evolution of diffus-
ible external goods. Evolution 64:2682–2687.
6. Duffy, B. K., and G. Defago. 2000. Controlling instability in gacS-gacA reg-
ulatory genes during inoculant production of Pseudomonas fluorescens bio-
control strains. Appl. Environ. Microbiol. 66:3142–3150.
7. Dulla, G., and S. E. Lindow. 2008. Quorum size of Pseudomonas syringae is
small and dictated by water availability on the leaf surface. Proc. Natl. Acad.
Sci. U. S. A. 105:3082–3087.
8. Fellay, R., J. Frey, and H. Krisch. 1987. Interposon mutagenesis of soil and
water bacteria—a family of DNA fragments designed for in vitro insertional
mutagenesis of Gram-negative bacteria. Gene 52:147–154.
9. Foster, K. R., G. Shaulsky, J. E. Strassmann, D. C. Queller, and C. R. L.
Thompson. 2004. Pleiotropy as a mechanism to stabilize cooperation. Nature
10. Gantner, S., et al. 2006. In situ quantitation of the spatial scale of calling
distances and population density-independent N-acylhomoserine lactone-
mediated communication by rhizobacteria colonized on plant roots. FEMS
Microbiol. Ecol. 56:188–194.
11. Greig, D., and M. Travisano. 2004. The Prisoner’s Dilemma and polymor-
phism in yeast SUC genes. Proc. R. Soc. Lond. Ser. B Biol. Sci 271:S25–S26.
12. Griffin, A. S., S. A. West, and A. Buckling. 2004. Cooperation and compe-
tition in pathogenic bacteria. Nature 430:1024–1027.
13. Han, B., A. Pain, and K. Johnstone. 1997. Spontaneous duplication of a 661
bp element within a two-component sensor regulator gene causes phenotypic
switching in colonies of Pseudomonas tolaasii, cause of brown blotch disease
of mushrooms. Mol. Microbiol. 25:211–218.
14. Harrison, F., L. E. Browning, M. Vos, and A. Buckling. 2006. Cooperation
and virulence in acute Pseudomonas aeruginosa infections. BMC Biol. 4:21.
15. Hassan, K. A., et al. 2010. Inactivation of the GacA response regulator in
Pseudomonas fluorescens Pf-5 has far-reaching transcriptomic conse-
quences. Environ. Microbiol. 12:899–915.
16. Heeb, S., and D. Haas. 2001. Regulatory roles of the GacS/GacA two-
component system in plant-associated and other Gram-negative bacteria.
Mol. Plant Microbe Interact. 14:1351–1363.
17. Hense, B. A., et al. 2007. Opinion—does efficiency sensing unify diffusion and
quorum sensing? Nat. Rev. Microbiol. 5:230–239.
18. Heydorn, A., et al. 2000. Quantification of biofilm structures by the novel
computer program COMSTAT. Microbiology 146:2395–2407.
19. Jousset, A., et al. 2009. Predators promote defence of rhizosphere bacterial
populations by selective feeding on non-toxic cheaters. ISME J. 3:666–674.
20. Klausen, M., M. Gjermansen, J. U. Kreft, and T. Tolker-Nielsen. 2006.
Dynamics of development and dispersal in sessile microbial communities:
examples from Pseudomonas aeruginosa and Pseudomonas putida model
biofilms. FEMS Microbiol. Lett. 261:1–11.
21. MacLean, R. C., A. Fuentes-Hernandez, D. Greig, L. D. Hurst, and I. Gudelj.
2010. A mixture of “cheats” and “co-operators” can enable maximal group
benefit. PloS Biol. 8:e1000486.
22. Maddula, V. S. R. K., E. A. Pierson, and L. S. Pierson III. 2008. Altering the
ratio of phenazines in Pseudomonas chlororaphis (aureofaciens) strain 30-84:
effects on biofilm formation and pathogen inhibition. J. Bacteriol. 190:2759–
23. Maddula, V. S. R. K., E. A. Pierson, and L. S. Pierson. 2004. Phenazines play
a role in adhesion/biofilm formation in Pseudomonas aureofaciens strain
30-84. Phytopathology 94:S64–S65.
24. Maddula, V. S. R. K., Z. Zhang, E. A. Pierson, and L. S. Pierson. 2006.
Quorum sensing and phenazines are involved in biofilm formation by Pseu-
domonas chlororaphis (aureofaciens) strain 30-84. Microb. Ecol. 52:289–301.
25. Miller, W. G., J. H. J. Leveau, and S. E. Lindow. 2000. Improved gfp and
inaZ broad-host-range promoter-probe vectors. Mol. Plant Microbe Inter-
26. Monds, R. D., and G. A. O’Toole. 2009. The developmental model of mi-
crobial biofilms: ten years of a paradigm up for review. Trends Microbiol.
27. Moons, P., C. W. Michiels, and A. Aertsen. 2009. Bacterial interactions in
biofilms. Crit. Rev. Microbiol. 35:157–168.
28. Morris, C. E., and J. M. Monier. 2003. The ecological significance of biofilm
formation by plant-associated bacteria. Annu. Rev. Phytopathol. 41:429–453.
29. Nadell, C. D., J. B. Xavier, and K. R. Foster. 2009. The sociobiology of
biofilms. FEMS Microbiol. Rev. 33:206–224.
30. Okasha, S. 2006. Evolution and the levels of selection. Clarendon Press,
Oxford, United Kingdom.
31. Pierson, L. S., III, V. D. Keppenne, and D. W. Wood. 1994. Phenazine
antibiotic biosynthesis in Pseudomonas aureofaciens 30-84 is regulated by
PhzR in response to cell density. J. Bacteriol. 176:3966–3974.
32. Pierson, L. S., III, and E. A. Pierson. 1996. Phenazine antibiotic production
in Pseudomonas aureofaciens: role in rhizosphere ecology and pathogen
suppression. FEMS Microbiol. Lett. 136:101–108.
33. Pierson, L. S., III, and L. S. Thomashow. 1992. Cloning and heterologous
expression of the phenazine biosynthetic locus from Pseudomonas aureofa-
ciens 30-84. Mol. Plant Microbe Interact. 5:330–339.
34. Platt, T. G., and J. D. Bever. 2009. Kin competition and the evolution of
cooperation. Trends Ecol. Evol. 24:370–377.
35. Poritsanos, N., C. Selin, W. G. D. Fernando, S. Nakkeeran, and T. R. de
Kievit. 2006. A GacS deficiency does not affect Pseudomonas chlororaphis
PA23 fitness when growing on canola, in aged batch culture or as a biofilm.
Can. J. Microbiol 52:1177–1188.
36. Rainey, P. B., and K. Rainey. 2003. Evolution of cooperation and conflict in
experimental bacterial populations. Nature 425:72–74.
37. Redfield, R. J. 2002. Is quorum sensing a side effect of diffusion sensing?
Trends Microbiol. 10:365–370.
7234DRISCOLL ET AL.APPL. ENVIRON. MICROBIOL.
38. Sandoz, K. M., S. M. Mitzimberg, and M. Schuster. 2007. Social cheating in Download full-text
Pseudomonas aeruginosa quorum sensing. Proc. Natl. Acad. Sci. U. S. A.
39. Schmidt-Eisenlohr, H., A. Gast, and C. Baron. 2003. Inactivation of gacS
does not affect the competitiveness of Pseudomonas chlororaphis in the
Arabidopsis thaliana rhizosphere. Appl. Environ. Microbiol. 69:1817–1826.
40. Thomashow, L. S., and D. M. Weller. 1988. Role of a phenazine antibiotic
from Pseudomonas fluorescens in biological control of Gaeumannomyces
graminis var. tritici. J. Bacteriol. 170:3499–3508.
41. van den Broek, D., G. V. Bloemberg, and B. Lugtenberg. 2005. The role of
phenotypic variation in rhizosphere Pseudomonas bacteria. Environ. Micro-
42. van den Broek, D., T. F. C. Chin-A-Woeng, G. V. Bloemberg, and B. J. J.
Lugtenberg. 2005. Molecular nature of spontaneous modifications in gacS
which cause colony phase variation in Pseudomonas sp. strain PCL1171. J.
43. Velicer, G. J. 2005. Evolution of cooperation: does selfishness restraint lie
within? Curr. Biol. 15:R173–R175.
44. Webb, J. S., M. Givskov, and S. Kjelleberg. 2003. Bacterial biofilms: pro-
karyotic adventures in multicellularity. Curr. Opin. Microbiol. 6:578–585.
45. West, S. A., S. P. Diggle, A. Buckling, A. Gardner, and A. S. Griffins. 2007.
The social lives of microbes. Annu. Rev. Ecol. Evol. Syst. 38:53–77.
46. Wilson, D. S. 1975. Theory of group selection. Proc. Natl. Acad. Sci. U. S. A.
47. Wilson, D. S., and E. O. Wilson. 2007. Rethinking the theoretical foundation
of sociobiology. Q. Rev. Biol. 82:327–348.
48. Yang, J. W., B. A. Evans, and D. E. Rozen. 2010. Signal diffusion and the
mitigation of social exploitation in pneumococcal competence signalling.
Proc. R. Soc. Lond. Ser. B Biol. Sci. 277:2991–2999.
VOL. 77, 2011 Gac REGULATORY MUTANTS ARE MUTUALISTS IN BIOFILMS7235