Cue-Driven Microbial Cooperation and Communication: Evolving Quorum Sensing With Honest Signaling
Cue-Driven Microbial Cooperation and Communication: Evolving Quorum Sensing With Honest Signaling
Cue-Driven Microbial Cooperation and Communication: Evolving Quorum Sensing With Honest Signaling
Abstract
Background Quorum sensing (QS) is the ability of microorganisms to assess local clonal density by measuring
the extracellular concentration of signal molecules that they produce and excrete. QS is also the only known way
of bacterial communication that supports the coordination of within-clone cooperative actions requiring a certain
threshold density of cooperating cells. Cooperation aided by QS communication is sensitive to cheating in two
different ways: laggards may benefit from not investing in cooperation but enjoying the benefit provided by their
cooperating neighbors, whereas Liars explicitly promise cooperation but fail to do so, thereby convincing potential
cooperating neighbors to help them, for almost free. Given this double vulnerability to cheats, it is not trivial why QS-
supported cooperation is so widespread among prokaryotes.
Results We investigated the evolutionary dynamics of QS in populations of cooperators for whom the QS signal
is an inevitable side effect of producing the public good itself (cue-based QS). Using spatially explicit agent-based
lattice simulations of QS-aided threshold cooperation (whereby cooperation is effective only above a critical cumu-
lative level of contributions) and three different (analytical and numerical) approximations of the lattice model, we
explored the dynamics of QS-aided threshold cooperation under a feasible range of parameter values. We demon-
strate three major advantages of cue-driven cooperation. First, laggards cannot wipe out cooperation under a wide
range of reasonable environmental conditions, in spite of an unconstrained possibility to mutate to cheating;
in fact, cooperators may even exclude laggards at high cooperation thresholds. Second, lying almost never pays off,
if the signal is an inevitable byproduct (i.e., the cue) of cooperation; even very cheap fake signals are selected against.
And thirdly, QS is most useful if local cooperator densities are the least predictable, i.e., if their lattice-wise mean
is close to the cooperation threshold with a substantial variance.
Conclusions Comparing the results of the four different modeling approaches indicates that cue-driven thresh-
old cooperation may be a viable evolutionary strategy for microbes that cannot keep track of past behavior of their
potential cooperating partners, in spatially viscous and in well-mixed environments alike. Our model can be seen
as a version of the famous greenbeard effect, where greenbeards coexist with defectors in a evolutionarily stable poly-
morphism. Such polymorphism is maintained by the condition-dependent trade-offs of signal production which are
characteristic of cue-based QS.
Keywords Microbial signaling, Quorum sensing, Threshold public goods game, Cue, Signal, Honesty, N person games
*Correspondence:
Szabolcs Számadó
[email protected]
Full list of author information is available at the end of the article
© The Author(s) 2024. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which
permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the
original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or
other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line
to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory
regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this
licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecom-
mons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
Fig. 1 A simplified representation of quorum sensing regulated cooperation for the production of common goods collectively utilized by bacteria
having access to them. The three basic genetic components of the QS-regulated cooperative system are (a) the signaling component (pink),
which comprises the genes for signal molecule production and excretion of signal molecules (stars); (b) the signal detection and transduction
system (green), including genes for the signal receptor and the second messenger system; (c) the cooperation genes (blue) which are transcribed
and expressed (and products externalized) if the extracellular concentration of signal molecules is sufficiently high. Red asterisks signify component
events of signal detection: signal capture and signal transduction
in synergy with each other. For example, the two princi- public good, a protease exoenzyme virulence factor [3].
pal QS systems (las and rhl) regulating the expression of The cheating mutant has been shown to enjoy a substan-
virulence factors in Pseudomonas aeruginosa have been tial reproductive advantage over cooperative strains [34]
shown to form a reciprocal signaling network that syner- for the obvious reason of not carrying the metabolic bur-
gistically enhances and tunes the strain’s reactivity to its den of cooperation (exoenzyme production).
physical and “social” environment. A more nuanced way of cheating is increasing signal
There are different interpretations of the function of production while evolving a higher threshold value for
quorum sensing (see [29]). The first interpretation is that cooperation. Brown and Johnstone in a seminal model
it aims at sizing up the local density of potential coopera- of QS were able to show that increasing conflict of inter-
tors [30–32], whereas the second one is that it serves to est (decreasing relatedness) favors such “coercive” vari-
assess the diffusibility of exoproducts in the environment ants [35]. These can manipulate older strains with lower
[33]. These two functions are difficult to disentangle as thresholds into increased production of the public good
both high cell density and low diffusion can lead to a high by mimicking a higher cooperator density with the
local concentration of the QS signal. There is some indi- increased signal production. In silico study of QS found
cation that some bacterial strains might be able to distin- that such coercive variants are more likely to emerge in
guish these two types of information using combinatorial genetically mixed populations with decreased relatedness
QS signals, i.e., by using two different signal molecules [36].
with covarying decay and autoinduction rates [29]. If, however, the signal cannot be switched off, cheat-
Either way, as QS is a communication system, it is ing is expected to be less deleterious for cooperation. For
prone to defection or deceit by another type of cheater: example, the signal may be the public good itself, as in
individuals with a silent set of cooperation genes but the case of lactic acid bacteria producing the bacteriocin
capable of sending out false signals of intent for coopera- called nisin. Nisin production is QS-regulated [37], but
tion (lying) may still enjoy the full benefit of the public the QS signal is nisin itself, so bacteria producing nisin
good produced by nearby cooperators responding to the are also signaling. Cooperators always produce nisin
false signal. A textbook example of such defectors (“liars”) at a low expression level, advertising their willingness
is the lasR mutant of Pseudomonas aeruginosa that does to cooperate, and they express and excrete nisin at an
not respond to QS signals of the wild type and, conse- elevated level when a quorum is reached. However, the
quently, does not cooperate in producing an important cooperation signal can be faked by non-cooperators: the
bacteriocin gene can be expressed constitutively at a low (4) spatial stochastic simulations of the threshold public
level to produce the signal, but cheaters never express it goods game (TPPG).
at sufficiently high levels to considerably contribute to
the common good. Therefore, individuals capable only of Model basics
low-level nisin expression are actually liars. We used four different model types: MF approximation,
Another type of cheater is a cooperator that expresses CF approximation, non-spatial and spatial (on-lattice)
more nisin in its signaling state than the normal signal agent-based models. They share the same basic assump-
level but less than the cooperation level. By issuing extra tions (explained below) but gradually relax crucial simpli-
signals above the basic expression level of honest coop- fications while also losing analytical tractability. For the
erators, these cheaters gain a fitness benefit by increas- specifics of the different approaches, see Methods and
ing the local signal concentration which, therefore, may models; for their parameters, see Table 1.
reach the quorum threshold and convince more coopera-
tors to join in the common effort, possibly in vain. This is Mean‑field model (MF)
equivalent to promising more cooperation than actually In Additional file 1: Appendix 1 [40], we construct the
provided—a milder version of the liar strategy. mean-field version of the model, with the assumptions
To study the dynamical and evolutionary properties of that population size ( P ), interacting group size ( N ) and
QS-regulated cooperation, it is sufficient to assume that the cooperation threshold κ are all very large, while
the individuals clonally inherit three fundamental QS- N /P ≪ 1 and κ/N → xc . In this limit, all the dynamical
related properties: (1) signaling (S) or not (s) the inten- effects are averaged across the entire habitat so that the
tion/ability to cooperate, (2) responding (R) or not (r) fitnesses of the strategies depend only on the average fre-
to above-quorum signal levels, and (3) cooperating (C) quencies of the strains present.
or not (c) when a quorum is reached (see Fig. 1). Each
of these properties may be determined by a number of Configuration‑field model (CF)
different genes, but the only trait we consider relevant Next, we assume that the population is still very large
is the functionality of the corresponding gene set as a (infinite), but individuals form random interacting
unit. Therefore, we assume three loci with two functional groups of finite size N . While in the previous section we
alleles on each, which allows for 23 = 8 different geno- considered N to be so large that each interacting group
types. For QS to hold, cooperation (i.e., the expression of consists of strategies in exact proportion to their global
the C allele) is assumed to be conditional on the presence frequencies in the population, now we assume that N is
of a critical number of signaler individuals (the quorum, smaller, and thus different interaction groups with differ-
i.e., those harboring S and/or C) within the interaction ent configurations of strategies are formed in an inher-
neighborhood of individuals possessing both C and R. In ently stochastic manner, due to sampling errors. The two
other words, cooperators capable of detecting the signal players participating in an elementary game step are ran-
will cooperate only above the critical local quorum and domly chosen members of their own interaction groups
mute their cooperation gene otherwise. Non-responder (both of size N) that are drawn at random from the popu-
cooperators cooperate unconditionally, and they may lation. The overall fitness for each of the eight strategies is
issue the signal either at the normal expression level or at calculated as the weighted average of its local fitness in all
an elevated one. possible configurations of the interaction group around
The question we aim to answer is whether cue-driven a focal individual of the given strategy (for more details,
(e.g., nisin type) cooperation and/or communication see Additional file 1: Appendix 1).
can be maintained in a population of quorum sensing
microbes in the face of all possible mutations allow- Agent‑based models
ing cheater strategies, assuming different costs of coop- To address the effect of finite interacting groups, along
eration, signaling, and signal detection/response. Here, with the spatial constraints arising from limited agent
we will scrutinize a family of models built on the above mobility and local (neighborhood-) interactions, we have
assumptions, allowing all three possible types of cheaters developed an agent-based simulation implementation of
to appear. The models are built on the individual-based the configuration-field model (agent-based nonspatial
approach of Czárán and Hoekstra [38], extending it both model) and a spatially explicit lattice version of it (agent-
in scope and methodology of representation. For a deeper based spatial model).
insight into the coexistence dynamics of the various Figure 2 depicts the relationship of these models. The
strategies, we derive analytical (1) mean-field (MF) and configuration field model is a technical tool that allows
(2) configuration-field (CF) [39] approximations besides the relaxation of constraining assumptions one-by-one
the corresponding individual-based (3) non-spatial and from the mean-field model to the lattice-based individual
Fig. 2 The relation between different model types: MF, CF, and agent-based (well mixed and lattice). The CF (compared to MF) represents
the relaxation of the assumption of infinitely large groups, hence the introduction of compositional variance. The agent-based lattice model
introduces spatial correlations
model. The CF (compared to MF) represents the relaxa- cooperator issuing extra signals but not listening to them.
tion of the assumption of infinitely large interaction We assume that the extra signal expression doubles the
groups and hence can enable compositional variance. The signal level so that a Bo individual counts as two signalers
agent-based lattice model introduces spatial correlations in its interaction group. “Smart” (Sm: CsR) is a coopera-
(that are missing both from MF and CF). tor that detects the quorum signal and cooperates if the
signal level in its immediate vicinity exceeds the quorum
Strategy set threshold. “Nerd” (Ne: CSR) is a quorum-sensitive coop-
Individual behaviors (strategies) are determined by three erator that always produces an extra signal dose. “Liar”
“functional genes” (heritable traits possibly encoded (Li: cSr) issues the quorum signal but never cooperates.
by a number of genes each) which control coopera- “Curious liar” (Cl: cSR) acts like Liar but also detects the
tion (C), extra signaling (S), and signal detection and signal. Finally, “Voyeur” (Vo: csR) only detects the signal
response (R) (see Fig. 3). Each of these genes can be in and never cooperates.
one of two states (i.e., they have two “alleles”): they may
be active (denoted by bold capitals: C, S, R) or inactive Metabolic costs
(bold minuscules: c, s, r). Note that an active coopera- Each player invests a fixed metabolic effort c0 into its own
tion gene (C) provides two things in the model: (i) the maintenance. This baseline metabolic burden is the same
public good and (ii) a baseline (cost-free) signal level for all strategies. Cooperation (C), the emission of extra
(hence “cue-based” cooperation). Accordingly, there are signal molecules (S), and the production, maintenance,
eight possible strategies (“phenotypes”). “Lazy” (La: csr) and operation of the signal response system (R) are all
never issues or detects the quorum signal and does not metabolically costly; the corresponding c , s, and r costs
cooperate. “Trusty” (Tr: Csr) cooperates unconditionally, are added to the baseline metabolic burden of the play-
as it does not communicate and neither gives nor listens ers expressing them, to yield the total metabolic cost of
to signals. “Bouncer” (Bo: CSr) is also an unconditional the corresponding genotype. Cooperators always express
Fig. 3 Strategy set of the QS model. Strategies are listed in boxes; their genotypes are denoted in bold typeface and their metabolic costs
in parentheses. Capital letters in genotypes indicate expressed “genes”; minuscules indicate inactive alleles. c0 is the baseline metabolic cost
paid by everyone, c is the cost of cooperation, θ = 1if the quorum threshold is reached (otherwise θ = 0), s is the signal production cost, and r
is the signal-detection cost. Underlined strategies are context dependent, capable of switching to cooperation when a signal quorum is reached.
The individual strategies are characterized as follows: Lazy does nothing; Voyeur detects signal but does not cooperate; Liar signals but does
not cooperate; Curious liar produces and detects signal but does not cooperate; Trusty does not communicate but always cooperates and thus
also issues the QS signal; Smart detects the signal and cooperates if the signal level exceeds quorum; Bouncy issues extra signal but does not listen
to it; Nerd produces the extra signal, detects the signal and cooperates if the signal level exceeds quorum
the public good at a low level, which constitutes an inevi- group of i , or if i is an unconditional cooperator; γi = 0
table cue of cooperation. The extra signal molecules and otherwise.
the signal response system are always expressed in all the
players carrying the active C and/or R genes for these The metabolic cost of various strategies
functions. The high-level expression of the cooperation Trusty always cooperates and signals, and thus it always
gene C means cooperation, which may be conditional pays the cost of cooperation (c ), but it enjoys the coop-
on the concentration of nearby signal molecules (i.e., eration benefit (b) only if the number of cooperators in
the number of signaling neighbors), provided that they its interacting group exceeds the critical threshold κ.
express the signal response system R and thus they are Bouncer behaves as Trusty, except that it produces twice
capable of detecting the signal. Cooperating players also as many signal molecules as Trusty at some extra cost s.
issue the cooperation cue at no additional cost, i.e., the Smart cooperates (and pays the cost c ) only if the num-
(honest) signal cost of cooperators is included in c. The ber of signal doses is at least Q in its interacting group
condition for the high-level expression of the coopera- but operating the response system listening to the signal
tion gene in conditional cooperators (the Smart and the of the others implies a small cost r . Nerd is a conditional
Nerd strategies) is that the number of signal doses within cooperator like Smart, producing an extra dose of signal
the QS neighborhood exceeds the QS threshold Q. Obvi- at the extra cost s. The strategy Liar never cooperates and
ously, non-expressed cooperation genes carry no meta- does not listen to the quorum signal, but it does produce
bolic cost. Figure 3 summarizes the total metabolic costs it, trying to induce conditional cooperators in its inter-
of the strategies with the local signal levels (i.e., the num- acting group, again at the cost s. This model assumes that
ber of signal doses within the QS neighborhood) below the quorum signal is the product of cooperation; thus,
and above the quorum signal threshold. the emission of a single dose of it is free of charge in all
potential and actual cooperators, but they can produce an
Cooperation benefit and fitness additional dose at a cost. Non-cooperators pay even for
The metabolic cost of a player with at least κ active coop- the first signal dose if they produce it, so that the Curious
erators in its own interaction group is reduced by a factor liar strategy both producing and detecting the signal pay
0 < b < 1, which is the cooperation benefit. The benefit the corresponding costs s and r ; the Voyeur strategy pays
reduces the metabolic cost of the individual in a multi- only r for detecting the signal. Figure 3 summarizes the
plicative manner. Notice that the cooperation threshold metabolic costs and benefits for all possible genotypes in
κ is not equivalent with the quorum signal threshold Q— interacting groups with the number of actual cooperators
these two thresholds are different, even if their values are below and above the cooperation threshold κ.
the same numerically. Q is the minimum number of quo-
rum signal doses within the interaction group of a con- Reproduction, mutation, evolution
ditional cooperator that is sufficient to switch its C gene Reproduction takes place during pairwise interactions
on, whereas κ is the minimum number of active coopera- between individuals, following the rules of probabil-
tors within a group necessary for members of the group istic imitation dynamics: player i has a chance of occu-
to enjoy the cooperation benefit. We assume κ = Q pying the site of its opponent j with its own offspring.
throughout this study, implying that this relation is the This has probability pij , proportional to the relative
Cj −Ci
evolutionary optimum for cooperators: any deviation by fitness (cost advantage) of i , as wij = C max
, where
switching on cooperation too late or too early is selected Cmax = c + s + r + bc0 is the largest possible cost dif-
against. The fitness wi of a player i is linearly decreasing ference (between an individual expressing all functional
with its actual metabolic cost Ci , that is: genes but not receiving benefit, and another one express-
ing none of the genes but receiving the full benefit). Then:
wi = w0 − Ci = w0 − (1 − θi b)(c0 + γi ci + si + ri ),
(1) pij = 0.5(1 + σ �wij ), (2)
where ci = c if the cooperation gene can be expressed in
player i , otherwise ci = 0; similarly, si and ri are the corre- where σ is the strength of selection. Obviously,
sponding costs of issuing an extra signal and responding pji = 1 − pij .
to a quorum of signals by player i . θi = 1 if the number of We assume that, during reproduction, any of the three
active cooperators in the interacting group that i belongs functional loci (C, S, and R) may mutate from its func-
to is at least κ, otherwise θi = 0, and γi = 1 if the num- tional to its inactive form in the offspring, and back-
ber of signal doses (i.e., the number of cooperators plus mutations are also allowed, with each of the six possible
the number of extra signalers) is at least Q within the mutation events occurring at its own specific rate (possibly
Fig. 4 Mean-field approximation, with the vector fields of the dynamics of the three feasible strategies (Lazy, trusty, Smart) in four different cases.
Parameters are as follows: A {r = 1/100, b = 1/10}, B {r = 1/100, b = 1/2}, C {r = 2/10, b = 65/100}, D {r = 1/100, b = 6/10}. Note that the parameter set
for A is not the same as in the other figures, as we wanted to show a case where Lazy does not win. In all cases, the rest of the parameters are {n = 9,
k = 0.3, c0 = 1, c = 0.3}
The configuration‑field (CF) model • Case 4: Finally, when the signal detection cost is low,
In the CF model, we also considered only the three fea- and the cooperation benefit is high enough, the sys-
sible strategies (Trusty, Smart, and Lazy), because none tem behaves like in case 2 except that an additional
of the others have a chance to persist in the presence of unstable fixed point and a saddle point occur on the
any of these three, just like in the MF approximation. The Lazy/Trusty polymorphic margin of the state space
replicator dynamics yields four characteristically differ- (background of Fig. 5D). This, however, does not
ent outcomes: change the final state of the system compared to case
2: it is either monomorphic Lazy or polymorphic
• Case 1: If cooperation is too costly compared to the Smart/Lazy, depending on initial conditions
benefit for both Trusty and Smart, and Lazy is the
only fixed point (stable one) of the system (back- For the detailed mathematical derivations of these
ground of Fig. 5A) results, see Additional file 1: Appendix 2 [21, 41].
• Case 2: With increasing benefit b, a stable and an
unstable polymorphic state of Smart and Lazy Agent‑based non‑spatial model
emerge. Which one of the two stable fixed points To address the effect of the finite size of interacting
(Smart/Lazy or Lazy), the system approaches depend groups, along with the spatial constraints arising from
on the initial frequencies (background of Fig. 5B) limited agent mobility and local (i.e., neighborhood-)
• Case 3: If the benefit of cooperation is high, but sig- interactions, we developed an agent-based implementa-
nal detection is costly, then Smart and Trusty swap tion of the CF model, with its basic assumptions as pre-
roles: a pair of stable and unstable polymorphic Lazy/ sented above.
Trusty fixed points emerge besides the pure Lazy Not surprisingly, the trajectories of the non-spatial
stable state, and the signal detection gene gets lost agent-based simulation model (Fig. 6, row 3 red lines)
(background of Fig. 5C) closely trace the vector fields of the CF approximation
Fig. 5 Configuration-field approximation, with the vector fields of the dynamics of the three feasible strategies (Lazy, trusty, Smart) in four different
cases. A Lazy is the only fixed point of the dynamics; parameters are b = 0.3, r = 0.01. B The monomorphic Lazy and the polymorphic Smart/
Lazy states are the stable fixed points of the dynamics; b = 0.5, r = 0.01. C The monomorphic Lazy and the polymorphic Trusty/Lazy states are
the stable fixed points of the dynamics; b = 0.65, r = 0.2. D The monomorphic Lazy and the polymorphic Smart/Lazy states are the stable fixed
points of the dynamics; b = 0.6, r = 0.01. All CF systems admit multiple unstable fixed points as well, including the Smart and the Trusty corners
of the state space. In all cases, the rest of the parameters are N = 9, κ = 3, c0 = 1, c = 0.3
Fig. 6 Sample trajectories of the configuration-field (CF) and the spatially explicit lattice model on the corresponding vector fields. Functional
mutations have not been considered here. Row 1: Spatially explicit simulations, no mixing (D = 0.0) (simulated vector fields). Row 2: Spatially
explicit simulations, strong mixing (D = 10.0) (simulated vector fields). Row 3: Non-spatial agent-based simulation model with CF approximations
(analytical vector fields in the background) D = 0.0. In all cases, parameters are as follows: N = 9, κ = 3, c0 = 1.0, c = 0.3; in the four columns: A
b = 0.3, r = 0.01; B b = 0.5, r = 0.01; C b = 0.65, r = 0.2; D b = 0.6, r = 0.01
(Fig. 6 row 3 backgrounds, where the parameters of three feasible strategies in all simulation runs. Results
the cases A–D correspond to Fig. 5A–D parameters) at differ from those of the lattice-based CF approximation
any parameter setting. The only difference between the (Fig. 6, trajectories of row 3, with the vector-field of the
assumptions of the two models is that the CF approxima- analytic CF as background), both at high and zero cell
tion assumes an infinite population size, whereas in the mobility ( D = 10.0 and D = 0.0, respectively, see Fig. 6,
non-spatial simulations, lattice size is P = 90.000. This rows 1–2). The empirical (simulated) vector fields reveal
accounts for the stochastic noise on the simulated trajec- that the spatially explicit model results in the least coop-
tories compared to the background CF vectors. erative steady-state populations (backgrounds of Fig. 6,
rows 1–2, columns B–D): almost all the fixed points are
Agent‑based spatial (lattice) model at, or very close to, the Lazy corner of the strategy sim-
Three‑strategy model without functional mutations plex. This means far worse conditions for cooperators
For comparative purposes, first we followed the trajecto- compared to the CF approximation (Fig. 6, row 3, col-
ries of the three feasible strategies (Lazy, Trusty, Smart) umns B–D) that allows for polymorphic steady states
in the lattice model, once again omitting those with at a considerable part of its parameter space. As the
no chance to persist in the MF and the CF approxima- only difference between the CF and the well-mixed lat-
tions (see Additional file 1: Appendix 1, 2) (Fig. 6, rows tice model is that in the lattice model, the competing
1–2). Mutations in any of the three functional loci were agents are immediate neighbors with overlapping coop-
ignored. The initial strategy distribution was even for the eration neighborhoods, this may seem a counterintuitive
Diffusion (D)
The most conspicuous trend is also the most obvious local neighborhoods are predictable enough to render QS
one: increasing agent motility ( D) benefits the parasitic a futile waste of resources. It is also worth noting that the
strategy (Lazy), giving it more access to the (undeserved) CF approximation (rightmost columns of panels in Fig. 8
cooperation benefit provided by unconditional and con- and Additional file 1: Appendix 3, Figure S6) behaves
ditional cooperators (Trusty and Smart, respectively; almost like the lattice model at low mixing (small D),
Additional file 1: Appendix 3, Figure S6). There are, how- once again underlining the negative effect of the overlap-
ever, a few more effects of increasing diffusion which ping cooperation neighborhoods of adjacent competitors.
are less obvious. One is the increasing proportion of the
Smart (quorum sensing) strategy among the decreas- Cooperation threshold (κ)
ing number of cooperators, which makes perfect sense The quorum signal response threshold (the minimum
in a population with local neighborhood configurations number of signals in a neighborhood that switches on
highly variable in terms of the number of cooperators in conditional cooperation in Smart agents) and the cooper-
them. It is in this case that the small cost of quorum sig- ation threshold (the minimum number of actual coopera-
nal detection pays off by sparing unnecessary cooperation tors providing the cooperation benefit for the focal agent
costs at low local cooperator density but switching on of the neighborhood) are assumed to be the same in all
cooperation in sufficiently cooperative neighborhoods. our models. For cooperation to be an option, κ = 2 is the
At very low and very high agent motilities, however, the
minimum requirement. The potential maximum number signals) appear only in a narrow section of the param-
of cooperators within a Moore neighborhood is κ = 9, eter space (at low costs for both QS signaling s and QS
but cooperation would be compulsory for all the agents at signal response r ), and even there they are present at
κ = 9, and we have found it almost impossible to evolve low frequencies (Additional file 1: Appendix 3, Figure
for κ > 6, so the feasible range to scan was κ = 2 to 6. S6). Cooperators trying to exaggerate their cooperative
Additional file 1: Appendix 3, Figure S6 clearly shows nature (Bouncer: CSr, and Nerd: CSR) do not appear
that, contrary to common intuition, cooperation evolves in the simulations in frequencies above their mutation-
to its highest frequency in the population at higher inter- selection balance.
mediate values of the cooperation threshold (at κ > 3)
[21, 42]. Approaching κ = 6 unconditional cooperators Discussion
(Trusty) often completely exclude parasites (Lazy) at low Cheap quorum sensing supports cooperation
agent motility ( D), whereas parasite/cooperator coexist- Since the functioning cooperation allele C is also a cue
ence is the generic outcome of spatial simulations at low for potential interacting partners on future cooperation,
κ and D, like in the CF model. Towards the high end of possession of a fully functional QS system means har-
the κ scale (upward from κ = 6), cooperation abruptly boring a functional response gene R and a conditionally
disappears in a phase-transition-like manner, leaving the expressed cooperation allele C. Figure 8 reveals that in
system in the monomorphic Lazy steady state across the spatially explicit, viscous systems a cheap response func-
whole parameter space with κ ≥ 7 (data not shown). tion (small r ) of QS substantially increases the overall
propensity for metabolically much more expensive coop-
QS response cost (r) eration. In this case, it is mostly conditional coopera-
The metabolic cost of QS signal response (signal detec- tors (Smart, CsR) that coexist with defectors (Lazy, csr).
tion and intracellular signal transduction) is assumed to More expensive R alleles prevent both signal response
be low compared to the cost of cooperation (r ≪ c ). This and cooperation, with only defectors surviving at mod-
is reasonable, given that cheap cooperation coordinated erate to high cell motility (D). In the CF approximation,
by expensive signal response is certainly a losing strategy cheap signal response (r = 1) allows the Smart strategy
against cheap cooperation with no QS at all because con- to attain high frequency in the steady state, but it does
stitutive cooperation would then be cheaper than listen- not noticeably contribute to the evolutionary success of
ing to the QS signal. If signal response is cheap (r = 1.0), cooperation: at higher response costs the Trusty popula-
then a considerable fraction of cooperating agents main- tion performs just as well without the R allele (see Fig. 5).
tain and use it within almost the entire range of the
parameter space, except where the parasitic Lazy popu- Spatial correlations do not favor cooperation in all contexts
lation goes extinct (Additional file 1: Appendix 3, Fig- The well-known prediction of almost any game theo-
ure S6). Obviously, with all parasites wiped out, it is not retical model on “uninstructed” cooperation (in which
worth listening to QS signals anymore, so Trusty takes individuals have no prior clue on the intent of coopera-
over. Also, at low agent motility (smaller D values), main- tive or defective behavior of their interacting partners)
taining a more expensive QS signal response (r = 5.0 is that the intensive mixing of cooperators and defectors
and 10.0) proves not to be feasible. An interesting effect destroys costly cooperation through the inevitable fitness
of cheap to moderately expensive QS signal response advantage of defectors in cooperator/defector encoun-
(r = 1.0 and 5.0) at D = 0.4 and κ = 6 is that it helps to ters. This universal conclusion holds true in the threshold
eliminate the parasite, even though it is ultimately not public goods game with competition restricted to neigh-
present in the steady-state population. More expensive bors, even if all cooperators are assumed to constitutively
signal response (r = 10.0, with all other parameters the broadcast their cooperative nature with a signal that their
same) results in parasite takeover. potential partners may detect at a small cost and switch
on or off their cooperation mechanism accordingly. This
QS signal cost ( s) is obvious from Fig. 6: defectors take over as the mix-
Recall that cooperation always means synchronized sign- ing parameter D increases in the agent-based, spatially
aling in our models, so an agent that issues an extra signal explicit simulations, irrespective of all other parameter
is a liar: it is either not a cooperator but pretends to be values. However, maintaining spatial correlations does
one, or is a cooperator, but promises more cooperation not always favor cooperation in any context. On the one
than it actually provides. Given that cheating is usually hand, spatial correlations always favor cooperation dur-
a profitable strategy in cooperative situations, it is quite ing the fitness acquisition phase, as it allows coopera-
surprising how efficient QS is in eliminating it. Liars (cSr tors to be more likely surrounded (and helped) by other
genotypes that do not cooperate and do not listen to QS cooperators; thus, cooperation is more likely to succeed.
On the other hand, spatial correlations favor defectors at cooperators are needed for harvesting the cooperation
the competitive reproduction stage as it hurts success- benefit), the higher the steady-state frequency of the
ful cooperators to compete against (potentially also suc- cooperators will be [21, 42], and at moderately high κ val-
cessful) defector neighbors. In other words, cooperators ues, the defectors may be eradicated altogether. On the
benefit from cooperating locally but competing globally one hand, this is a natural consequence of requiring more
for reproduction, suggesting an ambivalent effect of spa- cooperators per interaction neighborhood which trans-
tial mixing on cooperation in TPGG. This ambivalent lates to higher overall (regional) cooperator abundance
effect is dependent on the temporal scales of the two in mixed steady states, but on the other hand, it is sur-
types of interaction assumed. The advantage of coopera- prising that increasing κ does not favor defection—on the
tors in the CF model (as opposed to that of cheaters in contrary, it seals the fate of defectors. This effect shows
the well-mixed spatially explicit simulations) hinges on up at low agent mobility in the spatially explicit simula-
the implicit assumption of the instantaneous collection tions, but the trend of increasing cooperator frequency
of fitness by all agents in the cooperation phase and their with increasing k can also be observed in the CF approxi-
also instantaneous competitive interaction in the imita- mation, which does not provide the advantage of coop-
tion phase with a random opponent at a later moment erator clumping at all, as it assumes complete mixing of
in time. In both models, intensive population mixing is the population. At κ values too close to the neighborhood
assumed, but in the CF, the two interaction phases are size the steady-state populations are also monomorphic,
farther apart in time, whereas in the spatially explicit but then the defectors take over the lattice in a phase
simulations, they are simultaneous. It should also be transition-like manner with increasing κ.
noted here that mixing may also have a detrimental effect
on competing cooperators by allowing more frequent Liars are kept at bay if cooperativity is always visible, even
cooperator-parasite encounters in space. This and the if fake signaling is unconstrained
dissolution of cooperator clumps eventually always lead Yet another interesting prediction of the spatially explicit
to parasite takeover at fast population mixing in the spa- agent-based model is the limited range of the parameter
tially explicit lattice model. space in which cheating through issuing fake quorum
signal (lying) can evolve at all: the only section of the
Quorum sensing evolves if it conveys valuable information parameter space of the 8-strategy lattice model in which
QS is efficient at reducing the overall metabolic cost (and the Liar strategy shows up is at very low signal cost and
thus at increasing the average fitness of the population) low response costs (Fig. 6, s = 1, r = 1, and r = 5), and
if the expected number E(nc ) of cooperators per neigh- even there it attains almost negligible frequencies, in
borhood is close to κ (the cooperation threshold) in the spite of the fact that the S locus is free to mutate back
actual steady state of the population, and the variance and forth, just as the other two (C and R). The fact that
of the same variable, V (nc ), is relatively high. These are lying is always disadvantageous from the viewpoint of
the conditions at which it is difficult to predict if a par- evolving cooperation can be seen in the reduction of the
ticular neighborhood does or does not have the quorum frequency of cooperative strategies wherever Liars occur
of cooperators (nC ≥ κ); therefore, the information that in the population, but in all other parts of the parameter
the QS system provides is of the highest fitness advan- space, Liars are missing from the steady state. Lying is
tage. This is quite obvious in the spatially explicit lattice inefficient due to the fact that cooperators can produce
model versions but also applies to the CF approxima- a baseline signal for free in our model while liars have to
tion that explicitly considers the stochastic heterogene- pay for the same signal. In other words, there is a con-
ity of neighborhood composition, too. Of course, the MF dition-dependent trade-off that favors honesty. Theoreti-
approximation never yields QS because it assumes all cal models of honest signaling have shown that honesty
neighborhoods to be identical and thus completely pre- is maintained by such condition-dependent trade-offs
dictable, in which case maintaining the signal-detecting [43–48], instead of the equilibrium cost of signals (a.k.a.
R allele (i.e., expressing the signal receptor and the signal “handicaps”) as predicted by the erroneous Handicap
transduction system) would be a waste of resources and Principle [49] (see [50] for discussion), and later costly
is thus selected against. signaling models [51, 52]. While such trade-offs are dif-
ficult to measure, a recent experiment [53] supports the
High cooperation thresholds favor monomorphic key role of condition-dependent trade-offs in maintain-
equilibria ing honesty. In turn, our model gives an example of cheap
A counter-intuitive prediction of the spatially explicit (cost-free) and honest signaling under conflict of inter-
version of the threshold public goods game model is that est. This possibility was predicted long ago (e.g., [43, 45,
the higher the cooperation threshold κ (i.e., the more
54], but implementations of such systems are few and far approximations. This allowed us to resolve simple sce-
between. narios, and the approximations provided benchmarks
Our results can be seen as a version of the famous for the agent-based simulations. While this allows a thor-
greenbeard effect. This effect was proposed by Dawk- ough study of the cue-based QS system, there are several
ins (1976) based on the kin selection idea of Hamilton issues that merit further investigation. First, we assumed
(1962). The core of the idea is assortment based on a sig- that the cooperation threshold is the same as the activa-
nal called greenbeard that allows cooperators to recog- tion threshold for conditional cooperators. The reasoning
nize each other. There is a long debate in the literature behind this assumption is that conditional cooperators
whether greenbeards exist or not, reviewed by Madg- switching on too early or too late will be selected against.
wick et al. (2019). One obvious weakness of the idea is This assumption can be investigated with a slightly
the possible presence of cheaters called “fakebeards” that modified version of this model. The main reason for not
produce the signal yet do not cooperate. It was shown including this quite obvious and reasonable complication
that such fakebeards can destabilize the system result- in the current version is to keep the complexity of the
ing in dynamic polymorphic equilibria with rare-type investigation under control. Second, perhaps more inter-
advantage, in which mutation-selection balance will set estingly, our results beg the question: why is it that not all
the proportions of honest vs. dishonest types (Jansen & QS systems are cue-based? If cue-based systems promote
van Baalen, 2006). The key difference between our model honesty and cooperation, then one might expect such
and the previous models/ideas is that previous models systems to be widespread in nature. Yet, while there are
assumed cost-free signals, that is, each type could pro- cue-based systems besides the nisin system in Lactococ-
duce the signal at the same (no) cost. In contrast, as dis- cus lactis, like the mutacin 1140 system in Streptococcus
cussed above, we assume that different types pay different mutans [56], the mersacidin system in Bacillus sp. [57],
marginal costs when producing the signal. Our results or the listeriolysin S (LLS) system in Listeria monocy-
show that this is a crucial difference that can stabilize togenes [58, 59], there are many more examples in which
cooperation. This results in an evolutionarily stable coex- the signal molecule is different from the public good. It
istence of cooperative and defective types (Smart, Trusty, is conspicuous that most known cue-based QS systems
Lazy) as opposed to the dynamic coexistence identified regulate autoinduced bacteriocin (lantibiotic) produc-
by the multicolor chromodynamics model (Jansen & van tion, i.e., toxin excretion, possibly deployed against unre-
Baalen, 2006) that has to assume a steady influx of new lated competitor strains. Perhaps, cue-based systems are
honest types. also constrained by some other, still unknown, genetic, or
While several models have investigated potentially ecological mechanism. This leads to the issue of how cue-
deceptive QS strategies in bacteria [36, 38, 55], there are based QS systems had evolved in the first place—a poten-
key differences between the current investigation and the tially rewarding focus for future investigations.
previous ones. First of all, we investigate a cue-based sys- Our models provide several testable predictions.
tem, in which the signaling molecule is the same as the
public good. This provides dynamics different from those (i) Cue-driven systems will produce evolutionarily
of the traditional QS systems with separate signal and stable polymorphism. This implies that cue-driven
public good molecules, which, therefore, have to be syn- systems will consist of different types including self-
thesized along different metabolic pathways. Of course, ish and cooperator ones. On the practical level
not all QS systems are cue-based, but our results suggest this means that contribution to the public good is
that whenever it is cue-based, it strongly favors honesty expected to show a high variance (i.e., there should
and cooperation. Secondly, like Czárán and Hoekstra be a type consistently producing high level of public
(2009), we investigate the full range of 8 strategies pos- good vs. a different type consistently not producing
sible, given the assumption that each of cooperation, any) and this variance is expected to be stable on the
signaling, and signal detection may be ON or OFF, but long term.
in the present model, cooperators that are also signal- (ii) These systems will be dominated by honest coop-
ers are a new type of cheat: those who exaggerate their erators, unconditional cooperators and defectors
promise of cooperation. In spite of even more cheating (i.e., Smart, Trusty, Lazy). On the practical level, this
strategies being possible, we find a substantially reduced implies that there should be cooperators sensitive to
prevalence of cheats in the population. Last but not least, the level of the cue (honest cooperators), cooperators
we have decomposed and separated the effects of spatial insensitive to the cue (i.e., always producing the pub-
and temporal constraints present in the spatiotemporal lic good, unconditional cooperators). and finally self-
agent-based simulation model by partially adding or elim- ish individuals insensitive to the cue, i.e., never pro-
inating them in the mean-field and configurational-field ducing any public good.
(iii) The frequency of cheating strategies (fakebeards) averaged across the entire habitat so that the fitnesses of
will be low. Fakebeards in this system are individuals the strategies depend only on the average frequencies of
that produce the public good at a low level, sufficient the strains present. On the basis of simple fitness con-
to act as a cue (thus triggering conditional coopera- siderations, it holds that only combinations of the Lazy,
tors), yet they never up the production of the pub- Trusty, and Smart strategies can constitute any equilib-
lic good even if the concentration is high (i.e., even rium state in the MF approximation; thus, we have to
if there are other cooperators in the vicinity). Such analyze the dynamics only for these three strategies. To
difference in concentration between the “cue” and see this, let us assume indirectly that there is a fourth
the public good might be the most difficult thing to strategy S present in the equilibrium. Since at least one
identify. To simplify this prediction, one can say that of the strategies from among Lazy, Trusty, or Smart has
there should be (i) a type that is capable of increased a higher fitness than strategy S in any actual state of the
production of the public good (as a function of the frequency vector x , S can not be present in a dynamical
concentration of the cue, honest cooperators) vs. equilibrium. For the mathematical formulation of the MF
(ii) a different type that is only producing the pub- model, see Additional file 1: Appendix 1.
lic good (cue) at a much lower level and this type is
insensitive to the concentration of the cue, thus never Configuration‑field (CF) model
switching to an increased production. Next, we assume that the population is still very
large (practically infinite), but individuals form ran-
Conclusions dom interacting groups of finite size N . While in the
Explaining the evolutionarily stable maintenance of previous section we considered N to be so large that
cooperation in cooperative dilemma games, such as the each interacting group consists of strategies in exact
public goods game, remains a challenge. Our results proportion to their global frequencies in the popula-
demonstrate that cue-based quorum sensing can main- tion, now we assume that N is smaller, and thus dif-
tain cooperation at a high level, partly due to the con- ferent interaction groups with different configurations
ventional kin selection mechanism of cooperative games, of strategies are formed in an inherently stochastic
partly as the result of specific spatial effects in local public manner, due to sampling errors. The two players par-
goods games. The frequency of communication cheaters ticipating in an elementary game step are randomly
is constrained in such a system by the condition-depend- chosen members of their own interaction groups
ent trade-off in signal production (i.e., signaling is free (both of them of size N ) that are drawn at random
for honest signalers but costly for cheaters). Our results from the population. Note that assuming the overall
may apply to any other cue-based systems, i.e., any other fitness for each of the eight strategies is calculated as
threshold public goods game where the game a preceded the weighted average of its local fitness in all possible
by a signaling stage, where the signal is some contribu- configurations of the interaction group around a focal
tion to the public good. For example, cooperative fishing individual of the given strategy. Based on these fitness
in humans where the signal is the preparation, repair of formulae, it can be shown again that the feasible strat-
fishing equipment. Such cue-based system could provide egy set consists only of Lazy, Trusty, and Smart (for
a robust stepping stone for large-scale cooperation as more details, see Additional file 1: Appendix 2).
they do not require any accounting mechanism (cognitive
or otherwise) as they function without the punishment Agent‑based non‑spatial model
of cheaters. In turn, this implies that they do not require The agent-based non-spatial model is an accurate
any complex social norm that would regulate such action. computational realization of the CF approximation in
The simplicity of assumption behind the cue-driven sys- all respects, except the population is now finite (pop-
tem and the simplicity of the mechanism itself makes it ulation size P = 90.000 ). Two random samples of size
a prime candidate as driving force of cooperation from (N = 9), drawn from the population independently,
bacteria to humans. represent the interacting groups. One focal agent from
each group is chosen at random, and these two focal
Methods agents play out the imitation game, with the chance of
Mean‑field (MF) model winning for each dependent on the fitness payoff it has
In Additional file 1: Appendix 1, we construct the mean- realized within its own interacting group. Payoffs were
field version of the model, with the assumptions that assigned to each of the two players depending on their
population size ( P ), interacting group size ( N ), and the own cooperation costs and the number of actual coop-
cooperation threshold κ are all very large, while N /P ≪ 1 erators in their own group, which, in turn, depends
and κ/N → K . In this limit, all the dynamical effects are on the local number of signalers and conditional
3. Diggle SP, Gardner A, West SA, Griffin AS. Evolutionary theory of bacterial 29. Cornforth DM, Popat R, McNally L, Gurney J, Scott-Phillips TC, Ivens A,
quorum sensing: when is a signal not a signal? Philos Trans R Soc Lond B et al. Combinatorial quorum sensing allows bacteria to resolve their social
Biol Sci. 2007;362(1483):1241–9. and physical environment. Proc Natl Acad Sci U S A. 2014;111(11):4280–4.
4. West SA, Griffin AS, Gardner A, Diggle SP. Social evolution theory for 30. Fuqua WC, Winans SC, Greenberg EP. Quorum sensing in bacteria: the
microorganisms. Nat Rev Microbiol. 2006;4(8):597–607. LuxR-LuxI family of cell density-responsive transcriptional regulators. J
5. Nealson KH, Platt T, Hastings JW. Cellular control of the synthesis and Bacteriol. 1994;176(2):269–75.
activity of the bacterial luminescent system. J Bacteriol. 1970;104:313–22. 31. Swift S, Throup JP, Williams P, Salmond GP, Stewart GS. Quorum sensing: a
6. Karray F, Abdallah MB, Kallel N, Hamza M, Fakhfakh M, Sayadi S. Extracel- population-density component in the determination of bacterial pheno-
lular hydrolytic enzymes produced by halophilic bacteria and archaea type. Trends Biochem Sci. 1996;21(6):214–9.
isolated from hypersaline lake. Mol Biol Rep. 2018;45(5):1297–309. 32. Salmond GPC, Bycroft BW, Stewart GSAB, Williams P. The bacterial
7. van der Ploeg JR. Regulation of bacteriocin production in Streptococcus ’enigma’: cracking the code of cell–cell communication. Mol Microbiol.
mutans by the quorum-sensing system required for development of 1995;16(4):615–24.
genetic competence. J Bacteriol. 2005;187:3980–9. 33. Redfield R. Is quorum sensing a side effect of diffusion sensing? Trends
8. Fontaine L, Boutry C, Guedon E, Guillot A, Ibrahim M, Grossiord B, et al. Microbiol. 2002;10(8):365–70.
Quorum-sensing regulation of the production of Blp bacteriocins in 34. Chen R, Déziel E, Groleau M-C, Schaefer AL, Greenberg EP. Social cheating
Streptococcus thermophilus. J Bacteriol. 2007;189:7195–205. in a Pseudomonas aeruginosa quorum-sensing variant. Proc Natl Acad
9. Stintzi A, Evans K, Meyer JM, Poole K. Quorum-sensing and siderophore Sci U S A. 2019;116(14):7021–6.
biosynthesis in Pseudomonas aeruginosa: lasR/lasI mutants exhibit 35. Brown SP, Johnstone RA. Cooperation in the dark: signalling
reduced pyoverdine biosynthesis. FEMS Microbiol Lett. 1998;166:341–5. and collective action in quorum-sensing bacteria. Proc Biol Sci.
10. Popat R, da Silva AC, Easton SAS, McNally L, Williams P, Diggle SP. Environ- 2001;268(1470):961–5.
mental modification via a quorum sensing molecule influences the social 36. Wang Y, Rattray JB, Thomas SA, Gurney J, Brown SP. In silico bacteria
landscape of siderophore production. Proc R Soc B. 2017;284:20170200. evolve robust cooperation via complex quorum-sensing strategies. Sci
11. Rutherford ST, Bassler BL. Bacterial quorum sensing: its role in viru- Rep. 2020;10(1):8628.
lence and possibilities for its control. Cold Spring Harb Perspect Med. 37. Kleerebezem M. Quorum sensing control of lantibiotic production;
2012;2:a012427. nisin and subtilin autoregulate their own biosynthesis. Peptides.
12. Nadell CD, Xavier JB, Levin SA, Foster KR. The evolution of quorum sens- 2014;25(9):1405–14.
ing in bacterial biofilms. PLoS Biology. 2008;6:e14. Available from: https:// 38. Czárán T, Hoekstra RF. Microbial communication, cooperation and cheat-
doi.org/10.1371/journal.pbio.0060014 ing: quorum sensing drives the evolution of cooperation in bacteria. PLoS
13. Becker F, Wienand K, Lechner M, Frey E, Jung H. Interactions mediated One. 2009;4:e6655.
by a public good transiently increase cooperativity in growing Pseu- 39. Czárán T. Spatiotemporal models of population and community dynam-
domonas putida metapopulations. Sci Rep. 2018;8(1):4093. ics. Springer US; 1998. Available from: https://books.google.hu/books?id=
14. Chuang JS, Rivoire O, Leibler S. Cooperation and Hamilton’s rule in a HvIA28t6ZmoC
simple synthetic microbial system. Mol Syst Biol. 2010;6(1):398. 40. Hofbauer J, Sigmund K. Evolutionary games and population dynamics.
15. Rosenthal AZ, Qi Y, Hormoz S, Park J, Li SH-J, Elowitz MB. Metabolic Cambridge, MA: Cambridge University Press; 1998.
interactions between dynamic bacterial subpopulations. Ackermann M, 41. Phillips GM. Interpolation and approximation by polynomials. New York,
editor. eLife. eLife Sciences Publications, Ltd; 2018;7:e33099. Available NY: Springer; 2003.
from: https://doi.org/10.7554/eLife.33099 42. Archetti M, Scheuring I. Review: Game theory of public goods in one-
16. Veening JW, Stewart EJ, Berngruber TW, Taddei F, Kuipers OP, Hamoen LW. shot social dilemmas without assortment. J Theor Biol. 2012;299:9–20.
Bethedging and epigenetic inheritance in bacterial cell development. Available from: http://www.sciencedirect.com/science/article/pii/S0022
PNAS. 2008;105(11):4393–4398. Available from: https://doi.org/10.1073/ 519311003171
pnas.0700463105 43. Számadó S. The validity of the handicap principle in discrete action–
17. Strassmann JE, Queller DC, Avise JC, Ayala FJ. In the light of evolution V: response games. J Theor Biol. 1999;198(4):593–602.
cooperation and conflict. PNAS. 2011;108:10787–91. 44. Számadó S, Czégel D, Zachar I. One problem, too many solutions: how
18. Nowak MA, May RM. Evolutionary games and spatial chaos. Nature. costly is honest signalling of need? PLoS One. 2019;14(1):e0208443.
1992;359:826–9. 45. Lachmann M, Számadó S, Bergstrom CT. Cost and conflict in animal sig-
19. Kümmerli R, Griffin AS, West SA, Buckling A, Harrison F. Viscous medium nals and human language. Proc Natl Acad Sci USA. 2001;98(23):13189–94.
promotes cooperation in the pathogenic bacterium Pseudomonas 46. Bergstrom CT, Számadó S, Lachmann M. Separating equilibria in
aeruginosa. Proc Biol Sci. 2009;276(1672):3531–8. continuous signalling games. Philos Trans R Soc Lond B Biol Sci.
20. Vásárhelyi Z, Scheuring I. Invasion of cooperators in lattice populations: 2002;357(1427):1595–606 Johnstone RA, Dall SRX, editors.
linear and non-linear public good games. Biosystems. 2013;113(2):81–90. 47. Hurd PL. Communication in discrete action-response games. J Theor Biol.
21. Archetti M, Scheuring I. Coexistence of cooperation and defection in 1995;174(2):217–22.
public goods games. Evolution. 2010;65(4):1140–8. 48. Számadó S, Zachar I, Czégel D, Penn DJ. Honesty in signalling games is
22. Miller MB, Bassler BL. Quorum sensing in bacteria. Annu Rev Microbiol. maintained by trade-offs rather than costs. BMC Biol. 2023;21(1):4.
2001;55:165–99. 49. Zahavi A. Mate selection—a selection for a handicap. J Theor Biol.
23. Sturme MHJ, Kleerebezum M, Nakayama J, Akkermans ADL, Vaughan 1975;53(1):205–14.
EE, de Vos WM. Cell to cell communication by autoinducing peptides in 50. Penn DJ, Számadó S. The Handicap Principle: how an erroneous hypoth-
gram-positive bacteria. Antonie Van Leeuwenhoek. 2002;81:233–43. esis became a scientific principle. Biol Revi. 2020;95(1):267–290. Available
24. Papenfort K, Bassler BL. Quorum sensing signal–response systems in from: https://doi.org/10.1111/brv.12563
Gram-negative bacteria. Nat Rev Microbiol. 2016;14(9):576–88. 51. Grafen A. Biological signals as handicaps. J Theor Biol. 1990;144(4):517–46.
25. Schuster M, Sexton DJ, Diggle SP, Greenberg EP. Acyl-homoserine lactone 52. Godfray HCJ. Signalling of need by offspring to their parents. Nature.
quorum sensing: from evolution to application. Annu Rev Microbiol. 1991;352(6333):328–30.
2013;67(1):43–63. 53. Számadó S, Samu F, Takács K. Condition-dependent trade-offs maintain
26. Charlesworth JC, Beloe C, Watters C, Burns BP. Quorum sensing in honest signalling. R Soc Open Sci. 2022;9(10):220335.
archaea: recent advances and emerging directions. Biocommun Archaea. 54. Számadó S. The cost of honesty and the fallacy of the handicap principle.
2017:119–32. Animal Behav. 2011;81(1):3–10.
27. Paggi RA, Martone CB, Fuqua C, Castro RE. Detection of quorum sensing 55. Gurney J, Azimi S, Brown SP, Diggle SP. Combinatorial quorum sensing in
signals in the haloalkaliphilic archaeonNatronococcus occultus. FEMS Pseudomonas aeruginosa allows for novel cheating strategies. Microbiol-
Microbiol Lett. 2003;221(1):49–52. ogy. 2020;166(8):777–84.
28. Henke JM, Bassler BL. Three parallel quorum-sensing systems regulate 56. Kreth J, Merritt J, Shi W, Qi F. Co-ordinated bacteriocin production and
gene expression in Vibrio harveyi. J Bacteriol. 2004;186:6902–14. competence development: a possible mechanism for taking up DNA
from neighbouring species. Mol Microbiol. 2005;57(2):392–404.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in pub-
lished maps and institutional affiliations.
1. use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
control;
2. use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
3. falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
writing;
4. use bots or other automated methods to access the content or redirect messages
5. override any security feature or exclusionary protocol; or
6. share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at