competition sensing the social side

8/9/2019 Competition Sensing the Social Side

1/9

Ecological competition refers to the pro-cess by which one individual decreases thesurvival or reproduction of others, andecologists have long recognized that thiscompetition can be divided into two majortypes: exploitative competition and interfer-ence competition1,2. Exploitative competitionis indirect and occurs when one organismconsumes the resources of another, such asa polar bear eating a seal and thus deprivinganother bear of food. Exploitative competi-tion also occurs in microorganisms and isparticularly strong when they settle on sur-faces and form dense communities, such asbiofilms. In these cases, nutrient limitationis common, resulting in strong exploitativecompetition among both cells of the same

genotype and those of different genotypes3,4.When exploitative competition occurs

purely among cells of one genotype, there isno true competition in an evolutionary sensebecause all cells have the same evolutionaryinterests. Under these conditions, evolution-ary theory predicts that cells will invest ingroup-beneficial phenotypes that reducenutrient competition, such as the produc-tion of enzymes that release nutrients fromthe environment, and these phenotypes doemerge in nature4. However, cells also com-monly meet and mix with cells of different

genotypes, as evidenced by the incrediblediversity found in metagenomic analyses ofmicroorganisms in environments as variedas soil, stromatolites and the mammaliangut57. Although mixed-genotype groupscan sometimes engage in mutually beneficialcooperation8, genotypic mixing is expectedto limit the potential for group-beneficialphenotypes because of the potential forcheaters, which use the beneficial secre-tions of other genotypes without themselvesinvesting in the secretion4.

Further than just limiting cooperation,the strong natural selection that results fromexploitative competition among differentgenotypes is predicted to generate interfer-ence competition, the second form of ecologi-

cal competition. Interference competition isthe competition that arises when individualsdirectly harm each other2,911. In animals,this form of competition is exemplified byphysical fighting. In microorganisms, theemergence of interference competition iswell established in the large body of literatureabout the secretion of products that harmother cells, including antibiotic compounds(reviewed below) and smothering poly-mers12,13. Co-culture experiments show thatthese secreted factors often determine whichgenotype prevails in mixed cultures13,14. There

is also evidence that toxins drive the co-evolution of defensive measures. Antibioticresistance genes are commonplace in naturalisolates and, importantly, evolved many yearsbefore the clinical use of antibiotics15.

There is good evidence, then, that bothexploitative and interference competitionare prevalent in bacterial communities andstrongly influence the outcome of naturalselection on bacteria911. Although it is clearthat such natural selection can drive theevolutionary dynamics of social traits suchas the secretion of toxins or beneficial mol-ecules, little attention has been paid to theeffects of ecological competition on the reg-ulation of these traits. In this Opinion article,we argue that ecological competition haswide-ranging effects on the form and func-tion of bacterial regulatory networks and, inparticular, that these effects can be seen inseveral of the well-studied stress responses.

The competition-sensing hypothesis

Our hypothesis is that bacteria have evolvedways to directly detect and respond to eco-logical competition, and that these responsesare reflected in several of the major bacterialstress responses. We argue that by detectingharmful effects on the cell, stress responsesrobustly sense ecological competition andallow cells to respond. We call this competi-tion sensing: a physiological response thatdetects harm caused by other cells and thatevolved, at least in part, for that purpose. Inthis context, the word competition includesall examples of ecological and evolutionarycompetition. As discussed above, we takeecological competition to be any situation inwhich one bacterial cell has a negative effecton the survival and reproduction of anotherbacterial cell. By contrast, evolutionarycompetition is more restrictive and dependson whether the interacting cells have differ-ent evolutionary interests. When exploita-tive competition occurs among cells of onegenotype (hereafter referred to as self cells),evolutionary competition is absent becausethe cells have the same evolutionary inter-

ests. Conversely, when exploitative competi-tion comes from cells of a different strain orspecies (hereafter referred to as other cells),evolutionary competition also occurs4.

We do not restrict ourselves to evolution-ary competition here because it is impor-tant for cells to detect harm irrespective ofwhether it is caused by self or by other cells.The appropriate response to this competi-tion will depend on the source of the compe-tition. However, the genetic diversity foundin many microbial communities might meanthat sensing harm will typically indicate both

O P I N I O N

Competition sensing: the social sideof bacterial stress responsesDaniel M. Cornforth and Kevin R. Foster

Abstract | The field of ecology has long recognized two types of competition:

exploitative competition, which occurs indirectly through resource consumption,

and interference competition, whereby one individual directly harms another. Here,

we argue that these two forms of competition have played a dominant role in the

evolution of bacterial regulatory networks. In particular, we argue that several of

the major bacterial stress responses detect ecological competition by sensing

nutrient limitation (exploitative competition) or direct cell damage (interference

competition). We call this competition sensing: a physiological response that detects

harm caused by other cells and that evolved, at least in part, for that purpose. A key

prediction of our hypothesis is that bacteria will counter-attack when they sense

ecological competition but not when they sense abiotic stress. In support of this

hypothesis, we show that bacteriocins and antibiotics are frequently upregulated by

stress responses to nutrient limitation and cell damage but very rarely upregulated

by stress responses to heat or osmotic stress, which typically are not competition

related. We argue that stress responses, in combination with the various mechanisms

that sense secretions, enable bacteria to infer the presence of ecological competition

and navigate the microbe-kill-microbe world in which they live.

P E R S P E C T I V E S

NATURE REVIEWS |MICROBIOLOGY VOLUME 11 |APRIL 2013 |285

2013 Macmillan Publishers Limited. All rights reserved


2/9

Cellde

nsity

Genetic diversity

D, cell damage

N, nutrient stressQS, quorum selfQO, quorum other

QSQO

QSQO

QOQ

S

QODN

QSN N

D

D

D

ecological and evolutionary competition.How, though, can a cell assess the strength ofecological competition (FIG. 1)? One way is toindirectly infer the potential for competition

via compounds released by self and othercells (for example, via quorum sensing mol-ecules), and we discuss this further below1620. However, sensing the secretions of othersdoes not directly sense the harm caused by

ecological competition. By contrast, the well-known stress responses do directly detect awide range of harmful effects caused by theenvironment in which the cell exists. Below,we propose that many of the regulatoryeffects of stress responses are consistent withthe evolution of systems that both detect andrespond to competition.

Stress responses for defence and attack

In the effort to understand and controlbacteria, the responses of bacterial cells toharmful conditions have been intensively

studied. As might be expected, bacterialregulatory networks often show strong, clearshifts when cells face physical or chemicalchallenges, such as high temperatures, lownutrient levels or the presence of antibiotics.A large number of stress response geneshave been identified that typically displaydifferential regulation in the face of a par-ticular stressor and promote resistance to

that stressor (that is, inactivation of the generesults in decreased stress tolerance)21.

The potential for links between ecologi-cal competition and stress responses areclearest in the study of nutrient stress, forwhich the typical experimental model isnutrient competition during batch growthin pure culture. Moreover, we discuss inBOX 1how the major stress responses maponto the traditional distinctions made byecologists between exploitative competition,interference competition and the abioticenvironment. Although it is clear that many

stress responses will respond to ecologicalcompetition, it could be argued that this issimply a by-product of adaptation to abioticstresses. To date, the literature about stressresponses has focused on traits that help cellscope with stress, such as protection, repairand dormancy21. However, it is typically dif-ficult to know whether such coping traitshave evolved to deal with biotic or abioticthreats. One possible exception is the obser-

vation that stress responses of all kinds tendto promote antibiotic resistance22, whichwould be expected if stress were indicativeof the threat of toxin attack by other cells.But even in this case, there is ambiguitybecause a degree of cross-protection is com-mon to many stress responses. Except forhighly specific defences like -lactamasesecretion23, it is difficult to ascertain howoften this cross-protection is the result ofnatural selection for antibiotic resistance,

rather than a by-product of selection forsomething else.

Linking stress responses and competition

The topic of antibiotic resistance brings usto a somewhat neglected phenotype in thestudy of bacterial stress responses: antibioticproduction. Bacteria commonly releasetoxins that kill other bacteria, and thisrelease occurs both across the membraneand through lysis of a subset of cells of onegenotype. The limited overlap betweenthe literature about stress responses andthat about bacterial toxin production isillustrated in the recent edited volume onbacterial stress responses21, which, althoughimpressive, rarely mentions the regulationof antibiotic production. Toxins are par-ticularly useful in discerning evolutionaryfunction because one can typically be moreconfident that toxins have evolved to influ-ence a biotic target. There are exceptionsto this rule, such as pyocyanin, which hasmultiple potential roles in addition to anti-biosis, including effects on metabolism andnutrient acquisition24. Nonetheless, one canbe more confident of the role of other toxins.

This is particularly true for the bacteriocins,which are narrow-spectrum antibiotics thattarget other bacteria25. One prediction ofour hypothesis is that stress responses whichdetect ecological competition will often beassociated with the release of toxins, becausein bacterial communities, ecological com-petition normally implies the presence offoreign genotypes (and of evolutionary com-petition, as described above). By contrast, itshould be rare to find toxins that are regu-lated by responses to predominantly physicalstresses, like heat or osmotic pressure.

Figure 1 | The correlated information provided by competition sensing and quorum sensing. Four

scenarios are shown, with the top right representing the most intense ecological competition. There

are four potentially correlated cues that can provide information about the degree of ecological com-petition in a population (see Supplementary information S3(figure)): nutrient stress (N), cell damage

(D) caused by toxins produced by other cells, quorum-related responses to self-compounds (QS) and

quorum-related responses to compounds produced by other cells (QO

). Quorum-related responses

include recognition of the canonical autoinducers of quorum sensing, but also of any other secretory

product that correlates with cell density, such as peptidoglycan fragments45. If all genotypes use the

same compound (for example, autoinducer 2), QSand Q

Ocollapse into one information source. This

simplistic figure already reveals the potential complexities in the use of stress responses and

quorum responses to infer ecological competition. To some degree, the two major types of stress

response provide separate information; nutrient stress relates to cell density and exploitative com-

petition (moving from bottom to top between scenarios), whereas cell damage relates to genetic

diversity (moving from left to right between scenarios). However, for some transitions, such as the

diagonal move, the changes in both forms of stress are correlated. Despite this, the combination

of the two sources of stress can effectively distinguish the four scenarios. In principle, the quorum

responses can also distinguish the four scenarios if it is possible to differentiate the quorum mole-

cules of self and those of others. Finally, another option to distinguish all four conditions is to usecell damage combined with self quorum responses. These correlations are strong when tested with

the model described in BOX 2(see Supplementary information S3 (figure)).


286 |APRIL 2013 |VOLUME 11 www.nature.com/reviews/micro

http://www.nature.com/nrmicro/journal/v11/n4/suppinfo/nrmicro2977.htmlhttp://www.nature.com/nrmicro/journal/v11/n4/suppinfo/nrmicro2977.html


3/9

We carried out a literature search with thegoal of summarizing the key regulators ofantibiotic production. Regulators are typicallyinferred from the effects of gene knockoutsor overexpression, and it is worth noting thatboth types of mutation can have confound-ing pleiotropic effects. In addition, whereassome studies measure toxin levels directly,others infer these levels indirectly using theamount of mRNA from toxin-encoding loci.Furthermore, as we discuss below, we didnot include cases in which a specific nutrientsource affected toxin production, or thosein which certain non-stressful temperaturestriggered greater induction than other non-stressful temperatures, as these responsesare often difficult to interpret. Such casesand caveats aside, FIG. 2and Supplementaryinformation S1(table) summarize the resultsof our survey, which are in strong agreementwith our hypothesis. We find many links

between stress responses and toxin secre-tion, and we also find that the majority ofthese links are for responses associated withdetecting either nutrient stress or cell dam-age. Specifically, the survey found 85 differentassociations between a stress response regu-lator and a toxin (Supplementary informa-tion S1 (table)), 81 of which are associatedwith nutrient limitation and cell damage (thatis, ecological competition), compared withonly four associated with heat or osmoticstress (that is, abiotic stress). In eight cases,increased stress leads to decreased toxin pro-duction. We discuss these exceptions furtherbelow and in Supplementary information S1(table), but note here that there are only sixsuch exceptions out of 81 cases for bioticstresses, whereas these exceptions constitutetwo of the four abiotic stressor cases.

These data should be interpreted cau-tiously, as our survey is not exhaustive andthere are strong biases owing to the researchfocus on a few model species. Moreover,some of the data come from groups ofrelated bacteriocins, such as the S-typepyocins from Pseudomonas aeruginosa.This said, in each category of biotic stress

included in the survey, there are multipleindependently evolved associations betweena toxin and a stress regulator, as is made clearby the phylogenetic presentation in FIG. 2.These associations can be in the form of non-homologous toxins associated with one stressregulator (for example, both R-type pyocinsand S-type pyocins are associated with therecombinase gene recAin P. aeruginosa)or inthe form of non-homologous toxins associ-ated with non-homologous regulators (suchas the use of RNA polymerase -factors Wand Eto regulate diverse toxins in response

to cell wall damage in Bacillus subtilisandStreptomyces coelicolor, respectively, as Wand Eare not closely related homologues26).These examples of cell damage responses are

notable because treatment of these bacteriawith antibiotics results in the targeted cellsreleasing toxins2730. This suggests that thesecretion of toxins via stress responses can

Box 1 | Stress responses protect against ecological competition

It is possible to map the major stress responses onto the traditional distinctions made by ecologists

among exploitative competition, interference competition and the abiotic environment. In

particular, stress responses can be classified as leading to induction through nutrient limitation,

cell damage or abiotic stress. As for any such classification, our analysis is complicated by the fact

that responses to distinct stressors are often overlapping. For example, heat, osmotic stress and

antibiotics might all induce a particular regulator. Nevertheless, the literature normally points

towards a primary inducer for each stress response system, and we use this as the basis for ourfunctional interpretation21.

Responses associated with low nutrient levels

The stringent response. Nutrient limitation has widespread effects on gene regulation in bacteria

through the stringent response, which is controlled by the intracellular alarmones guanosine

pentaphosphate and guanosine tetraphosphate (collectively referred to as (p)ppGpp) and the

ribosome-associated enzymes RelA and SpoT (which are (p)ppGpp synthases and hydrolases).

These factors respond to low levels of specific nutrients including amino acids, glucose,

phosphate, iron and fatty acids and reduce the biosynthesis of macromolecules while

simultaneously attempting to restart metabolism.

General stress responses.Many bacteria also show a more general response as cell division

slows under nutrient limitation. In Escherichia coli, this response occurs through the RNA

polymerase -factor S, which is modulated by (p)ppGpp, and also through cyclic AMP levelsand leucine-responsive regulatory protein (Lrp), which respond to glucose and leucine levels,

respectively. The level and activity of S

can also be induced by other sudden environmental shifts,including DNA damage, an increase in temperature, an increase in osmolarity and a decrease in

oxygen availability, which is why Sis considered to be a general stress response regulator. Thewidespread effects of Son bacterial cells include a shift to fermentation, altered morphology,altered membranes and the synthesis of protective and storage carbohydrates. Regulators such as

Sare often known as stationary phase regulators because of the timing of their activation in batchculture. Importantly, however, they can also activate under continuous growth when there is a

continuous but limited nutrient supply (as may commonly occur in biofilms).

Responses to antibiotics and other cellular damageEnvelope stress. Bacteria respond strongly to cellular damage. Envelope stress responses result from

perturbations to the membranes or cell wall. In E. coli, two of the best characterized envelope stress

responses are the Eand Cpx responses. The results of induction are diverse but centre on reducingthe amount of protein in the periplasm and improving the folding of the protein that is there.

Antibiotics induce envelope stress responses in both Gram-negative and Gram-positive bacteria.

DNA damage. A second major class of damage responses occurs owing to DNA damage. The

canonical example is the SOS response of E. coli, which is commonly induced in the laboratory by

ultraviolet light or DNA-damaging antibiotics; these activate the recombinase RecA, which binds

to single-stranded DNA. Cell division is inhibited, and DNA repair is initiated by removal of the

damaged nucleotides followed by homologous recombination. The SOS response plays a part in

DNA repair in many species, including Gram-positive bacteria.

Oxidative stress.Other key inducers of the SOS response are ROS (reactive oxygen species) and

RNS (reactive nitrogen species). In addition to the SOS response, several dedicated responses to

ROS and RNS have been identified, including those mediated by the SoxR (redox-sensitive

transcriptional activator) and OxyR (hydrogen peroxide-inducible genes activator) regulons in

E. coli. Many antibiotics generate redox-active molecules and cause oxidative stress.

Responses to abiotic environmentHeat stress.The heat shock response is one of the most biologically conserved stress responses.

The E. coli heat shock response is highly specific and appears to be induced only by a high abundance

of misfolded proteins, the levels of which are controlled by the heat shock regulator 32. Theresponse of the Gram-positive bacterium Bacillus subtilisoverlaps with that of E. colibut is

controlled by a more diverse set of regulators.

Osmotic stress.Variability in the total concentration of dissolved solutes in the environment is

another form of stress experienced by bacteria. Cells deal with rapid external upshifts in osmotic

pressure by pumping in potassium ions or through the use of osmoprotectants like trehalose.

External downshifts in osmotic pressure are tolerated owing to mechanosensitive channels that

open in response to membrane strain and release solutes from the cell.



http://www.nature.com/nrmicro/journal/v11/n4/suppinfo/nrmicro2977.htmlhttp://www.nature.com/nrmicro/journal/v11/n4/suppinfo/nrmicro2977.htmlhttp://www.nature.com/nrmicro/journal/v11/n4/suppinfo/nrmicro2977.htmlhttp://www.nature.com/nrmicro/journal/v11/n4/suppinfo/nrmicro2977.html


4/9

S(pyrrolnitrin)

RecA (lectin-like putidacin A)

RecA (pyocins R1, R2, R3, F1, F2, F3, S1, S2, S3, S4, S5, AP41)

RdgBRecA (carotovoricin, pectin lysase, carocin D)

LexA (pesticin)

LexARegC (bacteriocin 28B)

LexA (alveicins A and B)

LexA (klebicins B and C)

LexA (klebicins B and D)

ppGpp (actinorhodin, CDA, undecylprodigiosin),

B(undecylprodigiosin), L(actinorhodin),

E(undecylprodigiosin, actinorhodin, CDA), SoxR (actinorhodin)

UviA (bacteriocin BCN5)

CodY (putative bacteriocin)

CcpA (H20

2)

VicRK (mutacins I, IV, V, VI)

ppGpp? (bialaphos)

ppGpp (formycin)

ppGpp (actinomycin)

ppGpp (streptomycin)

ppGpp (nikkomycin)

ppGpp (colicin K, microcin J), S(microcin C), IHF (microcin B),IHFLrp (microcin J), LexARecA (colicins A, B, D157, E1, E2, E3, E6, E7,

Js, N, S4, U, Y, Ia, Ib, 5, 10, K)

ppGpp (bacilysin), CodY (bacilysin), H(subtilin), Spo0A (SkfA),AbrBSpo0A (subtilosin, TasA), W(putative bacteriocin)

Pseudomonas fluorescens

Pseudomonas putida

Pseudomonas aeruginosa

Pectobacterium carotovorum

Yersinia pestis

Serratia marcescens

Hafnia alvei

Klebsiella pneumonia

Klebsiella oxytoca

Escherichia coli

Clostridium perfringens

Bacillus subtilis

Streptococcus pneumoniae

Streptococcus gordonii

Streptococcus mutans

Streptomyces hygroscopicus

Streptomyces lavendulae

Streptomyces antibioticus

Streptomyces griseus

Streptomyces coelicolor

Streptomyces tendae

sometimes be a direct reciprocation to beingattacked by toxins, and we return to suchtoxin kickback below.

Exceptions and future analyses

Although the majority of stress response-

regulated toxins fit with our expectations,there are also a few examples that we didnot predict (Supplementary information S1(table)). For example, colicins are releasedunder high temperature and only whenthe heat shock regulator 32(also known asRpoH) is present in a cell31. However, it isnot heat stress that promotes colicin release,but rather a temperature of 37 C or above.It is possible, then, that cells are respondingto being in the mammalian gut, where theymust compete for their niche. There is also alink between the production of microcin B

and the osmotic stress-induced gene ompR,but the function of ompRis unclear, as it is notassociated with protection against osmoticstress32,33. Finally, there are six cases in whichtoxins are downregulated by nutrient stress.Two of the six exceptions involve pyocyanin,

which has diverse functions in addition toantibiosis24. However, the other cases moreclearly involve antibiotics. For instance,despite conflicting reports, cephamycin Cseems to be expressed primarily under expo-nential growth in batch culture and to beunder negative regulation by the stringentresponse in Streptomycesclavuligerus 33. Suchexceptions are important, as they highlightthe fact that it is not universally true thatantibiotics are positively regulated by stressresponses. This is reasonable because somespecies might reliably meet competitors

under high-nutrient conditions. However,such cases of negative regulation of toxins bystress responses seem to represent a minorityof the total set of examples.

The data for the genus Streptomycesalsoillustrate the potential for complexity in themetabolic regulation of toxins. This complex-ity is also seen in cases in which limitation ofa specific nutrient promotes toxin production(cases that we did not consider in our survey).One major class of such examples consists ofcases for which the production of antibioticsis increased more by limitation of a preferredcarbon source of a species than by depletionof non-preferred carbon sources34. Theseexamples fit our basic hypothesis that cells willproduce toxins when ecological competitionis having the most harmful effects. But thereare other examples, such as iron and phos-phate depletion promoting pyocyanin pro-duction35, for which it is less clear why the

limitation of one nutrient but not anothershould promote toxin production. Oneexplanation put forward recently is that cellsuse prudent regulation, whereby they makeproducts for secretion only when they havean excess of the nutrients needed to makethe product (for example, because a limita-tion in another nutrient inhibits growth)36,but a full discussion of these interestingissues is outside the scope of this article.

Diverse regulation of defence and attack

Our general hypothesis that many stressresponses function as a way to sense ecologi-cal competition is supported by the manylinks between toxin secretion and nutrient-or antibiotic-induced stress. However, it isalso clear that not all biotic stress responsesin all species are associated with toxinsecretion and, moreover, there is diversityin the regulation of toxins. This includesan important role for quorum sensing andother mechanisms that sense secretions fromother cells (Supplementary information S1(table)). If stress responses simply detect thestrength of ecological competition, why istoxin regulation so rich and varied?

Evolution is more likely to co-opt anexisting regulator to control a new toxinthan to generate an entirely new regulatorypathway. Some of the diversity of regula-tion in Supplementary information S1(table), therefore, will be due to histori-cal constraints on molecular architecture.Nevertheless, every species has multipleregulators that can be used, and one canask why one type of regulation is used overanother. To begin to answer this question,one must consider that bacteria live inunpredictable environments and there is

Figure 2 | Cladogram of bacteria that induce toxin secretion as a result of stress response

activation. Stress response regulators that are known to induce toxin production are shown for each

species, with the induced toxin in parentheses. The four types of response shown (BOX 1)are nutrient

limitation (light blue), DNA damage (red), oxidative stress (green), and envelope stress (yellow). We

include only those cases from Supplementary information S1(table) for which a regulator is known.

Note that oxidative stress induces the production of bacteriocins in Gram-negative species

(Pseudomonas aeruginosa andEscherichia coli in Supplementary information S1 (table)), but these

cases are not shown, as they probably occur via DNA damage responses. The phylogenetic analysis

was carried out principally with SUPERFAMILY62

, a database of structural and functional annotationsof proteins and genomes. Hafnia alvei63and the majority of Streptomycesspp.64were placed according

to studies using 16S rRNA genes. Details of these associations, as well as associations between quorum

sensing and bacteriocin induction, are given in Supplementary information S1 (table). CcpA, catabolite

control protein A; CDA, calcium-dependent antibiotic; ppGpp, guanosine tetraphosphate; SkfA,

sporulation killing factor.



http://www.nature.com/nrmicro/journal/v11/n4/suppinfo/nrmicro2977.htmlhttp://www.nature.com/nrmicro/journal/v11/n4/suppinfo/nrmicro2977.html


5/9

therefore a large degree of uncertainty asto their physical and social surroundings.Bacteria can be thought of as decision-making machines: given a set of environ-mental cues, their genetic network canassess the likely environmental conditionsand respond appropriately (a process knownas Bayesian inference)37. These cues aresubject to noise and, sometimes, manipula-tion by competing genotypes19, and cellsmust therefore integrate information from

various sources to minimize the influenceof unreliable inputs while making use of theavailable information. Clearly, this decision-making process is not a conscious one. Cellswith certain genetic networks outcompeteothers over evolutionary time. As thesenetworks shape the strategy used by the cell,natural selection pushes genomes towardincreasingly advantageous strategies, asthough the cells themselves were intentionalagents38.

But in this noisy world, what is the bestregulatory mechanism for inferring ecologi-

cal and evolutionary competition? Thereare three broad classes of social informationthat are available to cells and are knownto affect cell physiology. The first two weconsider to be true competition-sensingsystems: the direct effects of nutrient limi-tation and damage to the cell. The third isthe sensing of molecules released by othercells, which does not directly detect ecologi-cal competition but can indirectly provideinformation about the potential for harmcaused by other cells (FIG. 1; TABLE 1). Thissensing includes the use of autoinducer

signal molecules, but also any indicatorof density, such as acidity, which corre-lates with density in lactic acid bacteria39(Supplementary information S1 (table)).

To understand the value of the differ-ent sources of information, it is first usefulto consider more specifically how theseinformation sources can help a cell. As acase study, in BOX 2(see also Supplementaryinformation S2(box)) we work through asimple model of nutrient stress-dependenttoxin regulation in which a fixed amount ofnutrients is used up (akin to a simple batchculture experiment). This model suggeststhat three key factors favour toxin produc-tion: a high density of self cells, a high den-sity of other cells and a late growth stage.These factors are all linked, but each is usefulfor its own reason. A high density of self cellsmeans that toxin release will be effective,because a high concentration of toxins canbe rapidly generated. A high density of othercells means that there are many target bacte-ria present that might otherwise poison and

steal. A late growth stage is most relevantto environments in which a fixed amountof nutrients is available for consumption;at a late growth stage, the most essentialearly growth phase is over and the costs ofinvesting in toxin secretion are relatively low(BOX 2). For our fixed-nutrient example, it isalso the case that if cells leave it too late therewill not be enough energy left for substantialtoxin secretion. Together, these predictionslead to the hypothesis that secreting toxins atsome intermediate level of nutrient stress isan effective strategy.

Our model in BOX 2considers speciesthat colonize and grow in a particular typeof environment, time and time again. Suchcells will experience a predictable reduc-tion in nutrients, increase in toxin levels andincrease in quorum sensing molecules suchthat any one of these information sourcescould be used to coordinate strategies ofdefence and attack. However, under lessregimented conditions, these correlationswill weaken (FIG. 1and Supplementary infor-mation S3(figure)), and one source ofinformation might become a more usefulpredictor than another when it comes to pre-dicting the density of self and other cells andthe growth stage.

The utility of quorum-related information

As we propose above, the diversity in toxinregulation might be explained by one sourceof information being more useful thananother in responding to adversarial pheno-types. But when would quorum-relatedinformation be more useful than a stress

response that senses ecological competi-tion directly? Most simply, if the density ofself cells is the key factor determining thebenefits of toxin secretion, then quorumregulation might be the best regulator, aslong as the effects of diffusion are not too

variable among different environments20,40.Indeed, this is the typical explanation ofquorum regulation: clone mates are sig-nalling each other to ensure that thereis enough of their own genotype in thepopulation for their secretions (be thesetoxins or growth-promoting products) to be

Table 1 | Examples of toxin regulation based on quorum-related sensing and competition sensing*

Example Classification Rationale for classification

Quorum sensing of BlpCinduces Blp bacteriocins inStreptococcus thermophilus65

QR sensing of self(and QR sensingof others)

Cells release and detect the autoinducer BlpC, which qualifies the process as QR sensing ratherthan CS

Other strains of the same species might make BlpC and cause bacteriocin induction in a focalstrain, hence the process might be QR sensing both of others and of self

The benefit of this regulation might be due to the strong correlation between a high density ofthe self genotype and the presence of competitors (FIG. 1; BOX 2; Supplementary information S3(figure))

Peptidoglycan inducespyocyanin production inPseudomonas aeruginosa45

QR sensing ofothers

Peptidoglycan is shed from cell walls of Gram-positive bacteria and might be a cue thatindicates their presence to another bacterium

It is because of examples like this that we use the term quorum-related sensing rather thansimply quorum sensing

Starvation induces bacilysinin Bacillus subtilis via both(p)ppGpp and GTP66

CS of nutrients General nutrient limitation is sensed by the lack of a range of metabolites, rather than thepresence of a particular secreted molecule

Depending on the ecology, this resource limitation can be caused by other lineages (FIG. 1) Note that even when a cell always faces the same competitors (so that the mechanism does

not detect genetic diversity in a community), nutrient-based induction allows cells to infer thestrength of interaction with other genotypes and other useful information (BOX 2)

DNA damageinduces colicins inEscherichia coli25,28,30,67

CS of damage DNA damage can be caused by other cells, and the SOS response that is triggered can thus beused as an indication that other cells are present, rather than the organism sensing a specificmolecule that other cells produce

CS, competition sensing; (p)ppGpp, guanosine pentaphosphate and guanosine tetraphosphate; QR, quorum-related. *Note that QR sensing encompasses bothstandard autoinducer-based quorum sensing and more general sensing of molecules produced by other organisms (such as in the pyocyanin example).



http://www.nature.com/nrmicro/journal/v11/n4/suppinfo/nrmicro2977.htmlhttp://www.nature.com/nrmicro/journal/v11/n4/suppinfo/nrmicro2977.htmlhttp://www.nature.com/nrmicro/journal/v11/n4/suppinfo/nrmicro2977.htmlhttp://www.nature.com/nrmicro/journal/v11/n4/suppinfo/nrmicro2977.htmlhttp://www.nature.com/nrmicro/journal/v11/n4/suppinfo/nrmicro2977.htmlhttp://www.nature.com/nrmicro/journal/v11/n4/suppinfo/nrmicro2977.htmlhttp://www.nature.com/nrmicro/journal/v11/n4/suppinfo/nrmicro2977.htmlhttp://www.nature.com/nrmicro/journal/v11/n4/suppinfo/nrmicro2977.htmlhttp://www.nature.com/nrmicro/journal/v11/n4/suppinfo/nrmicro2977.htmlhttp://www.nature.com/nrmicro/journal/v11/n4/suppinfo/nrmicro2977.html


6/9

0 2,500 5,000 7,500 10,0000

0.1

0.2

0.3

0.4

0.5

0.4

0.5

0.6

0.7

0.8

a

c

Time (hours)

Densityofbacte

ria(mgl1)

Densityofbacteria(mgl1)

Constitutive toxin production

Proportionoftoxinproducers

b

Facultative toxin production

0

200

400

600

800

1,000

0

500

1,000

1,500

0 20 40 60 80 100 120 140

Time (hours)

0 20 40 60 80 100 120 140

2,000

2,500

3,000

Starting nutrients (mg l1)

Growththreshold(/max)

Low starting nutrients High starting nutrients

Toxin producerSensitive strain

Toxin producerSensitive strain

Toxin producer

Sensitive strain

Toxin producer

Sensitive strain

Box 2 | A simple model of regulating toxins by nutrient limitation

The prevalence of stress-regulated toxins in bacteria (see Supplementary

information S1(table)) can be explained by correlations between stress

and competition with another genotype. Non-clonal competition is

greatest in high-density, high-diversity environments (top right panel

in FIG. 1), and this is where toxins are most effective. Here, we develop

this logic with a simple batch culture model of the gradual transition

from a high-nutrient, low-competition environment to a low-nutrient,high-competition environment, as caused by cell growth in a fixed

amount of nutrients. The evolution of toxin production can be affected by

the number of interacting strains59and the potential for strains that are

immune to toxins but do not produce them14. To focus our analysis here,

we consider a two-strain scenario in which a toxin-producing strain

interacts with a sensitive strain. Readers are referred to Supplementary

information S2(box) for details of the dynamics and parameters used in

the model.

Our key question is, how can variation in nutrient levels affect the

benefits and costs of toxin use? The variation in nutrient levels can be

caused in two ways: first, through variability in the initial conditions, and

second, through the progressive depletion of nutrients during growth.

We first consider the simple case of competition between a constitutive

toxin producer and a sensitive strain60,61(see the figure, part a). For

constitutive toxin producers that is, cells for which (the thresholdgrowth rate below which toxin production is active) is equal to

max(the

maximum specific growth rate) the key variable is nutrient level at the

start of growth, as toxin production in these strains does not respond to

changes in nutrient levels as the cells grow. If nutrients are so sparse to

begin with (2,000 mg per litre in this model) that cells cannot reach a high

density, toxins will never have a strong impact. Consequently, the cost of

toxin production is never overcome, and non-producing, toxin-sensitive

cells dominate (solid lines). But in high-nutrient environments (here,

10,000 mg per litre), cells can grow to high enough densities for toxins to

be decisive and tip the balance in favour of producers 60,61(dashed lines).

In contrast to the constitutive case, stress-regulated toxin producers

(facultative producers) here, cells for which = 0.1 per hour candefer production until nutrient depletion results in a low growth rate ()and thus avoid the high initial costs of production while still enjoying its

subsequent benefits when toxins are most effective (see the figure,part b). In this scenario, when starting nutrient concentration is low(2,000 mg per litre), a starvation-induced producer turns the tables on a

sensitive strain (dashed lines), whereas the sensitive strain would win if

toxin production were constitutive in the producing strain (solid lines).

Several model details can affect these results. For instance, if toxin

efficacy depends on growth rate, late-term secretion can be less effective.

In addition, if the toxin efficacy is sufficiently high, constitutive production

could in theory be optimal. Finally, the model assumes that no new

nutrients diffuse into the system, as this might also affect conclusions.

These and other details deserve more attention in future work, but the

adaptiveness of delayed induction seems robust to basic biological

assumptions.

We evaluated the model at time 1,000 hours with varying starting

nutrient (N0) levels and values, and plotted the results to show the final

proportions of producers as a function of N0and (plotted as a fractionof

max; see the figure, part c). The dividing line is the Monod equation,

describing the microbial growth rate in relation to nutrient concentration

(=max

N/ (KN+N), in which K

Nis the nutrient saturation constant, or the

value ofNwhen is half maximal). The region above the dividing linerepresents constitutive toxin production because there are not enough

nutrients to allow a growth rate above , the toxin production threshold.For any starting nutrient concentration in the figure (part c), thethreshold that maximizes fitness (that is, which results in the highest

proportion of toxin producers at 1,000 hours) is below this line, which

implies that producing toxin after a delay is typically the best strategy.

The figure shows that cells should turn on toxins at some intermediate

degree of nutrient stress. Activating at very low nutrient stress (that is,

when is high) means that production is always on and effectivelyconstitutive. Activating at extremely high stress (that is, when islower than for the facultative producer in part b) means that little toxincan be produced because the bacteria are running out of food. In

summary, responding to nutrient stress can allow cells to minimize the

fitness cost of toxin production relative to its benefits, when these

benefits are a function of the strength of interaction with competitors.



http://www.nature.com/nrmicro/journal/v11/n4/suppinfo/nrmicro2977.htmlhttp://www.nature.com/nrmicro/journal/v11/n4/suppinfo/nrmicro2977.htmlhttp://www.nature.com/nrmicro/journal/v11/n4/suppinfo/nrmicro2977.htmlhttp://www.nature.com/nrmicro/journal/v11/n4/suppinfo/nrmicro2977.htmlhttp://www.nature.com/nrmicro/journal/v11/n4/suppinfo/nrmicro2977.htmlhttp://www.nature.com/nrmicro/journal/v11/n4/suppinfo/nrmicro2977.htmlhttp://www.nature.com/nrmicro/journal/v11/n4/suppinfo/nrmicro2977.htmlhttp://www.nature.com/nrmicro/journal/v11/n4/suppinfo/nrmicro2977.html


7/9

effective1620. However, quorum informationcan also predict the strength of ecologicalcompetition and, importantly, might do soearlier than a stress response that detectsharm (FIG. 1and Supplementary informa-tion S3 (figure)). Therefore, the predictionof ecological competition might also explainwhy competition-associated phenotypeslike toxin production are under the regula-tion of the products secreted by self cells.This explanation is particularly relevant forregulated processes that do not need a highdensity of clone mates to function, such asrepair, protection and the production ofcontact-dependent toxins41.

The potential for diversity in quorum-related molecules means that quoruminformation is useful not only for signallingamong self cells, but also potentially for thedetection of specific cues produced by othergenotypes (that is, for detecting evolutionary

competition)17,19. A high degree of molecularspecificity has the potential to separatequorum information from stress informationin that it can identify a particular strain orspecies that is an evolutionary competitor.For instance, Proteus mirabilisstrains candetect other strains in the species and isolatethemselves from would-be competitors42,and there is evidence that other micro-organisms might tailor their responses toparticular competitors43. The existence ofstrains with receptors for quorum sensingmolecules that they do not produce, such asLuxR orphans, is also consistent with onestrain monitoring the cues produced byanother strain or species17,19. Similarly, theabundance of two-component signallingsystems in bacteria44might be partlyexplained by natural selection on cells tomonitor other strains and species.

The utility of competition sensing

Specificity allows a response to be finelytuned to the threat posed by a particularstrain or species, but can require specificdetection systems for each cue. Moreover,the producer of the cue has the potential to

change it to evade detection. For these rea-sons, general responses might often be moreuseful for inferring evolutionary competi-tion. Sometimes, quorum-related moleculescan be general. For example, self quorummolecules can predict the presence of othergenotypes in so far as density of self and theother cells are correlated (FIG. 1). It is alsopossible for sensing of quorum moleculesfrom other species to be general for a largeclass of competitors; for instance, P. aerugi-nosaproduces pyocyanin in response to thepresence of peptidoglycan, which might

indicate the presence of competing Gram-positive bacteria45. However, stress responsesmight often prove to be more useful againstcompetitors than other mechanisms of infer-ring competition because stress responsesmore frequently operate independently ofcompetitor identity and provide a directassessment of ecological competition. In thisway, stress responses can be robust, both to

variation in community species compositionand to evolutionary changes in the chemi-cal signatures of competitors. Furthermore,competition sensing can distinguish betweena mixture of harmful and harmless strainson the basis of the key evolutionary cur-rency: their capacity to reduce the fitnessof a focal cell.

In summary, we hypothesize that stressresponses have two key advantages overother potential mechanisms of inferringcompetition. First, stress responses detect

the property of competitors that will typi-cally be most important for fitness theirharmfulness and second, these responsesprovide a general detection system thatis not specific to the chemical cues of anyone genotype. These factors might oftenmake stress responses a better solution forresponding to foreign genotypes thanquorum responses. This said, it is alsoclear that detecting nutrient limitation anddetecting toxin-mediated damage are notequivalent. Sensing nutrient limitation isin some ways more similar to the detec-tion of self quorum molecules than to thedetection of cell damage (FIG. 1). And in anuncertain world in which other genotypescome and go, cell damage is likely to be thebest indicator of evolutionary competi-tion. This is because compared with celldamage, nutrient sensing is less able todistinguish between ecological competitionfrom self versus other genotypes (FIG. 1andSupplementary information S3 (figure)).Directly sensing damage might have addi-tional benefits. The SOS response reportson DNA damage that can limit the futureprospects of individual cells that is, their

reproductive value. Because some SOS-regulated toxins seem to require cell lysisfor their activity46, SOS regulation mightallow a strain to select the most damagedcells to respond to an incoming attack25.Finally, the fact that the toxin of one straincan make another strain respond in kind the toxin kickback mechanism mentionedabove has interesting implications.Classic evolutionary models show that suchreciprocation can lead to both parties livingin peace47; however, kickback also raises thepossibility of escalating reciprocal rounds of

attack and counter-attack28. Which scenariois most representative of bacterial systems isan open question.

Clarifications and caveats

We believe that many of the major stressresponses are better understood by consider-ing the central role of ecological competitionin bacterial evolution. Below, we offer someclarifications of our argument so that it isnot misinterpreted.

Not all stress responses involve competi-tion sensing.There are a number of casesin which stress responses are not associ-ated with ecological competition. Thisincludes heat and osmotic stress, as wellas pathogens that seem to use their stressresponses as an indication of an attack fromthe host immune system48. This use of stressresponses is likely to be a recent adaptation,

as the major stress responses seem to longpredate the evolution of multicellularity49.Other potential biotic stressors that arenot strictly competition based include thepresence of ciliate predators and phages9.However, it is not clear that the major stressresponses have a central role in these cases21.The key known bacterial responses to phagesare abortive infection50and the CRISPR(clustered regularly interspaced short palin-dromic repeats) system51. These responsesappear to function independently of themajor stress responses, with the possibleexception of some toxinantitoxin systemsthat are stress regulated and implicated inphage abortive infection52.

Competition-sensing systems will not exclu-

sively respond to other genotypes.We predictthat harm caused by other cells is often usedto detect and infer competition from othergenotypes, a hypothesis that is supported bythe data in Supplementary information S1(table). However, we do not predict thatcompetition sensing will respond exclusivelyto other genotypes. First of all, detectionand response to self cells is an important

competition-sensing response that probablyprovides information about competitionfrom both self and other cells (BOX 2). Moregenerally, harmful stimuli in the absenceof ecological competition can cause acompetition-sensing response to misfire; forexample, P. aeruginosaproduces bacteriocinswhen challenged by synthetic antibiotics27.But a stress response does not have to bea perfect discriminator to be useful forresponding to competition, it just has to bepositively correlated with competition acrossdifferent environmental conditions.





8/9

Toxin production is not the only responseregulated by competition sensing.Someresponses to competition involve attack, butothers involve defence. We have focused onthe attack response, toxin production, in thisarticle because it is difficult to know whethermost defensive responses are defences againsta competitor or defences against an asocialenvironment. However, the repair and pro-tective roles of stress responses22are just asimportant as toxin production in ecologicalcompetition. For example, in addition to reg-ulating colicin secretion, the SOS responseof E. coliregulates DNA repair and the use oferror-tolerant DNA polymerases that raisethe mutation rate53. Both these effects onDNA might help in the escape from strongecological competition. Biofilm formation isanother candidate response to competitionsensing, as biofilms are both induced by andprotective against antibiotics54. Conversely,

competition sensing can be a response tobeing in a biofilm in which cells are denseand nutrients are limited55.

Metabolic shifts might also help cellsto compete with other strains and species.Evolutionary theory suggests that cells shoulduse high-rate, low-yield metabolism, such asfermentation in the presence of oxygen, whenthey face competition from other genotypes56.It is interesting, then, that stress responsesregulate a range of metabolic changes, includ-ing a shift towards fermentation, in responseto antibiotics. These changes have been takenas evidence that antibiotics function more assignals than as toxins in nature57,58. Our inter-pretation is that antibiotics from other geno-types function as signals only in the sense thata cell can benefit from detecting ecologicalcompetition before it becomes bactericidal.However, in evolutionary terminology, thiswould make antibiotics cues rather than truesignals because true signals require that thesender benefits from signalling17,19.

Conclusions

Most bacteria face the potential for intenseecological competition from both their

own genotype and other genotypes911. Ourhypothesis is that this competition has hada strong effect on the evolution of bacterialregulatory networks. In particular, whenecological competition implies the pres-ence of other genotypes, these networksshould drive responses that both defendcells and allow them to counter-attack. Ourwork makes the distinction between twomajor classes of mechanism that cells useto detect ecological competition. The firstis the detection of specific secreted prod-ucts of cells, including quorum sensing and

related responses. This will often involvea specific receptor to detect other cells viatheir secreted compounds. The second classof mechanism is the stress responses thatdirectly detect ecological competition in aprocess we call competition sensing (TABLE 1).By responding to harm itself, competitionsensing is not specific to the competitor,which can be a strength in diverse microbialcommunities, as it will not misidentify aharmless foreign strain as a threat.

This view of bacterial responses to compe-tition hides many complexities that we haveonly touched on briefly (BOX 2). For example,it is clear that both quorum-related responsesand competition sensing might also provideinformation about when to secrete productsthat carry a density-dependent benefit, whichcan be independent of ecological competi-tion. Furthermore, we do not yet understandwhy species differ in the mechanisms they

use to regulate competition-associatedphenotypes like toxin secretion. Finally, wedo not know whether the original functionof nutrient or damage stress responses was inecological competition. It is conceivable thatstress responses originally arose for abioticfactors and were only subsequently co-optedfor competition sensing. As data from morespecies become available, phylogenetic analy-ses can look for evidence of ancient linksbetween stress response regulators and toxinsor toxin regulators, although the subsequentdiversification of toxins might obscure theancient character states. Another potentialapproach is experimental evolution follow-ing the evolution of stress response regula-tors under conditions of varying types anddegrees of ecological competition. Whateverthe precise relationship between the originand current function of stress responses, itseems inevitable in modern ecosystems thatkey stimuli for many of the stress responseswill be the effects of other cells, and the fre-quent association of stress responses withtoxin production suggests that modern bac-teria use these stimuli as information. In thisway, stress responses can function as social

barometers that enable bacteria to assess andrespond to ecological competition.

Daniel M. Cornforth and Kevin R. Foster are at the

Department of Zoology, University of Oxford, Oxford, UK.

Daniel M. Cornforth is also at the Centre for Immunity,

Infection and Evolution, School of Biological Sciences,

University of Edinburgh, UK.

Kevin R. Foster is also at the Oxford Centre for

Integrative Systems Biology, University of Oxford,

Oxford, UK.

e-mails:[email protected];

[email protected]

doi:10.1038/nrmicro2977

Published online 4 March 2013

1. Birch, L. C. The meanings of competition.Am. Nat. 91,518 (1957).

2. Case, T. J. & Gilpin, M. E. Interference competitionand niche theory. Proc. Natl Acad. Sci. USA 71,30733077 (1974).

3. Hansen, S. K., Rainey, P. B., Haagensen, J. A. J. &Molin, S. Evolution of species interactions in a biofilmcommunity. Nature 445, 533536 (2007).

4. Nadell, C. D., Xavier, J. B. & Foster, K. R. Thesociobiology of biofilms. FEMS Microbiol. Rev. 33,206224 (2009).

5. Gans, J., Wolinsky, M. & Dunbar, J. Computationalimprovements reveal great bacterial diversity andhigh metal toxicity in soil. Science 309, 13871390(2005).

6. Burns, B. P., Goh, F., Allen, M. & Neilan, B. A.Microbial diversity of extant stromatolites in thehypersaline marine environment of Shark Bay,Australia.Environ. Microbiol. 6, 10961101 (2004) .

7. Dethlefsen, L., Eckburg, P. B., Bik, E. M. &Relman, D. A. Assembly of the human intestinalmicrobiota.Trends Ecol. Evol. 21, 517523 (2006).

8. Foster, K. R. & Wenseleers, T. A general model for theevolution of mutualisms.J. Evol. Biol. 19, 12831293(2006).

9. Little, A. E. F., Robinson, C. J., Peterson, S. B., Raffa, K. F.& Handelsman, J. Rules of engagement: interspeciesinteractions that regulate microbial communities.

Annu. Rev. Microbiol. 62, 375401 (2008).10. Hibbing, M. E., Fuqua, C., Parsek, M. R. &

Peterson, S. B. Bacterial competition: surviving andthriving in the microbial jungle.Nature Rev. Microbiol.8, 1525 (2009).

11. Foster, K. R. & Bell, T. Competition, not cooperation,dominates interactions among culturable microbialspecies. Curr. Biol. 22, 18451850 (2012).

12. Xavier, J. B. & Foster, K. R. Cooperation and conflict inmicrobial biofilms. Proc. Natl Acad. Sci. USA 104,876881 (2007).

13. Nadell, C. D. & Bassler, B. L. A fitness trade-offbetween local competition and dispersal in Vibriocholeraebiofilms. Proc. Natl Acad. Sci. USA 108,1418114185 (2011).

14. Kerr, B., Riley, M. A., Feldman, M. W. &Bohannan, B. J. Local dispersal promotes biodiversityin a real-life game of rockpaperscissors.Nature418, 171174 (2002).

15. DCosta, V. M.et al.Antibiotic resistance is ancient.Nature 477, 457461 (2011).

16. Fuqua, W. C., Winans, S. C. & Greenberg, E. P.Quorum sensing in bacteria: the LuxR-LuxI family ofcell density-responsive transcriptional regulators.

J. Bacteriol. 176, 269275 (1994).17. Diggle, S. P., Gardner, A., West, S. A. & Griffin, A. S.

Evolutionary theory of bacterial quorum sensing: whenis a signal not a signal? Philos. Trans. R. Soc. B. 362,12411249 (2007).

18. Waters, C. M. & Bassler, B. L. Quorum sensing:cell-to-cell communication in bacteria.Annu. Rev. CellDev. Biol. 21, 319346 (2005).

19. Keller, L. & Surette, M. G. Communication in bacteria:an ecological and evolutionary perspective. NatureRev. Microbiol. 4, 249258 (2006).

20. Hense, B. A.et al.Does efficiency sensing unifydiffusion and quorum sensing?Nature Rev. Microbiol.5, 230239 (2007).

21. Storz, G. & Hengge, R. (eds)Bacterial Stress Responses2nd edn (American Society for Microbiology Press,2011).

22. Poole, K. Stress responses as determinants ofantimicrobial resistance in Gram-negative bacteria.Trends Microbiol. 20, 227234 (2012).

23. Bagge, N.et al.Pseudomonas aeruginosabiofilmsexposed to imipenem exhibit changes in global geneexpression and -lactamase and alginate production.

Antimicrob. Agents Chemother. 48, 11751187 (2004).24. Price-Whelan, A., Dietrich, L. E. P. & Newman, D. K.

Rethinking secondary metabolism: physiological rolesfor phenazine antibiotics. Nature Chem. Biol. 2,7178 (2006).

25. Chavan, M. & Riley, M. inBacteriocins: Ecology andEvolution(eds Riley, M. A. & Chavan, M. A.) 1943(Springer, 2007).

26. Staron, A. et al.The third pillar of bacterial signaltransduction: classification of the extracytoplasmicfunction (ECF) factor protein family.Mol. Microbiol.74, 557581 (2009).

27. Brazas, M. D. & Hancock, R. E. W. Ciprofloxacininduction of a susceptibility determinant inPseudomonas aeruginosa.Antimicrob. AgentsChemother. 49, 32223227 (2005).



mailto:dcornforth%40gmail.com?subject=mailto:dcornforth%40gmail.com?subject=mailto:kevin.foster%40zoo.ox.ac.uk?subject=mailto:kevin.foster%40zoo.ox.ac.uk?subject=mailto:dcornforth%40gmail.com?subject=


9/9

28. Majeed, H., Gillor, O., Kerr, B. & Riley, M. A.Competitive interactions in Escherichia colipopulations: the role of bacteriocins. ISMEJ. 5,7181 (2010). 29. Cao, M.et al.Defining theBacillus subtilisWregulon: a comparative analysis ofpromoter consensus search, run-off transcription/macroarray analysis (ROMA), and transcriptionalprofiling approaches.J. Mol. Biol. 316, 443457(2002).

30. Walker, D. et al.Transcriptional profiling of colicin-induced cell death of Escherichia coliMG1655

identifies potential mechanisms by which bacteriocinspromote bacterial diversity.J. Bacteriol. 186,866869 (2004).

31. Cavard, D. Effects of temperature and of heat shockon the expression and action of the colicin A lysisprotein.J. Bacteriol. 177, 51895192 (1995).

32. Wood, J. inBacterial Stress Responses 2nd edn (edsStorz, G. & Hengge, R.) 133156 (American Societyfor Microbiology Press, 2011).

33. Liras, P., Gomez-Escribano, J. P. & Santamarta, I.Regulatory mechanisms controlling antibioticproduction in Streptomyces clavuligerus.J. Ind.Microbiol. Biotechnol. 35, 667676 (2008).

34. Sanchez, S.et al.Carbon source regulation of antibioticproduction.J. Antibiot. 63, 442459 (2010).

35. Jensen, V.et al.RhlR expression inPseudomonasaeruginosais modulated by the Pseudomonasquinolone signal via PhoB-dependent and -independentpathways.J. Bacteriol. 188, 86018606 (2006).

36. Xavier, J., Kim, W. & Foster, K. A molecularmechanism that stabilizes cooperative secretions inPseudomonas aeruginosa.Mol. Microbiol. 79,166179 (2011).

37. Perkins, T. J. & Swain, P. S. Strategies for cellulardecision-making.Mol. Syst. Biol.5,326(2009).

38. Gardner, A. Adaptation as organism design.Biol. Lett.5, 861864 (2009).

39. Hindr, T., Pennec, J. P., Haras, D. & Dufour, A.Regulation of lantibiotic lacticin 481 production at thetranscriptional level by acid pH. FEMS Microbiol. Lett.231, 291298 (2004).

40. Darch, S. E., West, S. A., Winzer, K. & Diggle, S. P.Density-dependent fitness benefits in quorum-sensingbacterial populations. Proc. Natl Acad. Sci. USA109,82598263 (2012).

41. Hayes, C. S., Aoki, S. K. & Low, D. A. Bacterial contact-dependent delivery systems.Annu. Rev. Genet. 44,7190 (2010).

42. Gibbs, K. A., Urbanowski, M. L. & Greenberg, E. P.Genetic determinants of self identity and socialrecognition in bacteria. Science 321, 256259

(2008).43. Garbeva, P., Silby, M. W., Raaijmakers, J. M.,

Levy, S. B. & de Boer, W. Transcriptional and

antagonistic responses of Pseudomonas fluorescensPf0-1 to phylogenetically different bacterialcompetitors. ISME J. 5, 973985 (2011).

44. Laub, M. T. inBacterial Stress Responses2nd edn(eds Storz, G. & Hengge, R.) 4558 (American Societyfor Microbiology Press, 2011).

45. Korgaonkar, A. K. & Whiteley, M.Pseudomonasaeruginosaenhances production of an antimicrobial inresponse to N-acetylglucosamine and peptidoglycan.

J. Bacteriol. 193, 909917 (2011).46. Braun, V., Patzer, S. I. & Hantke, K. Ton-dependent

colicins and microcins: modular design and evolution.Biochimie 84, 365380 (2002).47. Maynard Smith, J. & Price, G. R. The logic of animal

conflict. Nature 246, 1518 (1973).48. McGannon, C. M., Fuller, C. A. & Weiss, A. A. Different

classes of antibiotics differentially influence Shigatoxin production.Antimicrob. Agents Chemother. 54,37903798 (2010).

49. Kltz, D. Evolution of the cellular stress proteome:from monophyletic origin to ubiquitous function.

J. Exp. Biol. 206, 31193124 (2003).50. Chopin, M.-C., Chopin, A. & Bidnenko, E. Phage

abortive infection in lactococci: variations on a theme.Curr. Opin. Microbiol. 8, 473479 (2005).

51. Marraffini, L. A. & Sontheimer, E. J. CRISPRinterference: RNA-directed adaptive immunity inbacteria and archaea. Nature Rev. Genet. 11, 181190(2010).

52. Hayes, F. & Van Melderen, L. Toxins-ant itoxins:diversity, evolution and function. Crit. Rev. Biochem.Mol. Biol. 46, 386408 (2011).

53. Roth, J. R., Kugelberg, E., Reams, A. B., Kofoid, E. &Andersson, D. I. Origin of mutations under selection:the adaptive mutation controversy.Annu. Rev.Microbiol. 60, 477501 (2006).

54. Hoffman, L. R.et al.Aminoglycoside antibiotics inducebacterial biofilm formation. Nature 436, 11711175(2005).

55. Collet, A. et al.Impact of rpoSdeletion on theproteome of Escherichia coligrown planktonicallyand as biofilm.J. Proteome Res. 7, 46594669(2008).

56. Pfeiffer, T., Schuster, S. & Bonhoef fer, S. Cooperationand competition in the evolution of ATP-producingpathways. Science 292, 504507 (2001).

57. Davies, J., Spiegelman, G. B. & Yim, G. The world ofsubinhibitory antibiotic concentrations. Curr. Opin.Microbiol. 9, 445453 (2006).

58. Gillor, O. inBacteriocins: Ecology and Evolution (edsRiley, M. A. & Chavan, M. A.) 135145 (Springer,2007).

59. Gardner, A., West, S. A. & Buckling, A. Bacteriocins,spite and virulence. Proc. Biol. Sci. 271, 15291535(2004).

60. Frank, S. A. Spatial polymorphism of bacteriocins andother allelopathic traits.Evol. Ecol. 8, 369386(1994).

61. Wloch-Salamon, D. M., Gerla, D., Hoekstra, R. F. &de Visser, J. A. G. M. Effect of dispersal and nutrientavailability on the competitive ability of toxin-producingyeast. Proc. Biol. Sci. 275, 535541 (2008).

62. Gough, J., Karplus, K., Hughey, R. & Chothia, C.Assignment of homology to genome sequences usinga library of hidden Markov models that represent allproteins of known structure.J. Mol. Biol. 313,

903919 (2001).63. Dauga, C. Evolution of thegyrBgene and themolecular phylogeny of Enterobacteriaceae : a modelmolecule for molecular systematic studies. Int. J. Syst.Evol. Microbiol. 52, 531547 (2002).

64. Labeda, D.et al.Phylogenetic study of the specieswithin the family Streptomycetaceae .Antonie VanLeeuwenhoek 101, 73104 (2012).

65. Fontaine, L. et al.Quorum-sensing regulation of theproduction of Blp bacteriocins in Streptococcusthermophilus.J. Bacteriol. 189, 71957205(2007).

66. Inaoka, T., Takahashi, K., Ohnishi-Kameyama, M.,Yoshida, M. & Ochi, K. Guanine nucleotides guanosine5-diphosphate 3-diphosphate and GTP co-operativelyregulate the production of an antibiotic bacilysin inBacillus subtilis.J. Biol. Chem.278, 21692176(2003).

67. Jerman, B., Butala, M. & gur-Bertok, D. Sublethalconcentrations of ciprofloxacin induce bacteriocinsynthesis in Escherichia coli.Antimicrob. AgentsChemother. 49, 30873090 (2005).

AcknowledgementsThe authors thank S. Brown, N. Davies, N. Oliveira,C. Kleanthous, J. Schluter, C. Maclean, P. Swain, R. Popat,D. Newman, W. Kim, S. Mitri, Z. Frankel, S. Diggle and theanonymous referees. D.M.C. acknowledges a Clarendon FundScholarship (University of Oxford, UK) and the University ofEdinburgh School of Biological Sciences (UK) for support.K.R.F. is supported by European Research Council grant242670.

Competing interests statementThe authors declare no competing financial interests.

FURTHER INFORMATIONKevin R. Fosters homepage:http://zoo-kfoster.zoo.ox.ac.uk

SUPPLEMENTARY INFORMATION

See online article:S1(table) |S2(box) |S3(figure)

ALL LINKS ARE ACTIVE IN THE ONLINE PDF


http://zoo-kfoster.zoo.ox.ac.uk/http://zoo-kfoster.zoo.ox.ac.uk/http://www.nature.com/nrmicro/journal/v11/n4/suppinfo/nrmicro2977.htmlhttp://www.nature.com/nrmicro/journal/v11/n4/suppinfo/nrmicro2977.htmlhttp://www.nature.com/nrmicro/journal/v11/n4/suppinfo/nrmicro2977.htmlhttp://www.nature.com/nrmicro/journal/v11/n4/suppinfo/nrmicro2977.htmlhttp://www.nature.com/nrmicro/journal/v11/n4/suppinfo/nrmicro2977.htmlhttp://www.nature.com/nrmicro/journal/v11/n4/suppinfo/nrmicro2977.htmlhttp://www.nature.com/nrmicro/journal/v11/n4/suppinfo/nrmicro2977.htmlhttp://www.nature.com/nrmicro/journal/v11/n4/suppinfo/nrmicro2977.htmlhttp://www.nature.com/nrmicro/journal/v11/n4/suppinfo/nrmicro2977.htmlhttp://www.nature.com/nrmicro/journal/v11/n4/suppinfo/nrmicro2977.htmlhttp://www.nature.com/nrmicro/journal/v11/n4/suppinfo/nrmicro2977.htmlhttp://zoo-kfoster.zoo.ox.ac.uk/

competition sensing the social side

Documents