Bacteria are more social than what had long been expected. While they can swim around on their own, most of them in fact settle down as part of a surface-bound community. The plaque on our teeth and the gooey deposit in our bathroom pipes are the visible results of this communal lifestyle. Inside this slimy ‘biofilm’, cells share resources and are protected from toxic substances such as antibiotics. However, being tied to one spot is not always a good thing: it may be advantageous for a cell in a biofilm to strike out on its own and resume ‘single life’ if local conditions deteriorate.

Caulobacter crescentus bacteria do not always have this choice, as adult cells in this species become permanently glued into place upon joining a biofilm. When these divide, however, their daughters have a choice: swim away, or stick with the group. Previous research has shown that this decision is influenced by the health of the community. Dying cells release DNA fragments which disable the structures allowing newborn cells to adhere to the environment, and a high mortality rate in the biofilm therefore forces unattached cells to leave the colony.

Berne et al. wanted to build on these results and examine how exactly cells die in the biofilm. In particular, the deaths could be sudden and random, with the bacteria succumbing to injury; or they could result from cells activating one of their built-in self-destruct programs. To investigate this question, genetically modified C. crescentus bacteria were grown in the laboratory and exposed to different environments. Combining genetic and microscopic approaches revealed that as a biofilm becomes too crowded, certain individuals self-destruct via a cell death program known as the toxin-antitoxin system. Further experiments showed that low oxygen availability was the signal that triggered self-destruction. Drops in oxygen levels can happen when the environment becomes hostile or when a colony is overpopulated. The results by Berne et al. therefore suggest that by triggering self-destruction in certain members of the community, reduced oxygen access leads to newborn cells swimming away, which in turn prevents further overcrowding and allows new, more hospitable locations to be colonized.

Biofilms are of growing interest in a wide range of human industries, but they also present formidable challenges. This is particularly the case in healthcare, as they tend to infest medical devices and help disease-causing species to resist treatments. Understanding how bacteria are encouraged to join or leave their colony is necessary to better control biofilms to our advantage.

Biofilms are multicellular communities attached to a surface, where complex exchanges and interactions occur between the different members. Biofilms first start with the attachment of individual bacterial cells to a surface and then grow into a more complex community when attached bacteria divide and new ones join. The biofilm lifestyle is considered beneficial for bacteria, as they usually provide protection from xenobiotic stresses and predators, and increase collective nutrient availability (Flemming et al., 2016). However, when conditions deteriorate, cells can leave the biofilm through a process referred to as dispersal, disseminate to new environments, and form new biofilms, enabling the colonization of new niches (Guilhen et al., 2017). Dispersal is triggered in response to various environment and biological cues, and understanding its regulation is important to determine how biofilms can be controlled.

Many Alphaproteobacteria use a strong polar adhesin to irreversibly attach to surfaces and form biofilms (Berne et al., 2015; Berne et al., 2018b), with Caulobacter crescentus holdfast being the best characterized example. C. crescentus has a dimorphic life cycle, where each division cycle yields a sessile mother stalked cell and a motile daughter swarmer cell. Newborn motile swarmer cells bear a single flagellum and multiple pili at the new pole. After the cell cycle progresses beyond a certain point, or upon contact with a surface, the newborn cells secrete a holdfast at the same pole and differentiate into stalked cells by retracting their pili, ejecting their flagellum, and synthesizing a thin cylindrical extension of the cell envelope called the stalk, which pushes the holdfast away from the cell body. While the chemical composition of the holdfast is not entirely elucidated, it is composed of polysaccharides with four different monosaccharide constituents, as well as DNA and peptide molecules of unknown nature (Merker and Smit, 1988; Hernando-Pérez et al., 2018; Hershey et al., 2019). Holdfast is an extremely strong bioadhesin (Tsang et al., 2006; Berne et al., 2013) crucial for irreversible cell adhesion to solid surfaces (Ong et al., 1990; Bodenmiller et al., 2004), colonization of air–liquid interfaces (Fiebig, 2019), and biofilm formation (Entcheva-Dimitrov and Spormann, 2004).

In some bacterial species, extracellular DNA (eDNA) plays a stabilizing role in the biofilm matrix (Okshevsky and Meyer, 2015; Campoccia et al., 2021). In contrast, we previously showed that C. crescentus eDNA produced via cell lysis negatively regulates biofilm formation and stimulates cell dispersal. eDNA binding to unattached holdfasts inhibits their adhesiveness, thereby inhibiting cell attachment to surfaces (Berne et al., 2010). In contrast, eDNA does not dislodge previously bound holdfasts. Therefore, eDNA prevents newborn swarmer cells from joining mature biofilms, but does not dissociate existing biofilms. Because inhibition by eDNA is proportional to its concentration, we proposed that eDNA serves as a rheostat-like environmental cue to trigger dispersal when conditions are detrimental and cause cell death. However, it was not known if eDNA release is the simple consequence of random cell death occurring in the biofilm as conditions worsen, or if it is the result of an active mechanism, such as programmed cell death (PCD) (Berne et al., 2010; Kirkpatrick and Viollier, 2010).

In this study, we demonstrate that cell death and eDNA release in a biofilm are regulated by a PCD mechanism that responds to oxygen availability. PCD in bacteria includes all genetically encoded mechanisms that lead to cell lysis (Lewis, 2000; Bayles, 2014). Toxin-antitoxin systems (TAS) are important regulators of PCD (Rice and Bayles, 2008; Peeters and de Jonge, 2018). These systems are comprised of a stable toxin and its unstable antitoxin cognate. The antitoxin molecule usually antagonizes the toxin under ‘steady state’ growth conditions; but, in PCD-triggering conditions, the antitoxin is inactivated, leading to an excess of free toxins that target key cellular processes in response to various environmental signals (Harms et al., 2018; Wang et al., 2021). There are currently eight types of TAS described in bacteria. The classification depends on the nature of the antitoxin (RNA in types I, III, and VIII, or small protein in the other TAS types), and the toxin (small protein in all but type VIII where the toxin is a small RNA), and how the antitoxin neutralizes the toxin activity (Song and Wood, 2020a, Singh et al., 2021; Srivastava et al., 2021).

TAS are widespread in bacterial and archaeal genomes, but despite their abundance, the biological relevance of most TAS is still elusive (Fraikin et al., 2020). TAS were first described as plasmid addiction modules that ensure plasmid stabilization via post-segregational killing of plasmid-free cells (Ogura and Hiraga, 1983; Gerdes et al., 1986). TAS have also been shown to promote addiction to certain chromosomally encoded elements such as integrative conjugative elements (Wozniak and Waldor, 2009) or CRISPR-cas loci (Li et al., 2021). In addition, TAS have been described as defense mechanism against phage infection where host translation is inhibited by the phage (Pecota and Wood, 1996; Fineran et al., 2009; Song and Wood, 2020b, LeRoux and Laub, 2022; Vassallo et al., 2022). In bacterial cells that lose their plasmid/chromosomal element encoding the TAS, or that are infected by phage, the amount of labile antitoxin rapidly decreases, leading to toxin activation and subsequent cell death. In addition to the well accepted role of TAS in plasmid addiction and phage exclusion, TAS have been linked to diverse physiological responses, such as biofilm formation, stress response, and persistence (Kamruzzaman et al., 2021), although this is still debated (Ronneau and Helaine, 2019; Wade and Laub, 2019; Song and Wood, 2020a, Jurėnas et al., 2022). Many TAS have been reported to be transcriptionally upregulated under environmental stress conditions (Jurėnas et al., 2022), but this increase does not necessarily trigger liberation of an active toxin (LeRoux et al., 2020).

Among the 18 TAS identified in the C. crescentus genome (Ely, 2021), 13 have been studied experimentally, and belong to four different groups: (1) four paralogous RelBE (Fiebig et al., 2010) operons and one HigBA (Kirkpatrick et al., 2016) operon, belonging to the type II systems where the toxins (RelE or HigB) are known to be mRNA endonucleases; (2) four type II systems belonging to the ParDE family (Fiebig et al., 2010), where ParE toxins are usually DNA gyrase inhibitors; (3) three paralogs of HipBA, also a type II system, where the HipA toxins inhibit protein synthesis (Huang et al., 2020; Zhou et al., 2021); and (4) SocAB, the only member of the type VI TAS described so far, where the SocB toxin directly inhibits DNA replication (Aakre et al., 2013). The environmental conditions that trigger any of these TAS and their biological function are not yet fully identified.

In this study, we show that the ParDE₄ TAS is involved in PCD and eDNA release in C. crescentus biofilms where it stimulates cell dispersal. We show that areas of a biofilm with decreased O₂ availability experience more cell death. Cell viability is improved in a ∆parDE₄ mutant biofilm, especially in areas of decreased O₂ availability, generating less cell lysis and less eDNA release. We also show that cell dispersal is decreased when parDE₄ is deleted, probably due to the lower local eDNA concentration. Expression of parDE₄ is positively regulated by O₂ and the expression of this operon is decreased in biofilms where O₂ availability is low. Thus, PCD by an O₂-regulated TAS stimulates dispersal away from areas of a biofilm with decreased O₂ availability.

While eDNA has been shown in the past to be a major component of the biofilm matrix that plays a role in overall biofilm architecture for different species (Okshevsky and Meyer, 2015; Campoccia et al., 2021), we previously reported a novel role for this molecule, as an environmental cue to sense deleterious environments in a biofilm and promote cell dispersion in C. crescentus (Berne et al., 2010). We showed that cell death and eDNA release increase during biofilm formation and that eDNA prevents newborn swarmer cells from attaching to surfaces and settling into a biofilm by binding specifically to and inhibiting the adhesiveness of the holdfast. Because stalked cells attached by their holdfast cannot be dislodged from a surface by eDNA, this mechanism promotes swarmer cell dispersal without causing a potentially undesirable dissolution of the existing biofilm. While our previous results support the role of eDNA as a cue that prevents settling of swarmer cells (Berne et al., 2010; Kirkpatrick and Viollier, 2010), it was not known if this mechanism was simply a consequence of random cell death or the result of an active mechanism.

In this work, we show that C. crescentus biofilms experience increasing cell death in regions of a mature biofilm where O₂ becomes limiting. We show that this increased cell death in biofilm regions of O₂ limitation depends on the parDE₄ TAS. We also show that parDE₄-dependent cell death is highest in conditions where O₂ is limited. Under conditions of maximal aeration, cell death and eDNA release are equivalent in WT and in the parDE₄ mutant, whereas they are higher in WT under conditions of limited aeration.

Interestingly, even if cells of the parDE₄ mutant increase their biofilm biomass more rapidly than WT, they have a reduced dispersal from the biofilm. This result is consistent with our previous findings that eDNA release increases as biofilms mature and that eDNA inhibits swarmer cell adhesion by binding to their holdfast, but does not dislodge previously bound cells (Berne et al., 2010). We suggest that, as the C. crescentus biofilm increases in biomass, O₂ limitation activates the ParDE₄ TAS, resulting in cell death and concomitant eDNA release, which promotes dispersal of swarmer cells by binding to their holdfast and inhibiting their adhesion. This process enables the colonization of new environments while preserving the survival of the initial biofilm (Figure 9).

Figure 9

Download asset Open asset

We showed that both WT and ∆parDE₄ strains react in a similar manner to the presence of excess eDNA (Figure 3—figure supplement 1), so how can we reconcile this with our observation that WT cells disperse more efficiently than ∆parDE₄ cells (Figure 4C)? C. crescentus biofilms are composed of discrete, clonal microcolonies (Figure 4A and Entcheva-Dimitrov and Spormann, 2004; Rossy et al., 2019). These microcolonies provide a microenvironment where conditions can differ from other microcolonies in the same biofilm. As cells die in biofilm microcolonies, the released eDNA will be at maximum concentration in the microcolony where it was released and will interact with nearby cells before diffusing to adjacent microcolonies where its concentration will be lower. Since eDNA inhibition of swarmer cell adhesion is proportional to its concentration, microcolonies with increased death rate will contain more eDNA and have a higher swarmer cell dispersal rate. Cells from WT microcolonies, which have a higher death rate in mature biofilms, should therefore have a higher rate of dispersal than cells from ∆parDE₄ microcolonies.

In addition to promoting dispersal, the PCD mechanism we describe here may also help preserve the initial biofilm. Indeed, once irreversibly attached via their holdfasts, stalked cells cannot disperse (Berne et al., 2010). In addition to stimulating dispersal of swarmer cells, PCD of some of the stalked cells in the biofilm may help reduce overcrowding, O₂ starvation, and the death of the entire colony. Moreover, O₂ limitation can result not only from overcrowding but also from environmental changes. Since C. crescentus is an obligate aerobe, this PCD mechanism can also act as a safety trigger that promotes the propagation of daughter cells to more hospitable environments when the colony starts to experience suboptimal O₂ levels resulting from an external modification of the environment. Since the magnitude of inhibition of swarmer cell adhesion is proportional to eDNA concentration, this PCD mechanism may act as a rheostat tying cell dispersal to the availability of O₂. Other PCD mechanisms in C. crescentus could similarly tie cell dispersal from the biofilm to other environmental conditions.

The involvement of PCD in biofilm regulation is well documented in different bacterial species and different TAS have been reported to play a role in biofilm regulation and cell dispersal (Wang and Wood, 2011; Wen et al., 2014; Kamruzzaman et al., 2021; Singh et al., 2021). In Escherichia coli, several TAS are involved in regulation of biofilm formation at different stages (Kim et al., 2009; Kolodkin-Gal et al., 2009; Yamaguchi and Inouye, 2011; Soo and Wood, 2013; Zhao et al., 2013). Other TAS have been reported to regulate biofilms in different species, such as in Vibrio cholerae (Wang et al., 2015), Staphylococcus aureus (Kato et al., 2017), or Pseudomonas aeruginosa (Wood and Wood, 2016; Shmidov et al., 2022; Song et al., 2022). O₂ limitation is one of the major environmental stresses experienced by bacteria in a mature biofilm (Stewart and Franklin, 2008; Kostakioti et al., 2013) and some species, such as Shewanella oneidensis or P. aeruginosa, respond to O₂ depletion by triggering single cell dispersion (Thormann et al., 2005; Barraud et al., 2009; Rumbaugh and Sauer, 2020). PCD is involved in biofilm regulation upon O₂ limitation in S. aureus, via the coordinated increased expression of the AltA murein hydrolase and decreased expression of the cell wall-teichoic acids (Mashruwala et al., 2017). In addition, the Hha/TomB TAS regulates cell death in E. coli biofilms in response to O₂ availability (Marimon et al., 2016): under high O₂ conditions, the toxin is inactivated by spontaneous oxidation enhanced by transient interactions with the antitoxin under those conditions. In O₂-limited regions of the biofilm, the TomB anitoxin fails to efficiently oxidaze the Hha toxin stimulates cell death and enables the formation of gaps in the biofilm matrix, thereby facilitating biofilm dispersal through those gaps (García-Contreras et al., 2008; Marimon et al., 2016). However, to our knowledge, none of the previously reported systems function as described in this study.

We showed that the O₂-limited conditions where parDE₄-dependent cell death occurs correspond to the lowest level of parDE₄ transcription. This result suggests the possibility that lowering the expression of the parDE₄ operon triggers cell death. During surface adhesion and in early stages of biofilm development, O₂ availability is good relative to later stages. When biomass increases, O₂ becomes limiting, and cell death is increased. It is interesting to note that the ‘neutralized’ steady state of the ParDE₄ TAS, when the toxin is inactivated, seems to be when O₂ is abundant, that is, when parDE₄ transcription is at its highest. In most studied TAS, stresses have been shown to induce transcription of TAS (LeRoux et al., 2020; Jurėnas et al., 2022), but here, the stress inflicted on the cells by O₂ limitation is accompanied by a lower expression of parDE₄. We are unaware of cases where reduced TAS expression is correlated with the condition that activates the PCD in biofilm regulation. This suggests a novel regulatory mechanism of PCD, in the context of biofilms, that remains to be understood. The system described in this study is conceptually similar to TAS involved in plasmid maintenance/postsegregational killing or phage defence mechanisms. Indeed, TAS encoded on a plasmid, such as the CcdAB (Ogura and Hiraga, 1983) or Hok/Sok (Gerdes et al., 1985) pairs in E. coli, ensure plasmid maintenance over generations: if a daughter cell fails to inherit the plasmid bearing the TAS upon division, the amount of labile antitoxins in this cell quickly drops, allowing free toxins to kill. A similar strategy is seen in instances of phage neutralization upon infection (LeRoux and Laub, 2022) by TAS such as ToxIN (Fineran et al., 2009) and DarTG (LeRoux et al., 2022) for example. Phage infection usually provokes inhibition of the host cell transcription, followed by degradation of the antitoxins and liberation of the toxins. Infected cells are rapidly killed, preventing massive phage replication. In this example, killing a few cells via PCD neutralizes phage infection and rescues the rest of the colony. We propose that the ParDE₄-mediated PCD we describe in this study might play a similar role: selectively killing a few cells in a crowded biofilm where conditions are detrimental to promote the dispersal of newborn cells from the deleterious conditions.

TAS have often been reported to play a direct and crucial role in adaptation to different stresses. However, the physiological role of TAS-regulated cell death has been challenged in recent years (Harms et al., 2017; Goormaghtigh et al., 2018; Wade and Laub, 2019; Fraikin et al., 2020; Rosendahl et al., 2020). Even if a TAS operon is transcribed under stress conditions, it does not necessarily mean that the toxin is liberated, and transcription might not always reflect TAS activity (LeRoux et al., 2020). Therefore, whether the decrease in parDE₄ transcription that correlates with increased cell death is mechanistically related to TA activation is still not known.

Our future work will focus on identifying the pathway that controls the activation of the ParDE₄-mediated PCD. Based on sequence similarity, ParDE₄ belongs to the type II TAS family (Fiebig et al., 2010). In C. crescentus, the ParDE₄ homolog ParDE₁ is thought to have a bacteriostatic effect, inhibiting cell division and causing cell filamentation by an unknown mechanism (Fiebig et al., 2010). Under conditions where ParDE₄-correlated cell death is observed (O₂-limited conditions), cell morphology does not change significantly (Figure 8 and Figure 7—figure supplement 2). In E. coli, the plasmid borne ParDE TAS inhibits DNA replication by targeting DNA gyrase, leading to cell death (Jiang et al., 2002). However, the E. coli ParE₂ toxin, whose structure is close to that of the predicted C. crescentus ParE₁ (Dalton and Crosson, 2010) does not bind to DNA gyrase and the molecular target of this toxin is still unknown (Sterckx et al., 2016). These results suggest that the ParDE family is not homogenous in terms of targets, mechanisms, and physiological outcomes. So far, we have failed to express the parE₄ gene both in E. coli and C. crescentus, suggesting that its product is highly toxic. Our future efforts will aim at engineering an attenuated version of parE₄. Such a tool will help to identify its target(s) and elucidate the pathway that leads to cell killing by this TAS.

Does the PCD-stimulated cell dispersal mechanism we have described constitute cooperation and even altruism? Altruism mechanisms in biofilms have been discussed previously (Penesyan et al., 2021; Sadiq et al., 2021). Altruism involves a trade-off between growth rate and growth yield in the biofilm to preserve resources at the cost of individual growth rate for the fitness of the entire population (Kreft, 2004), or the production of public goods that are used by the whole attached colony (Nadell et al., 2009). In addition, to be beneficial for the overall community while being costly for the individual, Hamilton’s rule states that altruism must promote only the success of the surviving kin (Hamilton, 1970; Ramisetty and Sudhakari, 2020). We showed previously that eDNA stimulation of dispersal through holdfast inhibition is highly specific to Caulobacter species: biofilms of other tested species were not impaired by addition of eDNA, including Caulobacter DNA, and Caulobacter adhesion is only inhibited by eDNA from Caulobacter (Berne et al., 2010). Our current results show that deletion of ParDE₄ decreases dispersal from the mature biofilm. Intuitively, stimulating dispersal away from a biofilm where environmental quality is declining to another, potentially better environment would appear to be advantageous for the dispersing kin in terms of fitness, but this has not been directly demonstrated in our case. Reducing the competing population in a mature biofilm with declining O₂ availability could also provide an advantage to the surviving attached cells, but this is difficult to determine experimentally. Thus, the PCD-mediated mechanism described in this study might confer a fitness advantage to the overall Caulobacter population under the appropriate environmental conditions.

All data generated or analyzed during this study are available on Dryad.

1. 1. Berne C
   2. Zappa S
   3. Brun Y

   (2022) Dryad Digital Repository

   Data from: eDNA-stimulated cell dispersion from Caulobacter crescentus biofilms upon oxygen limitation is dependent on a toxin-antitoxin system.

   https://doi.org/10.5061/dryad.4j0zpc8fc

1. 1. Aakre CD
   2. Phung TN
   3. Huang D
   4. Laub MT

   (2013) A bacterial toxin inhibits DNA replication elongation through A direct interaction with the β sliding clamp

   Molecular Cell 52:617–628.

   https://doi.org/10.1016/j.molcel.2013.10.014

   • PubMed
   • Google Scholar
2. 1. Barraud N
   2. Storey MV
   3. Moore ZP
   4. Webb JS
   5. Rice SA
   6. Kjelleberg S

   (2009) Nitric oxide-mediated dispersal in single- and multi-species biofilms of clinically and industrially relevant microorganisms

   Microbial Biotechnology 2:370–378.

   https://doi.org/10.1111/j.1751-7915.2009.00098.x

   • PubMed
   • Google Scholar
3. 1. Bayles KW

   (2014) Bacterial programmed cell death: making sense of a paradox

   Nature Reviews. Microbiology 12:63–69.

   https://doi.org/10.1038/nrmicro3136

   • PubMed
   • Google Scholar
4. 1. Berne C
   2. Kysela DT
   3. Brun YV

   (2010) A bacterial extracellular DNA inhibits settling of motile progeny cells within a biofilm

   Molecular Microbiology 77:815–829.

   https://doi.org/10.1111/j.1365-2958.2010.07267.x

   • PubMed
   • Google Scholar
5. 1. Berne C
   2. Ma X
   3. Licata NA
   4. Neves BRA
   5. Setayeshgar S
   6. Brun YV
   7. Dragnea B

   (2013) Physiochemical properties of Caulobacter crescentus holdfast: a localized bacterial adhesive

   The Journal of Physical Chemistry. B 117:10492–10503.

   https://doi.org/10.1021/jp405802e

   • PubMed
   • Google Scholar
6. 1. Berne C
   2. Ducret A
   3. Hardy GG
   4. Brun YV

   (2015) Adhesins involved in attachment to abiotic surfaces by gram-negative bacteria

   Microbiology Spectrum 3:e0018-2015.

   https://doi.org/10.1128/microbiolspec.MB-0018-2015

   • PubMed
   • Google Scholar
7. 1. Berne C
   2. Ellison CK
   3. Agarwal R
   4. Severin GB
   5. Fiebig A
   6. Morton RI
   7. Waters CM
   8. Brun YV

   (2018a) Feedback regulation of Caulobacter crescentus holdfast synthesis by flagellum assembly via the holdfast inhibitor hfia

   Molecular Microbiology 110:219–238.

   https://doi.org/10.1111/mmi.14099

   • PubMed
   • Google Scholar
8. 1. Berne C
   2. Ellison CK
   3. Ducret A
   4. Brun YV

   (2018b) Bacterial adhesion at the single-cell level

   Nature Reviews. Microbiology 16:616–627.

   https://doi.org/10.1038/s41579-018-0057-5

   • PubMed
   • Google Scholar
9. 1. Berne C
   2. Brun YV

   (2019) The two chemotaxis clusters in Caulobacter crescentus play different roles in chemotaxis and biofilm regulation

   Journal of Bacteriology 201:e00071-19.

   https://doi.org/10.1128/JB.00071-19

   • PubMed
   • Google Scholar
10. 1. Bodenmiller D
    2. Toh E
    3. Brun YV

    (2004) Development of surface adhesion in Caulobacter crescentus

    Journal of Bacteriology 186:1438–1447.

    https://doi.org/10.1128/JB.186.5.1438-1447.2004

    • PubMed
    • Google Scholar
11. 1. Campoccia D
    2. Montanaro L
    3. Arciola CR

    (2021) Tracing the origins of extracellular DNA in bacterial biofilms: story of death and predation to community benefit

    Biofouling 37:1022–1039.

    https://doi.org/10.1080/08927014.2021.2002987

    • PubMed
    • Google Scholar
12. 1. Crosson S
    2. McGrath PT
    3. Stephens C
    4. McAdams HH
    5. Shapiro L

    (2005) Conserved modular design of an oxygen sensory/signaling network with species-specific output

    PNAS 102:8018–8023.

    https://doi.org/10.1073/pnas.0503022102

    • PubMed
    • Google Scholar
13. 1. Dalton KM
    2. Crosson S

    (2010) A conserved mode of protein recognition and binding in A pard-pare toxin-antitoxin complex

    Biochemistry 49:2205–2215.

    https://doi.org/10.1021/bi902133s

    • PubMed
    • Google Scholar
14. 1. Ely B

    (2021) Evolutionary history of caulobacter toxin-antitoxin systems

    Current Microbiology 78:2899–2904.

    https://doi.org/10.1007/s00284-021-02549-y

    • PubMed
    • Google Scholar
15. 1. Entcheva-Dimitrov P
    2. Spormann AM

    (2004) Dynamics and control of biofilms of the oligotrophic bacterium Caulobacter crescentus

    Journal of Bacteriology 186:8254–8266.

    https://doi.org/10.1128/JB.186.24.8254-8266.2004

    • PubMed
    • Google Scholar
16. 1. Fiebig A
    2. Castro Rojas CM
    3. Siegal-Gaskins D
    4. Crosson S

    (2010) Interaction specificity, toxicity and regulation of a paralogous set of ParE/RelE-family toxin-antitoxin systems

    Molecular Microbiology 77:236–251.

    https://doi.org/10.1111/j.1365-2958.2010.07207.x

    • PubMed
    • Google Scholar
17. 1. Fiebig A

    (2019) Role of Caulobacter cell surface structures in colonization of the air-liquid interface

    Journal of Bacteriology 201:e00064-19.

    https://doi.org/10.1128/JB.00064-19

    • PubMed
    • Google Scholar
18. 1. Fineran PC
    2. Blower TR
    3. Foulds IJ
    4. Humphreys DP
    5. Lilley KS
    6. Salmond GPC

    (2009) The phage abortive infection system, toxin, functions as a protein-RNA toxin-antitoxin pair

    PNAS 106:894–899.

    https://doi.org/10.1073/pnas.0808832106

    • PubMed
    • Google Scholar
19. 1. Flemming HC
    2. Wingender J
    3. Szewzyk U
    4. Steinberg P
    5. Rice SA
    6. Kjelleberg S

    (2016) Biofilms: an emergent form of bacterial life

    Nature Reviews. Microbiology 14:563–575.

    https://doi.org/10.1038/nrmicro.2016.94

    • PubMed
    • Google Scholar
20. 1. Fraikin N
    2. Goormaghtigh F
    3. Van Melderen L

    (2020) Type II toxin-antitoxin systems: evolution and revolutions

    Journal of Bacteriology 202:e00763-19.

    https://doi.org/10.1128/JB.00763-19

    • PubMed
    • Google Scholar
21. 1. García-Contreras R
    2. Zhang XS
    3. Kim Y
    4. Wood TK

    (2008) Protein translation and cell death: the role of rare trRNAs in biofilm formation and in activating dormant phage killer genes

    PLOS ONE 3:e2394.

    https://doi.org/10.1371/journal.pone.0002394

    • PubMed
    • Google Scholar
22. 1. Gerdes K
    2. Larsen JE
    3. Molin S

    (1985) Stable inheritance of plasmid R1 requires two different loci

    Journal of Bacteriology 161:292–298.

    https://doi.org/10.1128/jb.161.1.292-298.1985

    • PubMed
    • Google Scholar
23. 1. Gerdes K
    2. Rasmussen PB
    3. Molin S

    (1986) Unique type of plasmid maintenance function: postsegregational killing of plasmid-free cells

    PNAS 83:3116–3120.

    https://doi.org/10.1073/pnas.83.10.3116

    • PubMed
    • Google Scholar
24. 1. Gober JW
    2. Shapiro L

    (1992) A developmentally regulated Caulobacter flagellar promoter is activated by 3’ enhancer and IHF binding elements

    Molecular Biology of the Cell 3:913–926.

    https://doi.org/10.1091/mbc.3.8.913

    • PubMed
    • Google Scholar
25. 1. Goormaghtigh F
    2. Fraikin N
    3. Putrinš M
    4. Hallaert T
    5. Hauryliuk V
    6. Garcia-Pino A
    7. Sjödin A
    8. Kasvandik S
    9. Udekwu K
    10. Tenson T
    11. Kaldalu N
    12. Van Melderen L

    (2018) Reassessing the role of type II toxin-antitoxin systems in formation of Escherichia coli type II persister cells

    MBio 9:e00640-18.

    https://doi.org/10.1128/mBio.00640-18

    • PubMed
    • Google Scholar
26. 1. Guilhen C
    2. Forestier C
    3. Balestrino D

    (2017) Biofilm dispersal: multiple elaborate strategies for dissemination of bacteria with unique properties

    Molecular Microbiology 105:188–210.

    https://doi.org/10.1111/mmi.13698

    • PubMed
    • Google Scholar
27. 1. Hamilton WD

    (1970) Selfish and spiteful behaviour in an evolutionary model

    Nature 228:1218–1220.

    https://doi.org/10.1038/2281218a0

    • PubMed
    • Google Scholar
28. 1. Harms A
    2. Fino C
    3. Sørensen MA
    4. Semsey S
    5. Gerdes K

    (2017) Prophages and growth dynamics confound experimental results with antibiotic-tolerant persister cells

    MBio 8:e01964-17.

    https://doi.org/10.1128/mBio.01964-17

    • PubMed
    • Google Scholar
29. 1. Harms A
    2. Brodersen DE
    3. Mitarai N
    4. Gerdes K

    (2018) Toxins, targets, and triggers: an overview of toxin-antitoxin biology

    Molecular Cell 70:768–784.

    https://doi.org/10.1016/j.molcel.2018.01.003

    • PubMed
    • Google Scholar
30. 1. Hernando-Pérez M
    2. Setayeshgar S
    3. Hou Y
    4. Temam R
    5. Brun YV
    6. Dragnea B
    7. Berne C

    (2018) Layered structure and complex mechanochemistry underlie strength and versatility in a bacterial adhesive

    MBio 9:e02359-17.

    https://doi.org/10.1128/mBio.02359-17

    • PubMed
    • Google Scholar
31. 1. Hershey DM
    2. Porfírio S
    3. Black I
    4. Jaehrig B
    5. Heiss C
    6. Azadi P
    7. Fiebig A
    8. Crosson S

    (2019) Composition of the holdfast polysaccharide from Caulobacter crescentus

    Journal of Bacteriology 201:e00276-19.

    https://doi.org/10.1128/JB.00276-19

    • PubMed
    • Google Scholar
32. 1. Huang CY
    2. Gonzalez-Lopez C
    3. Henry C
    4. Mijakovic I
    5. Ryan KR

    (2020) HipBA toxin-antitoxin systems mediate persistence in Caulobacter crescentus

    Scientific Reports 10:2865.

    https://doi.org/10.1038/s41598-020-59283-x

    • PubMed
    • Google Scholar
33. 1. Jiang Y
    2. Pogliano J
    3. Helinski DR
    4. Konieczny I

    (2002) ParE toxin encoded by the broad-host-range plasmid RK2 is an inhibitor of Escherichia coli gyrase

    Molecular Microbiology 44:971–979.

    https://doi.org/10.1046/j.1365-2958.2002.02921.x

    • PubMed
    • Google Scholar
34. 1. Johnson RC
    2. Ely B

    (1977) Isolation of spontaneously derived mutants of Caulobacter crescentus

    Genetics 86:25–32.

    https://doi.org/10.1093/genetics/86.1.25

    • PubMed
    • Google Scholar
35. 1. Jurėnas D
    2. Fraikin N
    3. Goormaghtigh F
    4. Van Melderen L

    (2022) Biology and evolution of bacterial toxin-antitoxin systems

    Nature Reviews. Microbiology 20:335–350.

    https://doi.org/10.1038/s41579-021-00661-1

    • PubMed
    • Google Scholar
36. 1. Kamruzzaman M
    2. Wu AY
    3. Iredell JR

    (2021) Biological functions of type II toxin-antitoxin systems in bacteria

    Microorganisms 9:1276.

    https://doi.org/10.3390/microorganisms9061276

    • PubMed
    • Google Scholar
37. 1. Kato F
    2. Yabuno Y
    3. Yamaguchi Y
    4. Sugai M
    5. Inouye M

    (2017) Deletion of MazF increases Staphylococcus aureus biofilm formation in an ica-dependent manner

    Pathogens and Disease 75:ftx026.

    https://doi.org/10.1093/femspd/ftx026

    • PubMed
    • Google Scholar
38. 1. Kim Y
    2. Wang X
    3. Ma Q
    4. Zhang XS
    5. Wood TK

    (2009) Toxin-antitoxin systems in Escherichia coli influence biofilm formation through YjbK (TabA) and fimbriae

    Journal of Bacteriology 191:1258–1267.

    https://doi.org/10.1128/JB.01465-08

    • PubMed
    • Google Scholar
39. 1. Kirkpatrick CL
    2. Viollier PH

    (2010) Cell dispersal in biofilms: an extracellular DNA masks nature’s strongest glue

    Molecular Microbiology 77:801–804.

    https://doi.org/10.1111/j.1365-2958.2010.07268.x

    • PubMed
    • Google Scholar
40. 1. Kirkpatrick CL
    2. Martins D
    3. Redder P
    4. Frandi A
    5. Mignolet J
    6. Chapalay JB
    7. Chambon M
    8. Turcatti G
    9. Viollier PH

    (2016) Growth control switch by a DNA-damage-inducible toxin-antitoxin system in Caulobacter crescentus

    Nature Microbiology 1:16008.

    https://doi.org/10.1038/nmicrobiol.2016.8

    • PubMed
    • Google Scholar
41. 1. Kolodkin-Gal I
    2. Verdiger R
    3. Shlosberg-Fedida A
    4. Engelberg-Kulka H

    (2009) A differential effect of E. coli toxin-antitoxin systems on cell death in liquid media and biofilm formation

    PLOS ONE 4:e6785.

    https://doi.org/10.1371/journal.pone.0006785

    • PubMed
    • Google Scholar
42. 1. Kostakioti M
    2. Hadjifrangiskou M
    3. Hultgren SJ

    (2013) Bacterial biofilms: development, dispersal, and therapeutic strategies in the dawn of the postantibiotic era

    Cold Spring Harbor Perspectives in Medicine 3:a010306.

    https://doi.org/10.1101/cshperspect.a010306

    • PubMed
    • Google Scholar
43. 1. Kreft JU

    (2004) Biofilms promote altruism

    Microbiology 150:2751–2760.

    https://doi.org/10.1099/mic.0.26829-0

    • PubMed
    • Google Scholar
44. 1. LeRoux M
    2. Culviner PH
    3. Liu YJ
    4. Littlehale ML
    5. Laub MT

    (2020) Stress can induce transcription of toxin-antitoxin systems without activating toxin

    Molecular Cell 79:280–292.

    https://doi.org/10.1016/j.molcel.2020.05.028

    • PubMed
    • Google Scholar
45. 1. LeRoux M
    2. Laub MT

    (2022) Toxin-Antitoxin systems as phage defense elements

    Annual Review of Microbiology 76:21–43.

    https://doi.org/10.1146/annurev-micro-020722-013730

    • PubMed
    • Google Scholar
46. 1. LeRoux M
    2. Srikant S
    3. Teodoro GIC
    4. Zhang T
    5. Littlehale ML
    6. Doron S
    7. Badiee M
    8. Leung AKL
    9. Sorek R
    10. Laub MT

    (2022) The DarTG toxin-antitoxin system provides phage defence by ADP-ribosylating viral DNA

    Nature Microbiology 7:1028–1040.

    https://doi.org/10.1038/s41564-022-01153-5

    • PubMed
    • Google Scholar
47. 1. Lewis K

    (2000) Programmed death in bacteria

    Microbiology and Molecular Biology Reviews 64:503–514.

    https://doi.org/10.1128/MMBR.64.3.503-514.2000

    • PubMed
    • Google Scholar
48. 1. Li M
    2. Gong L
    3. Cheng F
    4. Yu H
    5. Zhao D
    6. Wang R
    7. Wang T
    8. Zhang S
    9. Zhou J
    10. Shmakov SA
    11. Koonin EV
    12. Xiang H

    (2021) Toxin-Antitoxin RNA pairs safeguard CRISPR-Cas systems

    Science 372:eabe5601.

    https://doi.org/10.1126/science.abe5601

    • PubMed
    • Google Scholar
49. 1. Marimon O
    2. Teixeira JMC
    3. Cordeiro TN
    4. Soo VWC
    5. Wood TL
    6. Mayzel M
    7. Amata I
    8. García J
    9. Morera A
    10. Gay M
    11. Vilaseca M
    12. Orekhov VY
    13. Wood TK
    14. Pons M

    (2016) An oxygen-sensitive toxin-antitoxin system

    Nature Communications 7:13634.

    https://doi.org/10.1038/ncomms13634

    • PubMed
    • Google Scholar
50. 1. Marks ME
    2. Castro-Rojas CM
    3. Teiling C
    4. Du L
    5. Kapatral V
    6. Walunas TL
    7. Crosson S

    (2010) The genetic basis of laboratory adaptation in Caulobacter crescentus

    Journal of Bacteriology 192:3678–3688.

    https://doi.org/10.1128/JB.00255-10

    • PubMed
    • Google Scholar
51. 1. Mashruwala AA
    2. van de Guchte A
    3. Boyd JM

    (2017) Impaired respiration elicits srrab-dependent programmed cell lysis and biofilm formation in Staphylococcus aureus

    eLife 6:e23845.

    https://doi.org/10.7554/eLife.23845

    • PubMed
    • Google Scholar
52. 1. McDougald D
    2. Rice SA
    3. Barraud N
    4. Steinberg PD
    5. Kjelleberg S

    (2011) Should we stay or should we go: mechanisms and ecological consequences for biofilm dispersal

    Nature Reviews. Microbiology 10:39–50.

    https://doi.org/10.1038/nrmicro2695

    • PubMed
    • Google Scholar
53. 1. Merker RI
    2. Smit J

    (1988) Characterization of the adhesive holdfast of marine and freshwater Caulobacters

    Applied and Environmental Microbiology 54:2078–2085.

    https://doi.org/10.1128/aem.54.8.2078-2085.1988

    • PubMed
    • Google Scholar
54. Book

    1. Miller J

    (1972)

    Experiments in Molecular Genetics

    Cold Spring Harbor Laboratory.

    • Google Scholar
55. 1. Nadell CD
    2. Xavier JB
    3. Foster KR

    (2009) The sociobiology of biofilms

    FEMS Microbiology Reviews 33:206–224.

    https://doi.org/10.1111/j.1574-6976.2008.00150.x

    • PubMed
    • Google Scholar
56. 1. Nierman WC
    2. Feldblyum TV
    3. Laub MT
    4. Paulsen IT
    5. Nelson KE
    6. Eisen JA
    7. Heidelberg JF
    8. Alley MR
    9. Ohta N
    10. Maddock JR
    11. Potocka I
    12. Nelson WC
    13. Newton A
    14. Stephens C
    15. Phadke ND
    16. Ely B
    17. DeBoy RT
    18. Dodson RJ
    19. Durkin AS
    20. Gwinn ML
    21. Haft DH
    22. Kolonay JF
    23. Smit J
    24. Craven MB
    25. Khouri H
    26. Shetty J
    27. Berry K
    28. Utterback T
    29. Tran K
    30. Wolf A
    31. Vamathevan J
    32. Ermolaeva M
    33. White O
    34. Salzberg SL
    35. Venter JC
    36. Shapiro L
    37. Fraser CM

    (2001) Complete genome sequence of Caulobacter crescentus

    PNAS 98:4136–4141.

    https://doi.org/10.1073/pnas.061029298

    • PubMed
    • Google Scholar
57. 1. Ogura T
    2. Hiraga S

    (1983) Mini-F plasmid genes that couple host cell division to plasmid proliferation

    PNAS 80:4784–4788.

    https://doi.org/10.1073/pnas.80.15.4784

    • PubMed
    • Google Scholar
58. 1. Okshevsky M
    2. Meyer RL

    (2015) The role of extracellular DNA in the establishment, maintenance and perpetuation of bacterial biofilms

    Critical Reviews in Microbiology 41:341–352.

    https://doi.org/10.3109/1040841X.2013.841639

    • PubMed
    • Google Scholar
59. 1. Ong CJ
    2. Wong ML
    3. Smit J

    (1990) Attachment of the adhesive holdfast organelle to the cellular stalk of Caulobacter crescentus

    Journal of Bacteriology 172:1448–1456.

    https://doi.org/10.1128/jb.172.3.1448-1456.1990

    • PubMed
    • Google Scholar
60. 1. Pecota DC
    2. Wood TK

    (1996) Exclusion of T4 phage by the hok/sok killer locus from plasmid R1

    Journal of Bacteriology 178:2044–2050.

    https://doi.org/10.1128/jb.178.7.2044-2050.1996

    • PubMed
    • Google Scholar
61. 1. Peeters SH
    2. de Jonge MI

    (2018) For the greater good: programmed cell death in bacterial communities

    Microbiological Research 207:161–169.

    https://doi.org/10.1016/j.micres.2017.11.016

    • PubMed
    • Google Scholar
62. 1. Penesyan A
    2. Paulsen IT
    3. Kjelleberg S
    4. Gillings MR

    (2021) Three faces of biofilms: a microbial lifestyle, a nascent multicellular organism, and an incubator for diversity

    NPJ Biofilms and Microbiomes 7:80.

    https://doi.org/10.1038/s41522-021-00251-2

    • PubMed
    • Google Scholar
63. 1. Pfaffl MW

    (2001) A new mathematical model for relative quantification in real-time RT-PCR

    Nucleic Acids Research 29:e45.

    https://doi.org/10.1093/nar/29.9.e45

    • PubMed
    • Google Scholar
64. 1. Poindexter JS

    (1964) Biological properties and classification of the Caulobacter group

    Bacteriological Reviews 28:231–295.

    https://doi.org/10.1128/br.28.3.231-295.1964

    • PubMed
    • Google Scholar
65. 1. Ramisetty BCM
    2. Sudhakari PA

    (2020) “Bacterial programmed cell death”: cellular altruism or genetic selfism?

    FEMS Microbiology Letters 367:fnaa141.

    https://doi.org/10.1093/femsle/fnaa141

    • PubMed
    • Google Scholar
66. 1. Rice KC
    2. Bayles KW

    (2008) Molecular control of bacterial death and lysis

    Microbiology and Molecular Biology Reviews 72:85–109.

    https://doi.org/10.1128/MMBR.00030-07

    • PubMed
    • Google Scholar
67. 1. Ronneau S
    2. Helaine S

    (2019) Clarifying the link between toxin-antitoxin modules and bacterial persistence

    Journal of Molecular Biology 431:3462–3471.

    https://doi.org/10.1016/j.jmb.2019.03.019

    • PubMed
    • Google Scholar
68. 1. Rosendahl S
    2. Tamman H
    3. Brauer A
    4. Remm M
    5. Hõrak R

    (2020) Chromosomal toxin-antitoxin systems in Pseudomonas putida are rather selfish than beneficial

    Scientific Reports 10:1–13.

    https://doi.org/10.1038/s41598-020-65504-0

    • Google Scholar
69. 1. Rossy T
    2. Nadell CD
    3. Persat A

    (2019) Cellular advective-diffusion drives the emergence of bacterial surface colonization patterns and heterogeneity

    Nature Communications 10:2471.

    https://doi.org/10.1038/s41467-019-10469-6

    • PubMed
    • Google Scholar
70. 1. Rotman B
    2. Zderic JA
    3. Edelstein M

    (1963) Fluorogenic substrates for β-D-galactosidases and phosphatases derived from fluorescein (3, 6-dihydroxyfluoran) and its monomethyl ether

    PNAS 50:1–6.

    https://doi.org/10.1073/pnas.50.1.1

    • Google Scholar
71. 1. Rumbaugh KP
    2. Sauer K

    (2020) Biofilm dispersion

    Nature Reviews. Microbiology 18:571–586.

    https://doi.org/10.1038/s41579-020-0385-0

    • PubMed
    • Google Scholar
72. 1. Sadiq FA
    2. Burmølle M
    3. Heyndrickx M
    4. Flint S
    5. Lu W
    6. Chen W
    7. Zhao J
    8. Zhang H

    (2021) Community-Wide changes reflecting bacterial interspecific interactions in multispecies biofilms

    Critical Reviews in Microbiology 47:338–358.

    https://doi.org/10.1080/1040841X.2021.1887079

    • PubMed
    • Google Scholar
73. 1. Schneider CA
    2. Rasband WS
    3. Eliceiri KW

    (2012) Nih image to imagej: 25 years of image analysis

    Nature Methods 9:671–675.

    https://doi.org/10.1038/nmeth.2089

    • PubMed
    • Google Scholar
74. 1. Shmidov E
    2. Lebenthal-Loinger I
    3. Roth S
    4. Karako-Lampert S
    5. Zander I
    6. Shoshani S
    7. Danielli A
    8. Banin E

    (2022) PrrT/A, a Pseudomonas aeruginosa bacterial encoded toxin-antitoxin system involved in prophage regulation and biofilm formation

    Microbiology Spectrum 10:e0118222.

    https://doi.org/10.1128/spectrum.01182-22

    • PubMed
    • Google Scholar
75. 1. Singh G
    2. Yadav M
    3. Ghosh C
    4. Rathore JS

    (2021) Bacterial toxin-antitoxin modules: classification, functions, and association with persistence

    Current Research in Microbial Sciences 2:100047.

    https://doi.org/10.1016/j.crmicr.2021.100047

    • PubMed
    • Google Scholar
76. 1. Song S
    2. Wood TK

    (2020a) A primary physiological role of toxin/antitoxin systems is phage inhibition

    Frontiers in Microbiology 11:1895.

    https://doi.org/10.3389/fmicb.2020.01895

    • PubMed
    • Google Scholar
77. 1. Song S
    2. Wood TK

    (2020b) Toxin/antitoxin system paradigms: toxins bound to antitoxins are not likely activated by preferential antitoxin degradation

    Advanced Biosystems 4:e1900290.

    https://doi.org/10.1002/adbi.201900290

    • PubMed
    • Google Scholar
78. 1. Song Y
    2. Zhang S
    3. Ye Z
    4. Song Y
    5. Chen L
    6. Tong A
    7. He Y
    8. Bao R

    (2022) The novel type II toxin-antitoxin pacta modulates Pseudomonas aeruginosa iron homeostasis by obstructing the DNA-binding activity of fur

    Nucleic Acids Research 50:10586–10600.

    https://doi.org/10.1093/nar/gkac867

    • PubMed
    • Google Scholar
79. 1. Soo VWC
    2. Wood TK

    (2013) Antitoxin MqsA represses curli formation through the master biofilm regulator CsgD

    Scientific Reports 3:3186.

    https://doi.org/10.1038/srep03186

    • PubMed
    • Google Scholar
80. 1. Srivastava A
    2. Pati S
    3. Kaushik H
    4. Singh S
    5. Garg LC

    (2021) Toxin-Antitoxin systems and their medical applications: current status and future perspective

    Applied Microbiology and Biotechnology 105:1803–1821.

    https://doi.org/10.1007/s00253-021-11134-z

    • Google Scholar
81. 1. Sterckx YGJ
    2. Jové T
    3. Shkumatov AV
    4. Garcia-Pino A
    5. Geerts L
    6. De Kerpel M
    7. Lah J
    8. De Greve H
    9. Van Melderen L
    10. Loris R

    (2016) A unique hetero-hexadecameric architecture displayed by the Escherichia coli O157 paaA2-parE2 antitoxin-toxin complex

    Journal of Molecular Biology 428:1589–1603.

    https://doi.org/10.1016/j.jmb.2016.03.007

    • PubMed
    • Google Scholar
82. 1. Stewart PS
    2. Franklin MJ

    (2008) Physiological heterogeneity in biofilms

    Nature Reviews. Microbiology 6:199–210.

    https://doi.org/10.1038/nrmicro1838

    • PubMed
    • Google Scholar
83. 1. Thanbichler M
    2. Iniesta AA
    3. Shapiro L

    (2007) A comprehensive set of plasmids for vanillate- and xylose-inducible gene expression in Caulobacter crescentus

    Nucleic Acids Research 35:e137.

    https://doi.org/10.1093/nar/gkm818

    • PubMed
    • Google Scholar
84. 1. Thormann KM
    2. Saville RM
    3. Shukla S
    4. Spormann AM

    (2005) Induction of rapid detachment in Shewanella oneidensis MR-1 biofilms

    Journal of Bacteriology 187:1014–1021.

    https://doi.org/10.1128/JB.187.3.1014-1021.2005

    • PubMed
    • Google Scholar
85. 1. Tsang PH
    2. Li G
    3. Brun YV
    4. Freund LB
    5. Tang JX

    (2006) Adhesion of single bacterial cells in the micronewton range

    PNAS 103:5764–5768.

    https://doi.org/10.1073/pnas.0601705103

    • PubMed
    • Google Scholar
86. 1. Vassallo CN
    2. Doering CR
    3. Littlehale ML
    4. Teodoro GIC
    5. Laub MT

    (2022) A functional selection reveals previously undetected anti-phage defence systems in the E. coli pangenome

    Nature Microbiology 7:1568–1579.

    https://doi.org/10.1038/s41564-022-01219-4

    • PubMed
    • Google Scholar
87. 1. Wade JT
    2. Laub MT

    (2019) Concerns about “stress-induced MazF-mediated proteins in Escherichia coli.”

    MBio 10:e00825-19.

    https://doi.org/10.1128/mBio.00825-19

    • PubMed
    • Google Scholar
88. 1. Wang X
    2. Wood TK

    (2011) Toxin-Antitoxin systems influence biofilm and persister cell formation and the general stress response

    Applied and Environmental Microbiology 77:5577–5583.

    https://doi.org/10.1128/AEM.05068-11

    • PubMed
    • Google Scholar
89. 1. Wang Y
    2. Wang H
    3. Hay AJ
    4. Zhong Z
    5. Zhu J
    6. Kan B

    (2015) Functional relBE-family toxin-antitoxin pairs affect biofilm maturation and intestine colonization in Vibrio cholerae

    PLOS ONE 10:e0135696.

    https://doi.org/10.1371/journal.pone.0135696

    • PubMed
    • Google Scholar
90. 1. Wang X
    2. Yao J
    3. Sun YC
    4. Wood TK

    (2021) Type VII toxin/antitoxin classification system for antitoxins that enzymatically neutralize toxins

    Trends in Microbiology 29:388–393.

    https://doi.org/10.1016/j.tim.2020.12.001

    • PubMed
    • Google Scholar
91. 1. Wen Y
    2. Behiels E
    3. Devreese B

    (2014) Toxin-Antitoxin systems: their role in persistence, biofilm formation, and pathogenicity

    Pathogens and Disease 70:240–249.

    https://doi.org/10.1111/2049-632X.12145

    • PubMed
    • Google Scholar
92. 1. West L
    2. Yang D
    3. Stephens C

    (2002) Use of the Caulobacter crescentus genome sequence to develop a method for systematic genetic mapping

    Journal of Bacteriology 184:2155–2166.

    https://doi.org/10.1128/JB.184.8.2155-2166.2002

    • PubMed
    • Google Scholar
93. 1. Wood TL
    2. Wood TK

    (2016) The HigB/HigA toxin/antitoxin system of Pseudomonas aeruginosa influences the virulence factors pyochelin, pyocyanin, and biofilm formation

    MicrobiologyOpen 5:499–511.

    https://doi.org/10.1002/mbo3.346

    • PubMed
    • Google Scholar
94. 1. Wozniak RAF
    2. Waldor MK

    (2009) A toxin-antitoxin system promotes the maintenance of an integrative conjugative element

    PLOS Genetics 5:e1000439.

    https://doi.org/10.1371/journal.pgen.1000439

    • PubMed
    • Google Scholar
95. 1. Yamaguchi Y
    2. Inouye M

    (2011) Regulation of growth and death in Escherichia coli by toxin-antitoxin systems

    Nature Reviews. Microbiology 9:779–790.

    https://doi.org/10.1038/nrmicro2651

    • PubMed
    • Google Scholar
96. 1. Zhao J
    2. Wang Q
    3. Li M
    4. Heijstra BD
    5. Wang S
    6. Liang Q
    7. Qi Q

    (2013) Escherichia coli toxin gene hipA affects biofilm formation and DNA release

    Microbiology 159:633–640.

    https://doi.org/10.1099/mic.0.063784-0

    • PubMed
    • Google Scholar
97. 1. Zhou X
    2. Eckart MR
    3. Shapiro L
    4. Blaser MJ

    (2021) A bacterial toxin perturbs intracellular amino acid balance to induce persistence

    MBio 12:e03020-20.

    https://doi.org/10.1128/mBio.03020-20

    • PubMed
    • Google Scholar

Decision letter after peer review:

Thank you for submitting your article "eDNA-stimulated cell dispersion from Caulobacter crescentus biofilms upon oxygen limitation is dependent on a toxin- antitoxin system" for consideration by eLife. Your article has been reviewed by 2 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Naama Barkai as the Senior Editor. The reviewers have opted to remain anonymous.

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

1) The authors attempt to measure parDE expression in different areas of a biofilm with a promoter-lacZ reporter in which activity of the reporter is measured with a substrate that becomes fluorescent when cleaved by the lacZ gene product. This is a clever approach, but it lacks normalization by cell density, a typical feature of assessments of promoter-lacZ fusions. On first pass, the results presented are identical to what one would expect based on differences in cell density alone and I don't believe any conclusions can be drawn about expression of parDE4 across the biofilms grown in static cultures. Given that the authors have already built PparDE4-GFP fusions, it would be quite straightforward to repeat the experiment presented in Figure 8A by comparing PparDE4-GFP signal with a constitutively expressed red fluorescent protein for normalization of cell density. However, the use of stable reporter proteins poses a problem as these would accumulate more in non-growing or slow-growing cells and not necessarily read out cell density. The best solution here therefore would be to measure expression by directly measuring RNA levels (qRT-PCR or RNA-seq) not via stable protein reporters.

2) While the introduction presents background on TA types, and known TA involvement in biofilms, there is no discussion about other known roles of TA function. In particular, post-segregational killing seems like a relevant topic since this is presumably a similar mechanism as what is being observed here (in that case, TA transcription stops due to plasmid loss, while here TA transcription is reduced due to O2-dependent regulation). Several of the points made in the discussion about TA systems/PCD could also be made in the introduction to more clearly set the reader up with the necessary context to appreciate the impact/contribution of this work (e.g. some of the controversy in the PCD/TA literature, the inverse relationship between TA transcription and activity).

3) The choice of media in each experiment is not clear, but as the authors know, has important implications for biofilm formation in terms of baseline fraction of cells which elaborate holdfast and sensitivity to surface stimulation of holdfast synthesis. Furthermore, no rationale is given for media choices. The authors should be explicit in each experiment about the growth conditions including media, or at least state that all physiological experiments are were done in the same medium. The methods section ambiguously states that cells were cultures in either PYE or M2G.

Reviewer #1 (Recommendations for the authors):

Suggested experiments to further support the key conclusion that the ParE toxin is responsible for cell death:

– Make a catalytically inactive toxin variant to see whether cell death requires the toxin activity.

– Measure ParE activity directly e.g. assess whether cells have altered sensitivity to gyrase inhibiting drugs, image cells at higher resolution (possibly with DAPI staining) to determine whether changes to cell morphology is consistent with gyrase inhibition.

Suggested experiments to address whether transcriptional repression of ParDE4 leads to release of the toxin and thus cell death:

– Measure protein levels of toxin and antitoxin by western blot in O2-limited cells to see if there is a change in ratio of toxin and antitoxin

– Shut off parDE transcription either by placing the locus under control of a repressible promoter or using CRISPRi, then measuring growth (or evidence of ParE activity) after repression. In both cases, growth inhibition consistent with ParE activity should now be O2-independent.

Reviewer #2 (Recommendations for the authors):

Given the timing of biofilm formation, in the experiments involving growth in a flow cell, it seems the authors have used defined medium in which wild-type cells have a low propensity for holdfast synthesis and attachment. Though not essential for this publication, I wonder if parDE plays a role in PCD and dispersal when cells are grown in richer, more complex PYE media where attachment is more robust.

I have several questions/comments about the measurements and experimental conditions for the static growth experiments:

The authors claim to quantify "cell death and eDNA release in these biofilms" (line 148) but the measurements of live/dead cells and eDNA are from the planktonic phase not the surface attached cells. I suspect the phenomena that parDE4 contributes to cell death and eDNA release under limited oxygen conditions is not limited to surface attached biofilm conditions, thus I don't question the data, but rather how it is discussed. I recommend the authors focus on "culture conditions" rather than "biofilm condition" in the early part of the paper where they establish that parDE4 contributes to cell death under oxygen limitation. Certainly the culture conditions affect biofilm dynamics, but I suspect these phenotypes do not require the cells to be in a biofilm. If the authors wanted to formally examine this idea, and specifically test if the cells needed to be in a biofilm to 'feel' the effects of parDE4 and low oxygen, they could repeat the static culture experiments in a holdfast null strain. I predict that cell death and eDNA release would be comparable to holdfast competent cells, without the potentially confounding effects of cell attachment.

Figure S1A: Please be more explicit about how the growth measurement were made. Did you measure the same tube over time? Were the cultures mixed prior to measuring OD? If so, how did you maintain the aeration condition? Or did you set up parallel tubes and make terminal measurements at each time point? Growth medium?

Line 36: use of an acronym in abstract without definition

Line 79-80: perhaps emphasize "new-born swarmer" and consider "dissociate" instead of "disperse".

Line 150: 'viability' was not measured. Please rephrase to reflect the experimental measurement.

Please add scale bars to all microscopy images (Figure 2 and 4).

line 189: Similar to other comments above, the authors should rephrase point 1 as they don't measure cell death "in the biofilm". Perhaps instead: "ParD4 has a protective effect against cell lysis that enhances biofilm formation".

The red signal corresponding to dead cells is very difficult to see in Figure 2. From the images as presented, it doesn't seem that 40% of WT cells are dead. Consider showing the channels separately and merged. Also given the difficulty if seeing sparse signal on a black background, consider scaling the images differently (for example from 0=white to max=color, or present the single channels as inverted BW images).

Line 181: activation of TAS is not just about "production" of the proteins. More generally activation comes when there is an imbalance in steady state levels, and reduced stability of antitoxins compared to toxins is an important contributor to these imbalances. The discussion should be enhanced.

The authors should clarify metrics used to quantify microscopy images. It seems that the authors use thresholded areas in a binary fashion that ignores information from signal intensity. But it in these multidimensional structures, areas with thicker cell masses should have higher signal than areas with monolayers of cells. By simply thresholding the images and quantifying area, the authors loose a dimension of the data. Summing intensity in areas where signals overlap seems more informative than simply treating the pixels as binary.

Figure 6D, would be better presented more parallel to 6C such that the Y-axis reflected "relative biofilm formation (∆parDE4 / WT)" rather than "biofilm formation (% of WT)". This would also be more consistent with the language in line 280.

It is not clear why some experiments assess ccoN expression via a promoter reporter and others via transcript abundance by RT-PCR. Nevertheless, the RT-PCR experiments should be labeled as ccoN transcript levels rather than PccoN expression (line 273 and Figure 6B). Also I suspect that ccoN transcript is essentially undetectable under maximal aeration conditions leading to a denominator of practically 0 in the "fold-change". While I have no question about the veracity of the data, I prefer non-normalized (non-ratioed) data whenever possible, especially when one condition is essentially "off". Just my two cents.

Line 306-7: wording is somewhat confusing.

Figure 7 and S3: I have several suggestions for data presentation. Make the data points smaller in the scatter plots to better visualize the density of points near the axis. Consider plotting on a log scale rather than a linear scale to better spread the points with lower intensity and highlight the bulk of the cells rather than the handful of outliers which are what is primarily obvious on the linear scaling.

Please indicate how the threshold was determined to assess if each promoter was on or off for figure 7A. How many cells have "both"? Is "non-labeled" the best description of cells in which neither promoter was sufficiently active to be called "on"? Even "non-fluorescent" seems better than "non-labeled" as label implies staining.

Line 505: the authors report using 10 ug/ml of chloramphenicol to select for plasmids in Caulobacter. This seems extremely high and quite inhibitory even in the presence of a Cm resistance gene. Typically for Caulobacter 1-2 ug/ml chlor is sufficient to maintain a plasmid. Please check if this number is correct.

Essential revisions:

1) The authors attempt to measure parDE expression in different areas of a biofilm with a promoter-lacZ reporter in which activity of the reporter is measured with a substrate that becomes fluorescent when cleaved by the lacZ gene product. This is a clever approach, but it lacks normalization by cell density, a typical feature of assessments of promoter-lacZ fusions. On first pass, the results presented are identical to what one would expect based on differences in cell density alone and I don't believe any conclusions can be drawn about expression of parDE4 across the biofilms grown in static cultures. Given that the authors have already built PparDE4-GFP fusions, it would be quite straightforward to repeat the experiment presented in Figure 8A by comparing PparDE4-GFP signal with a constitutively expressed red fluorescent protein for normalization of cell density. However, the use of stable reporter proteins poses a problem as these would accumulate more in non-growing or slow-growing cells and not necessarily read out cell density. The best solution here therefore would be to measure expression by directly measuring RNA levels (qRT-PCR or RNA-seq) not via stable protein reporters.

We agree with that the experiment presented in Figure 8A was lacking some internal controls. While we agree that qPCR or RNA-seq would be good options to assess parDE₄ expression in the different portions of the biofilm (top, middle and bottom layers), technical reasons precluded this. First, our experimental set-up (about 1/3 of a 1.8 x 0.5 cm piece of plastic coverslip) didn't provide sufficient material for RNA extraction for suitable qPCR or RNA-seq experiments. Furthermore, both the toxin and antitoxin genes are short (303 and 291 bp respectively), providing limited options for qPCR primers. We tested many primer pairs but in each case, the denaturation curve showed a lot of non-specific amplification.

As an alternative approach, and as suggested by the reviewer, we constructed a strain constitutively expressing the red fluorescence protein Dsred (similarly to the one we use for the competition experiments in Figure 4) and harboring the PparDE₄-lacZ construct we are using throughout this work (Figures 3C and 6B). We repeated the experiments and monitored parDE₄ expression in different biofilm layers using our promoter-lacZ reporter and FDG (a fluorogenic dye cleaved by ß-galactosidase to release fluorescein) and the red fluorescence signal produced by Dsred. The red fluorescence produced by the constitutively expressed Dsred was used as a measure of cell density to normalize the results. These new results now replace the previous data in Figure 8A and lead to the same conclusions as in the initial submission. In addition, we repeated these experiments using a strain harboring the same constitutively expressed Dsred protein and the O₂ regulated PccoN-lacZ construct we used in Figure 3C to provide a measure of O₂ availability in the two lower biofilm areas.

2) While the introduction presents background on TA types, and known TA involvement in biofilms, there is no discussion about other known roles of TA function. In particular, post-segregational killing seems like a relevant topic since this is presumably a similar mechanism as what is being observed here (in that case, TA transcription stops due to plasmid loss, while here TA transcription is reduced due to O2-dependent regulation). Several of the points made in the discussion about TA systems/PCD could also be made in the introduction to more clearly set the reader up with the necessary context to appreciate the impact/contribution of this work (e.g. some of the controversy in the PCD/TA literature, the inverse relationship between TA transcription and activity).

We agree and we now discuss this in both the introduction and Discussion sections in this new version.

Lines 103-130 (introduction): " TAS are widespread in bacterial and archaeal genomes, but despite their abundance, the biological relevance of most TAS is still elusive (Fraikin et al., 2020). TAS were first described as plasmid addiction modules that ensure plasmid stabilization via post-segregational killing of plasmid-free cells (Ogura and Hiraga, 1983, Gerdes et al., 1986). TAS have also been shown to promote addiction to certain chromosomally-encoded elements such as integrative conjugative elements (Wozniak and Waldor, 2009) or CRISPR-cas loci (Li et al., 2021). In addition, TAS have been described as defense mechanism against phage infection where host translation is inhibited by the phage (Pecota and Wood, 1996, Fineran et al., 2009, Song and Wood, 2020, LeRoux and Laub, 2022, Vassallo et al., 2022). In bacterial cells that lost their plasmid / chromosomal element encoding the TAS, or that are infected by phage, the amount of labile antitoxin rapidly decreases, leading to toxin activation and subsequent cell death. In addition to the well accepted role of TAS in addiction and phage exclusion, TAS have been linked to diverse physiological responses, such as biofilm formation, stress response and persistence (Kamruzzaman et al., 2021), although this is still debated (Ronneau and Helaine, 2019, Wade and Laub, 2019, Song and Wood, 2020, Jurėnas et al., 2022). Many TAS have been reported to be transcriptionally upregulated under environmental stress conditions (Jurėnas et al., 2022), but this increase does not necessarily trigger liberation of an active toxin (LeRoux et al., 2020). "

Lines 519-544 (discussion): " The system described in this study is conceptually similar to TAS involved in plasmid maintenance / postsegregational killing or phage defence mechanisms. Indeed, TAS encoded on a plasmid, such as the CcdAB (Ogura and Hiraga, 1983) or Hok/Sok (Gerdes et al., 1985) pairs in E. coli, ensure plasmid maintenance over generations: if a daughter cell fails to inherit the plasmid bearing the TAS upon division, the amount of labile antitoxins in this cell quickly drops, allowing free toxins to kill. A similar strategy is seen in instances of phage neutralization upon infection (LeRoux and Laub, 2022) by TAS such as ToxIN (Fineran et al., 2009) and DarTG (LeRoux et al., 2022) for example. Phage infection usually provokes inhibition of the host cell transcription, followed by degradation of the antitoxins and liberation of the toxins. Infected cells are rapidly killed, preventing massive phage replication. In this example, killing a few cells via PCD neutralizes phage infection and rescues the rest of the colony. We propose the ParDE₄-mediated PCD we described in this study might fulfill a similar role: selectively killing a few cells in a crowded biofilm where conditions are deleterious, to promote the dispersal of newborn cells from the deleterious conditions."

3) The choice of media in each experiment is not clear, but as the authors know, has important implications for biofilm formation in terms of baseline fraction of cells which elaborate holdfast and sensitivity to surface stimulation of holdfast synthesis. Furthermore, no rationale is given for media choices. The authors should be explicit in each experiment about the growth conditions including media, or at least state that all physiological experiments are were done in the same medium. The methods section ambiguously states that cells were cultures in either PYE or M2G.

We apologize for the confusion. All experiments have been done in M2G. PYE was only used during strain construction and cloning procedures. We now provide this information in the "Material and Method section" (Lines 619-620: "All experiments were carried out using M2G, while C. crescentus was grown in PYE for strain construction."). M2G usage was already mentioned in the relevant subsection of the Material and Method section, and we now also added this information in figure legends when appropriate.

We used M2G in all our experiments for several reasons: Experiments in our previous related work on cell death in the biofilm (Berne et al. (2010) Mol Microbiol) were done in M2G. M2G is also the reference medium we used in our earlier works on holdfast regulation and biofilm formation (Berne et al. (2018) Mol Microbiol; Berne et al. (2019) J Bact).

Reviewer #1 (Recommendations for the authors):

Suggested experiments to further support the key conclusion that the ParE toxin is responsible for cell death:

– Make a catalytically inactive toxin variant to see whether cell death requires the toxin activity.

– Measure ParE activity directly e.g. assess whether cells have altered sensitivity to gyrase inhibiting drugs, image cells at higher resolution (possibly with DAPI staining) to determine whether changes to cell morphology is consistent with gyrase inhibition.

We thank the reviewer for these suggestions. While the present work focusses on the link between O₂ deprivation, parDE₄ expression, and biofilm control, our next step will be to dissect the mechanism of action of the ParD₄-ParE₄ pair in this context. We did assess cell morphology as suggested and can already say that we do not observe noticeable changes in cell morphology under conditions where ParDE₄-correlated cell death is observed (O₂-limited conditions). We have added these observations as Figure 7—figure supplement 2, and in the Discussion section (Lines 544-546: "Under conditions where ParDE₄-correlated cell death is observed (O₂-limited conditions), cell morphology does not change significantly (Figure 8 and Figure 7—figure supplement 2)").

Suggested experiments to address whether transcriptional repression of ParDE4 leads to release of the toxin and thus cell death:

– Measure protein levels of toxin and antitoxin by western blot in O2-limited cells to see if there is a change in ratio of toxin and antitoxin

– Shut off parDE transcription either by placing the locus under control of a repressible promoter or using CRISPRi, then measuring growth (or evidence of ParE activity) after repression. In both cases, growth inhibition consistent with ParE activity should now be O2-independent.

Here again, we think these are very interesting suggestions and will be considered in our future experiments, as the presented work here does not focus on the mode of action of the ParDE₄ TAS, but rather its relationship with biofilm formation and cell dispersal under deleterious conditions (aka O₂ deprivation).

In addition, we previously tried to obtain antibodies for both ParD₄ and ParE₄ without success. First of all, we had antibodies raised against the entire ParD₄ purified protein (in rabbit) produced and purified, but, despite our numerous efforts, we failed to detect a band in Western blots using them, even when using the overexpressing strain carrying the plasmid pMT686-parD₄ that clearly induces the expression of ParD₄ in our hands here and in previous work published in the Crosson laboratory (Fiebig et al. (2010) Mol. Microbiol.). Second, the company we used for antibody production failed to overexpress ParE₄, despite multiple tries over several months. We are currently exploring other routes for antibody production.

We also tried several strategies to overexpress ParE₄ and failed to do so in both E. coli and Caulobacter. This protein seems highly toxic, even for E. coli, and none of our efforts have been successful. We are currently designing new strategies and this will be part of our future exploration of ParDE₄ mode of action. This aspect is now included in the Discussion section (Lines 553-557: "So far, we have failed to express the parE₄ gene both in E. coli and C. crescentus, suggesting that its product is highly toxic. Our future efforts will aim at engineering an attenuated version of parE₄. Such a tool will help to identify its targets and elucidate the pathway that leads to cell killing by this TAS.").

Reviewer #2 (Recommendations for the authors):

Given the timing of biofilm formation, in the experiments involving growth in a flow cell, it seems the authors have used defined medium in which wild-type cells have a low propensity for holdfast synthesis and attachment. Though not essential for this publication, I wonder if parDE plays a role in PCD and dispersal when cells are grown in richer, more complex PYE media where attachment is more robust.

We address our choice of using defined M2G medium in the reply to essential revisions above. Nevertheless, we tested biofilm formation of the different single ParDE and RelBE mutants used in this study in PYE and obtained similar results. The ∆parDE₄ mutant is the only tested mutant to behave differently than WT in PYE, and, under moderate aeration, we can measure an about 15% increase of biofilm (compared to about 30% in M2G). This attenuated response is not that surprising, as previous work by us and other laboratories report a higher dynamic range of holdfast related phenotypes in defined minimal M2 media (M2G or M2X, see Fiebig et al. 2014, Berne et al. 2018, Berne and Brun 2019). We are reluctant to add these data to the present manuscript, as all the presented experiments were done in M2G and we think it could add confusion.

Author response image 1

Download asset Open asset

I have several questions/comments about the measurements and experimental conditions for the static growth experiments:

The authors claim to quantify "cell death and eDNA release in these biofilms" (line 148) but the measurements of live/dead cells and eDNA are from the planktonic phase not the surface attached cells. I suspect the phenomena that parDE4 contributes to cell death and eDNA release under limited oxygen conditions is not limited to surface attached biofilm conditions, thus I don't question the data, but rather how it is discussed. I recommend the authors focus on "culture conditions" rather than "biofilm condition" in the early part of the paper where they establish that parDE4 contributes to cell death under oxygen limitation. Certainly the culture conditions affect biofilm dynamics, but I suspect these phenotypes do not require the cells to be in a biofilm. If the authors wanted to formally examine this idea, and specifically test if the cells needed to be in a biofilm to 'feel' the effects of parDE4 and low oxygen, they could repeat the static culture experiments in a holdfast null strain. I predict that cell death and eDNA release would be comparable to holdfast competent cells, without the potentially confounding effects of cell attachment.

We agree with the reviewer that section was misleading and we and modified the text to be clearer regarding our conclusions. We first changed to "We tested the ability of the TAS mutants to form biofilms after 48 hours, and quantified cell death and eDNA release under these growth conditions"(line 179, instead of "…cell death and eDNA release in these biofilms"), and then rewrote the conclusion sentence to " These results suggest that ParDE₄ plays a role in cell death and eDNA release under our experimental conditions, and that the observed changes in eDNA concentration yield differences in biofilm regulation. " (lines 187-189, instead of "… a role in biofilm regulation and eDNA release under our experimental conditions"). We also modified the text in the end of section 2, to " Furthermore, these results indicate that ParDE₄ is involved in stimulating cell death and eDNA release, yielding a change in biofilm formation." (lines 220-226, instead of "… cell death and eDNA release in biofilms").

As suggested by the reviewer, we added a new supplementary figure (Figure 1—figure supplement 2) providing eDNA release and percentage of dead cells in the planktonic phase in a holdfast parDE₄ double mutant (∆parDE₄ ∆hfsDAB). As expected, this double mutant exhibits similar phenotypes compared to the parDE₄ single mutant (lower cell death and subsequent eDNA release compared to WT), showing that, indeed, cell death and eDNA release respond to oxygen availability and do not require cells to be able to adhere and form a biofilm. We added these observations in the result section "To test if this phenotype was specific for C. crescentus cells that are able to form biofilms, we knocked out the hfsDAB holdfast synthesis cluster in the ∆parDE₄ background to generate a strain unable to produce holdfast, and therefore unable to adhere to surfaces. The double mutant ∆parDE₄ ∆hfsDAB phenocopied the ∆parDE₄ strain, with lower eDNA and lower proportion of dead cells (Figure 1—figure supplement 2). These results indicate that the function of ParDE₄ does not require cells to be adhered to a surface and suggest that it might be responding to the differences in medium aeration as described in a later section." (lines 189-204).

Figure S1A: Please be more explicit about how the growth measurement were made. Did you measure the same tube over time? Were the cultures mixed prior to measuring OD? If so, how did you maintain the aeration condition? Or did you set up parallel tubes and make terminal measurements at each time point? Growth medium?

We now provide all this information in the new "Growth curve measurements." subsection of "Material and method": " Growth curve measurements. Mid-log C. crescentus cultures grown in M2G medium under standard shaking conditions (3 ml cultures in 15 ml tubes, agitation at 300 rpm) were diluted to OD₆₀₀ = 0.15 in the same culture medium and incubated at 30°C under moderate and maximal aeration conditions (see "Bacterial strains and growth conditions" subsection above). A volume of two µl was taken at different time points over the course of 48 h from each culture, and A₆₀₀ was measured using a Nanodrop 2000c spectrophotometer (Thermo Scientific). Readings were corrected by a factor 10, to adjust for a 1 cm pathlength (and as determined by a standard curve using a WT culture (2-fold dilutions in M2G) read once using the 2 µl drop option and a second time using the 1 ml (in a 1 cm plastic cuvette) option of the Nanodrop spectrophotometer)." (lines 691-708).

Line 36: use of an acronym in abstract without definition

Corrected.

Line 79-80: perhaps emphasize "new-born swarmer" and consider "dissociate" instead of "disperse".

Corrected.

Line 150: 'viability' was not measured. Please rephrase to reflect the experimental measurement.

Corrected to "The percentage of dead cells was lower". (line 181)

Please add scale bars to all microscopy images (Figure 2 and 4).

Done.

line 189: Similar to other comments above, the authors should rephrase point 1 as they don't measure cell death "in the biofilm". Perhaps instead: "ParD4 has a protective effect against cell lysis that enhances biofilm formation".

Corrected.

The red signal corresponding to dead cells is very difficult to see in Figure 2. From the images as presented, it doesn't seem that 40% of WT cells are dead. Consider showing the channels separately and merged. Also given the difficulty if seeing sparse signal on a black background, consider scaling the images differently (for example from 0=white to max=color, or present the single channels as inverted BW images).

We now added a supplementary figure (Figure 2—figure supplement 1), to show the separated green and red channels, as well as inverted signals converted in BW, for a better visualization.

Line 181: activation of TAS is not just about "production" of the proteins. More generally activation comes when there is an imbalance in steady state levels, and reduced stability of antitoxins compared to toxins is an important contributor to these imbalances. The discussion should be enhanced.

Corrected to "In TAS, cell death usually occurs when there is an imbalance in steady state levels of toxins and antitoxins produced in the cell". (lines 231-232)

The authors should clarify metrics used to quantify microscopy images. It seems that the authors use thresholded areas in a binary fashion that ignores information from signal intensity. But it in these multidimensional structures, areas with thicker cell masses should have higher signal than areas with monolayers of cells. By simply thresholding the images and quantifying area, the authors loose a dimension of the data. Summing intensity in areas where signals overlap seems more informative than simply treating the pixels as binary.

We now provide all the relevant information in the new "Epifluorescence microscopy and image analysis" subsection in the Material and Methods: " … Image analysis was performed using ImageJ functions and plug-ins (Schneider et al., 2012). For Live / Dead staining quantification, red and green images were analyzed separately. 16-bit images were thresholded using the B/W default setting and the total area of fluorescent signal was automatically determined using the Analyze Particle function, as the fraction area. For fluorescence intensity measurements (FDG labelling, GFP, mCherry, or DsRed signals), regions of interest (ROI) corresponding to single cells or clusters of cells (for biofilm images) were first determined using the phase images using the Analyze Particle function. ROI were then added to the ROI manager and transferred to fluorescence images (red and green separately). Finally, integrated fluorescence intensity was measured in each ROI. Background fluorescence was calculated as the average of 5 different ROI devoid of cells for each image and subtracted to the mean fluorescence intensity of each particle". (lines 830-846).

We used thresholded images when quantifying Live/Dead staining images: for those experiments we needed an overall response (what is the area representing the red or green signal over the total field of view?), and each channel (red or green) was treated independently. For experiments where we measured fluorescence intensity, we indeed measured the integrated fluorescence intensity signal. We added this information in the method section ("Epifluorescence microscopy and image analysis").

Figure 6D, would be better presented more parallel to 6C such that the Y-axis reflected "relative biofilm formation (∆parDE4 / WT)" rather than "biofilm formation (% of WT)". This would also be more consistent with the language in line 280.

Corrected. Figure 6D has been modified to represent "ratio biofilm formation (∆parDE₄ / WT).

It is not clear why some experiments assess ccoN expression via a promoter reporter and others via transcript abundance by RT-PCR. Nevertheless, the RT-PCR experiments should be labeled as ccoN transcript levels rather than PccoN expression (line 273 and Figure 6B).

In Figure 5, we assayed both ccoN and parDE₄ promoter activities using promoter-lacZ fusions to assay both promoter using the same method. As mentioned above in the "Essential revisions" section, our several attempts to measure parDE₄ expression by qPCR were unsuccessful despite our best efforts. The lacZ fusions provided a good option, especially as we are using them in Figure 8 to measure PparDE₄ expression in biofilms by microscopy. In addition, we now include the PccoN-lacZ fusion to measure ccoN expression in biofilms (in Figure 8).

In Figure 6, we wanted to assay ccoN and parDE₄ expressions in the same cultures. parDE₄ expression was assayed using lacZ fusion (our only current option, see above), and, as we couldn't use at the same time the PccoN-lacZ construct to quantify ccoN expression in the same tube, we used qPCR instead. We realize that this information was not clearly provided in the previous version of the manuscript and we have corrected it here (in the result and method sections, as well as in Figure 6 legend).

When ccoN expression was assayed using qPCR, we now corrected to "ccoN transcript levels" (Figure 6 and in the text).

Also I suspect that ccoN transcript is essentially undetectable under maximal aeration conditions leading to a denominator of practically 0 in the "fold-change". While I have no question about the veracity of the data, I prefer non-normalized (non-ratioed) data whenever possible, especially when one condition is essentially "off". Just my two cents.

We agree that representing of qPCR data can indeed be tricky for the reasons you explained. We chose to represent the data as classical qPCR ratios using a reference condition (Maximal aeration conditions). This choice was driven by the fact that we wanted to represent these data along the PparDE₄-lacZ data, and we wanted to have a value for the "maximal aeration" condition (reference condition, set to 1 as base ratio). We now added this information in the figure legend ("qPCR (calculated as described in the Material and Method section) were normalized to the on in "Maximal aeration" conditions set to 1").

This reference condition, while consisting in low transcription of ccoN, is still significant. The Ct of ccoN transcript is approximately 32 and reproducible for the reference condition and reaches 18-21 in the other tested conditions. The reason why we feel that the representation as ratios is valid is because our gene of interest (ccoN) and housekeeping gene (rpoD) are amplified with primers with rather important differences regarding their respective PCR efficiencies (≈80% and ≈96% respectively). As a result, the use of the Pfafl method and ratio representation (as mentioned in the method section) was appropriate for data analysis.

Line306-7: wording is somewhat confusing.

The sentence was deleted (it was redundant with the next one).

Figure 7 and S3: I have several suggestions for data presentation. Make the data points smaller in the scatter plots to better visualize the density of points near the axis. Consider plotting on a log scale rather than a linear scale to better spread the points with lower intensity and highlight the bulk of the cells rather than the handful of outliers which are what is primarily obvious on the linear scaling.

We modified these figures. The data points are now smaller for a better readability. While still being linear scale, we are now including, in Figure 7—figure supplement 1, representations using one-segment axes (like in the original submission), and two-segment axes for a better visualization of lower intensity datapoints. Anti-correlation curves are usually reported using linear scale graphs and we decided to keep linear scales axes. We felt that the anti-correlation of our data was better reflected using linear scale data compared to log scale ones, as illustrated in author response image 2.

Author response image 2

Download asset Open asset

Please indicate how the threshold was determined to assess if each promoter was on or off for figure 7A. How many cells have "both"? Is "non-labeled" the best description of cells in which neither promoter was sufficiently active to be called "on"? Even "non-fluorescent" seems better than "non-labeled" as label implies staining.

The images were not thresholded prior to quantification in this experiment. We now provide all the information in the method section ("Epifluorescence microscopy and image analysis" subsection), as explained above. A cell was considered expressing GFP or mCherry if the fluorescence signal was at least 1.2 times the background (information added in Figure 7 figure legend). If a cell expressed both GFP and mCherry signals, it was included twice in the quantification (once for each channel). Interestingly, a very small percentage cells carried both green and red fluorescence, as presented in Figure 6B and in the result section.

We agree with the reviewer that "non-labeled" was not the most appropriate word to use, and we changed the legend from "non-labelled" to "non-fluorescent" cells, as suggested.

Line 505: the authors report using 10 ug/ml of chloramphenicol to select for plasmids in Caulobacter. This seems extremely high and quite inhibitory even in the presence of a Cm resistance gene. Typically for Caulobacter 1-2 ug/ml chlor is sufficient to maintain a plasmid. Please check if this number is correct.

This was indeed a mistake and the correct Cm concentration is 1 µg/ml. It is now correct in the text.

Author details

1. Cecile Berne

   Département de microbiologie, infectiologie et immunologie, Université de Montréal, Montréal, Canada

   Contribution

   Conceptualization, Resources, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing - original draft, Writing – review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-3731-9317

2. Sébastien Zappa

   Département de microbiologie, infectiologie et immunologie, Université de Montréal, Montréal, Canada

   Contribution

   Formal analysis, Validation, Investigation, Methodology, Writing – review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-3190-9199

3. Yves V Brun

   Département de microbiologie, infectiologie et immunologie, Université de Montréal, Montréal, Canada

   Contribution

   Conceptualization, Supervision, Funding acquisition, Methodology, Project administration, Writing – review and editing

   For correspondence

   yves.brun@umontreal.ca

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-9289-1909

• 1,507

  Page views

• 286

  Downloads

• 1

  Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.