Metabolic exchange and energetic coupling between nutritionally stressed bacterial species, the possible role of QS molecules

To clarify the principles controlling inter-species interactions, we previously developed a co-culture model with two anaerobic bacteria, Clostridium acetobutylicum and Desulfovibrio vulgaris Hildenborough, in which nutritional stress for D. vulgaris induced tight cell-cell inter-species interaction. Here we show that exchange of metabolites produced by C. acetobutylicum allows D. vulgaris to duplicate its DNA, and to be energetically viable even without its substrates. Physical interaction between C. acetobutylicum and D. vulgaris (or Escherichia coli and D. vulgaris) is linked to the quorum-sensing molecule AI-2, produced by C. acetobutylicum and E. coli. With nutrients D. vulgaris produces a small molecule that inhibits in vitro the AI-2 activity, and could act as an antagonist in vivo. Sensing of AI-2 by D. vulgaris could induce formation of an intercellular structure that allows directly or indirectly metabolic exchange and energetic coupling between the two bacteria.


Introduction
Microbial communities are ubiquitous and exert a large influence in geochemical cycles and health 1-4 . In natural environments, stress factors such as nutrient deficiencies and the presence of toxic compounds can induce interactions between microorganisms from the same or different species and the establishment of communities which can occupy ecological niches otherwise inaccessible to the isolated species 5,6 . Interactions between microorganisms can affect the behavior of the community either positively or negatively 7 .
For studying ecological communities, it is crucial to understand how the different members communicate with each other and how this communication is regulated. Interactions may occur either by release of molecules into the environment 8 , or by direct contact between the microorganisms through structures like nanowires 9 or nanotubes 10 . Dubey and Ben-Yehuda 10 were the first to demonstrate a contact-dependent exchange of cytoplasmic molecules via nanotubes in Bacillus subtilis which contributes to proper colony formation 11,12 . The evolution of how metabolites came to be transferred between bacteria, and its functioning today, were both well described in a recent review 13 .
The type and extent of nutritional interactions between microbes partly determine the metabolism of an entire community in a given environment 14 . Very little is known about the molecular basis of interactions between species, as this is difficult to investigate, especially in Nature, on account of community complexity. The use of a synthetic microbial ecosystem has considerable interest because the reduced complexity means that the investigation is more manageable, allowing not only identification of the specific community response, but also description of the different events at the molecular and cellular level 15 . of great importance in the proper development of the community and in its stability in the long term 24 .
Here we partially answer these questions; in particular, we examine if the nutritional stress, which appears to be necessary, is also sufficient. We have investigated the possible role that quorum-sensing molecules could play in attaching the two bacterial cells involved in the consortia previously studied: D. vulgaris and C. acetobutylicum or D. vulgaris and E. coli, and we have examined how satisfactory the energetic state of D. vulgaris is in the co-culture when it is deprived of sulfate, and why the presence of nutrients prevents interaction between the bacteria.

Tight bacterial interaction in the co-culture allows D. vulgaris to be metabolically active and to grow by using carbon metabolites produced by C. acetobutylicum
Our previous results demonstrated that conditions of nutritional stress of D. vulgaris To evaluate the physiological impact that C. acetobutylicum has on D. vulgaris in coculture, we labelled D. vulgaris cells with Redox Sensor Green (RSG), a small molecule that can easily pass through the membranes of Gram-negative and Gram-positive bacteria, used as a respiration sensor to identify metabolically active cells. It has been successfully tested on numerous bacteria, as an indicator of active respiration in pure or co-cultures 9,25 . If D.
vulgaris became metabolically active due to its physical interaction with C. acetobutylicum, then a RSG fluorescence should be detected as for D. vulgaris cultivated in Starkey medium (lactate/sulfate respiration) (Fig. 1a). As expected, in respiration conditions i.e. in the presence of lactate and sulfate, substrates of D. vulgaris, D. vulgaris in pure culture grows well and all the cells show intense RSG fluorescence, in contrast to the culture in GY medium, where there is no growth and very few cells fluoresce ( Fig. 1a and b). As a control, we added CCCP, which by dissipating the proton gradient and abolishing the ATP synthesis, tightly impacts RSG fluorescence ( Supplementary Fig.1). When D. vulgaris is labelled with RSG, as above, and then cultivated for 20h with C. acetobutylicum in GY medium, despite the lack of sulfate it shows a significant RSG fluorescence indicating that the cells have enough reducing power to reduce redox green and are metabolically active (Fig. 1c-e). Furthermore, C. acetobutylicum, which was not labelled by RSG at the beginning, also fluoresce intensely, indicating that RSG has been transferred from D. vulgaris.
To detect carbon exchange between the two bacteria, we used Stable Isotope Probing, growing C. acetobutylicum, either alone or in co-culture with unlabelled D. vulgaris, on 13 Cglucose medium. Cells were collected at the end of the exponential phase and we separated 13 C-DNA (heavier) from 12 C-DNA (lighter) by density-gradient centrifugation (Supplementary Figure. 2). Analysis of the different fractions using specific gene markers for each bacterium shows that in the co-culture DNA from D. vulgaris is "heavy" ( 13 C-labelled), indicating that metabolites derived from 13 C-glucose were transferred between the two bacteria and used by D. vulgaris. Quantitative PCR emphasized the presence of D. vulgaris 13 C-labelled-DNA with C. acetobutylicum 13 C-labelled-DNA in the same fraction (Fig. 2). A small amount of 12 C-unlabelled-DNA (from D. vulgaris and C. acetobutylicum) can be detected in this high-density fraction when an unlabelled co-culture is used, but negligible in relation to the total DNA. As D. vulgaris cannot grow on glucose or other hexoses, the 13 Clabelled-DNA from D. vulgaris must have been formed using metabolites produced by C. acetobutylicum.

C. acetobutylicum produces AI-2
As nutritional restrictions of D. vulgaris appear indispensable for inducing physical interactions between C. acetobutylicum and D. vulgaris, we investigated whether in addition to being necessary they were also sufficient, or if another element, such as quorum-sensing (QS) molecules, often associated with bacterial communication was required. We first looked at to homoserine lactone molecules without any success. As the co-culture is composed by an association of Gram-positive and Gram-negative bacteria, AI-2, known to be involved in interspecies communication, appears as a good candidate 26 , and is widely accepted as the universal cell-to-cell signal in prokaryotic microorganisms 27,28 . Furthermore, although its production has not been described in C. acetobutylicum, it has been in E. coli DH10B, which can replace C. acetobutylicum in the co-culture 18 . Clostridium species are known to develop QS systems based on peptides, but QS remains relatively unknown in sulfate-reducing bacteria (SRB) although inferences on the presence of putative QS systems in them can be made 29 . However, no AI-2 signalling/sensing had been described in C. acetobutylicum or D.
vulgaris. To determine whether C. acetobutylicum could generate AI-2-like activity, a cellfree supernatant from C. acetobutylicum was tested for its ability to induce luminescence in Vibrio harveyi BB170 AI-2 reporter strain, a classical assay for AI-2 molecule detection. This cell-free supernatant stimulated luminescence in a similar manner to cell-free supernatant from E. coli DH10B (AI-2 producer) (Fig. 3a).
Furthermore, AI-2 activity was also detected in the co-culture C. acetobutylicum & D.
vulgaris. The last step of AI-2 biosynthetic pathway is catalysed by the luxS gene product 30 ,which is present in C. acetobutylicum genome (CA_C2942) annotated as Sribosylhomocysteinase and could encode for the LuxS protein. Genetic engineering on genus Clostridium remains limited and requires the utilization of specific genetic tools. To avoid this, and to test if this gene is involved in AI-2 production, the putative luxS gene from C.
coli DH5α (luxS) expressing the luxS gene of C. acetobutylicum has AI-2 activity (Fig. 3a). In contrast, D. vulgaris does not have a homolog of the luxS gene and the cell-free culture supernatant collected from D. vulgaris culture, grown in Starkey medium, does not have AI-2 activity, as with the cell-free culture supernatant of E. coli DH5α (Fig. 3a). As D. vulgaris does not produce AI-2, we propose that the AI-2 molecules present in the co-culture are produced by C. acetobutylicum. To verify whether C. acetobutylicum AI-2 production follows the growth, cell-free culture supernatant from C. acetobutylicum or from C. acetobutylicum & D. vulgaris co-culture taken at different times were used in the V. harveyi bioluminescence assay, as described in Materials and Methods. As shown in Fig. 3b and 3c, C. acetobutylicum can synthesize functional AI-2 molecule and its synthesis follows the growth in single culture as well as in co-culture, indicating that its production is independent of the presence of D.

vulgaris.
Cytoplasmic exchange of molecules between bacteria in the co-culture as well as metabolic activity of D. vulgaris depend on the presence of AI-2.
As C. acetobutylicum and E. coli DH10B, used in the previous studies 18 , both produce AI-2, this raised the question of the situation if AI-2 was not present, that is if E. coli DH5α, were used. So E. coli DH5α or E. coli DH10B harbouring a pRD3 plasmid containing the gene mCherry, was mixed with D. vulgaris cells lacking the mCherry gene but labelled with calcein, and the co-culture was analyzed by microscopy. When D. vulgaris cells were cultivated on GY medium with E. coli DH10B, more than 90% of D. vulgaris cells acquired a mCherry fluorescence signal after 20 h of culture ( Fig. 4a panel 2, Fig. 4b and c and Supplementary Fig. 3 panel 1).
In contrast, no mCherry fluorescence was observed in D. vulgaris cells when they were cultivated with E. coli DH5α (Fig. 4 panel 3, Fig. 4b and c and Supplementary Fig. 2 panel 3).
Taken together, these results imply that AI-2 is essential for cell-to-cell communication and exchange of cytoplasmic molecules in co-culture, but not sufficient, as nutritional stress is also required. These results may explain why in the work reported by Pande et al 20 nanotubes, used to transfer amino acids, were observed between E. coli auxotrophe mutants, and between E. coli and Acinetobacter baylyi mutants, as in both cases there is the possibility of AI-2 produced by E. coli. In contrast, no nanotubes were observed between mutants of A. baylyi in which luxS is absent, according to our genome bioinformatic analysis. So when AI-2 is not produced there may be no physical interaction even if there is a nutritional stress.
In view of the necessity of AI-2 to allow growth of D. vulgaris in the co-culture, we tested its effect on the energetic state of the cells with RSG, as in Fig. 1. Lack of AI-2 should prevent RSG fluorescence in D. vulgaris by preventing physical interaction between E. coli and D. vulgaris in GY medium. Cells were incubated with RSG as described above and mixed with E. coli DH5α or with E. coli DH10B harboring the gene of mCherry and the co-culture was analyzed by microscopy after 20h incubation at 37 °C. D. vulgaris cells displayed a significant RSG fluorescence when they are mixed with E. coli DH10B (Fig. 5 panel 1). In contrast, no RSG fluorescence was observed in D. vulgaris cells in the co-culture with E. coli DH5α (Fig. 5 panel 2). Moreover E. coli DH5 does not show RSG fluorescence as E. coli DH10B and C. acetobutylicum (Fig. 1), which supports the absence of cytoplasmic exchange.
In contrast, D. vulgaris cells co-cultivated with E. coli DH5α complemented with luxS gene displayed RSG fluorescence similar to that observed with E. coli DH10B (Fig. 5 panels 1 and   3). These results demonstrate that the AI-2 molecule is important for physical interaction between D. vulgaris and E. coli and thus in metabolic activation of D. vulgaris. All these results suggest that D. vulgaris can detect AI-2.

D. vulgaris produces an antagonist of AI-2 in the presence of sulfate and under respiratory conditions.
The effect of AI-2 in the co-culture suggests that D. vulgaris can detect it. However, lactate and sulfate in the co-culture medium prevent physical contact between the two bacteria and transfer of cytoplasmic molecules, despite the fact that C. acetobutylicum and E. coli can produce AI-2. This suggests a regulatory mechanism linked to the presence of lactate and sulfate in the culture medium and/or to the sulfate respiration metabolism of D. vulgaris.
Various hypotheses may explain this: (i) in the presence of lactate and sulfate, C.
acetobutylicum does not produce AI-2; (ii) AI-2 could be used as carbon source by D.
vulgaris; (iii) D. vulgaris in the presence of sulfate produces an antagonist of AI-2.
The addition of lactate and sulfate to pure culture of C. acetobutylicum does not modify the production of AI-2. (Fig. 6a). In contrast, in co-culture with D. vulgaris, the AI-2 activity greatly decreased, even at 5 mM lactate and sulfate, and was not detected by growing the coculture in 10 mM or higher concentrations (Fig. 6a), suggesting that in sulfate respiratory conditions D. vulgaris could produce one or more metabolites that inhibit the activity of AI-2.
An important point is that in these conditions butyrate was produced, indicating that C.
acetobutylicum is metabolically active and that the lack of AI-2 is not due to a metabolic inactivity of C. acetobutylicum.
To test the presence of molecules that could interfere with AI-2 activity in sulfate respiratory conditions, we followed the AI-2 activity present in the supernatants of E. coli or  (Fig. 6d). The sulfate respiration process is probably therefore associated with the production of an antagonist or antagonists that inhibit the AI-2 activity. This production requires to the presence of sulfate, and is independent of the presence of C. acetobutylicum or E. coli.
To identify the compounds that interfere with the activity of AI-2, cell-free supernatants of D. vulgaris, grown in Starkey medium for 30h, were analyzed by HPLC. Different peaks (P1-P5) were recorded and their ability to inhibit AI-2 activity was determined on the supernatant of E. coli DH10B or on the in vitro synthesized AI-2 ( Fig. 7a and b). Only the peak (P5) containing a 186-Da molecule ( Fig. 7a et 7b, indicated by black arrow) inhibits the activity of AI-2 significantly in a dose-dependent manner (Fig. 7c). Note that the inhibitor has a molecular mass equivalent to that of AI-2, suggesting a similar type of molecule that could act as a competitive inhibitor. Different strategies for obtaining the structure did not succeed, probably because as with AI-2, there is an equilibrium between various forms 31 .

Discussion
The established bacterial community constituted of C. acetobutylicum and D. vulgaris, or acetobutylicum may be acting as final electron acceptor through a mechanism not yet elucidated.
How these kinds of interactions are initiated and controlled is at present poorly understood.
Ben-Yehuda's group showed that YmdB is involved in the late adaptive responses of B.
subtilis in the early stage of nanotube development 12 . In various types of cells, it is the cell undergoing stress that develops nanotube formation, suggesting that this might be directly induced by stress and constitutes a defense mechanism 32 as apparently also in this consortium.
Surprisingly, the role and the consequence of the QS molecules, well described in pure culture, are poorly investigated and understood in bacterial consortia, closer to those found in Nature. Recently one study investigates this question using mathematical model to demonstrate/ propose how QS control the population trajectories in synthetic consortium 33 .
However, this has attracted the attention of researchers studying mixed cultures in bioreactors for treating waste water. Although the real mechanism involved in QS regulation, of complex microbial consortia, remained to be elucidated, studies of this type have shed light on it 34,35 .
The QS molecule AI-2 is crucial for metabolic interaction between the two bacteria of the co-culture, as its absence prevents the metabolic exchange. Thus C. acetobutylicum produces a molecule with AI-2 activity, which had not been described before; furthermore, the C. acetobutylicum gene luxS can restore the AI-2 production of an E coli DH5. Our results explain why E. coli can connect to other bacterial cells to exchange cytoplasmic molecules as Some organisms only produce AI-2 whereas others only sense the signal 36,37 . D. vulgaris appears to be in this last category, as it lacks the luxS gene, but appears to sense AI-2. This suggests the presence of AI-2 receptors in D. vulgaris. However, no genes similar to lsrB (coding for LsrB, protein receptor of AI-2 in enterobacteria) or to luxP (gene coding for LuxP, the protein receptor of AI-2 in vibrionaceae) are present in the genome of D. vulgaris 16 .
However, some bacteria can respond to exogenous AI-2 signal 38,39 despite not having the genes luxS, lsrB or luxP. Furthermore, two proteins can bind AI-2 in Helicobacter pylori 40  Although the original QS concept was focused on the detection of cell density for the regulation of gene expression, studies in microbial ecology suggest a wider function. For example, the efficient-sensing concept 41 assumes that the ecologically relevant function of AI-2 sensing is to pre-assess the efficiency of producing extracellular effectors or "public goods".
Cooperative genes regulated by QS molecules can also be sensitive to nutrient conditions, suggesting that metabolic information is integrated into the decision to cooperate. AI-2 molecules are involved in the mechanism stimulating viable but no cultivable cell exits from dormancy, perhaps signalling to dormant cells when conditions are now favorable for growth 42,43 . This supports the idea that AI-2-dependent signalling reflects the metabolic state of the cell, and can function as a proxy for the production of effectors such as enzymes, or the formation of nanotubes. Integrating metabolic information with QS offers a possible mechanism to prevent cheating, as cells can only cooperate when they have the appropriate nutritional resources to do so, reducing the cost of cooperation to the individual cell 44 .
We demonstrate the quenching of the AI-2 activity by a QQ molecule produced by D.
vulgaris in the presence of lactate/sulfate. Quorum quenching (QQ) has been suggested to be achieved in three ways: (i) blocking synthesis of autoinducers; (ii) interfering with signal receptors; and (iii) degrading the autoinducers [45][46][47][48] . As we show competition between AI-2 present in C. acetobutylicum or E. coli supernatant, or even synthetic AI-2 and an AI-2 quencher, a small molecule presents in the D. vulgaris supernatant, we can exclude the first and the third hypotheses. Only a few AI-2 interfering mechanisms have been reported and most of them include synthetic molecules as quencher 49,50 . Roy et al 31 proposed that the antagonist (C1 alkyl analogues of AI-2) could compete with AI-2 for the LsrR transcriptional regulator in the lsr system 31,51 and the presence of the competitor is linked to a decrease in AI-2 production.
As AI-2 consists of a group of molecules in equilibrium, not a unique defined structure 31 , analogy with enzymes and alternative substrates suggests that different types of AI-2 molecules may interact with a receptor, with only some of them inducing a response. We cannot discard the possibility that the QQ molecule identified in D. vulgaris supernatant could bind to an AI-2 receptor in C. acetobutylicum and induce an effect at the level of gene transcription that could be translated into metabolic modification. Moreover, we also cannot discard the possibility that the QQ molecule identified represents a QS signal for D. vulgaris.
Our analysis provides new insights into metabolic prudence 47,52 and bacterial communication, and about how metabolic signals influence social behaviour but many details of its molecular implementation remain to be discovered. Which proteins detect the metabolic signals? How do they interact with QS regulation at the molecular level? However, one should also be cautious in using the word "signalling" because every change in a living organism affects every other, and thus acts as a signal of some kind 53,54 . In all of these studies it is important to keep in mind the ecological context, but the analysis of how the components of an ecological system influence one another has barely begun 55 . Anyway, we can see in these microbial communities established, thanks to QS molecules, the preliminary steps in the evolutionary pathway of multicellular organisms and eukaryotes.

Media and growth conditions
Strains were grown to steady state in Hungate tubes under anaerobic conditions, in LB medium for E. coli DH10B and DH5, in Starkey medium (containing lactate and sulfate) for D. vulgaris 56 and 2YTG medium for C. acetobutylicum 57 . The growth medium (Glucose-Yeast extract (GY) medium) used for studying the consortium was prepared with glucose (14 mM), 0.1% yeast extract, and supplemented with the similar inorganic nutrients used for the Starkey preparation (but with MgCl 2 instead of MgSO 4 ). GY medium was inoculated with either washed D. vulgaris or C. acetobutylicum or E. coli or with the combination of different strains to constitute an artificial consortium in a 1:1 ratio according to the absorbance at 600 nm. In some cases, the growth medium was supplemented with 5 or 10mM lactate or/and 5 or 10 mM sulfate. The experiments were carried out at least in triplicate.

Construction of E. coli DH5α (luxS) strain
The luxS ORF (corresponding to gene CA_C2942) was amplified using the genome of C.
acetobutylicum as a template and oligonucleotide primers Next, the pRD4 was transformed in E.coli DH5α to obtain E.coli DH5α (luxS) strain.

Labeling of D. vulgaris with calcein-acetoxymethyl-ester (AM)
The labelling of D. vulgaris cells was carried out as described by Benomar et al. 18

Labelling of E. coli DH5α and E. coli DH10 with mCherry
The labelling was carried out as described by Benomar et al 18 .

Exchange of cytoplasmic molecules between D. vulgaris and E. coli
To study the exchange of molecules between the two bacteria, D. The presence and relative amount of DNA from each bacterium were followed. The primers used for the qPCR (dsrA and endoG) are listed in Supplementary Table 1. The reaction was performed with GoTaq mix and the PCR was carried out in a Techne Prime Elite thermal cycler as follows: 2min at 98°C for the initial activation of enzymes, 21 cycles of 30s at 98°C, 30s at 58°C and 2 min at 72°C. Experiments were made in triplicate.

AI-2 synthesis
AI-2 was obtained in a four step sequence starting from the commercially available methyl The organic layer was dried (Na 2 SO 4 ), filtered and the filtrate concentrated under reduced pressure. The crude diol was used in next step as such. To the diol in methanol (3.5mL) and water (1.5mL) was added sodium periodate (0.146g, 0.69 mmol) and stirred at RT for 30min.
The reaction mixture was diluted with water and extracted with dichloromethane. The organic layer was dried (Na 2 SO 4 ), filtered and the filtrate was concentrated under reduced pressure.     vulgaris was grown in GY medium. The aliquots of different cultures were taken at 30h and filtered to remove cells. AI-2 activity in the cell culture supernatant was measured using the V. harveyi BB170 bioassay as described in Material and Methods. Data are represented as mean ± SD with n = 3, in comparison to extracellular AI-2 activity from D. vulgaris wild type. p-values calculated in Tukey HSD tests,*p < 0.05; **p < 0.01; ***p < 0.001 (a). Time course of extracellular AI-2 accumulation in pure culture of C. acetobutylicum or in coculture C. acetobutylicum & D. vulgaris. Exponentially growing C. acetobutylicum in 2YTG medium and D. vulgaris in Starkey medium under anaerobic conditions were washed two times with fresh GY medium. Next, C. acetobutylicum was inoculated alone or mixed with D.
vulgaris into GY medium at time zero and the aliquots were taken at indicated times. Cell growth was monitored by measuring the optical density at 600nm (b), and AI-2 activity in cell-free culture fluids was measured in a pure culture of C. acetobutylicum or in co-culture, using the V. harveyi bioluminescence assay (c). AI-2 activity is reported as relative light unit    was grown in Starkey medium during 30h at 37°C and the filtered supernatant was analyzed by RP-HPLC. The peak P5 which presents an AI-2 inhibitor activity was analyzed mass spectrometry (A, indicated by black arrow) (a). The AI-2 inhibitor activity of different peaks collected (P1 to P5) was determined on synthetic AI-2 ((S)-4,5-Dihydroxy-2,3-pentandione, DPD) (2,5µM) or on AI-2 produced by E. coli strain DHI0B grown on GY medium (b). To test the hypothesis that the P5 can compete with AI-2, V. harveyi reporter strain BB170 was grown in AB medium (100µl) supplemented with 500nM of synthetic AI-2 and 5µL of P5 and Bioluminescence was measured each hour over the course of 4h (black circle) or of 5h (green circle) (c).      Scale bar, 2μm in all panels.