Cooperative Adaptation to Establishment of a Synthetic Bacterial Mutualism

To understand how two organisms that have not previously been in contact can establish mutualism, it is first necessary to examine temporal changes in their phenotypes during the establishment of mutualism. Instead of tracing back the history of known, well-established, natural mutualisms, we experimentally simulated the development of mutualism using two genetically-engineered auxotrophic strains of Escherichia coli, which mimic two organisms that have never met before but later establish mutualism. In the development of this synthetic mutualism, one strain, approximately 10 hours after meeting the partner strain, started oversupplying a metabolite essential for the partner's growth, eventually leading to the successive growth of both strains. This cooperative phenotype adaptively appeared only after encountering the partner strain but before the growth of the strain itself. By transcriptome analysis, we found that the cooperative phenotype of the strain was not accompanied by the local activation of the biosynthesis and transport of the oversupplied metabolite but rather by the global activation of anabolic metabolism. This study demonstrates that an organism has the potential to adapt its phenotype after the first encounter with another organism to establish mutualism before its extinction. As diverse organisms inevitably encounter each other in nature, this potential would play an important role in the establishment of a nascent mutualism in nature.


Introduction
Mutualism is based on a mutually beneficial interaction between two organisms and is ubiquitous in nature [1,2,3,4,5,6]. Mutualisms observed in nature are thought to be the result of adaptation of each organism to the existence of the partner after their first encounter. The genetic origin and trajectory of this adaptation has been investigated via a phylogenetic approach [2,7,8]. However, tracing back established mutualisms to their origin is challenging as no intermittent states are defined with which to measure the adaptation in terms of phenotypic traits, population size and local environment [1]. To investigate the environmental conditions required to establish a nascent mutualism, one study reported a synthetically designed mutualism using two species of bacteria [9]. The findings of that study clearly demonstrated the importance of spatially structured environments for the establishment of mutualism, providing proof of principle of natural selection of cooperative behavior that has been proposed by the theoretical studies [10,11,12,13]. These types of experimental studies using microbial ecosystems to test the theories of cooperative systems have recently been reported [14,15,16,17,18]. Most of these studies focused not on the adaptation of the organisms but on the environmental conditions required for the persistence of cooperative behavior in natural selection.
Some studies have characterized the behavior of organisms in nascent mutualisms. Wintermute et al. synthetically designed mutualisms comprising certain pairs of auxotrophs of Escherichia coli and found significant metabolic synergy in 17% of 1035 such pairs tested [19], although it was unclear if any adaptation of the bacteria contributed. Shou et al. synthetically designed an obligate mutualism composed of two yeast auxotrophs [20], each of which was genetically engineered to overproduce the metabolite essential for the growth of the partner. Both of the auxotrophic strains grew to saturation without the need for external supplementation of their essential metabolites compensating for the auxotrophy. Moreover, they showed adaptation in as little as one hundred generations, where they became capable of growing from diluted cell densities or ceased growth due to weakening of the beneficial interaction. Hillesland et al. demonstrated that the growth rate of microorganisms in another synthetic mutualism increased after serial passage, even in the absence of spatially structured environment, while the extent of the adaptation was increased in a spatially structured environment [21]. These adaptations of microorganisms occurred after the establishment of nascent mutualisms, strengthening their interactions.
Can adaptation occur before the establishment of a nascent mutualism, leading to its establishment? Here we show that a strain of bacteria became more beneficial to another strain before their population started growing and establishing a nascent mutualism. Specifically, we synthetically designed an obligate mutualism comprising two auxotrophs of Escherichia coli. We show that one of the two auxotrophs, upon encountering the partner strain before their own population growth, adapted by oversupplying the metabolite essential for growth of the partner, which in turn permitted its own growth, leading to the successive growth of both strains. This study therefore shows the potential of organisms to adaptively respond to the first encounter with another organism, which could lead to the establishment of nascent mutualisms. As diverse organisms inevitably encounter each other in nature, this potential would play an important role in the establishment of nascent mutualisms in nature.

Results
To create a synthetic model of obligate mutualism, we constructed two different types of nutrient auxotrophs of E. coli by genetic recombination (Fig. 1A): an isoleucine (Ile) auxotroph, designated I -, labeled with a red-fluorescent protein (dsred.T3), and a leucine (Leu) auxotroph, L -, labeled with a green-fluorescent protein (gfpuv5) [22] (see Methods). We were able to distinguish these two strains by flow cytometry (FCM). In minimal medium without amino acid supplements, neither strain was able to grow in monoculture. However, in coculture, if the two strains supplied a sufficient amount of the essential amino acids required by the other strain, they would successively grow and thereby establish mutualism.
We measured the supply of amino acids from each strain in monoculture to test whether the quantities were sufficient for the successive growth of both strains in coculture. Figs. 1B and 1C show the supply of Leu from Icells in monoculture, with and without the addition of 10 mM Ile, respectively, and Figs. 1D and 1E show the supply of Ile from Lcells in monoculture, with and without the addition of 10 mM Leu, respectively. Before inoculation of these cultures, we washed each strain with minimal media not containing amino acids to exclude the carry-over of supplements from preculture (see Methods). Obviously, both strains did not grow without the addition of amino acids ( Fig. 1C and 1E). We measured the concentrations of Leu and Ile in the culture media using a bioassay (see Methods), and expressed these as the cell concentration of Land Icells which can be produced by the amount of amino acid supplied, respectively. In every case ( Fig. 1B-E), the final concentrations of an amino acid in the recipient cell were always less than the maximum concentrations of the donor cell. That is, the nutrient supply from the donor cells was insufficient to produce an equal amount of nutrients in the recipient cells, and was therefore insufficient to sustain the net growth of both strains [20]. These results implied that any adaptation to the mutualism, such as an increase in the nutrient supply, needs to occur in coculture for the successive growth of both strains.
Despite the insufficient amino acid supply in monoculture, both strains grew to saturation (around 10 8 to 10 9 cells/ml) in coculture (Fig. 1F). Initially, Icells grew (red N at ,10 h), followed by Lcells (green m at ,20 h). Qualitatively, the initial growth of Icells was consistent with the results of amino acid supplementation in monoculture as follows. In monoculture, Icells supplied Leu only after growth and the uptake of Ile ( Fig. 1B and 1C), while Lcells supplied Ile regardless of growth ( Fig. 1D and 1E). These results suggested that initially Lcells supplied Ile promoting the growth of Icells. In addition to the initial growth of Icells, the amount of Leu was detected at time 0 in coculture, as shown in Fig. 1F (blue +). As Icells supplied Leu only when they consumed Ile, these results indicated that Icells consumed Ile supplied from Lcells and then supplied Leu just after mixing but prior to sampling. However, quantitatively, the initial growth of Icells was inconsistent with the results from monoculture. Icells grew to greater than twice the concentration of Lcells (Fig. 1F, red N at ,10 h). That is, the Ile supply from Lcells was sufficiently high to produce a greater concentration of Icells than Lcells, which was different from the results of monoculture described above ( Fig. 1D  and 1E). The final concentration of Ile was also significantly higher than that of Lcells in coculture (Fig. 1F, orange 6). The inconsistency in the quantity of Ile supplied by Lcells in coculture and monoculture suggested the enhanced supply of Ile from Lcells on encountering Icells. It should be noted that Lcells did not show significant growth before nine hours in coculture when the Ile supply from Lcells already appeared to be enhanced (Fig. 1F), which indicates that enhancement of the Ile supply from Lcells did not require the population growth of strain Litself. Also, enhancement of the Leu supply from Icells was detected in coculture ( Fig. 1F), as discussed later.
We investigated the growth kinetics of the cocultures at various initial cell concentrations of strains Iand L - (Fig. 2). The cells entered stationary phase at around 20-30, 40-120, and 300-600 h when the initial cell concentration of strain Lwas 10 7 , 10 6 and 10 5 /ml, respectively. Cell growth was not observed when the initial cell concentration of strain Lwas 10 4 /ml. On the other hand, clear dependency on the initial cell concentration of strain Iwas not observed. The difference in the dependencies on the initial concentration of Iand Lcells was consistent with the differences in the features of nutrient supply found in the monocultures: only Lcells supplied Ile even in the absence of amino acids in monoculture as described above (Fig. 1E). These results suggested that Lcells initiated the first steps towards establishing mutualism in coculture.
The time courses of the cocultures also showed another feature. In some cases, the initial growth of Icells reached a concentration of up to approximately 10-fold greater than that of Lcells (Fig. 2, depicted by arrows). This suggested that the Ile supply from Lcells was sufficiently high to produce this concentration of Icells. As the Ile supply from Lcells was insufficient to produce an Icell concentration equal to that in monoculture, which was equivalent to the initial lag time in coculture, the sufficient supply of Ile from Lcells in coculture suggested that Lcells changed to a "high supplier" phenotype prior to growth. Indeed, a mathematical model assuming the change in Lcell phenotype could explain the time course of coculture (Fig. S1). It should also be noted that the lag period was dependent on the initial cell concentration of Lcells in the coculture, which suggested that the interactions between cells were required for the change in Lcell phenotype.
To directly observe the change in Lcells to a high supplier phenotype, we tested reconstituted cocultures (re-coculture) using Lcells prepared from mid-coculture (Fig. 3). To prepare Lcells separately from Icells from mid-coculture, we inoculated each strain into media separated by a membrane, which was permeable to amino acids but not to E. coli (membrane coculture) (see Methods). The time course of the membrane coculture was almost the same as that of coculture without membrane separation (Fig. 3A). For the re-coculture, we used both strains harvested from three different culture conditions: (i) at the log phase in monoculture with the addition of the required amino acids, which is the same as the initial state (0 h) of membrane coculture (Iini and Lini ), (ii) at 23 h of membrane coculture, when Icells had grown to a concentration approximately 10-fold higher than that of the Lcells ( Fig. 3A) (Ico and Lco ), and (iii) at 23 h of membrane monoculture in the absence of amino acids, when both strains were not growing (Imono and Lmono ). Before inoculation of the re-cocultures, we washed each strain with minimal media to exclude supplements carried over from the first membrane cultures. Fig. 3B-E show the time courses of the re-cocultures comprising Iini and Lini , Ico and Lini , Iini and Lco , and Ico and Lco cells, respectively. Only the re-coculture containing Lco cells showed initial growth of Icells without a lag phase ( Fig. 3D and 3E, arrows). These results indicated that Lco cells were high suppliers of Ile at time 0 in the re-coculture, in contrast to Lini cells. It is worth noting that Lini cells represent the initial state of Lco cells in the first membrane coculture, i.e., Lcells change to a high supplier phenotype in the first membrane coculture. Recoculture containing Lmono cells exhibited a lag phase before the initial growth of Icells ( Fig. 3F and 3G). These results indicated that Lmono cells were not high suppliers, like Lco cells, and the change to a high supplier phenotype was dependent on coculturing. A significant change in Icells was not detected ( Fig. 3B and 3C) in the re-coculture, although the oversupply of Leu from strain Iwas observed in Fig. 1F. These results experimentally confirmed that Lcells changed to a high supplier phenotype in coculture prior to their own growth.
These findings raised the question: how does gene expression change in the two strains during coculture? To investigate this, we carried out a comprehensive analysis of gene expression in these two strains. To harvest each strain separately from coculture, we again employed membrane coculture. Using a DNA microarray, we measured and compared the expression intensities of all 4345 genes of each strain cultured under three different conditions. These three conditions were: (i) at the log phase in monoculture, which is the same as the initial state (0 h) of membrane coculture, as described above (Iini and Lini ), (ii) at the stationary phase (45 h) of membrane coculture (Ist,co and Lst,co ), and (iii) at the stationary phase (45 h) of monoculture after growth in the presence of the required amino acids (Ist,mo and Lst,mo ). As E. coli is known to substantially change its gene expression depending on the growth phase [23], samples were taken at 45 h (not 23 h) to identify coculture-specific changes by comparing samples at the same phase (stationary phase). In the Istrain, the changes in gene expression from Iini to Ist,co strongly correlated with those from Iini to Ist,mo (Fig. 4A), that is, the dominant changes were dependent on the growth phase. This correlation was also observed in Lcells and the slope of linear regression was smaller than of Icells (Fig. 4B), which may have been because the time after entering stationary phase was shorter in Lst,co cells than in Ist,co cells (Fig. 3A). More importantly, in strain L -, the correlation coefficient was smaller than that in strain I - (Fig. 4B). These results indicated that the change in gene expression of Lcells in coculture was more coculture-specific than that of Icells.
The next question raised was: which categories of genes were involved in the coculture-specific changes in expression in Lcells shown in Fig. 4B? Initially, we focused on genes that showed significantly induced or repressed expression in coculture compared to monoculture, i.e., those in which the ratio of gene expression in Lst,co cells to that in Lst,mo cells was greater than three or lower than one-third. We statistically screened the upregulated and down-regulated gene categories to which the significantly induced or repressed genes belonged (Table 1). For the gene categories, we adopted the ''cellular processes'' category in the Gene Ontology (GO) database [24], and the categories for genes regulated by sigma factors in the database, RegulonDB [25]. In the "cellular processes" category of the GO database, 14 categories out of 124 were found to be up-regulated, and most of these up-regulated categories were related to anabolism, such as the biosynthesis of amino acids (tryptophan, proline, methionine, phenylalanine, leucine, cysteine, and chorismate, which is a precursor of tyrosine, phenylalanine and tryptophan), polyamines and proteins. In contrast, of the nine down-regulated categories, most were related to catabolism, such as the various energy cycles (glyoxylate and tricarboxylic acid cycles), fatty acid oxidation, and the catabolism of amino acids, aminobutyrate and carbohydrates. Some of these up-and down-regulated categories were also identified when comparing Lst,co and Lini cells (Table 1, indicated by arrows). Although Lcells oversupplied Ile in coculture, no significant increase was found in the expression of genes related to Ile biosynthesis or Ile transport in Lst,co cells compared with both Lst,mo and Lini cells (Fig. S2). It is worth noting that the results of liquid chromatography showed that the predominant supplement from Lcells required by Icells was Ile (Fig. S3). Among the genes regulated by sigma factors, we detected the down-regulation of genes regulated by Sigma 70, the housekeeping sigma factor [26], during the change from Lini to Lst,co cells (arrows, Table 1) and the change from Lini to Lst,mo cells. These results were consistent with the change in growth phase to stationary phase. Although the down-regulation of Sigma 70 genes occurred in both Lst,co and Lst,mo cells, the gene expression significantly differed between Lst,co and Lst,mo cells (Fig. 4B). Therefore both up-and down-regulation was found when comparing Lst,co and Lst,mo cells. Down-regulation of the glutamine biosynthesis gene category correlated with the downregulation by Sigma 54, the sigma factor controlling nitrogen usage [26,27]. As glutamine biosynthesis opposes glutamate biosynthesis leading to the biosynthesis of other amino acids, the down-regulation of glutamine biosynthesis is not inconsistent with the up-regulation of the anabolic categories. As above, we found that the coculture-specific changes in gene expression in Lcells were not related to the local activation of the biosynthesis and transport of isoleucine, but were related to the global activation of anabolic metabolism. Discussion In our synthetic model of obligate mutualism comprising two auxotrophs of E. coli, strains Iand L -, the increase in the Ile supply from Lcells occurred before the population growth of Lcells, and both strains grew successively thereafter in coculture. We found that the increase in the Ile supply from Lcells depended on coculture with Icells and was accompanied by coculture-specific changes in the gene expression of Lcells. This change in Lcells in coculture was not related to the local activation of the biosynthesis and transport of isoleucine, but was related to the global activation of anabolic metabolism.
What is the mechanism behind the phenotypic change in Lcells to become ''high suppliers'' of isoleucine? There are two possibilities: (i) a fraction of high suppliers preexisted in the initial population and their fraction in the Lpopulation increased in the coculture (natural selection), (ii) Lcells changed their phenotypes in response to the changes in the environment from monoculture to coculture (phenotypic plasticity). We can rule out neither possibility completely.
Let us first assume that (i) is true, and estimate an approximate range of the fraction of preexisting high suppliers in the initial population ( f H0 ). At first, the supply of Ile from high suppliers was about 10-fold higher than that of ''normal'' Lcells. Therefore, f H0 would be less than 10%, otherwise the supply of Ile from Lcells at the initial state or in the monoculture would be higher than the experimental results. Second, the shortest time until the initial growth of Icells was about 10 hours. By then the high suppliers had already been the majority in the Lpopulation and their concentration was approximately 10 7 /ml (Fig. 1F). For the concentration of the high suppliers to become 10 7 /ml in 10 hours from the initial concentration 10 7 ?f H0 /ml, f H0 . exp(-10g H ) must be satisfied, where g H is the growth rate of the high suppliers. Even if only high suppliers grew at a maximum growth rate of Lcells (g H = 0.4/h, Table S1), f H0 .2% was required. f H0 thus can be estimated as 2%,f H0 ,10%. Note that there is no reason why only high suppliers grew at a maximum rate in the environment, where the normal Lcells did not grow. In the mixed liquid culture, all of the Lcells were considered to acquire Leu from the media (not from Icells directly) in a homologous environment (actually, physical contacts were negligible; Fig. 3A). Moreover, f H0 must be kept in this range in monoculture because the time until the initial growth of Icells were reproducible even when we used another clone of Lcell for the preparation of the initial population.
We then discuss about the possibility (ii). E. coli is known to alter its phenotype in response to environmental changes, such as amino acid starvation, and this is known as a stringent response [28] and represents a kind of phenotypic plasticity. As Lcells were subject to Leu starvation at the initial in coculture, they would have changed their phenotype as the stringent response. This response might have been preserved even after their growth in which they had already been released from Leu starvation. Indeed, the up-regulation of amino acid synthesis, which is known to occur in the stringent response [28], was observed in Lst,co relative to Lst,mo (Table 1). However, although Lcells in monoculture without Leu (Lmono ) were subject to Leu starvation, they did not change to the high supplier phenotype (Fig. 3F and 3G). In our experiments, Lcells changed to a high supplier phenotype only in coculture, and the genes related to Ile biosynthesis and transport were not significantly induced in these cells (Fig. S2), in contrast, these genes are induced during the stringent response [28]. It is known that Ile uptake is increased and amino acid permeability is decreased during the stringent response [28], which seems to oppose the extracellular leakage of Ile. Our results might therefore indicate that the phenotypic change in Lcells was related not only to the known stringent responses, but also to other responses due to the interaction among strains via the media. As both the Iand Lstrains were constructed by a single gene deletion from the same original strain, DH1 (see Methods), the substances supplied by them via cell leakage would be expected to be almost the same. Thus, the interaction between these strains is unlikely due to the expression of a specific substance, as is the case in quorum sensing [29], but is more likely due to a global change of the composition of multiple substances [30,31]. This might be consistent with the observed global activation of expression of genes involved in anabolic metabolism (Table 1). It is worth noting that a similar phenomenon was observed in a synthetic mutualism comprising Icells and an uracil auxotroph (Fig. S4), therefore, the change in Lcell phenotype is not due to the similarities between the metabolism of Ile and Leu. Further studies are required to fully elucidate the mechanism behind the phenotypic change in strain L -.
It is unknown to what extent such an adaptation to a first encounter contributes to the establishment of a nascent mutualism in nature. However, this potential would provide insight into the positive factors required for the establishment of natural mutualisms. We do not believe that an adaptation, such as that described in this study, results in the establishment of nascent mutualism in every case, because we actually failed to establish mutualism with some combinations of auxotrophs (such as a glutamine auxotroph and an uracil auxotroph). Synthetic mutualism also failed to be established in other types of organisms without the introduction of metaboliteoverproducing mutations [9,20]. However, due to the great variety of organisms in nature [3,32], organisms inevitably encounter other kinds of organisms and have the opportunity to establish a nascent mutualism, where such an adaptation to this first encounter can facilitate this process. It is worth noting that such an adaptation to first encounter might be a kind of the phenotypic plasticity in response to a new environment [33,34,35,36,37,38,39,40], and the adaptation in response to a first encounter might have evolved because the organisms possessing this potential should survive in the bio-diversified nature. Field studies are required to fully investigate the contribution of such an adaptation to the establishment of nascent mutualism in nature.
The simplicity of the synthetic model of mutualism used in this study enabled us to identify unexpected and quantitative changes in the organisms. Experimental ecosystems not only provide empirical proof of theories but also highlight unexpected phenomena, such as the unknown potential of organisms which may lead to novel theories. For example, in another bacterial system, Fiegna et al. found that a single point mutation changed a cheater into a cooperator with a tolerance to exploitation by the cheater [41,42]. Without the simplicity of the system, it would have been impossible to detect such a phenomenon. For future studies, synthetically-constructed experimental ecosystems combining naturally non-interacting species [21,43,44,45,46] and reconstructing interactions using genetic modifications [16,19,20,47,48,49,50,51,52,53], would be invaluable for the detection of other unexpected phenomena.
The simplicity of our synthetic model of mutualism will enable further studies to experimentally resolve some of the remaining questions, such as the molecular mechanisms behind the observed adaptation in Lcells and the evolutionary pathway of this mutualism. Our findings may also contribute to the study of mutualism in other organisms, including higher organisms, and in field studies investigating natural ecosystems.

Construction of E. coli strains
The E. coli strain DH1DilvE::(dsred.T3-cat), designated I -, was constructed from the E. coli strain DH1 (National BioResource Project, National Institute of Genetics, Shizuoka, Japan), by replacing the chromosomal ilvE gene with a foreign DNA fragment, P tetA -dsred.T3-P cat -cat, comprised of a reporter gene (dsred.T3) and the chloramphenicol resistance gene (cat). The dsred.T3 gene and the cat gene were transcribed from their promoters in opposite directions.

Culture conditions
All cultures were grown at 37uC in well-mixed minimal media modified with M63 (pH 7.0, 62 mM K 2 HPO 4 , 39 mM KH 2 PO 4 , 15 mM ammonium sulfate, 1.8 mM FeSO 4 -7H 2 O, 15 mM thiamine hydrochloride, 0.2 mM MgSO 4 -7H 2 O and 22 mM glucose; mM63 [56]). Amino acids were added to the media when appropriate. Before culturing, we washed E. coli strains with the minimal media without amino acids to exclude the carry-over of supplements from preculture. For the membrane culture, we used cell culture inserts with a pore size of 0.45 mm at a density of 10 8 /cm 2 , and used six-well cell culture companion plates for the inserts (BD Falcon, Franklin Lakes, NJ, USA). The initial concentration of Iand Lcells are depicted at time 0 in the figures or described in the figure legends.

Measurement of cell concentrations
We measured the cell concentration relative to a known concentration of fluorescent beads (Fluoresbrite YG Microspheres, 3 mm; Polysciences Inc., Warrington, PA, USA) using a Cytomics TM FC500 Flow Cytometer (Beckman Coulter, Inc, CA, USA) by loading culture samples mixed with the beads. A 488 nm argon excitation laser was employed and band-pass filters of 515-535 and 610-630 nm were used to measure green and red fluorescence, respectively. Clusters of red and green cells, and the fluorescent beads, were clearly segregated (Fig. S5), and each cell concentration was calculated from these counts.

Measurement of amino acid concentrations using a bioassay
To measure the Ile concentration of a culture, the culture was passed through a 0.2 mm filter and the supernatant was supplemented with a one-sixth volume of the mM63 media and inoculated with Icells at 10 4 /ml. Then the Ile concentration was obtained by multiplying the saturation concentration of Icells (.48 h) by six. The Leu concentration of a culture was obtained using the same method for strain L -. It is worth noting that when Ile was added to the monoculture of Icells in mM63 media, the concentration of added Ile and the saturation concentration of Icells was proportional, with a constant of 9.8610 5 (cells/ml)/(mM). The same was also true for the added Leu concentration and the saturation concentration of Lcells, with a constant of 1.8610 6 (cells/ml)/(mM) (both correlation coefficients: R.0.98).
Gene expression analysis E. coli gene expression was examined using a GeneChipH E. coli Genome Antisense Genome Array according to the Expression Analysis Technical Manual (Affymetrix, 2004). The expression analyses of co-culture samples were performed with two technical replicates using two different target cDNAs separately prepared for each sample. The expression level of each gene was computed according to the FH model [58]. The estimated expression levels were normalized using a quantile normalization method [59]. For the analysis of the gene categories, we used three as the threshold for the ratio of gene expression to determine whether the expression of a gene had changed. When we calculated the ratio of gene expression for each of the 4345 genes between the two replicates of Lst,co cells in individual cocultures, the ratios were less than three for 98% of genes. To screen the categories that were significantly up-or down-regulated, we used a one-side binomial test at the significance level of 0.01. indicates the concentration in the culture media of X. We defined active cells of strains I 2 and L 2 as I 2 act and L 2 act , respectively, for the following reason. When we measured the cell concentration as colony forming units (cfu) under starvation conditions, the concentration determined by cfu was less than the concentration determined by analysis of fluorescent particles by flow cytometry (Fig. S5B and S5C). Although it was difficult to determine whether cells were alive or dead, we defined a cell being able to form a single colony as an active cell. This model neglects the decrease in [I 2 ] and [L 2 ] because it was slow ( Fig.  S5B and S5C). The symbol S represents glucose as a carbon source in the minimal media, which only determines the saturation concentration (we set 10 9 /ml for the simulation in Fig. S1B). The explanations and the values of the parameters are shown in Table S1. This model is based on the Monod model with the maintenance rate [60]. The specific character of this model is the heterogeneity of the supply function of the amino acid between I 2 and L 2 cells, as experimentally shown in Fig. 1. A mathematical model assuming these two types of nutrient supply has been reported for another obligate mutualism comprising two bacteria isolated from soil microcosms [61]. As I 2 cells supplied Leu only after growth, a L was defined as the number of L 2 cells produced in the presence of Leu from a single new I 2 cell until its death in the culture. In this model, a mathematically solved necessary condition for the stable growth of both strains is a L k L /m L .1. k L /m L represents the number of I 2 cells produced in the presence of Ile from a single new L 2 cell until its death in the culture. In our experiments, k L /m L was less than one in monoculture ( Fig. 1D and 1E) and at the lag phase in coculture ( Fig. 2 and 3), but was nearly 10 after the lag phase in coculture ( Fig. 2 and 3). (B) Comparison between the simulation results of the model and the experimental results shown in Fig. 2. The parameters used for the simulation (Table S1) assumed the cooperative change of L 2 cells to be k L = 0.4/h, which was determined from the experimental results of coculture (Fig. 3D). When k L = 0.007/ h, which is the value determined from the experimental results of monoculture (Fig. 1E), both strains must not grow successively because a L k L /m L ,1. As the model neglects the lag time until the change in L 2 cells, the deviations of the simulation results from the experimental results for the initial growth of I 2 cells are shown. (C) Comparison between the simulation results of the model and the experimental results shown in Fig. 1F. The simulation also roughly fit to the experimental results regarding the amino acid concentrations. (EPS) Figure S2 The change of the expression of genes related to Ile biosynthesis and transport in L 2 cells. The black and gray bars show the ratio of the expression of each gene in L 2 st,co and L 2 st,mo cells compared to that in L 2 ini cells, respectively, where L 2 st,co , L 2 st,mo and L 2 ini represent the state of L 2 cells at the stationary phase of coculture, the stationary phase of monoculture and the growth phase as their common initial state, respectively (see text for details). Genes related to Ile biosynthesis and transport are depicted as ''Biosynthesis'' and ''Transport,'' respectively, under their gene names. None of the Ile biosynthesis-related genes were induced in L 2 st,co cells compared with L 2 ini cells or in L 2 st,co cells compared with L 2 st,mo cells, using three-fold (red line) as the significant threshold (see Methods). None of the Ile transportrelated genes were significantly changed between any two of the three conditions. (EPS) Figure S3 The ratio of the concentration of amino acids determined by HPLC to those determined by a bioassay. We determined the quantity of Ile and Leu in the supernatants of the cultures in which each strain reached saturation phase in the presence of the required amino acid (1 mM Ile or 1 mM Leu for I 2 or L 2 cells, respectively) using two different methods: HPLC and a bioassay. The results indicated that the supplied nutrient from L 2 cells that compensated for the Ile auxotrophy of I 2 cells consisted mainly of Ile, while the supplied nutrient from I 2 cells that compensated for the Leu auxotrophy of L 2 cells consisted mainly of substances other than Leu. Methods for the bioassay are described in the text and the methods for HPLC are described below. We added Ile and Leu to the supernatants (both 0.05 mM) to raise the concentrations in the supernatants above the detection range of HPLC. As an internal standard, norleucine was also added to 0.25 mM in the supernatant. The resultant solutions were derivatized by phenylisothiocyanate (Wako, Osaka, Japan) and applied to a reverse phase HPLC on a Waters LC Module 1 (Waters Corporation, MA, USA) with a column of Wakosil-PTC (4.06250 mm, Wako, Osaka, Japan). The column was soaked in a circulating water bath at 40uC. The mobile phase comprised 60 mM sodium acetate (pH 6.0) and acetonitrile (94:6) as eluant A; eluant B consisted of 60 mM sodium acetate (pH 6.0) and acetonitrile (40:60). Gradient elution was employed according to the following linear program: time 0, 0% eluant B; 20 min, 70% eluant B; 21 min, 100% eluent B. The flow rate was 1 ml/min. Amino acid derivatives were detected by their absorbance at 254 nm. (EPS) Figure S4 Basic design and cell growth of the synthetic mutualism comprising I 2 cells and a Ura auxotroph (U 2 ). (A) Schematic diagram of the synthetic mutualism. Two auxotrophs of E. coli, strains I 2 and U 2 , supply nutrients to each other to form a potential mutualism. (B-E) Cell growth and nutrient release properties of the monocultures. The concentration of Ura was determined by a bioassay, as was the concentration of Ile and Leu. The concentration of Ura or Ile is indicated as the density of U 2 or I 2 cells which can be produced by that amount of Ura or Ile, respectively. When the nutrient concentration was not detected (under the detection limit 10 5 /ml), we plotted it at 10 5 /ml. (B and C) The time courses of the concentration of Ura (blue square) and I 2 cells (red circle) in monoculture. (B) 10 5 /ml of I 2 cells were inoculated into minimal media along with 10 mM of Ile. (C) 10 7 /ml of I 2 cells were inoculated into minimal media along with Ile. (D and E) The time courses of the concentration of Ile (orange square) and U 2 cells (green circle). (D) 10 5 /ml of U 2 cells were inoculated into minimal media along with 10 mM of Ura. (E) 10 7 /ml of U 2 cells were inoculated into minimal media without the addition of Ura. (F) The time courses of the concentration of I 2 cells (red symbols) and U 2 cells (green symbols) in coculture. 10 7 /ml of both I 2 and U 2 cells (N and m, respectively), 10 7 /ml of I 2 and 10 6 /ml of U 2 (# and D), or 10 6 /ml of I 2 and 10 7 /ml of U 2 (6 and +) were inoculated into minimal media in the presence of Ile and Ura. These results were similar to the results of the mutualism with I 2 and L 2 cells shown in Fig. 1. The final concentrations of nutrients were always less than the maximum concentrations of the donor cell in monoculture (B-E), which meant that the nutrient supplies from these strains in monoculture were insufficient for the continuous growth of both strains in coculture. Despite the insufficient level of nutrient supply in monoculture, both strains grew to saturation in coculture with all of the initial cell concentrations used (F). Strain U 2 (DH1DleuB::(gfpuv5-Km r )) was constructed from DH1 cells, as was strain L 2 , by replacing the chromosomal pyrE gene with a foreign DNA fragment comprising a reporter gene (gfpuv5) and the kanamycin resistance gene (Km r ). (Log 10 x) 2 (yellow solid line) for the beads. (B and C) The difference in the cell concentrations determined by FCM (closed symbols) and colony forming units (cfu) (open symbols) for I 2 (B) and L 2 cells (C). 10 7 /ml (black circles) or 10 5 /ml (blue squares) of the cells were inoculated into minimal media without the addition of any amino acid, or 10 5 /ml of the cells were inoculated into minimal media with 10 mM of the required amino acid (Ile for I 2 cells and Leu for L 2 cells) (red triangles). Although there was little difference between the concentration determined by FCM and the concentration determined by cfu at time 0, the concentration determined by cfu decreased more quickly than the concentration determined by FCM. Therefore, although it is difficult to determine whether a cell is alive or dead, we defined an active cell as a cell that was able to form a single colony in the mathematical model in Fig. S1A. (EPS)