Strains and media
The set of 16 strains used in this experiment contains environmental isolates along with strains from the ATCC collection (Supplementary Table 1). The strains were chosen based on two criteria: a distinct colony morphology that would enable visual identification when plated on an NB agar plate; and ability to coexist for ~60 generations with at least two other strains in our collection.
All cultures were grown in M9 minimal salts media containing 1X M9 salts, 2 mM MgSO4, 0.1 mM CaCl2, 1X trace metal solution (Teknova), supplemented with 3 mM galacturonic acid (Sigma), 6.1 mM Serine (Sigma), and 9.1 mM sodium acetate as carbon sources, which correspond to 16.67 mM carbon atoms for each compound and 50 mM overall. We chose a combination of carbon sources representing three chemical groups—a carbohydrate, an amino acid, and a carboxylic acid—in order to promote the survival and coexistence of a diverse set of species. The media was prepared on the day of each transfer. A carbon source mixture was prepared ahead at 10X, and was kept in aliquots at 4 °C for up to four weeks.
Evolution experiment
Frozen stocks of individual species were streaked out on nutrient agar Petri plates and grown at 28 °C. After 48 h single colonies were picked and inoculated into 15 ml falcon tubes containing 3 ml nutrient broth (5 g/L peptone BD difco, BD Bioscience; 3 g/L yeast extract BD difco, BD Bioscience), and were grown overnight at 28 °C shaken at 250 rpm. Initial mixtures were prepared by diluting each species separately to an OD of ({10}^{-2}) and mixing the normalized cultures at equal volumes. OD measurements were done using a Epoch2 microplate reader (BioTek) and were recorded using the Gen5 v3.09 software (BioTek). After mixing, the cocultures were aliquoted to replicates and further diluted to a final OD of ({10}^{-4}), at which the evolutionary experiment was initialized. The number of replicates for each community varied between 3 and 18 (Supplementary Data 1).
Communities were grown in 96-well plates containing 200 µl M9 at 28°C and were shaken at 900 rpm. Every 48 h cultures were diluted by a factor of 1500 into fresh M9 media, and OD600 was measured. For this dilution factor, each cycle corresponds to ~10.5 generations. As 1 OD600 ~ of ({10}^{9}) C.F.U/ml, and communities reached ~ 0.5 OD600 and were grown in 200 µl and was diluted by 1500, ~({10}^{5}) cells were transferred each dilution. To avoid cross contaminations, cultures were grown in a checkerboard formation, meaning that each community was surrounded by wells containing media but no bacteria.
At transfers 0, 2, 5, 7, 10, 14, 19, 30, and 38 community composition was measured by plating on nutrient agar plates (5 g/L peptone BD difco, BD Bioscience; 3 g/L yeast extract BD difco, BD Bioscience, 15 g/L agar Bacto, BD Bioscience) and counting colonies. For that, the cultures were diluted to an OD of (2.4* {10}^{-8})− (1* {10}^{-8}) and 100 µl of the diluted culture was plated on NB plates and spread using glass beads. Plates were incubated at 28 °C for 48 h and colonies were counted manually. The distribution of the number of colonies counted at each plate to infer community composition is found in Supplementary Fig. 11.
We chose the communities based on a preliminary experiment that was conducted by the same protocol for six transfers. In this experiment, 114 of 171 possible pairs of a set of 19 strains (3 strains were not included in the evolution experiment) were cocultured. Pairs that had coexisted for the duration of this experiment, and were confidently distinguishable by colony morphology, and trios that are composed of these pairs, were used for the coevolutionary experiment. We started the evolutionary experiment with 51 pairs and 51 trios, and removed communities that did not coexist for the first ~70 from the final analysis. If a replicate was suspected to be contaminated it was also excluded from further analysis.
Ecological experiments
We supplemented the data of the evolutionary experiment with two ecological competition experiments with the same experimental condition. In order to assess whether communities typically reach an ecological equilibrium within ~50–70 generations (Supplementary Fig. 3), we cultured eight of the pairs that were used in the evolutionary experiment. This experiment was initiated in the same way as the evolutionary experiment, only that after the species’ starters were normalized they were inoculated at the varying initial fractions – 9:1, 5:5, 1:9. Because the normalization depended on optical density, there is a variation in the actual initial fractions between different pairs. Community composition was then measured on six transfers during this experiment: 0, 1, 2, 4, 5, and 6.
In order to assess whether changes in composition are due to heritable changes in species’ phenotypes, we used strains that were re-isolated from 31 evolved pairs, and 13 pairs of ancestral strains (Supplementary Figs. 6, 7). Strains were replicated from glycerol stocks into the experimental media and grown for 24 h. The starters were normalized to initiate the competition assay at ({rm{OD}}={10}^{-4}) in fresh M9 media. Species were mixed at equal volume and were propagated for five cycles. community composition was measured at initial conditions, and at the end of the final cycle (5).
Quantification of repeatability
In order to quantify the qualitative repeatability of different replicate communities we first identified which species was the maximally increasing member at each replicate, that is, which species had increased its abundance by the largest factor between generation 70 and 400. Then, we quantified the frequency of the replicates that had the same maximally increasing member for each community. This measure always produces a value between 1 and 1/n where n is the number of species in the community. We checked the distribution of the repeatability scores against the null hypothesis that the factor by which a species’ abundance increases during evolution is independent of the species or the community. For this, we shuffled the factor of change in relative abundance across all samples, for pairs and trios separately, and quantified the new repeatability scores of the shuffled data. Data of the null hypothesis were generated over 2000 times, and the p value was given by the probability to get a mean equal or above the real data mean.
We used the average Euclidean distance of replicates from the median replicate in order to quantify the variability between replicate communities. In order to check whether the distribution of variabilities is similar to what can be expected of random communities, in which each species in the community is just as likely to have any relative abundance, we replaced the real fractions with fractions drawn from a uniform Dirichlet distribution with (underline{{boldsymbol{alpha }}}=underline{1}). We then checked the statistical difference between the two distributions using one-sided Mann–Whitney U test.
Trio composition predictions
We used the formerly established method for predicting the composition of trios from the composition of pairs that was developed by Abreu et al.14 In this approach the fraction of a species when grown in a multispecies community is predicted as the weighted geometric mean of the fraction of the species in all pairwise cultures. The accuracy of the predictions was measured as the Euclidean distance between the prediction and the mean composition of the observed trio, normalized to the largest possible distance between each two communities, (sqrt{n}), where n is the number of species.
We used the factors by which species increased their abundance during coevolution in pairs (between generations ~70 and ~400) to predict which species would increase by the largest factor in trios. The maximally increasing member in a given community was assigned to be the one that was the maximally increasing member in the most replicates of that community. If the same species was the maximally increasing member in both pairs it was a member of, then this species was predicted to be the maximally increasing member of the trio. If in every pair a different species was the maximally increasing member, then we predicted that the maximally increasing member of the trio would be the one with the highest mean increase. Only two trios had such transient topology, where in each pair a different species increases, thus we are unable to determine the general utility of the latter approach.
Re-isolation
Each ~50 generations all communities were frozen at −80 °C with 50% glycerol in a 96-deep well plate. In order to re-isolate strains, stocks were inoculated to a 96-well plate containing the experimental media using a 96-pin replicator, and grown for 24 h at 28 °C. After growth, cultures were diluted by a factor of (2.4* {10}^{-8}) and 100 µl were spread on a nutrient agar plate using glass beads. Plates were kept at room temperature for at least two days and no longer than a week before re-isolations. 5-15 colonies of each strain were picked using a sterile toothpick, and pooled together into 200 µl M9. Re-isolated strains were incubated at 28 °C and shaken at 900 rpm for 24 h and kept in 50% glycerol stock at −80 °C until further use.
Growth rates and carrying capacities of individually evolved strains
Re-isolated strains were replicated from glycerol stocks into the experimental media and grown for 24 h. The starters were normalized to initiate the growth assay at (OD={10}^{-4}) in fresh M9 media. The optical density was measured in two automated plate readers simultaneously, Epoch2 microplate reader (BioTek) and Synergy microplate reader (BioTek), and was recorded using Gen5 v3.09 software (BioTek). Plates were incubated at 28 °C with a 1 °C gradient to avoid condensation on the lid, and were shaken at 250 cpm. OD was measured every 10 min. Each strain was measured in four technical replicates, evenly distributed between the two plates, and 2–3 evolutionary replicates were measured for each species (replicates that evolved separately for the duration of the experiment). Growth rates were quantified as the number of divisions it takes a strain to grow from the initial OD of ({10}^{-4}) to an OD of (8* {10}^{-2}) (({log }_{2}frac{0.08}{{10}^{-4}})) divided by the time it took the strain to reach this OD. This measure gives the average doubling time during the initial growth and also accounts for the lag times of the strain. The growth rates of evolutionary replicates were averaged after averaging technical replicates.
Carrying capacity was defined as the OD a monoculture reached at the end of each growth cycle of the evolutionary experiment averaged across replicates. These measurements were done in an Epoch2 microplate reader (BioTek). In order to reduce noise, the trajectories of OD measurements were smoothed for each well using moving mean with an averaging window of three.
Carrying capacities of coevolved strains
Re-isolated strains were replicated from glycerol stocks into the experimental media and grown for 48 h in M9 media at 28 °C. Cultures were then diluted by 1500 into 3 technical replicates in fresh M9-media, and were given another 48 h to reach carrying capacity. The strains used in this experiment were isolated from 1-3 evolutionary replicates (Supplementary Data 2).
Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this article.
Source: Ecology - nature.com
