Adaptation of a Cyanobacterium to a Biochemically Rich Environment in Experimental Evolution as an Initial Step toward a Chloroplast-Like State

Chloroplasts originated from cyanobacteria through endosymbiosis. The original cyanobacterial endosymbiont evolved to adapt to the biochemically rich intracellular environment of the host cell while maintaining its photosynthetic function; however, no such process has been experimentally demonstrated. Here, we show the adaptation of a model cyanobacterium, Synechocystis sp. PCC 6803, to a biochemically rich environment by experimental evolution. Synechocystis sp. PCC 6803 does not grow in a biochemically rich, chemically defined medium because several amino acids are toxic to the cells at approximately 1 mM. We cultured the cyanobacteria in media with the toxic amino acids at 0.1 mM, then serially transferred the culture, gradually increasing the concentration of the toxic amino acids. The cells evolved to show approximately the same specific growth rate in media with 0 and 1 mM of the toxic amino acid in approximately 84 generations and evolved to grow faster in the media with 1 mM than in the media with 0 mM in approximately 181 generations. We did not detect a statistically significant decrease in the autotrophic growth of the evolved strain in an inorganic medium, indicating the maintenance of the photosynthetic function. Whole-genome resequencing revealed changes in the genes related to the cell membrane and the carboxysome. Moreover, we quantitatively analyzed the evolutionary changes by using simple mathematical models, which evaluated the evolution as an increase in the half-maximal inhibitory concentration (IC50) and estimated quantitative characteristics of the evolutionary process. Our results clearly demonstrate not only the potential of a model cyanobacterium to adapt to a biochemically rich environment without a significant decrease in photosynthetic function but also the properties of its evolutionary process, which sheds light of the evolution of chloroplasts at the initial stage.

. Preculture for the evolution experiment in TCM0. We transferred the cells from BG-11 to TCM0 and transferred them 4 times in TCM0. The colors show the cultures with different inoculation concentrations as plotted at time 0 (8.8×10 4 , 3.8×10 5 , 8.3×10 5 , and 3.9×10 6 cells/mL for blue, green, red, and black, respectively) and we used the red line for the preculture of the evolution experiment.
Growth was not stable before approximately 15 days, but later stabilized (black and red lines), showing initial adaptation from BG-11 to TCM0. We found no mutations in the genome of the initially adapted cells (red line, see Table S2). Figure S3. The specific growth rates (μ) as a function of the relative concentration of the toxic amino acids (x) in each transfer round. μ were determined from the growth curve as a slope of the linear regression of the natural log of the cell concentration when the number of data points for the round was greater than 2 and as ln(C f /C 0 ) when the number of the data points for the round was 2. C f and C 0 are the final and initial cell concentrations for the round. The red curves show the fitting of the experimental data to the equation μ = μ max /(1+x/IC 50 ) = e β1 /(1+x/e β2 ), where β1 and β2 are the fitting parameters that correspond to ln[μ max ] and ln [IC 50 ], respectively. The fitting results are summarized in  The transfers of the culture by dilution are shown as the vertical decrease in cell concentration. The initial cell concentration of each transfer was varied (mostly approximately 10 5 cells/mL). The cell concentration affected the growth, although the basis is still unclear. For example, less than 10 5 cells/mL seemed to make the culture unstable after day 79, and less than 10 6 cells/mL seemed to make the culture unstable at the first culture of the initial adaptation ( Figure S2). Thus, we compared the growth in TCM1 to that in TCM0 at the same initial cell concentration (dashed redline and black line, respectively). Figure S5. The difference between the approximate analytical solution of Eq. 2 and the numerical simulation according to Eq. 1. We calculated S IC50 by both Eq. 2 and the numerical simulation using the various parameters of r, d z , and N. The range used for r and d z was 10 -5~0 .1 and 0.01~1, respectively, the same as the range shown in Figure 3C, and N = 10 5 (black), 10 7 (blue), or 10 9 (red).
The gray solid line shows where the points should fall if Eq. 2 and the numerical simulation were equal. The dotted lines show the experimentally determined value of S IC50 . The deviation from the gray line becomes rather large for S IC50 <0.001 and 1<S IC50 (more than 10 fold). Because we determined S IC50 from the numerical simulation for generation until 100, S IC50 becomes too small to accurately quantify for S IC50 <0.001. For 1<S IC50 , the value of the probability r used for this range was large (almost 0.1), which is out of our approximations.  Text S1 Here we derive Eq. 1 and 2 shown in the main text. We used an approximate continuous derivation for simplicity. We first assume a cell population with the frequency h(z,t) in which each cell has a trait z (a variable that represents ln[IC 50 ]) and a specific growth rate μ(z), with a variable population size , . When a cell with z produces an offspring, the trait of the progeny becomes z+d z with a probability r, or else becomes z (the same as the parent) (1−r). Then, we derive the rate equation Eq. S1 Here, μ(z) is described as μ(z)=μ max e z /(e z +x) along with the main text definition (μ(z) = μ max /(1+x/IC 50 ) = μ max IC 50 /(IC 50 +x)), where x is the toxic amino acid concentration. We approximated it as μ(z) ≈ ce z , where c = μ max /x, assuming e z +x≈x in the transferred line during the experimental evolution. Note that the evolutionary properties with respect to generation (not time) do not depend on the absolute fitness (i.e., c) but only depend on the relative fitness (c is canceled out below), and we ignored the fact that we changed the amino acid concentration x (thus c) in the experimental evolution.
Both terms on the right hand side are positive, with 0<r<1 and μ(z)>0, and the population size increases over time as Eq. S2 derived by taking the integral of both sides of Eq. S1 with respect to z. Then, we derived the rate equation for the frequency , with the fixed population size N, i.e., , (Eq. 1) from Eqs. S1 and S2. The generation is determined from the time variation in N h (t). From Eq. S2, N h (t) is solved as exp ̅ , where is the total frequency at time 0. The generation g satisfies 2 , and is solved as ̅ / 2. Thus 2 ln Eq. S3 which is required to obtain the evolutionary rate per generation (see below).
We roughly obtain an approximated analytical solution of the evolutionary rate of the mean of z (designated as M) per generation dM/dg (=S IC50 , when it is constant) in the model shown in Eq. 1.