**Barcoding indicated that the RO5AC_GP_C bottle was physically cross-contaminated at the start of the experiment and rapidly outcompeted.
Figure legends
Figure 1 : Population-wide characterization of copper tolerance in the two populations (mining and reference inlet), based on mono-clonal acute 72-h dose-response curve exposure. A) mining and the reference inlets population distribution of copper tolerance estimated based on EC50, the effective concentration inhibiting growth rate by 50% (N=30 and 28 strains, respectively). B) and C) are examples of full dose-response curves for the most sensitive, tolerant, and median strain (see symbol legend for strain names) per population based on ranked EC50. N=5 for control and 18 copper treatments for dose-response curves. Strain RO5AC is included as an internal control between populations. Error bars/shaded areas show 95% confidence intervals, and the median as well as RO5AC’s EC50 is indicated with a vertical line.
Figure 2: Changes in copper tolerance during the artificial evolution experiment in a copper-toxic (Copper) and non-toxic environment (Control). At the end of the selection phase, copper toxicity was relaxed for three days to test if the developed tolerance was reversible (plastic) or permanent (evolved). Panel B shows a mono-clonal culture of RO5AC which was run in parallel with the reference (Rep. 1) and the mining population (Rep. 2). Metabarcoding showed that RO5AC was cross-contaminated with the GP-Control treatment from day nine and onwards (Table 1). These samples are omitted in panel B. Error bars show 95% confidence intervals based on dose-response curve data gathered as shown in Fig. 1.
Figure. 3. Photosynthetic activity was inhibited at a slower rate in evolved or pre-exposed (Mining population) cultures, compared with naïve strains (Reference population). The yield of Photosystem II (PSII) was monitored during a short exposure (3h) to lethal copper concentrations. Shown are means with 95% confidence intervals with N=3 replicates per time-point (N=1-2 for single strain RO5AC). Note that both populations require more than three days of selection to develop tolerance in contrast to the single strain RO5AC, which reaches its minimum inhibition rate within three days.
Figure. 4. Changes in strain abundances based on metabarcoding during the artificial evolution experiment. Shown are time-point averages based on quantification of strain abundances recorded at days 0, 9, and 42 (N range between 1 and 5). Cross-contaminated samples (Fig S4) have been excluded from the Mining population at day 42. Strain GP2-4_45 and 46 are clones and cannot be distinguished. Strain VG1-2_65 and VG1-2_99 lack strain-specific DNA sequencing data but correspond to two pairs of unknown amplicons tightly linked at a 1:1 ratio across the metabarcoding data (Pinder et al. in prep). They have been annotated as belonging to either of these strains. Abundant strains are highlighted in plots with the strain names according to the legend, and the sequential order of strains follows the legend top-down and left-to-right.
Figure 5. Comparison of metabarcoded fitness estimates during the selection experiment (Observed growth rate) against mono-clonally predicted ones (Modeled). Growth rate was modeled for the copper treatment using mono-clonal 72-h dose-response curves (N=23 total replicates) and the applied 8.65 µM copper exposure, or the specific growth rate in the control environment. The observed growth rate corresponds to mean growth rates (N=1 to 5 depending on strain and treatment) during the first nine days of the selection experiment. Extinct strains (missing in all selection experiment replicates) for each panel are highlighted in the legend and not included in plots. Colors of the strains are the same as in the legend of Fig. 4, with key strains from Fig. 1 and 4 shown with names. Error bars show 95% confidence intervals when N>1. Bolded strain names highlight cases when modeled selection outcome agrees with the selection experiment at day 42.