Discussion
Phytoplankton communities respond quickly to environmental changes across seasons largely by a dynamic relationship among species, however, the evolutionary potential and adaptability of individual species are still not well understood. Past experimental evolution studies have shown that strong directional selection pressures can lead to adaptation through de novo mutations, often within a few hundred generations (Collins & Bell, 2004; Collins et al., 2014; Malerba et al., 2020; Schaum et al., 2018). However, the role of selection on standing genetic variation has to date not been directly assessed. In this study, we utilized a newly developed strain-specific metabarcoding approach to track selection among strains in two populations of a common pelagic diatom, and we were able to assess how rapidly selection acts on standing genetic variation. Our first hypothesis was that the diatomS. marinoi had evolved elevated tolerance to copper at a mining-exposed inlet. We found some support for this hypothesis from the observation that a small subset (10-15%) of the strains were much more tolerant to copper stress than any of the strains from a reference inlet. Secondly, we hypothesized that genetic variation would allow the population exposed to copper close to a mining site to evolve copper tolerance more rapidly, and with a larger amplitude, than from a single strain genotype, or from an unexposed population. This hypothesis was also supported by the outcome of the artificial evolution experiment where the mining population rapidly selected for the most tolerant strain in less than 50 generations and recovered more than three times more fitness (growth rate) than the mono-clonal RO5AC strain and the reference population.
Metabarcoding revealed that strain selection drove the evolutionary responses of our artificial populations, providing empirical support to the hypothesis that standing genetic variation within phytoplankton populations can support rapid adaptation (Godhe & Rynearson, 2017). In contrast, our results did not suggest that de-novo mutations were involved. Mutations are, of course, the ultimate source of genetic variation, and that we observed no tolerant strains at the reference site suggests that the trait had evolved locally during centuries of mining exposure, presumably through mutations and/or recombination. Acquiring de-novo mutations is generally a slow process because the vast majority of mutations are near neutral or deleterious (Kimura, 1983; Ohta, 1992) and restricted to affecting one locus at a time (Karve & Wagner, 2022; Tupin et al., 2010), which is why such processes could not rival selection from standing variation over relatively short experimental periods, such as in our experiment (50-100 generations).
In one isolated replicate, sexual recombination and loss of heterozygosity caused a strain to develop a hyper copper tolerant phenotype. This observation suggests that much of the fitness variability that de-novo mutations could rapidly generate may already reside in natural populations and be available for selection to act upon, especially if recombined into diverse combinations. This is a reasonable expectation as phytoplankton populations harbor up to three percent single nucleotide polymorphic diversity across the genome (Flowers et al., 2015; Mock et al., 2017), contain large-scale re-arrangements across species pan genomes (Blanc-Mathieu et al., 2017; Kashtan et al., 2014; Osuna-Cruz et al., 2020; Read et al., 2013), and can facilitate substantial rates of horizontal gene transfer (Vancaester et al., 2020). Such high standing genetic variation, combined with recombination during meiosis or horizontal gene transfer, should enable phytoplankton populations to rapidly combine favorable alleles from distant loci, strains, or even separate taxa. It is therefore reasonable to expect that similar to the situation in macroorganisms (Barrett & Schluter, 2008), outcrossing and selection from standing genetic variation should provide the primary potential for evolutionary change in phytoplankton populations, and provide short-term adaptations to both seasonal changes, spatial heterogeneity, and anthropogenic stressors such as metal pollution.
Strong toxic selection pressures are expected to purge sensitive genotypes from a population (Blanck, 2002). In contrast, our results showed that sensitive strains persisted in the mining exposed population but that a small subset of strains (three out of 30) had evolved, and retained, permanently high tolerance to copper. This suggests that the mining site population is currently not experiencing a constant strong selection pressure from copper, or that a trade-off between toxic copper tolerance and other components of fitness exists, such as nutritional copper uptake (Sunda, 2012). The lack of tolerant strains at the reference site, and inhibition of adapted cultures by growth media with low/regular copper concentrations, support the notion of a fitness cost associated with high copper tolerance. There is unfortunately no available timeseries of monitoring data of water concentrations of metals from the mining area around Västervik Gåsfjärden, and we can only speculate on the selective processes that shaped copper tolerance of the two S. marinoi populations.
Since the mining activity ceased ca. 1920, and because metal concentrations in the sediment have declined since the 1980s (Ning et al., 2018), toxic exposure could be a historical event dating back to the active mining period. Alternatively, the mining inlet population may still experience fluctuating selection pressures between toxic and non-toxic copper conditions, as the mining tailings are still exposed to varying degrees of weathering. The fact that we chose to assay copper tolerance in the resting stage population rather than the actively growing planktonic population may also have influenced the copper tolerance trait distribution. Resting stages can remain viable for at least a decade (Lewis et al., 1999), potentially even centuries (Härnström et al., 2011), providing an evolutionary buffer against loss of diversity during periods of strong directional selection on the planktonic population (Sundqvist et al., 2018). Therefore, our populations likely contain resting stages from multiple bloom seasons, and tolerant strains may have been deposited during phases of high copper concentration, and sensitive ones during ambient conditions. However, we also cannot exclude the possibility that the larger variation in the mining population stems from spurious effect from having only two sampling locations, or other sampling artifacts. Because of the 10-fold slower sedimentation rate at the mining inlet (Supplemental Method and Results), we germinated resting stages from sediment deposited between 1995-2010, compared with 2012-2015 at the reference inlet. The potentially older and more extended deposition range at the mining inlet may have captured a more diverse set of strains than at the reference inlet. Irresectable of the driver of the larger variation in copper tolerance in the mining exposed population, our finding highlights the important of incorporate large amounts of strain diversity in adaptive studies of phytoplankton.
The variation in copper tolerance among distinct experimental strains raises the question of how much variation there is in a natural population. With a few notable exceptions (Ajani et al., 2020; Gross et al., 2017; Schaum et al., 2016), artificial evolution experiments and monoculture phenotyping studies incorporate less than ten strains to represent an entire population or species (Lohbeck et al., 2012; Ribeiro et al., 2011; Sassenhagen et al., 2015; Wolf et al., 2019). This is likely insufficient to represent the actual diversity of most phytoplankton populations, not least during blooms when they are predicted to contain thousands to millions of unique clones (Sassenhagen et al., 2021). Furthermore, pan-genome studies show that even if hundreds of strains are sequenced, they generally do not saturate (Kashtan et al., 2014). Mesocosm experiments using natural communities can incorporate sufficiently large population sizes to represent clonal population diversity, yet such experiments are challenging to analyze and maintain, and still rare [but see (Schaum et al., 2017; Scheinin et al., 2015)].
Using a strain-specific metabarcoding approach (Pinder et al., 2023) improve our experimental analysis in several ways. First, strain-specific metabarcoding estimates of fitness (growth rate and copper tolerance) provided, on average, three times higher precision than mono-clonal experiments. Second, the reduced number of experimental bottle replicates of mixed-culture experiments enabled longer and more complex experiments to be performed (Gresham & Dunham, 2014). Running selection experiments for a long time added the benefit that it also resolved if plastic responses develop over time, and if the plastic potential differed between strains. In our experiments, metabarcoding revealed that certain strains could develop a high degree of plasticity relatively slowly (over 2-10 generations), which the mono-clonal dose-response curves failed to capture. This was somewhat surprising since the 72-hrs dose-response assay proceeded over multiple generations (3-7 at <EC50), which is often sufficient to capture the complete acclimation response in phytoplankton (Falkowski & LaRoche, 1991).
Third, with the strain-specific metabarcoding we could incorporate more strains without added experimental effort. Since we have developed three additional barcode loci for S. marinoi (Pinder et al., 2023), multiplexing of several barcodes should enable separation between even more strains, or strains that are not clones, but homologous in the locus Sm_C12W1 . However, such approaches will only be possible if the allelic genotypes of all strains in the selection experiment are known, something that in diploid taxa requires strain isolation and genotyping with molecule resolution (through molecular resolution sequencing) to parse out the alleles. Yet with this added sequencing effort, strain-specific metabarcoding experiments should be able to incorporate much higher amounts of diversity than the 58 strains used in our study.
Finally, a strain-specific metabarcoding approach is arguably more accurate in determining the relative fitness of strains, compared with mono-clonal fitness estimations. In part, this is because growth in mixed population removes bottle effects and other experimental artifacts associated with mono-clonal phenotyping (Robinson et al., 2014). More importantly, mixed culture experiments can incorporate interactions between strains and their shared aqueous environment. Strain-specific metabarcoding should therefore be compatible with mesocosm experiments on natural plankton communities (Scheinin et al., 2015; Tatters et al., 2013), or experiments investigating the evolutionary effects of predation (Sjöqvist et al., 2014), nutrient competition within (Collins, 2011) and between species (Descamps-Julien & Gonzalez, 2005), or other fitness traits that are challenging to determine using mono-clonal assays.