Biodiversity currently faces unprecedented threats owing to species extinctions. Ecologically, compensatory dynamics can ensure stable community biomass following perturbation. However, whether there is a contribution of genetic diversity to community responses is an outstanding question. To date, the contribution of evolutionary processes through genotype shifts has not been assessed in naturally co-occurring multi-species communities in the field. We examined the mechanisms contributing to the response of a lake phytoplankton community exposed to either a press or pulse acidification perturbation in lake mesocosms. To assess community shifts in the ecological response of morphospecies, we identified taxa microscopically. We also assessed genotype shifts by sequencing the ITS2 region of ribosomal DNA. We observed ecological and genetic contributions to community responses. The ecological response was attributed to compensatory morphospecies dynamics and occurred primarily in the Pulse perturbation treatment. In the Press treatments, in addition to compensatory dynamics, we observed evidence for genotype selection in two species of chlorophytes, Desmodesmus cuneatus and an unidentified Chlamydomonas. Our study demonstrates that while genotype selection may be rare, it is detectable and occurs especially when new environmental conditions are maintained for long enough to force selection processes on standing variation.
Human activity has become a major geophysical force reshaping the environment . Anthropogenic perturbations, varying in strength and duration, are globally considered to be a greater challenge to species than are natural perturbations . The ecological responses of aquatic ecosystems to perturbations of short (pulse) and longer (press) duration have been examined, but the potential for a genetic contribution to community responses remains largely unexplored. There is mounting evidence of a role for rapid evolution [3–7], including indirect evidence for phytoplankton . However, little work has been done to date to assess eco-evolutionary change in multi-species contexts. Rapid evolution raises the possibility that a genetic response could occur in communities impacted by sudden environmental shifts, through genotype selection among standing genetic variation (e.g. [9,10]). However, among studies examining eco-evolutionary responses in a multi-species context , there is very little actual genetic evidence. Plankton communities are not only vulnerable to sudden perturbations, but also provide an easily manipulated natural community with fast generation times, allowing an exploration of the contribution of both ecological and evolutionary processes in response to perturbation .
Community stability can be promoted by compensatory ecological dynamics (changes in relative species frequencies) among the constituent taxonomic morphospecies following perturbation. These ecological interactions can buffer density, total biomass and functional fluctuations potentially arising following species extinction [12–14]. Compensatory dynamics occur when two species covary negatively, with species being asymmetrically affected by the perturbation . Compensation occurs via the insurance effect when competing taxa are functionally redundant [16–18].
Following a perturbation, in addition to changing taxon relative abundances (community response), adaptive genetic responses of populations can also occur, enabling extinction avoidance (e.g. ‘evolutionary rescue’; [19–24]). Such genotype selection is often on standing genetic variation in species—variation that is extremely common in phytoplankton populations . To date, there have been no estimates of the contribution of evolutionary processes through genotype shifts to the response of naturally co-occurring multi-species communities following pulse and press perturbations (sensu ). Because of the different timescales involved, one might expect ecological ‘morphospecies' (taxonomic) responses such as compensatory dynamics to be most observable following pulse perturbations, while genetic ‘phylotype’ shifts should contribute more following press perturbations. We examined these processes using lake phytoplankton communities exposed to pulse and press acidification. While an easy perturbation to apply, acidification is also a common perturbation to lakes through accidental spills and longer-term environmental pollution, with substantial effects on plankton community composition, biomass and function [26,27]. Acidification can be applied in press but also in pulse mode because photosynthesizing phytoplankton can neutralize waters through the carbon cycle .
2. Material and methods
(a) Experimental design
The mesocosm experiment took place in Lake Hertel, Mont St-Hilaire (45°32′ N, 73°09′ W) near Montreal, QC, Canada. The lake is naturally meso-eutrophic (total phosphorus = 20 µg l−1) with pH of 7.5–8.5. Mesocosms consisted of 1 m diameter × 4.5 m deep clear plastic bags suspended from a large floating dock and open only at the air–water interface. Mesocosms were filled on 25 May 2012 by pumping water with phytoplankton directly from the lake through a 53-µm mesh to remove zooplankton. Zooplankton were subsequently collected from the lake with vertical net hauls (53 µm) and mixed in a large container, then repeatedly introduced in small aliquots into each mesocosm to randomize the initial communities. On the same date, we added 10 µg l−1 of phosphorus (as NaPO4) and 85 mg l−1 of nitrogen (as KNO3) (half the normal concentration in Lake Hertel) to ensure biomass persistence through the experiment. Zooplankton were not the focus of this study, but their responses are detailed in electronic supplementary material, appendix A.
Two weeks later (5 June 2012), acid perturbation to pH 5.0 was accomplished by slowly adding 37% HCl to treated mesocosms as monitored using a YSI6600 sonde. Treatments, each replicated three times, were (i) Press perturbation with pH maintained at 5.0 for the remainder of the experiment, (ii) Pulse perturbation with only one acidification event and neutral pH re-established through photosynthesis, and (iii) Control with no manipulation. The experiment was terminated five weeks later (17 July 2012) during which weekly sampling occurred.
(b) Assessing morphospecies responses
After mixing each mesocosm with a paddle, phytoplankton were sampled from 0 to 3 m depth using an integrated 1.5 cm diameter PVC tube sampler and preserved with Lugol's solution. Subsamples (10 ml) were later identified microscopically to the genus or species level using the Utermöhl method  with an inverted microscope (480×; Diavert, Leica). A minimum of 200 cells of the most abundant morphospecies and a maximum of 400 total cells were counted per sample and 10 individuals of each morphospecies were measured for biovolume calculations based on geometric equations . Summed morphospecies biovolumes provided total community biovolume, an estimate of total biomass.
We used principal response curve (PRC) analysis  to assess treatment effects on phytoplankton community (morphospecies) composition through time using the vegan package in R . PRCs are based on redundancy analysis modified to produce a more intuitive visual of time-dependent treatment effects through response curves. The primary axis of variation is plotted, showing treatment variation from the Control. The significance of this axis was tested with 1000 permutations. Only species that were present throughout the experimental period were included. PRCs also provide a score for each taxon (between −1 and 1), indicating the relative strength and the direction of the response to the treatment.
The variance ratio (VR) was used to quantify compensatory ecological dynamics, the degree of community synchrony among morphospecies. The index summarizes the ratio between the variance of the aggregate community biovolume (sum of the component populations, C) and the variances of the n individual population biovolumes (Pi) as follows: [33,34]. A VR of 0 indicates complete community asynchrony (negatively covarying morphospecies) and 1 indicates complete synchrony. The metric was calculated for all sampling dates.
(c) Assessing genotype responses
To determine genetic diversity and to identify genotypes in mixed communities, the eukaryote second internally transcribed spacer (ITS2) provides a potential barcode for chlorophyte algae  and is a complement to the COI animal barcode . ITS2 has been used as a barcode for diatoms , for cryptophyte phylogenies  and psychrophilic chlamydomonads . Intraspecific variation in the ITS2 region has been noted in a substantial number of algal taxa [40–45].
Genetic samples were collected from each mesocosm using a dark 1 l narrow-mouth Nalgene bottle at 0.5 m depth on two dates pre-acidification: PRE1 on 25 May (day 146), PRE2: on 5 June (day 157), and on two dates post-acidification: POST1 on 26 June (day 178); and POST2 on 10 July (day 192). A 400 ml (150 ml on 25 May) subsample was then filtered onto a 0.22 µm nitrocellulose membrane (0.22 µm GSWP, Millipore) to retain phytoplankton. Filters were folded closed and kept in foil at −20°C until DNA extraction (less than seven months). Because of a loss of some frozen samples, molecular analyses could only be done on two Control replicates. Details on the DNA extraction, primer design and phylogenetic analysis methods are detailed in electronic supplementary material, appendix B. We designed primers targeting the ITS2 of Scenedesmus spp., the taxon identified microscopically as one of the most responsive chlorophytes, driving the morphospecies response following acidification.
From the genus-specific alignments in the phylogenetic analyses, we assigned ITS2 sequences from each sample to operational taxonomic units (OTU; clusters of related sequences) using either more than or equal to 90% or more than or equal to 97% identity as the clustering cut-off. We used both cut-offs because it has not yet been determined at which level phytoplankton species are most readily identified using ITS2 for different genetic similarities. We refer to these 90% and 97% OTUs as phylotypes and genotypes, respectively, to clearly differentiate between them. To determine the relationships between our phylotypes and previously described morphospecies, we constructed phylogenetic trees using a representative sequence for each phylotype (removing phylotypes detected in less than or equal to two samples or comprised less than five sequences) with reference ITS2 database sequences, aligned with Geneious. Maximum-likelihood trees were built in PhyML  using generalized time reversible substitution model with among-site rate variation (using a gamma distribution), obtaining bootstrap values from the same program. Phylotypes were assigned to previously described species if they formed a well-supported clade with ITS2 sequences from those species (e.g. Dcun-0386 is a phylotype of the Desmodesmus cuneatus species complex). Phylotypes that were only distantly related to known species were assigned a general numerical ‘species' designation (e.g. Chlamydomonas sp. 11.).
(a) Community and morphospecies responses
Time series of the total phytoplankton biovolume in each mesocosm showed an important peak on the first dates (figure 1a), likely owing to the nutrients added during set-up. We thus considered the date just prior to the acidification (third date; Julian day 157) as the ‘initial’ condition. Later peaks in biomass were observed mainly in the acidified treatments (figure 1a). The first peaks in total phytoplankton biovolume were primarily composed of diatoms, with later peaks in the Press and Pulse mesocosms composed of chlorophytes, at the expense of cryptophytes (Control; figure 1b). Cyanobacteria, chrysophytes and dinoflagellates contributed only marginally. The dominant chlorophyte identified microscopically was Scenedesmus spp., revealed following molecular analyses, to be primarily Desmodesmus spp. (electronic supplementary material, appendix C). These community shifts were supported by the PRC results, which showed variation between treatments after acidification based on losses in cryptophyte and gains in chlorophyte morphospecies (electronic supplementary material, appendix C).
Mean community synchrony among morphospecies through time was greatest in the Control (VR = 0.63 ± 0.05; mean ± s.e.), followed by the Press (VR = 0.42 ± 0.02) and least in the Pulse (VR = 0.23 ± 0.05). All means differed significantly (ANOVA p = 0.00147, Tukey HSD all p < 0.05).
Zooplankton densities, especially those of pelagic cladocerans were reduced in the acidified treatments (detailed results in electronic supplementary material, appendix A).
(b) Genotype responses
Molecular analyses revealed that the majority of the Scenedesmus morphospecies actually belonged to the morphologically cryptic Desmodesmus genus. Because the PCR primers were not designed to specifically amplify Scenedesmus, they also identified nine other genera, for a total of 73 phylotypes (defined as clusters of ITS2 sequences exhibiting more than or equal to 90% identity; electronic supplementary material, appendix table D1).
Ten of the 73 phylotypes responded to the perturbation (figure 2 and electronic supplementary material, appendix figure D1). Five phylotypes responded negatively (phylotypes within Chlamydomonas, Coelastrum and Oedogonium): being present in the two time points before the perturbation, but not in the two points after, or being present only in the Controls. Five showed a positive response (phylotypes within Desmodesmus, Chlamydomonas and Oedogonium): appearing after, or remaining present, post-acidification.
For the phylotypes exhibiting responses to acidification, we determined whether that response involved differential shifts in the abundance of phylotypes that could be considered members of the same species (clustered together in the phylogenetic trees; figure 2), which would provide evidence for an evolutionary response. Because our mesocosms were mostly closed to immigration from the lake (only open at the top to rare airborne dispersal), all phylotypes present later would likely have been present from the start, although some were too rare to be sequenced, becoming detectable only later in the experiment, when their abundances increased. Thus, the strongest evidence for genetic shifts resulting from perturbation occurred when one phylotype was replaced by another within a treatment, but not in the Controls.
Two sets of phylotypes showed genetic responses (figure 2). The first case was for phylotypes Dcun-036 and Dcun-038 within the D. cuneatus species group (figure 2a). The second was within Chlamydomonas sp.11: phylotypes Csp-070 and Csp-049 (figure 2b). Within D. cuneatus, a shift from Dcun-038 prior to perturbation, to Dcun-036 post-perturbation occurred in all Press mesocosms (figure 3a). Furthermore, Dcun-036 was present in one Control replicate early on, indicating that it was likely present at the start to some degree in all mesocosms. It also appeared on one post-acidification date in one Pulse mesocosm, but did not persist as a dominant to the last date sampled.
Within Chlamydomonas sp.11, Csp-070 dominated half the Press mesocosms on the two post-perturbation dates (figure 3b). The Csp-070 phylotype was also present in one Pulse mesocosm, but only on the first date post-perturbation (POST1), being fully absent later. When Chlamydomonas sp.11 was present on the POST2 date in the Pulse and Control treatment, it consisted entirely of Csp-049, which had only been observed in one Press mesocosm prior to perturbation (figure 3b).
Defining phylotypes at 90% identity is very conservative and may not fully reflect intraspecific variation, which could be better captured at a higher % identity cut-off. Therefore to further elucidate the variation within Chlamydomonas sp.11 and D. cuneatus, we also performed a deeper analysis of genetic variation within the phylotypes belonging to these two species complexes at 97% similarity. We refer to these sequence clusters as genotypes. Any cut-off higher than 97% is not recommended as artefacts due to sequencing error can arise. In the case of D. cuneatus Dcun-036 had responded positively only in the Press treatment. Dcun-036 itself was dominated by a single genotype (red bars in electronic supplementary material, appendix figure E1-A), which was not observed in the Control, nor prior to acidification. Another genotype (grey bars in electronic supplementary material, appendix figure E1-A) of Dcun-036 was present in the Control and to a lesser degree also in treated replicates, post-acidification. Finally, several other genotypes of Dcun-036 were detected only in the Press treatments post-perturbation.
For Chlamydomonas sp.11, the positively responsive Csp-070 in the Press treatments was found to consist of many genotypes, with only one of these (beige bars in electronic supplementary material, appendix figure E1-B) occurring in the one Pulse mesocosm where Csp-070 had been observed (figure 3b). Csp-049 continued to be represented by a single genotype (data not shown) indicating no deeper genetic variation. There was a great deal of genetic variation at 97% similarity in two of three Press perturbation mesocosms, undetected elsewhere (electronic supplementary material, appendix figure E1-B).
Our study shows that under relatively natural conditions, with naturally co-occurring and co-evolved species in plankton communities, compensatory dynamics are critical following acidification. This was especially the case under a Pulse perturbation treatment, where little genetic response was observed, and may have been prevented because of the asynchronous compensatory response of the different phytoplankton taxa. However, we did find evidence for a role of evolutionary processes through genotype shifts following a Press perturbation. Longer-term environmental change thus appears to provide conditions for evolution to play an increased role in conjunction with ecological dynamics (some compensation) in community responses, including extinction risk .
The phytoplankton communities responded to perturbation with compensatory dynamics that enabled a greater overall biomass to persist through time. This greater biomass occurred via the dominance of larger chlorophytes associated with increases in several colonial morpho-taxa (Gloeocystis, Ulothrix, Mougeotia) and very large unicells (Closterium) in the Pulse, as well as one large diatom (Tabellaria floculosa) in the Press, all with higher individual biovolume relative to the chlorophyte taxa (Coelastrum, Scenedesmus armatus, Schroderia) and the cryptophytes (Cryptomonas, Rhodomonas) that dominated the Controls. Many chlorophytes are known to be acid-tolerant [47–51], explaining their positive responses to perturbation. Cryptomonads are also generally thought to be acid-tolerant , but our PRC results suggest that these single-celled morphospecies were outcompeted by acid-resistant chlorophytes that otherwise dominated the perturbed mesocosms. Compensatory dynamics have previously been shown to play an important role in community recovery [34,53], although on longer timescales (inter-annual variation) synchronous dynamics may be more prevalent . Among phytoplankton morpho-taxa specifically, as we also observed, compensatory dynamics can regulate community biomass in response to acid perturbation via opposing interactions of diatoms (or cryptophytes), with chlorophytes ; with larger-bodied chlorophytes being more acid tolerant. In the Pulse treatment in particular, wherein recovery of initial environmental conditions were permitted, such compensatory dynamics were most prevalent. Compensation may have been aided by the pH effect on zooplankton, especially the pelagic Cladocera, which experienced declines as well. Reduced zooplankton grazing pressure represents an additional effect of pH perturbation to aquatic ecosystems that could enable the ecological response observed in phytoplankton communities.
Our results indicate that community responses in the Press mesocosms likely included an evolutionary mechanism, as per our original hypothesis. The clearest genetic pattern was in D. cuneatus at 90% similarity in which Dcun-036 appeared to be outcompeted by Dcun-038 under neutral pH, but was able to dominate in acidic waters. These results were substantiated by deeper analyses at 97% identity with one genotype of Dcun-036 dominating all acidified samples, especially in the Press treatment, but being absent from the Control. These shifts represent genotypic variation in response to perturbation and indicate the selection of particular genotypes, especially through persistent acidification.
A less clearly defined, genetic shift was observed for Chlamydomonas sp.11 which was too rare for detection pre-acidification. However, following acidification, Csp-070 dominated the two POST dates in the Press treatments (with the exception of one replicate where Chlamydomonas sp.11 was entirely absent), while in Pulse mesocosms, Csp-049 dominated. Thus, Csp-070 is presumably more acid-tolerant, allowing it to dominate with persistent acidification, but being outcompeted under neutral pH (Pulse and Control treatments). Note however that there was one Pulse replicate where Csp-049 was absent and Csp-070 was present on the first of the two post-perturbation dates. Although this replicate differs from the others, it does not contradict our conclusion: in a mesocosm where Csp-049 was apparently absent, Csp-070 may have benefitted during the initial acidified phase in the Pulse treatment being then outcompeted by another species entirely when a neutral pH was re-established (by POST2). Finally, at 97% identity, Csp-070 displayed a great deal of genetic variation, while the two declining phylotypes (Csp-20 and Csp-49) showed no finer diversity, indicating a trend towards much greater genetic variation with acidification in the genotype favoured by the Press perturbation.
Our analyses and attempts to identify genotypic changes in a natural community reveal an important difficulty in defining species, strains and genotypes when dealing with clonal organisms. Defining phytoplankton species based only on sequence similarity is not necessarily most appropriate  because of properties such as asexuality and lateral gene transfer in some clades, including Chlamydomonas. Thus, it is not possible for us to determine with certainty whether the clades we identified at 90% or 97% of genetic identity are genotypes, species or even genera, and whether the shift observed in Chlamydomonas sp.11 represents a species or a genotype shift. On the other hand, it seems fairly safe to affirm that changes in D. cuneatus at 97% represent a genotype shift, indicating the presence of at least one important genetic event in our study, using a probe specific to only a portion of the overall community: the chlorophytes. We cannot tell whether other taxa from other non-target groups also displayed genetic responses and thus our conclusions are very conservative with respect to the overall potential response of natural phytoplankton communities.
Genetic variation in our species was quite low, with a maximum of two to three phylotypes at 90% identity per group, and between four and 15 genotypes at 97% for the one target taxon examined. This could be due to the fact that phytoplankton are predominantly clonal, possessing less genetic diversity than in a sexual population with recombination. That said, both Desmodesmus  and Chlamydomonas  have the capacity to reproduce sexually, which could have contributed to the small amount of genetic variation observed. Sexual reproduction may have furthermore allowed the chlorophytes to dominate the post-acidified communities as observed in the community results. A low level of measured genetic variation might also have been a result of our decision to disregard rare phylotypes.
Despite the low measured genetic variation in our study, minimal immigration, and genetic probes for only one clade, we nevertheless observed events indicative of evolutionary shifts in genotype frequencies in at least one, and possibly two taxa exposed to sustained habitat alteration, with demonstrated shifts in dominating genotypes following perturbation. Although the marker we chose for genotyping is a neutral marker, thus precluding certainty as to whether this event occurred through natural selection or genetic drift, the fact that the same genotype arose in all three Press replicates, in the strongest case of Dcun-036, points to a selection process (as per ). As mentioned, this genotype likely possesses a mutation that makes it more tolerant of low pH. Because our communities were only exposed to acidic waters for five weeks, it is unlikely that this was a mutation arose de novo with acidification [58,59] and favoured genotypes would need to be present a priori for evolutionary shifts to occur at ecologically relevant timescales.
The datsets supporting this article have been uploaded as part of the electronic supplementary material.
G.T. carried out the experimental work, molecular laboratory work, carried out the data analysis and sequence alignments, participated in the design of the study and drafted the manuscript; D.A.W. oversaw the genetic work and analyses and helped draft the manuscript; B.E.B. conceived the study, coordinated the study and helped draft the manuscript. All authors gave final approval for publication.
We declare no competing interests.
This work was supported by graduate scholarships to G.T. from the Natural Science and Engineering Council of Canada (NSERC), Fonds de Recherche en Nature et Technologies (FRQNT) and Groupe de Recherche Interuniversitaire en Limnologie (GRIL), and from NSERC Discovery grants to B.E.B. and to D.A.W.
We thank Travis Dawson, Heba El-Swais and Ting Ting Cui for their help with DNA sequencing; Gaël Lainé-Panet, Thibault Veyret and Julien Arsenault for field and laboratory assistance and Serge Paquette help with phytoplankton identification. Graham Bell, Steven Kembel and two anonymous reviewers provided important feedback on an earlier version of this manuscript. Finally, we thank the David Manelli, Charles Normandin and the Gault Natural Reserve.
- Received May 21, 2015.
- Accepted July 27, 2015.
- © 2015 The Author(s)
Published by the Royal Society. All rights reserved.