Understanding population-level responses to novel selective pressures can elucidate evolutionary consequences of human-altered habitats. Stream impoundments (reservoirs) alter riverine ecosystems worldwide, exposing stream fishes to uncommon selective pressures. Assessing phenotypic trait divergence in reservoir habitats will be a first step in identifying the potential evolutionary and ecological consequences of stream impoundments. We tested for body shape divergence in four stream-adapted fishes found in both habitats within three separate basins. Shape variation among fishes was partitioned into shared (exhibited by all species) and unique (species-specific) responses to reservoir habitats. All fishes demonstrated consistent significant shared and unique morphological responses to reservoir habitats. Shared responses were linked to fin positioning, decreased body depths and larger caudal areas; traits likely related to locomotion. Unique responses were linked to head shape, suggesting species-specific responses to abiotic conditions or changes to their trophic ecology in reservoirs. Our results highlight how human-altered habitats can simultaneously drive similar and unique trait divergence in native populations.
Revealing how populations respond to rapid environmental change will lend insights into ways human-altered habitats may modify the evolutionary trajectories of populations. Anthropogenic habitat degradation can destine populations to extinction , but altered habitats may also bound future evolutionary processes by limiting available genetic diversity , or modify traits of populations that mediate ecosystem-level processes . Aquatic ecosystems worldwide have been disturbed by stream impoundments [4–6]. Although reservoirs threaten numerous aquatic organisms with extinction [7,8], they are a good system to assess population-level responses to human-altered habitats because they are widespread, can be treated as replicated units and impact a wide-range of taxa.
Lentic habitats created upstream of impoundments are relatively aberrant environments, and likely exert a suite of selective pressures not experienced by stream fishes during their evolutionary history . The physical impoundment, and the standing pool above, can disrupt environmental processes that usually occur along river courses , altering biotic and abiotic components of stream habitats. Strong, atypical selective pressures in these new habitats are evidenced by changes to native stream fish communities (e.g. obligate stream fishes are usually extirpated from reservoir habitats, increased abundances of piscivorous fishes; [11,12]). However, in spite of these pressures, some stream fishes can persist in these man-made habitats, and initial investigations suggest reservoirs may drive contemporary phenotypic divergence in native, resident populations [13,14]. Understanding how phenotypic traits of populations are affected by reservoir habitats will be a first step in identifying potential evolutionary consequences of anthropogenically altered habitats.
The sheer abundance and global distribution of reservoirs  necessitates the potential for these habitats to influence an immense diversity of fishes. A suite of environmental pressures (biotic and abiotic) acting on diverse taxa with various ecologies and evolutionary histories make predicting phenotypic responses of fishes to reservoirs problematic. However, characteristic changes to flow regimes combined with tight linkages between morphology and performance [15–18] may alter phenotypes of stream-adapted reservoir fishes in predictable ways. Fishes in lotic habitats tend to have fusiform body shapes that reduce drag and enable prolonged swimming, whereas shallower anterior/head regions and increased caudal areas in lentic waters facilitates faster burst speeds and increased manoeuverability [15,16,19,20]. Indeed, intraspecific body shape variation investigated in reservoirs and nearby streams corroborated this pattern [13,14]. Yet, these relationships are not ubiquitous as some fishes can exhibit the opposite pattern: more streamlined body shapes in lentic (natural lakes) habitats compared with streams [21–23]. Therefore, species' ecologies and their evolutionary histories will likely regulate how traits respond to selective pressures.
While reduced flow velocities in reservoirs are an obvious abiotic change, there are likely numerous pressures affecting phenotypic variability of stream-adapted fishes in reservoirs. Just as different fishes demonstrate variable phenotypic responses to flow velocity, fishes may have unique responses to other selective pressures. For example, contrary to most streams, reservoirs can experience vertical depth gradients in light availability, temperature and dissolved oxygen concentrations , abiotic factors that can drive phenotypic change in fishes [25–27]. Additionally, changes to substrate sizes coupled with reduction of flow velocities can alter invertebrate fauna in reservoirs , altering prey assemblages available to invertivorous fishes. Thus, different species exposed to the same suite of selective pressures may display both shared and unique responses owing to their individual ecologies and evolutionary histories .
Observed phenotypic shifts in reservoir habitats are potentially attributable to environmental-induced morphological responses (i.e. phenotypic plasticity; [29,30]). The relative contribution of phenotypic plasticity and ‘genetic components’ can be nearly equivalent in explaining phenotypic variation in stream fishes . Nonetheless, environmentally contingent traits can become canalized, where the previous environmental stimulus is no longer required to produce the trait [31,32]. Even plastic responses to altered habitats may then facilitate evolution of resident populations [33,34]. Assessing how fishes respond to reservoir habitats, regardless of whether trait divergence is ‘genetic’ or plastic, will lend insight into the potential evolutionary consequences of impoundments.
Here, we tested for phenotypic divergence in four native stream-adapted fishes established in reservoirs, quantifying the relative amount of morphological diversification seen in each. We partitioned the observed morphological variation between reservoir and stream habitats into shared (i.e. responses exhibited by all species) and unique (species-specific) morphological responses. We predicted all fishes in reservoir habitats would exhibit replicated trait shifts often associated with changes to flow variation (i.e. smaller anterior/head regions, larger caudal areas), whereas within species shifts would reflect responses to changes in the trophic ecology of fishes (i.e. deformations in the head).
2. Material and methods
(a) Study sites and collections
We investigated shape variation in fishes from three basins with impounded rivers in the Hilly Gulf Coastal Plains in northwest Mississippi, USA (see the electronic supplementary material, appendix S1: study map and sample sizes). Impoundment of the Little Tallahatchie River in 1940, the Yocona River in 1952 and the Yalobusha River in 1954 created Sardis, Enid and Grenada reservoirs, respectively (see the electronic supplementary material, appendix S1: study map and sample sizes). All systems historically flowed into the Yazoo River in western Mississippi, and the three basins contain similar fish faunas. Fishes were collected between December 2011 and January 2012 from reservoir habitats by seine and a barge electro-fisher, whereas stream habitats were only sampled with a seine. All stream habitats sampled were upstream of each reservoir with no known physical barrier obstructing migration between reservoir and stream habitats. Fishes were euthanized with an overdose of MS-222, preserved and stored in 10 per cent formalin on site and returned to the laboratory for data acquisition. One or two sites were sampled in each reservoir and several stream sites were sampled in each basin (see the electronic supplementary material, appendix S1: study map and sample sizes), but fishes were opportunistically collected and individuals from sites within each basin and habitat (i.e. stream or reservoir) were combined (see the electronic supplementary material, appendix S1: study map and sample sizes). The distance between reservoir and stream collections within each basin was at least 35 km (Euclidean distance).
We assessed shape variation in four stream-adapted fishes from three families: blacktail shiner Cyprinella venusta, emerald shiner Notropis atherinoides (Cyprinidae), brook silverside Labidesthes sicculus (Atherinopsidae), and bluegill sunfish Lepomis macrochirus (Centrarchidae). We chose these species because of their densities in reservoir habitats, and their disparate evolutionary histories, ecology and morphology (figure 1). In addition, these species are not sexually dimorphic during non-breeding months (i.e. winter); therefore, effects of sex on shape were not investigated. Only juvenile L. macrochirus were collected while all other individuals were adults. All four species were collected from all basin–habitat combinations. We restricted our analyses to individuals of each species that had overlapping size distributions between habitat types and among basins (electronic supplementary material appendix S1: study map and sample sizes).
(b) Geometric morphometrics
We quantified body shape of all specimens using geometric morphometric analyses  with tps software (http://life.bio.sunysb.edu/morph/) and R . We digitally photographed (Canon PowerShot A1100) the left lateral side of each individual with a reference scale, randomized the order of digitized photographs (to reduce potential biases associated with the sequence specimens were photographed and subsequent landmarks placed on them), and set 11 homologous landmarks on each photograph using tpsDig2 software . The 11 landmarks set on photographs for geometric morphometric analyses included: (i) tip of the snout, (ii) corner of the mouth, (iii) centre of the eye, (iv) posterior tip of the supraoccipital process, (v) anterior terminus of the dorsal fin base, (vi) insertion of the last dorsal ray on the caudal fin, (vii) insertion of the last ventral ray on the caudal fin, (viii) anterior terminus of the anal fin base, (ix) anterior terminus of the pelvic fin base, (x) anterior terminus of the pectoral fin base, and (xi) posterior border of the bony opercle and the body outline (see the electronic supplementary material, figure S2). We resized landmark coordinates using the reference scale, and aligned landmark coordinates using a general procrustes analysis (GPA) to remove the effects of scale, translation and rotation on shape variation. Relative warps (hereafter, referred to as shape variables) were calculated and reserved for analyses. Hereafter, all analyses using the aligned specimens of all species will be referred to as ‘global dataset’. We then realigned landmarks (i.e. GPA) for each species separately before calculating shape variables for species-specific analyses. Variation in shape was visualized using thin-plate spline transformation grids in tpsRegr .
(c) Data analyses
We developed our analyses to test specific predictions by assessing the relative contribution of species, reservoir basins and habitats to shape variation. We then visualize shared and unique morphological responses exhibited by fishes to reservoir habitats. We first tested the prediction that species would demonstrate a shared morphological response to reservoir habitats (i.e. some portion of observed shape changes in reservoir habitats would be similar among species), and then visualized this shared morphological response. Second, we tested the prediction that each species would show significant body shape changes in reservoir habitats. Third, we isolated and visualized the nature of species-specific morphological responses to reservoir habitats (i.e. unique responses). Finally, we investigated the magnitude of morphological divergence exhibited by each species in reservoir habitats by assessing how well we could predict the inclusion of individuals into reservoir and stream habitat types based on shape variation.
(d) Shared morphological variation
We tested for shared morphological divergence among species between stream and reservoir habitats with a global model of multivariate analysis of covariance (MANCOVA). All MANCOVA models assume multivariate normality, homogeneity of covariance matrices, independence of observations, linear relationships between covariates and dependent variables, and homogeneity of slopes among groups . The MANCOVA model here included the 18 shape variables (global dataset) as dependent variables, standard length (SL) as a covariate (to test for effects of allometry), habitat type (to test for effects of stream or reservoir habitats), basin (to test for basin level effects) and species (to test for effects of species) as fixed factors. Heterogeneity of slopes was tested among species and basins, and between habitat types by including SL in the respective interaction terms. All non-significant interaction terms were removed from the final model. F-values were approximated using Wilk's lambda. Because of the statistical power associated with using MANCOVA with shape data, we focused our interpretation of model results on effect strengths by use of partial eta squared (ηp2) rather than p-values. We calculated the relative variance as the partial variance for a given term divided by the maximum partial variance value in the model.
To assess the nature of shared morphological divergence among species in reservoir habitats, we calculated a morphological divergence vector as defined by Langerhans  between the two habitat types using the global dataset. This morphological divergence vector does not distort morphological space and summarizes the linear combination of shape variables that contribute to the greatest difference in body shape for a given term of interest (here, reservoir and stream habitats) after controlling for other effects . To quantify this shared divergence vector, we multiplied the eigenvector of the habitat term's sums of squares and cross products matrix from the global MANCOVA (final model same as above) by the shape variables matrix to yield a column of shared divergence vector scores for each individual. Thus, this shared divergence vector summarizes the shape variation that all species elicited in reservoir habitats compared with stream habitats. The nature of shared divergence in reservoir habitats among species was visualized along this vector using thin-plate spline transformation grids.
(e) Unique morphological variation
To test for effects of reservoir habitats on body shape variation within each species, we used four separate MANCOVA models (one for each species). Similar to the global MANCOVA above, 18 shape variables were dependent variables, basin and habitat were included as fixed factors, and SL was a covariate in each model. Heterogeneity of slopes was tested between habitats and among basins by inclusion of SL in each respective interaction term. While these models tested for species-specific effects of reservoir habitats, quantification of a habitat divergence vector from each would not be an assessment of unique divergence of each species because shared variation among species would also be present in each vector. Therefore, to extract only each species' unique morphological divergence to reservoir habitats, we conducted four additional MANCOVA models (similar to above) where the shared divergence vector score of each individual was also included as a covariate in each model. The subsequent habitat divergence vector scores calculated from each model reflected each species' unique response to reservoir habitats (the shared axis of divergence to reservoir habitats was removed via its inclusion as a covariate in each model). We assessed each species' unique divergence in reservoir habitats by visualizing shape deformations along each unique divergence vector using thin-plate spline transformation grids.
To contrast the amount of morphological divergence demonstrated between the two habitat types by each species, we used discriminate function analysis (DFA) to predict inclusion of individuals into reservoir and stream habitat types based on shape variation. Four species-specific preparatory MANCOVAs with 18 shape variables as dependent variables, basin as a fixed factor, and SL as a covariate were used to obtain size- and basin-independent shape variables for each species (i.e. the unstandardized residuals from each MANCOVA). The shared divergence vector was not included in these models. These size- and basin-independent shape variables were then used as the independent variables in each DFA (n = 4). The percent of correctly classified individuals from jackknifed (i.e. leave one out) predictions served as an estimate of the degree of morphological divergence each species demonstrated between stream and reservoir habitats. All analyses were conducted in R, unless otherwise stated .
(a) Shared morphological variation
When testing for morphological divergence in reservoir habitats among all species, all terms in the global MANCOVA had significant effects on body shape of fishes. Of the major terms of interest, species had the strongest effect (ηp2 = 0.94), followed by SL (ηp2 = 0.43), habitat (demonstrating significant shared morphological divergence; ηp2 = 0.22) and the species–habitat interaction (unique morphological divergence; ηp2 = 0.16; table 1). The shared morphological habitat divergence vector among species demonstrated consistent divergence between habitat types in the replicate reservoir basins (figure 1). In most cases, mean shared divergence vector scores of reservoir populations were larger than scores of stream populations in each replicate basin (figure 1). The exception was N. atherinoides in the Sardis reservoir basin where the mean score of reservoir individuals was less than the mean score from stream individuals. In addition, within basin divergence between habitats was generally less variable in C. venusta and L. macrochirus compared with N. atherinoides and L. sicculus. The shared variation among species in response to reservoir habitats resulted in an upturn and decreased depth of the head, posterior and dorsal movement of the pectoral fin, decreased body depth via ventral and anterior movement of the dorsal fin, and increased caudal area via anterior movement of the anal fin (figure 1).
(b) Unique morphological variation
When testing for species-specific responses to reservoir habitats via four individual MANCOVAs, all terms had significant effects on body shape variation in each species (table 1). Habitat had the strongest effect on shape in all species (range of ηp2 0.30–0.51) except for L. macrochirus where SL had a stronger effect (ηp2 = 0.71) than habitat (ηp2 = 0.30; table 1), probably due to the size range of L. macrochirus used compared with the other species (see the electronic supplementary material, table S1). After controlling for shared morphological divergence, unique divergence vectors of each species demonstrated consistent divergence between habitats within the replicate basins (figure 2). Unlike the shared morphological response of all species, unique responses of each species were mostly limited to shape changes in the head. Generally, C. venusta and L. macrochirus had more elongate and shallower heads in reservoir habitats compared with streams. Conversely, N. atherinoides had a slightly more upturned mouth and L. sicculus had a more downturned mouth in reservoir habitats compared with streams (figure 2).
There were substantial and similar amounts of divergence between reservoir and stream habitats among species. The percent of correctly classified individuals between the habitat types was 79.5 per cent , 72.5 per cent , 79.5 per cent , and 81.0 per cent for C. venusta, N. atherinoides, L. sicculus, and L. macrochirus, respectively.
Our results indicated that the four stream-adapted fishes demonstrated consistent and replicated phenotypic divergence in the anthropogenically altered habitats. The shared responses of all species were linked to fin positioning, shallower heads and larger caudal areas, whereas unique responses (species-specific) involved shape deformations in the anterior portion of the body.
The shared morphological shifts observed among species (i.e. decreased head depth and increased caudal area) is consistent with morphologies associated with lower sustained swimming abilities and higher burst-swimming performance in lentic habitats . These shape changes were also qualitatively similar to morphological shifts in reservoirs observed by Haas et al.  and Franssen  in C. venusta and red shiner C. lutrensis, respectively. Based on the diversity of fishes investigated here, such a relatively strong, shared response by all species was rather unexpected (habitat explained nearly twice as much variation as basin in the global MANCOVA). Indeed, variation in body morphologies among species suggests they likely have different swimming mechanics associated with their variable trophic and behavioural ecologies . Thus, our data indicate that most stream fishes likely need to overcome similar resistances to locomotion in standing water in reservoirs versus flowing water in natural stream habitats.
Notropis atherinoides was the only species that did not demonstrate consistent morphological shifts in reservoir habitats (i.e. only two out of the three reservoirs showed consistent divergence; figure 1). The lack of observed divergence in the Sardis basin may also explain why N. atherinoides demonstrated the least amount of overall diversification between the two habitat types as evidenced by the DFA. While we cannot be certain, gene flow between the two habitats may be eroding morphological divergence in this basin. On the other hand, the stream habitats in this basin may have environmental conditions that resemble reservoirs, leading to similar morphologies in the two habitats. More probably, however, the different evolutionary histories of populations in the different basins may influence their phenotypic responses to similar selective pressures among basins . Indeed, the basin × habitat term was significant in three of the four species-based MANCOVAs (indicating unique responses of populations to reservoir habitats in the different basins; table 1).
Most unique morphological responses of stream-adapted fishes in reservoirs suggest they are eliciting different phenotypic responses to similar selective pressures. Many of the phenotypic changes between the two habitat types were linked to shape deformations in or near the head. Changes to the anterior/head of the body could be related to locomotion performance, but may also be affected by abiotic components of the environment (e.g. dissolved oxygen concentrations, light availability; [25–27]). Yet, these fishes are likely restricted to the littoral portions of reservoirs, where oxygen depletion is likely rare and light availability would be comparable to stream habitats. We suspect unique divergences in the anterior portion of fishes may be linked to species' trophic ecology in reservoirs, where the types and availability of prey can vary substantially compared with stream habitats . Morphological divergence linked to changes in trophic ecology has been observed in threespine stickleback Gasterosteus aculeatus populations in lake and stream pairs . Investigations of the trophic ecologies of these fishes may lend insight into the observed unique morphological response of reservoir–resident fishes.
The nature of our data precludes us from quantifying how much phenotypic variability was due to plastic effects. Fishes can demonstrate considerable phenotypic variation in response to different flow velocities [43–46] and plasticity certainly contributed to the some of the patterns observed here. However, natural selection can operate on reaction norms, facilitating adaptive phenotypic change [29,30,33,34]. Further research, potentially through common garden experiments, will be needed to quantify the relative contribution of plasticity versus genetic components to elucidate the contemporary evolution occurring in reservoir–resident populations.
Within basins, any known physical barrier that would hinder gene flow was not present between reservoir and stream populations. High rates of gene flow between reservoir and stream habitats should degrade morphological differences between habitats, if selective pressures are not strong enough to maintain trait divergences [21,47]. The relatively high morphological divergence here suggests either plasticity plays a large component in phenotypic variation or gene flow is not high enough to degrade differences in body shape between habitat types. Reservoir habitats themselves may reduce migration of individuals between stream and reservoir populations owing to their drastically different biotic and abiotic environmental conditions [48,49]. Detailed and replicated assessment of gene flow between reservoir and stream populations would provide insight into the role migration may play in limiting morphological divergence.
Assessing human-facilitated rapid evolution has gained interest in recent years, but understanding the ecological consequences of contemporary evolution on populations and other components of ecosystems has received little attention . To fully understand the potential ecological consequences of rapid evolution of traits, we first need to understand how multivariate phenotypes produced from different genetic architectures respond to human-induced selective pressures. Unique responses by taxa to similar selective pressures may make predicting the ecological consequences of trait evolution in altered habitats difficult. Nonetheless, understanding these dynamics will be a prerequisite for ameliorating the effects of habitat alteration on multiple levels of biological organization and their interactions with the environment.
We thank M. Tobler for thoughtful discussion on statistical analyses. This work was approved by USM IACUC no. 12020902. Collecting permit was issued to N.R.F. by the Mississippi Department of Wildlife, Fisheries, and Parks. Landmark coordinates of all species and habitats are available as electronic supplementary material.
- Received November 15, 2012.
- Accepted November 22, 2012.
- © 2012 The Author(s) Published by the Royal Society. All rights reserved.