Rapid morphological divergence of a stream fish in response to changes in water flow

Abstract

Recent evidence indicates that evolution can occur on a contemporary time scale. However, the precise timing and patterns of phenotypic change are not well known. Reservoir construction severely alters selective regimes in aquatic habitats due to abrupt cessation of water flow. We examined the spatial and temporal patterns of evolution of a widespread North American stream fish (Pimephales vigilax) in response to stream impoundment. Gross morphological changes occurred in P. vigilax populations following dam construction in each of seven different rivers. Significant changes in body depth, head shape and fin placement were observed relative to fish populations that occupied the rivers prior to dam construction. These changes occurred over a very small number of generations and independent populations exhibited common responses to similar selective pressures. The magnitude of change was observed to be greatest in the first 15 generations post-impoundment, followed by continued but more gradual change thereafter. This pattern suggests early directional selection facilitated by phenotypic plasticity in the first 10–20 years, followed by potential stabilizing selection as populations reached a new adaptive peak (or variation became exhausted). This study provides evidence for rapid, apparently adaptive, phenotypic divergence of natural populations due to major environmental perturbations in a changing world.

1. Introduction

A fundamental concept in evolutionary biology is that organismal phenotypes change in response to changes in their environment. The perception that evolutionary change is relatively slow, taking hundreds to thousands of generations is yielding to evidence of phenotypic changes on contemporary time scales [1–3]. Although examples of rapid evolution are increasingly common, only rarely is the precise timing of environmental change and organismal response well resolved outside of the laboratory. Consequently, details of phenotypic divergence at the upper end of the evolutionary rate scale remain unclear in natural populations.

In nature, the strength and direction of selection can vary with annual cycles [4], and phenotypic plasticity may contribute to abrupt patterns of phenotypic change [5]. Habitat alterations, such as dam-formed reservoirs, provide excellent opportunities to study temporal aspects of divergence because they cause temporally defined and permanent shifts from lotic (riverine) to lentic (lake) habitats [6]. In some fishes, populations sampled from lotic and lentic habitats exhibit significant morphological differences: development of a posteriorly deeper body in lake-dwellers than stream-dwelling conspecifics [3,7]. The deeper body phenotype is presumed to be adaptive because it can enhance swimming burst speed and manoeuverability relative to a more streamlined phenotype which may better maintain position in steady current [8,9]. Such adaptive changes could presumably evolve quickly if there is strong selection on one or more of the major or minor quantitative trait loci (QTL) that underlie body shape [10].

We investigated body shape change before and after river impoundment in multiple populations of a widespread stream fish (Pimephales vigilax, Cyprinidae) and assessed the timing of phenotypic divergence in a lentic reservoir environment. Morphometric analyses revealed similar rapid changes in body shape after reservoir construction across all populations. These results highlight the potential for significant phenotypic change over only a few generations in nature.

2. Material and methods

We sampled P. vigilax in the Sam Noble Oklahoma Museum of Natural History (OKMNH) from collections made prior to and after impoundment of seven Oklahoma streams. Samples consisted of collections of fish from reservoirs paired with collections from free-flowing adjacent reaches of the same stream (table 1). We photographed the left lateral side of each specimen and assigned 14 homologous landmarks to each photograph with tpsDIG [11]. We removed variation due to specimen rotation, transformation and scaling using a General Procrustes Analysis, condensed landmarks into fewer meaningful variables using a principal components (PC) analysis and determined the number of relevant PC using the broken stick method [12].

Table 1.

Collection localities, stream–reservoir pair, collection numbers (OKMNH), year of collection (Year Col), number of years following impoundment (Year Imp), rate of change in Haldanes (H), and sample size (n). A negative number in the ‘Year Imp’ column indicates the collection occurred prior to reservoir construction.

locality	pair	OKMNH	Year Col	Year Imp	H	n
Lake Eufaula	1	52419	1992	28	0.043	5
Canadian River	1	36101	1962	−2	n.a.	14
Hulah Reservoir	2	28868	1956	5	0.081	17
Caney River	2	64543	1995	44	n.a.	8
Lake Thunderbird	3	82210	2007	43	0.014	20
Little River	3	31898	1962	−2	n.a.	11
Lake Grand	4	26137	1948	8	0.111	6
Neosho River	4	67419	2001	61	n.a.	30
Lake Wister	5	27119	1955	6	0.160	15
Poteau River	5	75074	2004	49	n.a.	8
Lake Texoma	6	39442	1950	6	−0.068	3
Lake Texoma	6	28108	1953	9	0.058	11
Lake Texoma	6	27631	1954	10	0.029	15
Lake Texoma	6	27782	1955	11	0.130	9
Lake Texoma	6	30155	1956	12	0.055	9
Lake Texoma	6	30031	1958	14	0.077	19
Lake Texoma	6	31145	1959	15	0.063	25
Lake Texoma	6	39699	1962	18	0.027	19
Lake Texoma	6	39513	1963	19	0.038	17
Lake Texoma	6	40615	1971	27	0.039	20
Lake Texoma	6	44149	1989	45	0.024	18
Lake Texoma	6	48068	1993	49	0.027	11
Lake Texoma	6	63013	1995	51	0.020	23
Lake Texoma	6	60315	1999	55	0.017	20
Lake Texoma	6	62031	2000	56	0.019	23
Red River	6	80558	2010	66	n.a.	30
Lake Oologah	7	54061	1993	30	0.061	10
Verdigris River	7	28614	1956	−7	n.a.	20

We compared the first PC, which was identified as the only PC of interest (electronic supplementary material, figure S1), using a mixed-effects model. Since we were specifically interested in comparing body shape of fish from stream and reservoir habitats, we treated habitat (stream or reservoir) as a fixed factor and location (stream–reservoir pair) and the habitat × location interaction as random factors after accounting for size allometry (centroid size) [13]. Because of the relatively large number of Lake Texoma samples relative to the other populations, we used only one randomly selected sample to represent Lake Texoma (OKMNH no. 40615) in this model. Significance of each term was determined using a χ²-test and the amount of change of each reservoir population was quantified relative to the stream population in Haldanes [2].

Because we had multiple reservoir samples for Lake Texoma, we compared PC1 of these samples across years (treated as a random effect) after accounting for allometric effects using a mixed-effects model. The overall rate of change in Lake Texoma was estimated in Haldanes using the regression approach [2]. To assess the timing of phenotypic change in the Lake Texoma population, body shape was related to the amount of time the population experienced lentic conditions (time since dam construction). We accounted for unequal sample sizes among collections by bootstrapping the body shape data for each population 100 times and performed regressions on each bootstrap replicate using the average F-statistic, p-value and correlation coefficient. We calculated the standardized selection differential (i) to determine whether the resulting pattern was due to a reduction in the strength of directional selection over time [14]. Finally, we estimated the index of stabilizing selection (j) to determine whether a reduction in directional selection was associated with potential stabilizing selection (j < 0) ([14], but see [15]). All statistical analyses were performed in R v. 3.0.2 [16].

3. Results

PC1 explained 20.2% of the variation in landmarks with notable shifts in the terminality of the head, the location of the dorsal and pelvic fins, and body depth (figure 1). Size did not account for a significant portion of the variation in body shape (χ² = 0.155, p = 0.694). There was a significant habitat × location interaction (χ² = 7.673, p = 0.006), but reservoir pair was not a significant random effect (χ² = 0.265, p = 0.607). PC1 varied in parallel across all stream–reservoir pairs (χ² = 16.450, p < 0.001), with reservoir populations having larger scores than stream populations (figure 2a). The rate of change varied from −0.068 to 0.160 H across reservoirs, but tended to decline over time in the Lake Texoma population (table 1).

Figure 1.

Illustration of body shape differences among stream (a) and reservoir (b) fish by thin plate spline visualization of variation along PC1. Stream fish tended to have more negative PC1 scores than reservoir fish with notable change in the terminality of the head, the location of the dorsal and pelvic fins and body depth.

Figure 2.

(a) Comparison of morphological differences (PC1) for the stream (filled circles) and reservoir (open circles) fish for seven stream–reservoir pairs. The circle with cross indicates the reservoir samples from Lake Texoma. Error bars are 1 s.e. (b) Average PC1 scores of Red River and Lake Texoma populations plotted by the time under reservoir conditions. The regression line (solid) and upper and lower 95% CIs from bootstrap resampling (dotted lines) are shown. (c) The standardized selection differential (filled circles) and index of stabilizing selection (empty circles) and their regression lines for each Lake Texoma population plotted by the time under reservoir conditions.

After accounting for allometric effects in the Lake Texoma populations (χ² = 49.124, p < 0.001), body shape still varied significantly across collection years (χ² = 10.900, p = 0.001). Specifically, PC1 increased logarithmically with the number of years following river impoundment, as body shape was deeper in fish experiencing lentic conditions (F_1,13 = 6.485, p = 0.043, R² = 0.319; figure 2b). The average rate of change in the Lake Texoma population was 0.0129 ± 0.0125 (2 s.e.), an estimate lower than the rate of change at any particular time in Lake Texoma. Standardized selection differentials indicated that directional selection tended to increase through the first 20 generations before levelling off (log-transformed: F_1,13 = 7.336, p = 0.018, R² = 0.312; figure 2c). The index of stabilizing selection exhibited a marginally significant quadratic relationship with time suggesting that potential disruptive selection increased from 10–30 years and weakened thereafter (F_1,13 = 3.542, p = 0.062, R² = 0.266; figure 2c).

4. Discussion

We investigated the pattern and tempo of body shape divergence following river impoundment in P. vigilax. We show that morphological changes, including head size and shape, dorsoventral body depth, fin positions and caudal peduncle thickness, occurred in all populations after abrupt changes in flow regime. The response of each population was similar in direction, indicating common solutions to a similar selective pressure [3]. The highest rate of change observed in this study (0.160 H, Lake Wister) is comparable to rates of change observed in quantitative traits of other organisms, e.g. bill length in American house sparrows [2]. Time-series data for Lake Texoma indicate the highest rate of change occurred in the sample at year 11, after which rates of change declined substantially. The decline could be indicative of exhaustion of genetic variation in QTLs that underlie body shape. Alternatively, inference of standardized selection differentials suggest this pattern may be due to directional selection in the first 10–30 years followed by potential stabilizing selection which maintained the population on a new adaptive peak in the lotic environment [17].

Phenotypic plasticity may have contributed to the initial shape change. In a related study, we demonstrated that stream-derived juvenile P. vigilax, when experimentally raised in standing water, develop significantly deeper body morphs than adults from their source population (electronic supplementary material, figure S2). However, the magnitude of that change was only a small fraction of the total change observed in this study. Phenotypic plasticity is the ability of one genotype to produce more than one phenotype under different environmental conditions [5] and, by itself, would not account for incremental change observed over several generations. However, phenotypic plasticity accompanied by directional selection may best explain this pattern of divergence. We note that in this case positive selection could be acting on DNA sequence variation or it could act to increase phenotypic plasticity in the direction of selection [18] (possibly including epigenetic variation).

The timing and pattern of morphological divergence in P. vigilax appears to be best explained by an initial shift due to phenotypic plasticity followed by rapid but none-the-less incremental and adaptive change in response to the shift to a standing-water environment. Differences in the rate and magnitude of phenotypic responses, as well as variation among stream–reservoir pairs, may have been influenced by the extent of genetic variation present in each population as well as the strength of selection based on the local flow rate, predators and food types present at each locality. The retrospective analysis is consistent with plasticity initially maintaining viable populations in the standing-water environment until adaptive evolution can proceed. Characterization of QTL or epigenetic loci that underlie body shape variation in P. vigilax, and fishes in general, will have important implications for our understanding of adaptive evolution and illuminate the potential responses of organisms to a rapidly changing world.

Data accessibility

Data deposited in the Dryad Digital Repository: http://dx.doi.org/10.5061/dryad.s90j5.

Funding statement

This study was funded by the Graduate Student Senate and a Department of Biology Adams Scholarship at the University of Oklahoma, and a Rosemary Grant award from the Society for the Study of Evolution.