Connectivity of river habitats: population genetic survey on the effect of river fragmentation on Swiss brown trout
Hitoshi Araki, Armin Peter, Julian Junker, Laura Langeloh EAWAG, Seestrasse 79, Kastanienbaum
Summary
Habitats of wild organisms have been largely disturbed by human activities. River fragmentation is one of them, and Swiss rivers are very often fragmented by artificial barriers including hydroelectric power dams and weirs. Tributaries are often isolated because of bed incision of the main stem due to channelization.
Here we address the effects of artificial barriers on the population genetic structure of two populations of brown trout in Switzerland (Buechwigger in LU and Muemliswilerbach in SO). Based on fine-scale samplings in the two river systems (100–300 m apart from a subpopulation to the other), we found that above-barrier populations are genetically most differentiated from the others, and that they tend to have low genetic variations comparing with below-barrier populations. Altered migratory patterns were ob- served at the individual level, showing little evidence of inter-crosses between above- and below-barrier populations. The Ne/Nc ratio (Ne=effective, Nc=census population size) was 88 % and 50 % smaller for the above-barrier populations than for the below-barrier populations in Buechwigger and Muemliswiler- bach, respectively. However, a similar pattern was also observed in subpopulations without any artificial barrier in Buechwigger, most likely because of a small population size in the above-control population.
Thus, while the results are consistent with the barrier-mediated genetic isolation, further comprehensive investigations are required to generalize our conclusion.
State of the art (pre-ENHANCE)
In literature there is little information available on the population genetics effect as a result of artificial barriers for fish species also occurring in Swiss streams. But there are two prominent studies with salmo- nids, which are relevant for Swiss streams. In Japan, Yamamoto et al. (2004) investigated the effect of dams, which represented absolute barriers to upstream migration, on populations of white-spotted char (Salvelinus leucomaenis) and found consistently lower genetic diversity in populations upstream of dams compared to those downstream of the barriers. As far as population differentiation was concerned, they detected higher levels of differentiation between upstream and downstream populations than between downstream populations in neighbouring rivers. Deiner et al. (2007) similarly found lower levels of allelic richness/genetic diversity in above-barrier populations of rainbow trout (Oncorhynchus mykiss). Compar- ing levels of pairwise differentiation between above-barrier, between-barrier and below-barrier populations showed higher levels of divergence between above- and below-barrier populations within the same tributary than between above- and between-barrier populations. For North American nonsalmonid fish species Bessert and Orti (2008), looking at the effects of dams on populations of the blue sucker (Cyclep- tus elongatus), were also able to detect a significant effect of the impoundments on the genetic structure in the upper Missouri; despite more or less uniform levels of gene diversity, allelic richness was lower in populations above the barriers located at the lowest position in the system. They were, however, unable to find signs of population differentiation between populations on opposite sites of the barriers.
Motivation and research questions
Because of the high fragmentation of Swiss rivers (average migration reach for fishes is only about 650 m long) we were highly motivated to detect genetic differences in populations below and above artificial barriers.
Previous studies indeed suggest that salmonid species in small tributaries with artificial/natural barriers tends to exhibit low levels of genetic variation, although they generally suffer from the lack of appropriate control for proper comparisons to disentangle the effect of physical barriers from that of different size of habitats. Furthermore, there are additional issues that make the evaluation of the effects of artificial barri- ers difficult: Artificial barriers in river systems are often uni-directional, meaning that they are efficient bar- riers against up-migration, but not necessarily down-migration, of the fish species; Intensive fish stocking is occurring world-wide, which might alter the original (natural) population genetic structure. To overcome these difficulties, we compared brown trout populations with artificial barriers to those without in the same river systems, where stocking of trout was abandoned at least 8–10 years ago (3–4 generations for brown trout).
In addition, we took not only a genetic measure of population size (i.e. “effective” population size) but also an ecological measure (i.e. “census” population size) in order to control the effect of different population sizes among the subpopulations under comparisons.
The most important research questions were:
– what is the population characteristics of trout above and below the barriers (density, populations structure)
– what is the genetic variation of the populations above and below the barriers
Foto 1. artificial weir; © Armin Peter.
Foto 2. Brown trout; © Armin Peter.
Materials and Methods
Study populations
Two river systems were selected, Buechwigger (BU) in Canton Luzern and Muemliswilerbach (MU) in Canton Solothurn (Fig. 1). Both systems are tributaries of the Aare River, and they have no record of hatchery fish stocking in the last decade. BU is located upstream of Willisau. A river section is fragmented by a 100m-long culvert, over which paved roads and houses were built. We conducted four sets of sam- pling in BU on July 14th and 15th, 2010, based on the locations of the barrier and a reference stretch within the tributary. BU_AB, BU_BB, BU_AC and BU_BC represent above-barrier, below-barrier, above-control and below-control, respectively (Fig. 1). Each sampling site was c. a. 150 m along the river. Sampling sites were blocked with nets to conduct a quantitative sampling, and two sampling attempts were made per sampling site in BU. In total we collected 801 samples in BU, from which 38–47 individual samples per site were randomly selected for genetic analyses with adipose-fin clip (Table 1).
Buechwigger (BU) in Canton Luzern
25m
25m
Muemliswilerbach (MU) in Canton Solothurn
25m BU_AB BU_BB
BU_CB
BU_CA MU_AB
MU_BB MU_C
1850m apart
River flow
Artificial barrier Sampling site Fig. 1. Sampling sites.
Table 1. Summary statistics. nc: sample size for census estimate; ng: sample size for genetic analyses; Ar: allelic richness;
Ho: observed heterozygosity; He: expected heterozygosity; Fis: Inbreeding coefficient; Ne: estimated effective population size with 95 % Confidence interval based on a jackknife method. * P < 0.05, ** P < 0.01 by Fisher’s exact test of Hardy- Weinberg equilibrium.
nc ng Ar Ho He Fis Nc (95%CI) Ne (95%CI) Ne/Nc
Buechwigger
BU_AB 236 44 5.95 0.674 0.671 -0.0048 387 (352-422) 18.1 (15-22) 0.047
BU_BB 212 38 7.36 0.711 0.724 0.0169 271 (212-343) 109.8 (60-385) 0.405
BU_AC 111 47 4.80 0.679 0.657 -0.0332** 119 (112-127) 4.8 (4-7) 0.040
BU_BC 242 46 7.68 0.741 0.723 -0.0242 260 (243-279) 40.9 (30-59) 0.157
BU-total 801 175 7.69 0.701 0.736 0.0486 1037 173.6 0.167
Muemliswilerbach
MU_AB 319 50 7.40 0.651 0.677 0.0381* 352 (328-376) 49.4 (39-66) 0.140
MU_BB 181 48 7.91 0.686 0.681 -0.0068 211 (181-248) 59.3 (44-85) 0.281
MU_C 55 50 7.79 0.723 0.692 -0.0443** 60 (55-72) 41.7 (33-55) 0.695
MU-total 555 148 8.20 0.687 0.691 0.0063 623 150.4 0.241
Total 1356 323 8.55 0.694 0.729 0.0477 1660 324 0.195
MU is located near Balsthal. A river section is fragmented by a 1 m-tall artificial water drop. Brown trout can pass it only from up- to down-stream. We conducted three sets of sampling on November 11th, 2010 (Fig. 1). MU_AB, MU_BB and MU_C represent above-barrier, below-barrier and control, respectively.
Each sampling site was 123–150 m along the river. Stream reaches were blocked to conduct a quantita- tive sampling, and three sampling runs were made per sampling site. In total we collected 555 samples in MU, from which 48-50 individual samples per site were randomly selected for genetic analyses. Overall, we obtained 1356 fish samples for the census surveys from which 323 fish samples were used for the genetic surveys. Body size distribution is shown in Fig. 2.
No. of samples
Total length (mm)
No. of samples
0 10 20 30 40 50 60 70 80 90
30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200 210 220 230 240 250 260 270 280 290 300 310 320 330 340 350
BU_AB (ave.=78.4 sd=49.5) BU_BB (ave.=102.1 sd=59.6) BU_CA (ave.=153.5 sd=62.5) BU_CB (ave.=109.0 sd=67.4)
0 10 20 30 40 50 60
30 50 70 90 110 130 150 170 190 210 230 250 270 290 310 330 350 370 390 410 430 450 470 490 510
MU_AB (ave.=136.5, sd=53.3) MU_BB (ave.=164.1, sd=92.7) MU_C (ave.=192.7, sd=97.5)
Fig. 2. Body size distribution.
Census population size estimation
Trout population estimation was carried out by the program microfish. Based on the 2 or 3 runs the pro- gram calculates the population estimation and the 95 % confidence interval. We carried out the population estimation for 0+-trout and trout older than 0+ separately. The two groups were identified on a frequency histogram plot.
Genetic analyses
The Chelex extraction method was used for DNA extraction from adipose fin-clip from fish samples, which were collected in a non-invasive manner on the rivers. 14 microsatellite loci were originally used to char- acterize the detailed population genetic structure of brown trout in the two populations. They are Str60, SL438, Ssa110, Ssa85, Str543, Strutta12, Str73, Ssa197, Str591, Str85, SsoSL417, T3_13, Str15 and Str2 (primer information available at www.qub.ac.uk/bb-old/prodohl/TroutConcert/TroutConcert.htm). The forward primers were labeled using four different fluorescent dyes and the primers were divided into two multiplex mixes. The PCR amplification was accomplished in a reaction volume of 10ul containing, 3.5ul molecular grade water 0.2ul primer mix, 5ul Quiagen® Multiplex Mastermix and 1.5 ul DNA. The PCR cycling protocol for both multiplexes was composed of an initial denaturation for 15 min at 94 °C, followed by 35 cycles with 30s at 94 °C, 90s at 54 °C, 90s at 72 °C and a final extension of 30 min at 60 °C. The PCR products were diluted 1:70 and run on an ABI 3130 Genetic analyzer with GenscanTM 600-LIZTM Size Standard (Applied Biosystems) as a reference. Genotypes were determined with the ABI Prism®
Gene MapperTM version 3.0.
Among these loci, Str543 was excluded in the following analyses because in a pilot analysis using CER- VUS v3.0 , this locus showed a significant deviation from the global Hardy-Weinberg equilibrium (HWE, P < 0.001) and a high frequency expectation of null allele (13 %). In all the remaining 13 loci, the expected frequencies of null alleles were <10%, and average frequency of null alleles were 2.7 %. We also ex- cluded 11 individual samples so that all the samples in the following analyses contain the genetic informa- tion from 9-13 loci. Eventually the sample size (n) became 323, and the overall average of the fraction of successfully genotyped samples was 97.6 %.
Population genetic analyses were conducted using Genepop v4.0.10 (available at http://genepop.curtin.
edu.au/), Fstat v2.9.3, Arlequin v3.5.1.3 and STRUCTURE v. 2.3.3 (Prichard et al. 2000). The estimation of effective population size was based on the linkage disequilibrium (LD) and we used LDNe software (Waples and Do 2008). We used the 2 % minimum allele frequency criterion (Waples 2006) and a jack- knife method with 1500 iterations for the 95 % confidence interval measures.
Innovations and main results
Intra-population genetic variation
Summary statistics of the population genetic parameters within each sampling site and population are listed in Table 1. The observed levels of heterozygosity were generally high in both populations (Ho = 0.65–0.72), and the lowest level was observed for above-barrier subpopulation in both systems. When we compare the allelic richness (Ar) and the expected level of heterozygosity, however, the level of genetic variation was the lowest in the above-control subpopulation in BU (BU_AC). The reduced Ar in BU_AC was consistently observed among the loci, suggesting a genome-wide reduction in the genetic diversity in this subpopulation (Table 2). Estimated inbreeding coefficients (Fis) ranged around 0 (-0.044–0.038), in- dicating no strong signature of inbreeding within each subpopulation. However, a significant heterozygote deficiency was detected in MU_AB (P = 0.014 by Fisher’s exact test of HWE), indicating a slight increase of inbreeding in this subpopulation. The other cases of significant deviation from HWE showed excess, rather than deficiency, of heterozygotes (in BU_AC and MU_C, Table 1).
Table 2. Allelic richness.
loci BU_AB BU_BB BU_AC BU_BC BU-total MU_AB MU_BB MU_C MU-total grand-total
Str60 2.00 2.00 2.00 2.00 2.00 2.99 3.49 3.20 3.25 2.73
SL438 5.00 5.78 4.64 4.65 5.29 4.84 4.00 4.94 4.86 5.14
Ssa100 4.97 6.78 4.87 7.59 7.53 6.44 7.11 6.80 6.88 7.59
Ssa85 5.67 4.79 5.28 5.60 6.12 6.98 6.31 5.44 6.70 6.44
Strutta12 11.90 14.94 5.59 14.06 14.32 15.54 14.31 14.86 16.73 17.32
Str73 3.00 3.00 3.00 4.00 3.65 3.00 3.86 3.84 3.60 3.59
Ssa197 5.68 6.54 5.00 6.33 6.59 6.52 7.28 6.39 7.40 7.33
Str591 2.00 2.99 3.00 3.64 3.11 3.00 3.25 2.99 3.16 3.11
Str85 5.67 6.93 5.67 7.61 7.88 5.72 6.59 7.44 6.74 7.66
SsoSL4 6.89 6.79 6.52 9.17 8.41 11.63 11.37 10.97 11.39 10.72
T3_13 8.66 12.65 6.39 12.71 13.02 11.88 13.00 13.49 13.88 14.47
Str15 3.90 5.37 3.99 5.00 5.49 3.63 4.68 3.98 4.19 5.05
Str2 11.99 17.14 6.42 17.45 16.54 14.03 17.55 16.90 17.87 19.97
average 5.95 7.36 4.80 7.68 7.69 7.40 7.91 7.79 8.20 8.55
Inter-population genetic variation
Fst values were calculated to illustrate the inter-population genetic variation (Table 3). Although Fst was less than 0.01 in one case (MU_BB vs. MU_C), the level of inter-population genetic differentiation was all significant by Fisher’s exact test (P < 0.001, meaning that these populations are all genetically distinct).
The pairwise Fst between the two river systems ranged 0.045–0.089, suggesting the moderate-high level of genetic differentiation between the two systems. Within each river system, the Fst distribution varied:
In BU, the pair-wise Fst ranged 0.030–0.128, whereas the range was much lower in MU (Fst = 0.01–0.02, Table 3). When BU_AC and BU_BC are considered as references, the above-barrier subpopulation was consistently more differentiated than the below-barrier subpopulation. Interestingly, however, the level of genetic differentiation between above- and below-control was twice as high as that between above- and below-barrier in BU (Fst = 0.075 vs. 0.037).
Table 3. Between-population genetic differentiations.
Fst BU_AB BU_BB BU_AC BU_CB MU_AB MU_BB
BU_BB 0.037
BU_AC 0.128 0.083
BU_BC 0.084 0.030 0.075
MU_AB 0.089 0.056 0.078 0.063
MU_BB 0.070 0.055 0.077 0.065 0.021
MU_C 0.070 0.045 0.072 0.053 0.020 0.009
Genetic differentiation between all the pairs above was statistically significant (P<0.0001).
A Bayesian analysis of individual assignments using STRUCTURE provided further insights into the population structure of these river systems (Fig. 3). We first conducted the individual assignments in- cluding all genetic samples, which yielded a clear distinction of the genetic composition of the two river populations (Delta-K value best supported number of genetic groups [K] =2, Evanno et al. 2005). We then analyzed the two river populations separately to maximize the power to identify the genetic structure within each system. In BU, the best supported number of genetic group (K) by Delta-K was two, which generally distinguish BU_AB and BU_BB from BU_AC and BU_BC (Fig. 3a). It suggests that c.a. 2 km of the distance has been efficiently isolating these two pairs of subpopulations. The second supported K was four, in which BU_AC and BU_BC were further genetically distinguished. The genetic differentia- tion of BU_AB and BU_BB was rather moderate in K=2-4 (Fig. 3a). At the individual level, however, the STRUCTURE plots suggest that BU_BB has been receiving many down-migrants from BU_AB, but their inter-cross reproduction is either rare or unsuccessful. BU_AB did not contain almost any individual from BU_BB. In the control region, on the other hand, we found up-migrants than down-migrants more often, indicating that without an artificial barrier, brown trout rather tend to recolonize from the main-stem of the river to its branches.
In MU, the best supported K was four, but the overall low genetic differentiation prevented us to draw a clear conclusion from the STRUCTURE plot (Fig. 3b). Nevertheless, it is obvious that MU_AB is geneti- cally most distinct from the others even at the individual level, and that there is little sign of inter-crosses happening between MU_AB and the other subpopulations in MU.
K=3
K=4
BU_AB BU_BB BU_CA BU_CB
BU_AB BU_BB BU_CA BU_CB
BU_AB BU_BB BU_CA BU_CB
artificial barrier
K=2 (statistically best supported) (overall Fst = 0.078)
Fig. 3a: BU.
K=2
K=3
artificial barrier
MU_C
MU_AB MU_BB
MU_C
MU_AB MU_BB
MU_C
MU_AB MU_BB
K=4 (statistically best supported)
(overall Fst = 0.017)
BU & MU MU
BU & MU
BU MU
K=2 (statistically best supported)
0 10 20 30 40 50 60 70 80 90 100
1 2 3 4 5 6 7 8 9 10
K
Delta-K
Fig. 3b: MU, BU & MU.
Effective and census population size
Estimates of effective population size (Ne) and census population size (Nc) are listed in Table 1. Ne var- ied widely in BU (Ne = 4.8–109.8), whereas Ne in MU was rather similar to each other (Ne = 41.7–59.3).
The observed pattern in Ne was consistent with the genetic variation measure using allelic richness (Ar) in general, with an exception of BU_BC in which Ar was the highest among BU subpopulations yet the Ne estimate was intermediate (Ne = 40.9). Results from the pairwise comparisons for above- and below- barriers in the two systems showed reduced effective population sizes above the barriers. But the same pattern was also observed when we compared control-above and -below in BU, and the point estimate of Ne was only 4.8 for BU_AC.
Because the small Ne can be a simple consequence of a small Nc in a subpopulation, we compared Ne and Nc by taking their ratio (Table 1). The general trend was unchanged from the comparisons of Ne among subpopulations. However, in each system the above-barrier and/or the above-control subpopula- tions showed the smallest Ne/Nc ratio (0.04–0.14). Again, the smallest value was observed for BU_AC, but when BU_BC was considered as a control of BU_AC, the reduction in Ne/Nc in BU_AC (74.5 %,
= 1–0.04/0.157) was rather smaller than that in BU_AB for which BU_BB was considered as a control (88.4 %, = 1–0.047/0.405), which may reflect the secondary effect of the artificial barrier between BU_AB and BU_BB. The reduction rate for the above-barrier subpopulation was 50.2 % in MU (MU_AB vs.
MU_BB).
Discussion
Our surveys on brown trout populations in Buechwigger and Muemliswilerbach yielded three major con- clusions: First, the amount of genetic variation in above-barrier populations was consistently lower than below-barrier populations. Second, the above-barrier populations were genetically most distinct from the other populations. Finally, the fine-scale sampling scheme (100–300 m distance between above- and below-subpopulations) enables us to understand the recent events of migration and its consequences:
the STRUCTRE plots revealed that the 100 m-long culvert in BU has been an efficient barrier against the up-migration of brown trout, whereas down-migration is certainly and frequently occurring; nevertheless, we have rather scarce evidence of interbreeding between BU_AB and BU_BB; without an artificial barrier, on the other hand, there was a sign of up-migration activities. Although we admit that our statistical power to draw general conclusions is still limited and we need further surveys on more river systems, our results illustrate the advantages of genetic measures in the field surveys at the fine scale.
One limitation of our survey was that we could not determine the age of the fish samples, which might bias our estimates of the genetic variations and the effective population size. For example, the lowest es- timate of Ne in BU_CA might be caused by a different age structure of the subpopulation, which showed the largest average total length in the system at the sampling dates (Fig. 2). In BU, however, the body size distribution was quite distinct, and we can unambiguously distinguish 0+ fishes from >1+ fishes. And even after eliminating the presumably 0+ fish (< 90 mm in BU), the estimates of the effective population size were very similar to our estimates in Table 1, suggesting that at least the different fraction of 0+ fish in each subpopulation is not the cause of the different Ne among populations.
The result from the two study sites is an interesting contribution how river fragmentation is affecting population genetics of brown trout. We are currently preparing to write a paper for submitting to a peer- reviewed journal like Conservation Genetics”.
References
Bessert, M. L., and G. Orti, 2008 Genetic effects of habitat fragmentation on blue sucker populations in the upper Missouri River (Cycleptus elongatus Lesueur, 1918). Conservation Genetics 9: 821–832.
Deiner, K., J. C. Garza, R. Coey and D. J. Girman, 2007 Population structure and genetic diversity of trout (Oncorhynchus mykiss) above and below natural and man-made barriers in the Russian River, California. Conservation Genetics 8:
437–454.
Evanno, G., S. Regnaut and J. Goudet, 2005 Detecting the number of clusters of individuals using the software STRUC- TURE: a simulation study. Molecular Ecology 14: 2611–2620.
Pritchard, J. K., M. Stephens and P. Donelly, 2000 Inference of population structure using multilocus genotype data. Genet- ics 155: 945–959.
Waples, R. S., 2006 A bias correction for estimates of effective population size based on linkage disequilibrium at unlinked gene loci. Conservation Genetics 7: 167–184.
Waples, R. S., and C. Do, 2008 ldne: a program for estimating effective population size from data on linkage disequilibrium.
Moleclar Ecology Resources 8: 753–756.
Yamamoto, S., K. Morita, I. Koizumi and K. Maekawa, 2004 Genetic differentiation of white-spotted charr (Salvelinus leuco- maenis) populations after habitat fragmentation: Spatial-temporal changes in gene frequencies. Conservation Genetics 5: 529.
