Biological Invasions

, Volume 10, Issue 8, pp 1229–1242

Red shiner invasion and hybridization with blacktail shiner in the upper Coosa River, USA

  • David M. Walters
  • Mike J. Blum
  • Brenda Rashleigh
  • Byron J. Freeman
  • Brady A. Porter
  • Noel M. Burkhead
Original Paper

DOI: 10.1007/s10530-007-9198-6

Cite this article as:
Walters, D.M., Blum, M.J., Rashleigh, B. et al. Biol Invasions (2008) 10: 1229. doi:10.1007/s10530-007-9198-6

Abstract

Human disturbance increases the invasibility of lotic ecosystems and the likelihood of hybridization between invasive and native species. We investigated whether disturbance contributed to the invasion of red shiner (Cyprinella lutrensis) and their hybridization with native blacktail shiner (C. venusta stigmatura) in the Upper Coosa River System (UCRS). Historical records indicated that red shiners and hybrids rapidly dispersed in the UCRS via large, mainstem rivers since the mid to late 1990s. We measured the occurrence and abundance of parental species and hybrids near tributary-mainstem confluences and characterized populations at these incipient contact zones by examining variation across morphological traits and molecular markers. Red shiners represented only 1.2% of total catch in tributaries yet introgression was widespread with hybrids accounting for 34% of total catch. Occurrence of red shiners and hybrids was highly correlated with occurrence of blacktail shiners, indicating that streams with native populations are preferentially colonized early in the invasion and that hybridization is a key process in the establishment of red shiners and their genome in new habitats. Tributary invasion was driven by post-F1 hybrids with proportionately greater genomic contributions from blacktail shiner. Occurrence of red shiners and hybrids and the relative abundance of hybrids significantly increased with measures of human disturbance including turbidity, catchment agricultural land use, and low dissolved oxygen concentration. Red shiners are a significant threat to Southeast Cyprinella diversity, given that 41% of these species hybridize with red shiner, that five southeastern drainages are invaded, and that these drainages are increasingly disturbed by urbanization.

Keywords

Land use Turbidity Hybrid swarm Introgression Southeastern fishes Disturbance 

Abbreviations

UCRS

Upper Coosa River System

GMNH

Georgia Museum of Natural History

PCR-RFLP

Polymerase chain reaction restriction fragment length polymorphism

mtDNA

Mitochondrial deoxyribonucleic acid

bp

Base pair

Introduction

Habitat disturbance facilitates fish invasion in lotic ecosystems (Gido and Brown 1999; Marchetti et al. 2004; Moyle and Light 1996) and increases the likelihood of hybridization between fishes (see studies reviewed in Rhymer and Simberloff 1996). Hybridization is often an overlooked or under-appreciated aspect of species invasion in aquatic ecosystems, but the process can lead to the loss of locally adapted gene complexes and the genetic extinction of native species (Hitt et al. 2003; Rhymer and Simberloff 1996). Hybridization can have the perverse effect of enhancing invasion success (Allendorf et al. 2001; Ellstrand and Schierenbeck 2000; Hitt et al. 2003; Rhymer and Simberloff 1996) by mitigating the constraint of propagule pressure, a key factor in successful colonization by invasive fishes and other species (Kolar and Lodge 2001; Ruesink 2005).

Red shiners, Cyprinella lutrensis, have been introduced into at least five Southeastern United States drainages since the 1970s, presumably through bait-bucket or aquarium releases (Fuller et al. 1999). Red shiners thrive under harsh conditions (e.g., low flow, high turbidity, poor water quality) and aggressively colonize severely degraded habitats (Cross and Cavin 1971; Matthews 1985; Matthews and Hill 1977, 1979). Introduced populations spread rapidly, often displacing native congeners and other Cyprinids (Greger and Deacon 1988; Minckley and Deacon 1968; Moyle 2002). They readily hybridize with congeners, sometimes causing widespread displacement of native species (Larimore and Bayley 1996; Page and Smith 1970). Hybridization is a significant threat to Southeastern Cyprinella diversity, as red shiner hybrids have been reported for nine native species and subspecies (DeVivo 1996; Hubbs and Strawn 1956; Johnson 1999; W.C. Starnes personal communication; Burkhead unpublished data; Page and Smith 1970; Wallace and Ramsey 1982).

Prior studies hypothesized that habitat disturbance increases both the likelihood of red shiner colonization and the likelihood of hybridization with congeners (Hubbs et al. 1953; Larimore and Bayley 1996; Page and Smith 1970). Examples of degraded lotic habitats colonized by red shiners include drainage ditches and severely altered agricultural and urban streams (DeVivo 1996; Moyle 2002; Page and Smith 1970). Likewise, hybrid swarms involving red shiner in Texas and Illinois were attributed to poor water quality and high turbidity (Hubbs et al. 1953; Hubbs and Strawn 1956; Larimore and Bayley 1996; Page and Smith 1970). However, these prior accounts linking disturbance with red shiner invasions were observational, and disturbance hypotheses remain untested.

We investigated red shiner colonization and hybridization with native blacktail shiner, C. venusta stigmatura, in the Upper Coosa River System (UCRS, Georgia, Alabama and Tennessee). We have observed rapid dispersal of red shiners and hybridization with blacktail shiners in mainstem rivers of the UCRS since the 1990s. Along with establishing a timeline of invasion, we measured the occurrence and abundance of parental species and hybrids near tributary-mainstem confluences. We then characterized the genetic composition of populations at these incipient contact zones by examining variation across morphological traits and molecular markers. This approach enabled us to determine whether the occurrence of native congeners inhibits (e.g., via competition) or enhances (e.g., via hybridization) dispersal of red shiners into new habitats. Finally, we tested the hypotheses that disturbance (assessed at the basin and reach scales) promotes red shiner dispersal and hybridization with congeners.

Methods

Invasion history and site selection

We used an extensive database of fish distributions maintained by the Georgia Museum of Natural History (GMNH) to develop a chronology of red shiner and C. lutrensis × C. venustastigmatura hybrids (hereafter hybrids) dispersal in the UCRS. All specimens, including putative hybrids, were identified on the basis of morphological characters.

Collection records indicated that red shiners or hybrids were distributed in mainstem rivers of the UCRS including Coosa River upstream of Weiss Lake, Oostanaula River, and downstream reaches of the Etowah, Conasauga and Coosawattee rivers (Fig. 1a). The sample population of tributaries entering these mainstem rivers included 43 second, third, and fourth order streams. Sample reaches within streams were located at the first or second most downstream road crossing, since reaches closer to mainstem populations are most likely to be colonized first. Sites on 10 streams sites were inaccessible (e.g., too deep for wading), so 33 of the 43 available streams were sampled.
Fig. 1

(a) Historical spread of red shiners and C. lutrensis × C. venusta hybrids in the Upper Coosa River System. Key locales and dates illustrate the earliest known collections in the system and the rapid upstream migration of phenotypic red shiner and hybrids. For illustration purposes, 2001–2003 records for the Conasauga River only include the upstream-most site where red shiner or hybrids were collected. (b) Extent of fish collections 1993–1998 (when the hybrid swarm was first discovered) and 1998–2001 (corresponding with the period of rapid upstream dispersal). Blacktail shiners are widely distributed in 3rd order and larger streams, but red shiners and hybrids were collected at only three locales upstream of Lake Weiss prior to 1998 (see a). No red shiners or hybrids were collected in the Coahulla Creek or Conasauga River systems from 1993 to 1998, suggesting that the red shiners and hybrids did not radiate southward from a northern introduction. Samples collected from 1998 to 2001 confirm that red shiners and hybrids upstream of Weiss Lake were limited to the mainstem Oostanaula and Conasauga rivers

Collection and characterization of fishes

Reach-length for fish sampling was scaled to 25× stream width, and all reaches included both riffle and pool habitats. Representative habitats were sampled using a total of 30 kick-sets and/or seine hauls along with a backpack electrofisher. Samples were collected between June and August 2005, and all Cyprinella were anesthetized and preserved in 95% ethanol.

Genetic characterization

DNA extraction and mtDNA RFLP assay

Genomic DNA was extracted from ∼0.05 g of preserved fin tissue from each specimen using DNeasy kits (Qiagen, Inc., Valencia, CA). Approximately 10–50 ηg of DNA was then used as template for 15 μl polymerase chain reaction (PCR) mixtures that also included 2.5 mM MgCl2, 2.5 mM each dNTP, 0.5 units Taq DNA polymerase (Invitrogen, Carlsbad, CA), 0.5 μM each of a pair of oligonucleotide primers and PCR buffer (Invitrogen, Carlsbad, CA) to a final 1× concentration. The complete cytochrome b gene (1,140 bp) was amplified with primers HA and LA as described in Schmidt et al. (1998) under a thermal regime of 35 cycles of 94°C for 30 s, 49°C for 30 s, and 72°C for 90 s, followed by a final extension stage at 72°C for 5 min with a MJ Dyad thermocycler (MJ Research, Inc., Waltham, MA).

Prior to PCR amplification of the cytochrome b gene from specimens in this study, we examined cytochrome b sequence variation in each species from an alignment of sequences obtained from 34 blacktail shiners from the upper Conasauga River (TN), Raccoon Creek (Etowah River, GA) and Sipsey Creek (Black Warrior River) as well as sequences from 26 red shiners from Peachtree Creek (Chattahoochee River, GA), and from the Canadian River (OK) (Blum et al. unpublished data). All individuals included in this alignment came from areas unaffected by hybridization. These data revealed that HinfI restriction of cytochrome b amplified from blacktail shiners generate ∼130, 480 and 530 bp fragments versus ∼95, 130, 350 and 570 bp fragments from red shiners. These differences formed the basis for a PCR-RFLP approach to establish species-level mtDNA ancestry of specimens in this study. The PCR-RFLP approach involved HinfI restriction digestion of each cytochrome b PCR amplicon as recommended by the enzyme manufacturer (New England Biolabs) and scoring fragment size profiles by agarose gel electrophoresis of the restricted amplicons. To ground truth the approach, fragment size profiles were validated against cytochrome b sequences from 20 blacktail shiners from the upper Conasauga River and 20 red shiners from Peachtree Creek. Fragment size profiles were subsequently obtained for all specimens collected for this study.

Microsatellite PCR amplification and analysis

Individuals were genotyped at seven polymorphic microsatellite markers developed for other target species. We used the following loci (modified annealing temperature given in parentheses): Can6EPA (53°C) developed for Campostoma anomalum (Dimsoski et al. 2000); Nme 25C8.208 (55°C), Nme 18C2.178 (52°C), Nme 24B6.191 (57°C), and Nme 24B6.211 (57°C) developed for Notropismekistocholas (Burridge and Gold 2003); and Rhca20 (54°C) and Rhca24 (52°C) developed for Rhinichthys cataractae (Girard and Angers 2006). PCR mixtures for amplifying microsatellite loci were identical to those designed for amplifying the cytochrome b gene. The PCR regime for all loci was 25 cycles of 60 s at 95°C, 60 s at the locus-specific annealing temperature, and 90 s at 72°C, followed by a final extension stage of 7 min at 72°C. All reactions were run on an MJ Research Dyad with fluorescently labeled forward primers. Labeled PCR amplicons were characterized using a MJ Research Basestation Genetic Analyzer and Cartographer© software.

Microsatellite allelic variation was analyzed using Structure v2.2 to construct a multi-locus admixture profile for all specimens (Falush et al. 2007). Few putative C. lutrensis specimens were collected, so admixture profiles were based on an expanded dataset that included additional “learning samples” (Montana and Pritchard 2004) of 50 C. lutrensis from Proctor Creek (Chattahoochee River) and genotyped for a related study (Blum et al. unpublished data). For additional comparison, 34 C. venusta collected from the upper Conasauga River were also included as learning samples. After several intermediate-length trial runs, we chose a burn-in period of 30,000 iterations and collected data from an additional 106 iterations for five replicate runs where K (the number of populations) was set at two, which is representative of the two parental species potentially contributing to the ancestry of each specimen. Each run was parameterized following a model of admixture and correlated allele frequencies, and average assignment values to each cluster were subsequently calculated for all specimens. Admixture categories (which loosely conform to expected admixture proportions of Mendelian inherited traits) reflected average assignment values to the first cluster based on the following ranges of values: (i) red shiner, 0.90–1.0; (ii) backcross to red shiner (BCr) for values 0.76–0.89; (iii, iv) F2 hybrids for values of 0.26–0.39 and 0.61–0.75; (v) F1 hybrids for values 0.40–0.6; (vi) backcrosses to blacktail shiner (BCbt), 0.11–0.25; and (vii) blacktail shiner, 0–0.10.

Morphological characterization

We characterized individuals as red or blacktail shiner or hybrids based on three traits, caudal spot intensity, number of lateral line scales, and ratio of body standard length to depth (length:depth), that distinguish red and blacktail shiners (Boschung and Mayden 2003). The caudal spot is large and intense in blacktail shiners but absent in red shiners. Blacktail shiners have more lateral line scales (36–48 vs. 33–36 for red shiners), and a higher length:depth ratio. Caudal spot intensity was assessed on a phenetic scale scored from zero (absent) to two (intense), with a faint or muted spot scored as a one. Hybrids showed morphological intermediacy or incongruency (i.e., morphological traits of both parental species). For example, a hybrid might have a high lateral line scale-count and length:depth ratio (blacktail shiner traits), but lack a caudal spot (red shiner trait). We did not assess morphology of individuals <30 mm SL (n = 10) due to the difficulty of counting scales on juveniles.

Integrative characterization

Summary determinations of mixed ancestry for individuals ≥30 mm SL reflected incongruence or intermediacy of an individual’s morphology, mtDNA haplotype and microsatellite genotype. Individuals <30 mm SL were characterized using only mtDNA haplotype and nuclear genotype. Individuals were classified as a red or blacktail shiner if they demonstrated complete agreement across mtDNA haplotype, nuclear genotype, and phenotype for that species. Individuals were classified as hybrids if they demonstrated incongruence across these three categories. For example, a hybrid individual could exhibit blacktail shiner morphology and multi-locus microsatellite genotype, but express a red shiner mtDNA haplotype. Similarly, hybrid classification extended to individuals with a hybrid phenotype, but with blacktail mtDNA profile and microsatellite genotype. We separated hybrids into three categories: putative F1 hybrids, and later generation hybrids (e.g., F2 and backcross categories) with microsatellite asymmetry favoring red (Hr) or blacktail shiner (Hbt) traits.

Environmental characterization

Environmental variables and summary values are provided in Table 1. Baseflow turbidity was measured on three occasions using a hand-held turbidity meter (Hach 2100P). Baseflow conditions were met if no rain was observed in the region for the preceding 72 h and if gauges on nearby streams indicated stable, low flow. Turbidity measurements were made at least 2 weeks apart from May through October, which overlaps with the spawning seasons for red and blacktail shiner. Conductivity, pH and dissolved oxygen concentration were measured once using a Hydrolab Datasonde 4a.
Table 1

Environmental variables collected from mainstem tributaries in the Upper Coosa River System

Acronym

Variable description

Units

Transform

Range

Mean (1SD)

Landscape

DTMS

Distance to mainstem

m

log

0.2–7.8

2.3 (1.6)

DA

Drainage area

km2

log

4.6–456.3

46.7 (80.6)

URB

Basin urban area

%

Arcsine square-root

3.1–56.1

15.3 (13.2)

FOR

Basin forested

%

Arcsine square-root

18.7–86.7

55.8 (16.3)

AG

Basin agriculture

%

Arcsine square-root

2.9–67

21.5 (12.9)

IA

Impervious area, 1-km radius

%

Arcsine square-root

0.2–26.1

4.1 (6.5)

Reach habitat

GR

Gradient

%

Arcsine square-root

0.01–1.4

0.3 (0.3)

W

Average width

m

log

2.8–12.5

6.4 (2.5)

V

Maximum velocity

m s−1

log

0.03–1.12

0.57 (0.26)

V-m

Mean velocity

m s−1

log

0–0.3

0.11 (0.1)

D

Average depth

m

log

0.2–1

0.4 (0.2)

PS

Average particle size

phi

log2

1.6–8

4.1 (1.3)

SAND

Proportion sand

%

Arcsine square-root

0–59

12.8 (15.1)

GRAV

Proportion gravel

%

Arcsine square-root

4.8–96.3

62.5 (22.8)

COBB

Proportion cobble

%

Arcsine square-root

0–74

9.5 (18.2)

BR

Proportion bedrock

%

Arcsine square-root

0–67.9

15.2 (18.5)

LWD

Proportion large woody debris

%

Arcsine square-root

0–66

13.3 (13.7)

RH

Proportion of riffle habitat

%

Arcsine square-root

6–56.2

28.1 (14.7)

CV-D

CVa depth

%

None

31–99.5

61.2 (18.2)

CV-V

CV velocity

%

None

0–612.3

147.2 (117.1)

OC

% open canopy

%

Arcsine square-root

1.8–72.7

27.1 (19.5)

Water quality

CON

Conductivity

μS cm−2

log

88–573

232.8 (84.8)

NTU

Mean turbidity

NTUb

log

1.4–25.1

6.6 (5.3)

DO

Dissolved oxygen

mg l−1

log

2.4–9.7

7.1 (1.8)

pH

pH

NA

None

3.8–8.7

7.8 (0.8)

aCoefficient of variation

bNephelometric turbidity units

Depth, velocity, particle size, and large woody debris (LWD, >10 cm diameter) were measured along the centerline of the stream at intervals equal to 0.25 times the channel width (Walters et al. 2003b). The centerline transect length was 25× channel width, with n = 100 observations per site. Mean width was calculated from five randomly selected locations within the first 100 m of the reach. Velocity was measured at 60% depth using a velocity meter (Marsh-McBirney Flo-Mate). Maximum velocity was also measured in the area of highest flow observed within the reach. Substrate particle size was determined by visually estimating the dominant particle size class in a 50 cm diameter patch at each sample point (Walters et al. 2003b). Size classes were based on the phi scale and values were recorded as whole phi intervals (−log2 of intermediate axis in mm). These size classes were used to calculate the percentage of major substrate types (e.g., sand) for the reach. Each sampling point was categorized as either riffle or pool habitat. The presence of LWD in the cross section of the stream perpendicular to the centerline at the sampling point was also noted. Percentage open canopy was measured using a spherical densiometer at every 20th sampling location. Stream gradient was measured using an electronic total station.

Spatial analysis

Land use for the basin upstream of sites was calculated from a 1998, 18-class land-use layer (NARSAL 2005a). Classifications were further grouped into six classes: open water, urban (including high- and low-intensity urban and transportation areas), forest (including evergreen, deciduous and mixed forests), cleared land (including clear cuts, bare rock and quarries), agriculture (including pasture and row crops), and wetlands (including forested and non-forested wetlands). Only the three dominant land use types, urban, forest, and agriculture, were considered for analyses. For basins that extended beyond the Georgia border, data were patched in from the 1992 National Land Cover Dataset (Multi-Resolution Land Characteristics Consortium 1992). Percent of impervious surface area (ISA) in a 1 km radius surrounding the sampling location was also measured to provide an indicator of local land use conditions. ISA was calculated from 2001 color infrared photos with 1 m resolution (NARSAL 2005b). Basin area and distance from the sample reach to mainstem tributaries were calculated using ArcMap 9.0 (ESRI, Redlands, CA).

Statistical analysis of distributional patterns

Exploratory analysis was conducted to prepare data for regression analysis. We examined variables for normality and skewness prior to regression analysis (SAS, SAS Institute, Cary, NC), and non-normal variables were transformed accordingly (Table 1). Three variables (pH, CV-V, IA) were excluded from analyses due to extreme outliers (>3 interquartile ranges away from either the sample 25 or 75th percentiles, (Jongman et al. 1995)). We calculated Spearman correlations among variables and retained only those correlated at |ρ| < 0.80 in order to reduce multicollinearity (Glantz and Slinker 1990). Drainage area (DA) and width (W) were highly correlated (ρ = 0.89), and W was retained. Particle size (PS) was highly correlated with percent sand (SAND, ρ = −0.85), and SAND was retained.

We used the remaining 20 environmental variables to develop multiple logistic regression models predicting the probability of occurrence for red shiner, blacktail shiner, and hybrids. We considered all models containing ≤3 variables to avoid overfitting models relative to sample size (Burnham and Anderson 2002). Models were constructed using variables within each of three variable classes (landscape, reach, water quality) both alone and in combination, for a total of seven categories of models. We used Akaike’s Information Criterion corrected for small sample size (AICc) to assess model goodness-of-fit, where smaller AICc indicates a more parsimonious model (Burnham and Anderson 2002). For each model i, difference (Δi) was calculated between the model’s AICc and the minimum AICc value within the set. Values of Δi < 2 are considered the most parsimonious (Burnham and Anderson 2002). A weight (wi) was calculated for each model according to Burnham and Anderson (2002). The weight can be interpreted as the probability that model i is the best model within the set of models considered. We calculated the percent of correctly predicted presence, absence, and overall occurrences. In order to examine the intensity of hybridization relative to environmental variables, we used multiple linear regression to relate hybrid shiner abundance to environmental variables for the subset of sites that contained hybrids, blacktail shiners, or both (n = 18). We used the AIC approach for model selection and reported the r2 of the final model.

Results

Time-line of red shiner invasion of the Upper Coosa River System (UCRS)

Red shiners were first collected in Weiss Lake in 1974 (Fig. 1a) with additional populations collected in Terrapin Creek, a tributary of the “Dead River” arm of the Coosa River, in 1982. The earliest records upstream of Lake Weiss were two hybrids collected in 1992 in Coahulla Creek and a single red shiner collected in 1993 from the lower Etowah River. Collections in the Oostanaula River in 1998 revealed an extensive hybrid swarm extending upstream from Lake Weiss to the confluence with the Conasauga River. The hybrid swarm rapidly dispersed into the Conasauga River between 2000 and 2003. We conducted annual surveys of the Conasauga River from 2000 and 2003 to map the upstream extent of the hybrid swarm and documented that it extended 31 river km upstream August between August 2000 and August 2001 alone. Four fish collections were made in the Oostanaula and lower Conasauga rivers between 1993 and 1997 (Fig. 1b). Blacktail shiners were collected at three of these sites, but neither red shiners nor hybrids were present, suggesting that most of the upstream dispersal of red shiners in the UCRS has occurred since the mid-to-late 1990s. A 2005 survey confirmed that a hybrid swarm currently extends from Lake Weiss north to Dalton (M. Blum and B. Porter unpublished data).

The absence of red shiners and hybrids from localities between Coahulla Creek and Lake Weiss from 1974 to 1992 suggests the potential for separate introductions in the southern and northern parts of the UCRS. Samples collected from 1993 to 1998 from Coahulla Creek, Conasauga River, and their tributaries failed to uncover additional specimens in the northern part of the UCRS prior to the upstream expansion of the swarm from the Oostanaula River in 2000 (Fig. 1b). We draw four conclusions from historical records: (1) the spread of red shiners and hybrids in the UCRS likely began from the southern part of the UCRS in the vicinity of Weiss Lake; (2) red shiners and hybrids are dispersing upstream in the system via large, mainstem rivers; (3) much of the dispersal has occurred since the mid-to-late 1990s; and (4) the rate of dispersal is high (up to 31 km y−1).

Genetic and morphological assessment of populations

Phenotype often conflicted with mtDNA haplotype (results not shown) or microsatellite genotype. Six individuals had red shiner phenotype, but only five exhibited a red shiner mtDNA haplotype. A slightly smaller proportion of mismatches was found for blacktail shiner. Of the 384 specimens with blacktail shiner phenotype, 40 exhibited red shiner mtDNA haplotype. This is suggestive of marginal asymmetry (in terms of genomic contribution) in favor of blacktail shiners although only eight of 19 individuals with hybrid phenotype exhibited a blacktail shiner mtDNA haplotype. A stronger trend toward asymmetry favoring blacktail shiners was found when comparing phenotype to multi-locus genotype (Fig. 2a). All individuals with red shiner phenotype fell within the parental red shiner genotype cluster (RS), but phenotypic hybrids and blacktail shiner specimens also fell within this category. A comparatively smaller proportion of individuals with hybrid phenotypes exhibited parental blacktail shiner genotypes, and a disproportionate number of individuals with blacktail shiner phenotypes had intermediate genotypes (e.g., F1).
Fig. 2

Individual-based comparison of (a) phenotype to microsatellite multi-locus genotype and (b) mtDNA haplotype to microsatellite multi-locus genotype. Categorical assignment of genotypes shown along the x-axis are red shiner (RS), backcross to red shiner (BCr), F2, F1, backcross to blacktail shiner (BCbt), and blacktail shiner (BTS)

Comparison of mtDNA haplotype to microsatellite genotype found additional evidence of asymmetric blacktail shiner genomic contributions. Three of 13 (23%) individuals with parental red shiner genotypes exhibited a blacktail shiner mtDNA haplotype (Fig. 2b). Only 8% of individuals exhibiting a parental blacktail shiner genotype had a red shiner mtDNA haplotype. Most individuals with intermediate genotypes exhibited a blacktail shiner haplotype, and the proportion of blacktail shiner mtDNA haplotypes steadily increased as genotypic profile progressed towards parental blacktail shiner (Fig. 2b).

Red shiners, blacktail shiners, or hybrids were collected at 18 of 33 sites (Table 2, Fig. 3). Red shiners were rarely collected (n = 5 sites and 1.2% of the total catch) and only occurred at sites that contained both hybrids and blacktail shiners. Hybrids and blacktail shiners each occurred at sixteen sites. Hybrid occurrence was significantly correlated with blacktail shiner occurrence (Χ2 = 18.9, P < 0.0001) with hybrids occurring at all but two of the sites occupied by blacktail shiners and vice-versa. Hybrids accounted for 36% of the total catch, with the vast majority exhibiting asymmetry favoring blacktail shiners (Hbt, Table 2). Hybrids were well distributed among sites, with no clear trend in relative abundance with respect to upstream distance from Weiss Lake (Fig. 3).
Table 2

Percent population contribution of red shiner (RS), hybrids exhibiting asymmetry favoring red shiner (Hr), F1 hybrids, hybrids exhibiting asymmetry favoring blacktail shiner (Hbt), and blacktail shiner (BTS) collected in the study area

Site code

Site name

n

RS

Hr

F1

Hbt

BTS

1

Coahulla Cr.

73

0

9.6

0

12.3

78.1

2

Little Cr.

2

0

0

0

50

50

3

Drowning Bear Cr.

1

0

0

0

0

100

4

Jobs Cr.

70

0

2.9

4.3

28.6

64.3

5

Swamp Cr.

61

0

11.5

4.9

23

60.7

6

Polecat Br.

5

20

20

0

20

40

7

Town Cr.

47

4.3

2.1

6.4

34

53.2

8

Dry Cr.

35

2.9

5.7

5.7

25.7

60

9

Oothkalooga Cr.

11

0

9.1

9.1

27.3

54.5

10

Snake Cr.

26

0

3.8

3.8

19.2

73.1

11

Bow Cr.

6

16.7

16.7

0

0

66.7

12

Rocky Cr.

6

0

16.7

33.3

0

50

13

Johns Cr.

29

0

0

3.4

20.7

75.9

14

Lovejoy Cr.

6

0

0

33.3

66.7

0

15

Woodward Cr.

16

0

0

6.3

25

68.8

16

Silver Cr.

12

0

0

0

8.3

91.7

17

Webb Cr.

2

0

0

0

0

100

18

Kings Cr.

1

0

100

0

0

0

Totals

 

409

1.2

6.1

4.6

22.7

65.3

Categories were determined through integrative comparisons of phenotype, mtDNA haplotype, and microsatellite multi-locus genotype data described in Methods. Sample locales are shown in Fig. 3

Fig. 3

Collection locales in tributary streams. Relative abundance (%) of red shiner, blacktail shiner, and hybrids are provided in pie charts. Three categories of hybrids are pooled for illustration purposes. Open circles indicate locales where both species and hybrids were absent. Site codes are in Table 2

Environmental correlates of occurrence and hybridization

Overall percent correctly predicted occurrences was >75% for the best-fit models (Table 3), which is considered satisfactory model performance (Hurley 1986). Models constructed using variables from multiple environmental categories (i.e., landscape, reach geomorphology, and water quality) consistently outperformed single category models. Land use variables were generally absent from best-fit models, except for the negative relationship between hybrid occurrence and relative abundance with basin agriculture. Turbidity (NTU) was consistently selected in occurrence models that included water quality variables. However, NTU was not selected for models of hybrid relative abundance. Stream size and water velocity were consistent predictors of parental species and hybrids. All shiners were more likely to occur in larger, higher velocity streams.
Table 3

Predictive habitat models for presence of blacktail shiner (BTS), hybrids, and red shiner (RS) as well as relative abundance of hybrids presented in rank order

Model

Predictor variables

K

AICc

ΔAICc

w

% Correctly predicted

Presence

Absence

Overall

BTS presence=

(reach + wq)

V, W, NTU

4

29.04

0

0.19

81.3

82.4

81.8

(reach)

V, V-m, COBB

4

29.37

0.33

0.19

75.0

76.5

75.8

(reach + land)

W, V, DTMS

4

32.95

3.91

0.16

81.3

82.4

81.8

(reach + land + wq)

W, NTU, DTMS

4

33.00

3.96

0.16

68.8

76.5

72.7

(land + wq)

NTU, URB, CON

4

37.55

8.51

0.12

75.0

82.4

78.8

(wq)

NTU, CON, DO

4

38.62

9.58

0.12

81.3

70.6

75.8

(land)

AG, FOR, DTMS

4

50.47

21.43

0.07

43.8

47.1

45.5

Hybrid presence=

(land + wq)

AG, NTU, DTMS

4

25.00

0

0.18

87.5

88.2

87.9

(reach + wq)

D, NTU, CON

4

26.64

1.64

0.17

75.0

70.6

72.7

(reach)

V, EH, GRAV

4

27.55

2.55

0.16

81.3

82.4

81.8

(reach + land + wq)

W, NTU, DTMS

4

29.11

4.11

0.15

81.3

88.2

84.8

(wq)

NTU, CON

3

32.54

7.54

0.13

81.3

88.2

84.8

(reach + land)

D, W, DTMS

4

32.69

7.69

0.13

75.0

70.6

72.7

(land)

URB, DTMS

3

41.06

16.06

0.08

75.0

76.5

75.8

RS presence=

(reach + wq)

V-m, NTU, CV-D

4

15.49

0

0.21

60.0

89.3

84.8

(reach + land + wq)

V-m, FOR, NTU

4

18.94

3.45

0.17

40.0

92.9

84.8

(reach + land)

V-m, FOR, W

4

22.67

7.18

0.14

60.0

92.9

87.9

(reach)

D, V-m

3

23.44

7.95

0.14

40.0

96.4

87.9

(land + wq)

DTMS, NTU

3

25.32

9.83

0.13

20.0

89.3

78.8

(wq)

NTU, DO

3

26.49

11.0

0.12

40.0

96.4

81.8

(land)

FOR, DTMS

3

32.27

16.78

0.09

0.0

92.9

78.8

Hybrid abundance=

(reach + land + wq)

DO, AG, GR

4

−5.9

0

0.17

   

(reach + wq)

GR, DO, RH,

4

−4.4

1.5

0.15

   

(land + wq)

FOR, URB, DO

4

−3.9

2.0

0.15

   

(reach + land)

GR, FOR, V-m

4

−3.4

2.5

0.15

   

(land)

AG, FOR

3

−2.4

3.5

0.14

   

(reach)

LWD, RH, GR

4

−1.1

4.8

0.13

   

(wq)

DO, CON

3

1.4

7.3

0.12

   

K is the number of estimable parameters in the model; AICc is Akaike’s Information Criterion corrected for small sample size (smaller is better); Δi is the difference in AICc from the best-fit model; and w is the model weight (models with higher weights are more parsimonious). Explanations of predictor variables are given in Table 1. Variable categories: reach = reach habitat; wq = water quality; land = landscape

The most parsimonious model for blacktail shiners showed that they were more likely to occur in wider, more turbid streams with higher velocity (Table 4). None of the confidence intervals for the model parameters contained zero. Despite the similarity of the distribution patterns of blacktail shiners and hybrids, their best-fit models differed. Hybrids were more likely to occur at sites with higher turbidity and agricultural land cover that were closer to mainstem rivers. Only the confidence intervals for the intercept parameter contained zero (Table 3). Hybrid relative abundance was positively related to agriculture and negatively related to dissolved oxygen concentration. Relative abundance was also positively associated with gradient in this model. However, the confidence interval for this parameter contained zero, so there is uncertainty about this relationship. The AIC approach identified parsimonious models for red shiner occurrence, but even the most parsimonious model had poor explanatory power and confidence intervals on all model parameters contained zero. Low power for this model was related to infrequent occurrence of red shiners (n = 5 sites).
Table 4

Parameter estimates for best-fit models from Table 3

Logistic model

N

Wald χ2

P

Parameter

Parameter estimate

Standard error

95% Confidence interval

BTS

33

8.1

0.0448

     
    

Intercept

−17.3

6.3

−29.6

−5.0

    

NTU

6.6

3.2

0.3

12.8

    

V

7.7

3.3

1.2

14.2

    

W

10.7

5.1

0.7

20.6

Hybrid

33

6.9

0.0760

     
    

Intercept

15.0

11.0

−6.6

36.5

    

DTMS

−13.5

5.7

−24.7

−2.2

    

AG

20.9

8.9

3.4

38.4

    

NTU

13.7

5.3

3.3

24.1

RS

33

1.8

0.6176

     
    

Intercept

−22.3

19.5

−60.7

16.0

    

V-mean

140.1

120.7

−96.4

376.6

    

NTU

27.8

27.7

−24.4

82.0

    

CV-D

−0.3

0.3

−0.7

0.3

Linear model

 

r2

F, P

     

Hybrid

18

0.50

4.6, 0.02

     
    

Intercept

0.9

0.5

−0.3

2.0

    

DO

−1.5

0.5

−2.6

−0.4

    

AG

0.9

0.4

0.1

1.6

    

GR

0.4

2.9

−5.9

6.7

Logistic and linear regression models are based on occurrence and relative abundance, respectively

Discussion

Dispersal of invasive red shiner

Biological invasions generally follow a pattern of introduction, permanent establishment, and range expansion (Kolar and Lodge 2001). Prior reports of red shiner introductions compared pre- and post-introduction data collected decades apart (Moyle 2002; Page and Smith 1970), and thus provide limited information on early stages of invasion and factors contributing to establishment and spread. We identified patterns of red shiner dispersal in the UCRS that may be generalized to other red shiner invasions in other systems with native congeners. Red shiners and hybrids rapidly disperse (up to 31 river km y−1) via large, mainstem rivers with subsequent invasion of tributaries. Tributaries with populations of native congeners were colonized first, and the incipient invasion of tributary streams was driven primarily by hybrids. The strong affinity of red shiners and hybrids for streams with blacktail shiner populations indicates that hybridization plays a key role in the establishment and expansion of red shiners and their genome into new habitats, supporting prior findings that introgressive hybridization facilitates the establishment and spread of invasive species (Allendorf et al. 2001; Ellstrand and Schierenbeck 2000; Hitt et al. 2003; Rhymer and Simberloff 1996). This mechanism of dispersal is particularly relevant for fishes, which are more prone to hybridization than other vertebrates (Allendorf et al. 2001; Rhymer and Simberloff 1996).

Hybridization complicates taxonomic identification, hindering the assessment of native and introduced populations (Allendorf et al. 2001; Rhymer and Simberloff 1996). Populations in the UCRS expressed a mosaic of parental phenotypes and genotypes and are best characterized as hybrid swarms (Allendorf et al. 2001). The majority of individuals with hybrid genotypes had blacktail shiner phenotypes. Since these hybrids were morphologically indistinguishable from the parental species, the degree of hybridization in the UCRS could be underestimated without complementary multi-locus genetic analyses. Thus, our reconstructed timeline conservatively estimates the geographic extent and expansion of the hybrids, since it reflects morphological identification of specimens.

The nature of pre- and postzygotic reproductive isolation between red and blacktail shiners may influence patterns in phenotypic and genetic variation observed in the UCRS. Artificial crosses (where eggs and sperm were mixed to produce hybrids) and backcrosses of laboratory reared hybrids to red and blacktail shiners have produced viable and fertile hybrids (Hubbs and Strawn 1956). The UCRS hybrid swarm is dominated by later generation hybrids. F1 hybrids are relatively uncommon, suggesting that interactions between red and blacktail shiners are infrequent. If so, then prezygotic reproductive isolation is stronger than postzygotic isolation between red and blacktail shiners (Mendelson 2003). The predominance of Hbt individuals among hybrids is also suggestive of asymmetric pre- or postzygotic barriers that promote introgression. The comparatively low abundance of red shiners likely augments asymmetric backcrossing between hybrids and blacktail shiners (Taylor and Hastings 2005). Thus, undiminished hybrid fitness and persistent backcrossing under demographic conditions that favor introgression could be driving progressive genetic assimilation of blacktail shiners with few (if any) external indications of hybridization.

The role of disturbance in red shiner invasion and hybridization

Regression models supported prior hypotheses that disturbance increases colonization by red shiners and hybridization with congeners (e.g., Hubbs et al. 1953; Larimore and Bayley 1996; Page and Smith 1970) and other studies linking disturbance with fish invasion of lotic ecosystems (Gido and Brown 1999; Marchetti et al. 2004; Moyle and Light 1996). Best-fit models indicated that hybrid and red shiner occurrence increased with turbidity and agricultural land use. These findings have two limitations. First, red shiner models had low power due to small sample size, limiting their interpretation. Second, we cannot separate the role of disturbance from the presence of blacktail shiners in determining the distribution of red shiners and hybrids. That is, red shiners and hybrids both had strong affinity for blacktail shiner streams, but blacktail shiners were also positively correlated with turbidity, an indicator of stream disturbance. Hybrid relative abundance was also positively related to disturbance, particularly lower dissolved oxygen concentration and increased agricultural land use. This suggests that rates of hybridization or hybrid survival increase with disturbance, thereby facilitating the spread of red shiner genome into native populations. Occurrence and hybrid relative abundance models using variables from multiple environmental categories (i.e., landscape, reach geomorphology, and water quality) consistently outperformed single-category models. Superior performance of these models indicate that red shiner colonization and hybridization are influenced by factors operating at multiple spatial scales and that habitat and water quality are both influential at the reach-scale.

We expected positive correlations between turbidity and hybrid relative abundance since turbidity has been associated with other red shiner hybrid swarms (Hubbs et al. 1953; Hubbs and Strawn 1956; Larimore and Bayley 1996). Turbidity is thought to weaken prezygotic reproductive barriers by impairing visual recognition and assortative mating among Cyprinella species (Page and Smith 1970; Hubbs et al. 1956). This hypothesis is consistent with other studies demonstrating that turbidity can weaken sexual selection and promote hybridization (Candolin et al. 2007; Seehausen et al. 1997). Contrary to our expectations, turbidity was not a leading predictor of hybrid relative abundance. The influence of turbidity on hybridization rates may be obscured in the UCRS because blacktail shiners appear to be moderately tolerant to elevated turbidity (i.e., blacktail shiner occurrence was positively correlated with turbidity in this study). Alternatively, other aspects of mating behavior unaffected by turbidity could counterbalance the negative effects of turbidity on assortative mating. Controlled spawning experiments under different turbidity regimes are needed to test the hypothesis that light limitation increases hybridization between red shiners and congeners.

Conservation implications for Southeastern Rivers and Cyprinella diversity

Hybridization with red shiners is a serious threat to Southeastern Cyprinella diversity. Twenty-two Cyprinella occur in the Southeast (Warren et al. 2000), and nine (41%) hybridize with red shiners in the wild (C. analostana, C. camura, C. callitaenia, C. spiloptera, C. venusta cercostigma, C. v. stigmatura, C. v. venusta, and C. whipplei; (Hubbs and Strawn 1956; Johnson 1999; W.C. Starnes personal communication; Page and Smith 1970; Wallace and Ramsey 1982)) or under laboratory conditions (C. caerulea, Burkhead unpublished data). Southeastern rivers colonized by red shiners include the Pee Dee, Roanoke, Mobile, Apalachicola, and Altamaha river drainages (Fuller et al. 1999), and hybrids have been documented in three of these systems (DeVivo 1996; W.C. Starnes personal communication, this study; Wallace and Ramsey 1982). Southeastern fishes are characterized by high rates of endemism and imperilment, and nine Cyprinella are endemic to a single Southeastern river drainage (Warren et al. 2000). These species are particularly vulnerable to red shiner invasion considering their limited distribution and relatively small populations (Allendorf et al. 2001; Rhymer and Simberloff 1996).

While the blacktail shiner is not a protected species, their displacement via hybridization in the Coosa system underscores the conservation challenges facing other Southeastern Cyprinella species. The prevalence of post-F1 hybrids in the UCRS indicates that hybrids are fertile and capable of backcrossing with parental taxa. Native populations decline rapidly under these conditions as hybrid swarms form within a few generations and the proportion of parental individuals progressively declines in successive generations (Allendorf et al. 2001; Rhymer and Simberloff 1996). Through this process, native species are lost or relegated to isolated parts of their former range. Our data support this model of native species decline, with widespread introgression between native and invasive species in mainstem tributaries of the UCRS.

Urbanization will likely play an increasing role in red shiner invasions of Southeastern rivers. The Southeast is among the most rapidly developing regions of the U.S. (U.S. Department of Agriculture 2000), and the UCRS epitomizes development pressures facing these river systems. Human population in the UCRS increased 37% (270,000–370,000) from 1990 to 2006 with 7,400 residential building permits issued from 2000 to 2006 in one county (Bartow) centrally located in the UCRS (U.S. Census Bureau 2007). Red shiners thrive in Southeastern urban streams (DeVivo 1996) whereas endemic and other native fishes decline (Walters et al. 2003a). It is reasonable to assume that urbanization will increase vulnerability of Southeastern streams and native Cyprinella to red shiner invasion, barring management actions to mitigate the detrimental effects of urbanization on stream ecosystems.

Acknowledgements

We thank D. Homans for assisting in study design and supervising sample collection, T. Crum, B. Dakin, A. Kuenzi, and C. Tepolt for generating genetic data, C. Straight and M. Reif for compiling and mapping collection data, and K. Oswald for reviewing the manuscript. Although this work was reviewed by US EPA and approved for publication, it may not necessarily reflect official Agency policy. Mention of trade names or commercial products does not constitute endorsement or recommendation for use.

Copyright information

© Springer Science+Business Media B.V. 2007

Authors and Affiliations

  • David M. Walters
    • 1
  • Mike J. Blum
    • 1
    • 2
  • Brenda Rashleigh
    • 3
  • Byron J. Freeman
    • 4
  • Brady A. Porter
    • 5
  • Noel M. Burkhead
    • 6
  1. 1.U.S. Environmental Protection Agency, National Exposure Research LaboratoryCincinnatiUSA
  2. 2.Department of Ecology and Evolutionary BiologyTulane UniversityNew OrleansUSA
  3. 3.U.S. Environmental Protection Agency, National Exposure Research LaboratoryAthensUSA
  4. 4.Georgia Museum of Natural History and Odum School of EcologyUniversity of GeorgiaAthensUSA
  5. 5.Department of Biological SciencesDuquesne UniversityPittsburghUSA
  6. 6.U.S. Geological Survey, Florida Integrated Science CenterGainesvilleUSA

Personalised recommendations