Gene Flow Among Modoc Sucker and Sacramento Sucker Populations in the Upper Pit River, California and Oregon

Gene flow is one of the four fundamental forces that drives evolution in Mendelian populations (Wright 1930). Intraspecific gene flow (i.e., migration) and interspecific gene flow (i.e., hybridization) both facilitate maintenance of genetic variability and can introduce alleles into novel environments (Dowling and Secor 1997; Smith et al. 2003). However, excessive amounts of gene flow can prevent or limit adaptation to local environments and ultimately lead to reduction in genetic variability or extinction (Rhymer and Simberloff 1996). In cases where hybrid fitness is comparable to parental fitness in the parental environments, introgression (back-crossing of hybrid individuals into one or both parental populations) can lead to local loss of the parental types (Hegde et al. 2006). In cases where hybrid fitness is lower than that of the parental types in parental environments, parental types may continue to dominate in their respective environments, and stable hybrid zones may develop, often in intermediate environments (e.g., Wang et al. 1997).

The Modoc sucker Catostomus microps (Figure 1) is a small (usually <18-cm standard length) catostomid fish whose distribution is confined to two tributaries of the Pit River, a tributary of the Sacramento River, and one tributary of Goose Lake (Thomas Creek), a now disjunct upstream subbasin of the Pit River, in California and Oregon, USA (Figure 2). Modoc suckers generally occupy low energy pool habitats in low-to-moderate gradient reaches of small arid second- to fourth-order streams, which are often reduced to a series of isolated pools in the summer (Moyle 2002). However, they do disperse relatively well within occupied tributaries, as evidenced by their recolonization of Turner Creek after a rotenone treatment in 1983 and their frequent occurrence 1–2 km above permanent habitat in the early summer before seasonal upstream reaches dry up. Ecologically, Modoc suckers occupy an intermediate faunal zone between the high-gradient, cold water trout Oncorhynchus spp. zone and the low-gradient, warm water Sacramento sucker Catostomus occidentalis–pikeminnow Ptychocheilus grandis–hardhead Mylopharodon conocephalus zone of the mainstem Pit River and lower reaches of its primary tributaries (Moyle and Daniels 1982).

The Sacramento sucker is a similar, although larger, species that also occurs naturally in the upper Pit River (including Goose Lake) and is sympatric with Modoc suckers in the lower reaches of tributaries inhabited by the latter. Adult Sacramento suckers in the Pit River make spawning migrations up into tributaries in the early spring and then generally return downstream to larger habitat within a few weeks, with juveniles remaining resident in the smaller streams. Unlike Modoc suckers, Sacramento suckers are abundant throughout the Pit River, Goose Lake, and Sacramento River drainages and seem to flourish in a range of modified habitats (Moyle 2002).

The two sucker species differ in their sizes at maturity and core spawning times. Modoc suckers are small and relatively short lived, typically maturing at 8–12 cm and rarely exceeding 18 cm in length. In contrast, Sacramento suckers do not mature until at least 20 cm and reach more than 50 cm in length. Modoc suckers typically spawn from mid-April to early June, after flows have settled, whereas Sacramento suckers in the upper Pit River typically spawn in March–April (into May in Goose Lake), during higher spring flows that allow access to and return from upstream spawning areas. Nevertheless, the spatial proximity and partial temporal overlap between spawning populations of the two species led to early speculation that hybridization was occurring between the two species and to concerns that anthropogenic habitat alteration resulting in removal of barriers and increased contact between the two species might further promote hybridization (Moyle and Marciochi 1975).

Hydrological connectivity of Goose Lake to the upper Pit River has been intermittent in recent history, and opportunities for gene flow between Modoc suckers in tributaries of the Pit River and Goose Lake have been limited. Although Goose Lake historically drained southward through the Pit River and is thus considered part of the Pit River hydrologic system, the lake has rarely overflowed in recent times, and continuous outflow has probably not occurred since periods of substantially higher rainfall in the Pleistocene, effectively isolating Goose Lake from the Pit River (Laird 1971; Phillips and Van Denburgh 1971; Johnson et al. 1985). Goose Lake has apparently not overflowed into the North Fork of the Pit River since 1868 and a few hours in 1881 and before that in 1832 and 1856. Separation of Goose Lake from the Pit River, punctuated by occasional connectivity in particularly wet years, thus predates substantial European settlement of the region.

In 1985, the Modoc sucker was listed as endangered under the Endangered Species Act of 1973 (16 U.S.C. § 1531 et seq.). Habitat alteration, small population size, isolation of populations, and hybridization with Sacramento sucker were cited as the primary threats to the species (U.S. Federal Register 50, 112:24526–24530). Genetic data were not available for Modoc suckers at the time the species was listed as endangered. Here, we use genetic markers to evaluate the degree of connectivity within and among Modoc and Sacramento sucker populations, including the level of interspecific hybridization and introgression between them.

Samples of Modoc suckers were taken from 10 sites, representing all known populations, and samples of Sacramento suckers were taken from 17 sites in the upper Pit River, including Goose Lake (Table 1; Figure 2). Each individual was identified to species, and a small (<5 mm²) fin clip was taken and preserved in 95% ethanol. Fin clips were stored at ambient temperature before genetic analysis.

To reduce handling and take of Modoc suckers, specimens from streams not known to contain Sacramento suckers were presumed to be Modoc suckers and were generally fin clipped and released, including the following sites Washington, Hulbert, Coffee Mill, Turner-upper, Johnson, Dutch Flat and Thomas, as well as Garden Gulch, although some specimens from the latter two sites also were preserved and identified to voucher the new localities. Specimens from lower Turner Creek (Sacramento suckers), Garden Gulch (in part), Willow Creek (Ash), and all Goose sites except Thomas Creek itself (in part) were preserved and identified in the laboratory through a combination of dorsal ray, lateral line, and vertebral counts. Specimens from Ash Valley and middle Turner, as well as most Sacramento suckers from Pit River sites, were field identified as Modoc suckers if they had a combination of 10 or 11 dorsal rays and lateral line counts of 80–92 (field counted) or Sacramento suckers if they had 11–13 dorsal rays and fewer than 76 lateral line scales following Kettratad (2001) and Moyle (2002). Field counts were all by the same individual (Reid), and although counting on live fish is somewhat problematic, actual scale counts were typically well within the range for a given species, rather than in the narrow range (∼76–79 scales) where a misidentification might occur. Vertebral counts (total) used for identification were 40–42 for Modoc sucker and 43–45 (rarely 42 or 46) for Sacramento sucker (Kettratad 2001; observed ranges in this study).

Sample collection spanned 1–4 y per site, resulting in 57 collections in total. We use “site” to denote a locality (typically an entire stream, except for Turner Creek itself, which was separated into reaches), “collection” to denote individuals of a morphological species obtained from one site in 1 y, and “sample” to denote all collections taken at a site. Turner Creek was separated into reaches due to the presence of a higher gradient upper reach (Turner-upp) and two distinct meadow reaches (Turner-mid and Turner-low) separated by a short high gradient reach and a gabion barrier, consisting of metal cages filled with rocks, which was installed in 1982. The gabion barrier was originally thought to exclude Sacramento suckers from migrating upstream, but it has failed to do so since at least 1997. The upper meadow site was further separated into two reaches above (Turner-mid-a) and below (Turner-mid-b) the confluences of Turner Creek's principal tributaries (Washington and Hulbert creeks), which enter at the same point, to better resolve the zone of sympatry.

Deoxyribonucleic acid was extracted from a small piece (∼1 mm²) of tissue from each individual by using Chelex as described by Small et al. (1998). Extracted DNA was genotyped using eight microsatellite loci: Dlu26, Dlu243, Dlu476, Dlu488, Dlu4183, Dlu4184, Dlu4296 (Tranah et al. 2001), and Dlu4158 (forward primer, CTCTTCATCCTCAGCTCGACACA and reverse primer, GTGGCATGGCATGTTTTTAGATTT; J. Roach, Louisiana State University, personal communication). Loci were amplified using polymerase chain reaction (PCR) in the following solution: 2 µL of DNA template, 0.1 unit of Taq DNA polymerase (Promega, Madison, WI), 1.5 µL of 10× buffer (10 mM Tris-HCl [pH 9.0 at 25°C], 50 mM KCl, and 0.1% Triton® X-100), 0.3 µL of 10 mM dNTPs, 0.75 µL each of 10 µM reverse and forward primers, and a final concentration of 2 mM MgCl₂ in a 15-µL reaction volume. Template DNA was arrayed into 384-well plates, and PCR cocktail was added to each well by using a JANUS® Automated Workstation (PerkinElmer Life and Analytical Sciences, Boston, MA). Thermal cycling was carried out in an AB9700 thermal cycler (Applied Biosystems, Foster City, CA) as follows: 94°C (1 min), followed by 38 cycles of 94°C (1 s) + X°C (30 s) + 72°C (30 s), where X is an annealing temperature (Table 2), and a final extension at 72°C (7 min). Polymerase chain reaction products were size-fractionated using an AB3130 DNA sequencer (Applied Biosystems), and raw data (i.e., electropherograms) were analyzed using GeneMapper 4.0. All genotypes were scored by two independent readers. To exclude duplicate samples (multiple samples taken from the same individual), we used Microsatellite Toolkit (Park 2001) to screen for individuals with identical genotypes. Where groups of individuals with identical genotypes were detected, the first individual captured was retained, and all others were excluded from subsequent analyses.

We tested for genotypic ratios that departed from Hardy–Weinberg Equilibrium (HWE) using Fisher's exact tests in GENEPOP version 4.0 (Rousset 2008). The log-likelihood ratio statistic (G test) was used to test for genotypic disequilibrium between each pair of loci in each collection. Critical values for both of these tests were adjusted for the number of simultaneous tests (i.e., for HWE, α  =  0.05/57 collections per locus; for genotypic disequilibrium, α  =  0.05/28 pairwise comparisons).

Allelic richness (number of alleles observed per sample, corrected for unequal sample sizes) was calculated for each sample using FSTAT (Goudet 2001). The number of alleles observed per locus was quite variable (Table 2), so instead of presenting average allelic richness values we ranked the samples by allelic richness at each locus. The sum of ranks across loci for each sample was then used as a measure of the relative amount of diversity observed in that sample (with 1 being the highest diversity and 27 being the lowest).

Divergence among samples was estimated based on allele identity under an infinite allele model using F_ST (θ; Weir and Cockerham 1984) and based on allele size under a generalized stepwise model using R_ST (Michalakis and Excoffier 1996). To determine whether mutation played a substantial role (relative to migration and drift) in population divergence and thus whether F_ST or R_ST was more appropriate for evaluating population divergence, we used the permutation test described by Hardy et al. (2003). In brief, allele names were shuffled and R_ST was recalculated 10⁴ times to generate a distribution (pR_ST) against which the observed value of R_ST could be evaluated. If gene flow between two taxa has been limited to the extent that mutation has played a relatively substantial role in shaping divergence between them, then we expect R_ST > pR_ST. Global and pairwise estimates of R_ST and distributions of pR_ST were generated using the program SPAGeDi (Hardy and Vekemans 2002). Significance of pairwise F_ST also was evaluated by comparison with a distribution of 10⁴ permutations; in this case, individual genotypes were shuffled among samples.

Analysis of molecular variance (AMOVA; Excoffier et al. 1992) was used to partition variance among hierarchical levels of the population structure. First, partitioning of genetic variance among the two species was examined by dividing all samples into categories representing the two species (Modoc suckers [samples 1–9] in one group and Sacramento suckers [samples 10–27] in another group). Second, partitioning of genetic variance in Modoc suckers was evaluated by dividing samples of this species into three groups representing the three tributaries (Table 1). Null distributions for each AMOVA were generated based on 2 × 10⁴ permutations. Based on the results from the comparison of R_ST, pR_ST, and F_ST, variance in allele size was used for the first (interspecific) AMOVA, and variance in allele identity was used for second (intraspecific). Calculations were performed using Arlequin version 3.11 (Excoffier et al. 2005).

Principle component analysis (PCA) was conducted on the allelic matrix of individuals from each tributary to facilitate visual evaluation of the distinctiveness of the two species in each tributary. This was done using the program PCAGEN (Goudet 2009).

We used model-based assignment to hybrid classes as described by Anderson and Thompson (2002) and Anderson (2009) to compute the posterior probability that each individual was either Modoc sucker, Sacramento sucker, an F₁ (first-generation hybrid; an individual with one Modoc sucker parent and one Sacramento sucker parent), an F₂ (second-generation hybrid; an individual with two F₁ parents), or a back-cross (a hybrid individual with one parent that is either a Modoc sucker or Sacramento sucker and a hybrid parent). To focus on hybrids between species (interspecific) rather than hybrids between populations (intraspecific), individuals were grouped by tributary (Table 1) for this analysis. Calculations were performed using the program NewHybrids (Anderson and Thompson 2002). Uniform (uninformative) priors on allele frequencies and mixing proportions seemed most appropriate to us, however, to evaluate the impact of this assumption on our results we repeated the analysis using Jeffrey's-like priors. For each run, averages were collected over 4 × 10⁴ Markov chain Monte Carlo steps, following a 10⁴ step burn-in.

Our ability to accurately assign individuals to hybrid classes using the above method was assessed using a simulation approach. We first simulated 10³ individuals of each parental species and hybrid class using Hybridlab (Nielsen et al. 2006). Simulated individuals were generated using allele frequencies observed in samples in the upper sites of each tributary (site 2 in Ash Creek, site 9 in Turner Creek, and site 3 in Thomas Creek) and on Sacramento sucker samples from the Pit River (for Ash Creek and Turner Creek simulations) and Goose Lake (for Thomas Creek simulations). Simulated individuals were assigned to hybrid class by using the method described above. Initial runs revealed that parental individuals generally had most of their posterior distributions in a single parental class but that hybrid individuals generally had posterior distributions spread across several hybrid classes. This result was not unexpected, based on the modest number of available markers and the inherent difficulty in distinguishing among hybrid classes. Because of this, we binned individuals into three classes (Modoc, Sacramento, and hybrid). If the posterior probability of an individual being one of the parental classes was greater than our cut-off value, that individual was labeled as the corresponding parental class. Otherwise the individual was labeled a hybrid. Rates of Type I errors (assignment of a parental individual to a hybrid class) and Type II errors (assignment of a hybrid individual to either parental class) were evaluated in simulated individuals for posterior probability cut-off values of 90, 80, 70, 60 and 50% in each tributary.

For the 789 individuals included in the present analysis, we were able to obtain genotypes for at least seven of eight loci per individual (overall PCR failure rate of 0.8%). The loci examined here exhibited between 7 and 42 alleles each, with 194 alleles observed across loci in total (Table 2; Smith et al. 2011).

Departures from HWE expectations were observed in between two and seven collections per locus (out of 57 collections). After a correction for multiple comparisons, the only significant departure was at Dlu4158 in the Willow (Ash) 2008 collection. Genotypic disequilibrium was observed between zero and seven pairs of loci per collection (out of 28 pairwise comparisons; Table 1). One pair of loci (Dlu4183 and Dlu4296) exhibited genotypic disequilibrium across populations (P  =  0.037); however, this result was not significant after correction for multiple comparisons (corrected α  =  0.002). No loci were excluded from further analysis based on the results of HWE or genotypic disequilibrium tests.

Results of several analyses revealed a strong similarity between the Modoc sucker and Sacramento sucker samples from Ash Valley (samples 10 and 11, respectively). Therefore we performed another test for departures from HWE in which these two samples were pooled. No significant departures were detected.

Allelic richness was generally lower in Modoc sucker samples than in Sacramento sucker samples (Table 1). The 12 collections with the highest allelic richness values (ranked 1–12) were all collections of Sacramento suckers.

Comparison of divergence among samples using R_ST, pR_ST, and F_ST revealed that R_ST was greater than pR_ST for interspecific comparisons but not for comparisons among collections of Modoc suckers (Table 2). At the individual locus level, R_ST was greater than pR_ST for six of eight loci when comparing species but for only one of eight loci among Modoc sucker samples. These results indicate that mutation has had a large impact relative to migration among species but not within species. It was further noted that the two loci for which interspecific R_ST was not greater than pR_ST (Dlu26 and Dlu243) were also the two loci with the smallest number of alleles. Based on these results, we concluded that R_ST was appropriate for interspecific comparisons and that F_ST was appropriate for intraspecific comparisons.

One clear exception to the above-mentioned pattern was the Ash Valley Modoc sucker (sample 10), for which R_ST was greater than pR_ST in all comparisons with Modoc sucker collections but not for any comparisons with Sacramento sucker (see Supplemental Material, Appendix S1; http://dx.doi.org/10.3996/022010-JFWM-003.S1). Other exceptions to R_ST > pR_ST for interspecific but not intraspecific comparisons included one comparison between Modoc sucker samples in the Pit River (samples 5 and 2) and several comparisons among Goose Lake Sacramento sucker samples. Greater overlap in allele size ranges among the two species in the Goose Lake tributary than in the Pit River tributaries was also apparent upon visual examination of the allele size frequencies of some loci. For example, at the locus Dlu4296, collections of Sacramento suckers contained very few alleles larger than 215 nucleotides, whereas collections of Modoc suckers contained a high frequency of alleles larger than 215 nucleotides.

Pairwise estimates of F_ST were significant (i.e., were >95th percentile of null distribution) between all pairs of Modoc sucker samples (see Supplemental Material, Appendix S1; http://dx.doi.org/10.3996/022010-JFWM-003.S1). Most pairwise estimates of F_ST between Sacramento sucker samples were also significant; however, some groups of samples taken from relatively broad regions exhibited low or nonsignificant values. For example F_ST estimates were nonsignificant or low among samples from the west side of Goose Lake (samples 13–15, 17, and 19), as well as among samples from the Pit River mainstem (samples 21–24). Sacramento sucker samples exhibited greater average divergence between Goose Lake and the Pit River (F_ST  =  0.159) than among tributaries within the Pit River (F_ST  =  0.077), as might be expected based on physical separation of Goose Lake from the Pit River over most of the past two centuries. In contrast, Modoc sucker samples exhibited greater divergence among tributaries within the Pit River (F_ST  =  0.287) than between Goose Lake and the Pit River (F_ST  =  0.193). Estimates of R_ST between Modoc suckers and Sacramento suckers captured in each tributary were larger than estimates between Modoc suckers from different tributaries (see Supplemental Material, Appendix S1; http://dx.doi.org/10.3996/022010-JFWM-003.S1). Divergence between the samples of Modoc sucker and Sacramento sucker from Ash Valley (samples 10 and 11; F_ST  =  −0.004) was not significant.

Analysis of molecular variance revealed population structure within and among the Modoc sucker samples from all tributaries examined and confirmed that divergence among species was much greater than the divergence among Modoc sucker populations (Table 3). Comparison of Modoc suckers to Sacramento suckers revealed that the percentage of genetic variance between species (68.4%) was an order of magnitude greater than that between samples within species (6.2%). Grouping Modoc sucker samples into the three tributaries explained a large percentage (14.9%) of the observed genetic variance, however a substantial percentage (11.3%) also was observed among sample sites within tributaries. The percentage of variance among individuals within samples was small (1.0%) and nonsignificant (P  =  0.087).

Principle component analysis of individuals resulted in the primary (x) axis for each tributary corresponding to divergence between the two species (Figures 3A–3C). The primary axis accounted for 17.4% of the inertia in Ash Creek, 19.4% in Turner Creek, and 19.7% in Thomas Creek. Every individual in sample 10 clustered with Sacramento suckers (Figure 3A).

Assignment of simulated individuals to hybrid classes revealed that increasing the posterior probability cut-off for assigning individuals to the parental classes resulted in increased Type I error rates (calling a parental individual a hybrid) and decreased Type II error rates (assigning a hybrid individual as one of the parental species; Figure 4). Type I and Type II error rates were fairly consistent among Ash Creek and Turner Creek but seemed higher in Thomas Creek. Minimum combined error rates were observed with cut-off values of 0.6 to 0.7. To reduce Type I error rates, we chose to use a cut-off value of 0.6 for assigning samples to parental classes. Applying a cut-off of 0.6 to our simulated individuals resulted in Type I error rates of 0.6, 0.4, and 4.6% in Ash Creek, Turner Creek, and Thomas Creek, respectively.

Assignment of actual samples to hybrid classes revealed hybrid individuals in all three tributaries examined (Table 1). Individuals identified as hybrids had been classified based on phenotype as both Modoc suckers and Sacramento suckers (Figure 3). Although individuals were only required to have 0.6 posterior probability in a parent class to be assigned to that class, observed posterior probabilities for such assignments were generally higher (mean >0.97 for both species in all tributaries; Figure 5). Individuals assigned as “hybrids” generally had some posterior probability of belonging to one or both parental species and thus had lower mean posterior probabilities of belonging to the defined hybrid classes (Figure 5).

Hybrid individuals were identified among samples from each of the three tributaries. The proportion of individuals from Goose Lake that were identified as hybrids (10.9%) was several times that observed in collections from the Pit River tributaries (2.3% in Ash Creek, 1.6% in Turner Creek). In each tributary, the proportion of individuals identified as hybrids was greater than the Type I error rate predicted by analysis of simulated individuals. In all three tributaries, individuals identified as hybrids seemed to be associated with specific sampling areas. In the Pit River, hybrids were observed in two tributaries of Ash Creek (sites 1 and 12) and within a short (2.5-km) meadow reach in Turner Creek or just below (sites 7, 26, and 27). In Goose Lake, all hybrid individuals were collected from Thomas Creek (1) and its lower tributaries Auger and Cox creeks (sites 3, 16, and 18). No hybrids were detected in our samples of Pit River mainstem Sacramento sucker. Use of Jeffreys-like prior made no difference in assignment of individuals from Ash Creek, Turner Creek, or the Pit River mainstem, but it resulted in identification of additional hybrids in the Goose Lake collections. Using a Jeffreys-like prior resulted in 17.4% of all individuals from the Goose Lake collections being identified as hybrids (vs. 10.9% when uniform priors were applied).

Using the 60% posterior distribution cut-off described above, every individual in our sample of Ash Valley Modoc suckers (site 10) was identified as a parental Sacramento sucker. Posterior probabilities for these individuals ranged from 96.6 to 100.0%. These were the only individuals in the data set for which morphological traits resulted in identification as one parental species and genetic assignment resulted in identification as the other parental species.

Individuals classified as hybrids were generally associated with points near the center of the PCA plots (Figures 3A–3C), however, concordance was not perfect. For example, at least one individual near the center of Figure 3a was not classified as a hybrid. It is important to recall when comparing these plots to the model-based classification that the former is only using a subset of the information that the latter is using (Anderson and Thompson 2002).

The samples of Modoc suckers examined here exhibited substantial population structure among the Pit River tributaries. Of particular interest was divergence between Ash Creek and Turner Creek, which was greater than that between either of these and Thomas Creek (see Supplemental Material, Appendix S1; http://dx.doi.org/10.3996/022010-JFWM-003.S1). This suggests that the Pit River mainstem is as effective a barrier to gene flow among Modoc sucker populations as the dry land that has separated the Pit River from Goose Lake for much of recent history. Pairwise F_ST and allelic richness estimates both indicate that Modoc sucker populations are isolated and small relative to Sacramento sucker populations.

In addition to divergence among the three tributaries, our results revealed evidence of a finer level of structure within Turner Creek. Turner Creek was the only tributary in which we had more than two sample sites to compare, and examination of pairwise F_ST estimates revealed that one sample (Garden Gulch, site 5) was very divergent from adjacent sites (see Supplemental Material, Appendix S1; http://dx.doi.org/10.3996/022010-JFWM-003.S1). Given the fine geographic scale (hundreds of meters), it is possible that part of this divergence was due to family structure. Garden Gulch is a very small tributary to lower Turner Creek, with fewer than 10 occupied pools in a short occupied reach (∼100 m). Furthermore, the very small proportion of genetic variance observed among individuals within collections (∼1%; Table 3) is concordant with a hypothesis that individuals within collections may be closely related. If samples from site 5 were omitted, the proportion of variance among samples within tributaries drops from 11.3 to 5.9% but is still highly significant (P < 0.01). Additional sample sites in Ash and Turner creeks would be required in order to determine whether or not comparable population structure exists in those tributaries.

Results of the comparison of R_ST, pR_ST, and F_ST suggested that the rate of hybridization between Modoc sucker and Sacramento sucker was low relative to the rate of microsatellite mutation. The two loci for which R_ST was not greater than pR_ST were also the two loci with the smallest numbers of alleles (Table 2). Relatively low mutation rates at these two loci could explain these results. It seems reasonable to expect a mutation rate of no greater than 10⁻³ for most of our loci (reviewed by Jarne and Lagoda 1996) and thus a historical hybridization rate somewhat less than this. These results suggest that hybridization between Modoc suckers and Sacramento suckers has been limited over the broadest temporal scale examined here. In contrast, low R_ST (relative to pR_ST) for intraspecific comparisons suggests that gene flow among populations within each species is greater than the rate of microsatellite mutation.

Partitioning of genetic variance between the two species provided further indication that introgression between Modoc suckers and Sacramento suckers has been limited. The much greater proportion of variance among species than within species (Table 3) suggests greater gene flow among populations within species than between Modoc suckers and Sacramento suckers. This conclusion is further supported by divergence estimates between the two species within tributaries that were greater than those among Modoc sucker populations from different tributaries (see Supplemental Material, Appendix S1; http://dx.doi.org/10.3996/022010-JFWM-003.S1). Thus, the rate of hybridization has been substantially lower than the rate at which Modoc suckers from Ash Creek or Turner Creek migrate (over what has been dry land for most of the past two centuries) to Thomas Creek.

Although genetic data revealed substantial divergence between Modoc suckers and Sacramento suckers, model-based assignment to hybrid classes revealed hybrid individuals in all three tributaries. It has been suggested that habitat alteration may have allowed Sacramento suckers to invade habitat historically occupied by Modoc suckers and that hybridization and introgression may have resulted (Moyle and Marciochi 1975; U.S. Federal Register 50, 112:24526–24530); however, we are not aware of any studies that have demonstrated this. From a conservation perspective, it is important to distinguish between hybridization with introgression versus hybridization without introgression (Allendorf et al. 2001). The observation of recent-generation hybrid individuals between what otherwise seem to be two very distinct species could reflect 1) an increase in hybridization rates in recent generations or 2) a low level of historical hybridization that has been countered by natural selection against hybrids. These two scenarios would have different implications for conservation, especially in the case that the former was resulting in introgression. The magnitude of observed R_ST values (Table 2) indicates that rates of introgression were extremely small for many generations before sample collection. Interpretation of our data as supporting a recent increase in hybridization with introgression would thus require the assumption that the increase occurred in the generations that we sampled. This is possible but seems unlikely. Temporal replicate samples would be required to address this issue with certainty.

Hybrids were generally detected at sites with intermediate stream characteristics (depth and cover), where one might expect both parental species to be found (e.g., no hybrids were detected in the upstream reaches of Turner Creek [sites 6, 8, and 9] or in the mainstem Pit River [sites 21–25]). A possible explanation is that Modoc suckers and Sacramento suckers are species that are separated by postzygotic selection against hybrids in the parental environments. Hybrids may be produced in intermediate environments, but as long as the parental environments are intact, natural selection will act to keep the two species separated. Both species were found together only in middle Turner Creek (sites 7 and 27), although the probable area of contact in the lower reach of Thomas Creek above the confluence with Auger and Cox creeks could not be sampled.

A second interesting result of the model-based hybrid class assignment was the higher rate of hybridization detected in the Goose Lake collections (10.9%) than in the Pit River collections (1.6–2.3%). This result was complicated by greater error rates for hybrid assignment in Goose Lake (Figure 4), but the observed rate was still greater than the expected Type I error rate (4.6%). A higher level of historical hybridization could help explain greater similarity of the two species in the Goose Lake collections, but high and statistically significant R_ST values between the species support the notion that two species remain distinct. Finally, the high rates of hybrid detection were confined to two tributaries, within which no Modoc suckers were detected. If we ignore those two sites, then the rate of hybridization becomes fairly consistent among tributaries (∼1–2%).

Greater dependence of hybrid detection in the Goose Lake collections on the priors used in the Bayesian model probably reflects weaker divergence between the species there than in the Pit River collections. Although our results suggest that the rate of contemporary hybridization is higher in Goose Lake that in the Pit River, additional data would be required to make precise and accurate estimates of that rate. Specifically, it would be interesting to analyze additional samples from Auger and Cox creeks and to apply additional microsatellite markers developed for catostomids to these samples.

There are several possible explanations for conflicting morphological and genetic assignment to species of the individuals from sample 10. Given the nonsignificant F_ST between samples 10 and 11, and the lack of departures from HWE when these two samples were merged, we have no genetic basis for thinking they are not two morphologically divergent samples from a single breeding population. One might speculate that there are two separate populations (one Sacramento sucker population and one hybrid population) that our markers simply failed to differentiate and that the event which created the hybrid population resulted in a swamping of the neutral genetic characteristics (e.g., the microsatellites examined here) without loss of the alleles responsible for the phenotype (lateral scales and dorsal rays) used to morphologically assign the species. Another possibility is that Sacramento suckers in Ash Creek have either retained or secondarily acquired (via introgression) the ability to express the phenotype that we used to morphologically assign individuals as Modoc. Both of these hypotheses are challenged by the lack of distinction between samples 10, 11, and 22 and the fact that the morphological character states used to identify Modoc suckers have never been documented in any other Sacramento sucker population. The above-mentioned hypotheses would thus require us to assume that either the full range of phenotypes available to Sacramento suckers has been inadequately documented or else there are selective factors in the environment at site 10 that induce Sacramento suckers to adopt this morphology. A set of common garden experiments or else a broader characterization of Sacramento sucker morphology would be required to provide a better explanation for the conflicting morphological and genetic assignments of the samples from this location.

In conclusion, we find that Modoc sucker populations in Ash Creek, Turner Creek, and Thomas Creek are highly distinct from one another. The Pit River mainstem seems to be as effective a barrier between Ash Creek and Turner Creek as the now dry connection between the Pit River and Goose Lake is between these two and Thomas Creek. In contrast, Sacramento suckers seem to exhibit greater gene flow between Ash Creek and Turner Creek than between either of these and Goose Lake. Hybridization between the species seems to occur where the two species are found together, but we speculate that significant introgression (i.e., loss of parental genotypes) is limited by differences in environmental factors in the optimal habitats of these two taxa. This speculation is based on three results: 1) low levels of hybrid detection in all three tributaries, 2) the absence of hybrids in the “typical” parental habitats (upstream habitats for Modoc suckers and Pit River mainstem for Sacramento suckers), and 3) highly significant R_ST values between the species (i.e., very strong evidence that introgression has not occurred). Such natural hybrid zones commonly rely on natural selection against hybrids in the parental environments (Allendorf et al. 2001), however, and increased hybridization and introgression may result where the habitat of one or both of these species has been eliminated or modified.

Please note: The Journal of Fish and Wildlife Management is not responsible for the content or functionality of any supplemental material. Queries should be directed to the corresponding author for the article.

Appendix S1. Pairwise estimates of F_ST (above diagonal) and R_ST (below diagonal) between 27 samples of Catostomus spp. (sample numbers correspond to Table 1). Highlighted cells denote statistical significance (i.e. values >95th percentile of null distributions; see Methods for additional details). Dashed lines separate samples identified in the field as Modoc suckers Catostomus microps (1–10) from those identified as Sacramento suckers Catostomus occidentalis.

Found at DOI: http://dx.doi.org/10.3996/022010-JFWM-003.S1 (1361 KB DOC).

We are grateful to numerous landowners who provided access to properties and for technical assistance provided by Matt Diggs. Denise Hawkins, Patty Crandell, Doug Markle, an anonymous reviewer, the Subject Editor, and the Editor-In-Chief provided helpful discussion and suggestions on earlier drafts of this manuscript. We thank Eric Anderson for insightful discussions regarding hybrid detection.

This work was funded by the U.S. Fish and Wildlife Service. Any use of trade, product, website, or firm names in this publication is for descriptive purposes only and does not imply endorsement by the U.S. Government.