Broadscale genetic divergence patterns
Walleye populations from the unglaciated southern portion of the range (W–Y) are the most divergent, possessing unique alleles and haplotypes. The earliest and most divergent haplotype (19) occurs in the Ohio and New Rivers (W–X), and is very genetically distinct ([30, 62] this study). Other fishes, including brook trout Salvelinus fontinalis (Mitchill 1814) , smallmouth bass Micropterus dolomieu Lacepède 1802 , greenside darter Etheostoma blennioides Rafinesque 1819 , rainbow darter Etheostoma caeruleum Storer 1845 , and yellow perch , likewise demonstrate pronounced population divergence of Atlantic coastal groups, indicating long-term isolation.
The North River (Y), which drains south into the Gulf of Mexico, is dominated by an endemic walleye haplotype (20). The population is small, very isolated, and relatively low in diversity, as previously documented [25, 30]. Boschung and Mayden  noted that this indigenous North River population is at risk of colonization from northerly walleye due to connections through the Tombigbee–Tennessee River waterway, which may threaten localized adaptations. Likewise, yellow perch from the Gulf Coastal drainage possessed high endemism, unique haplotypes, and extensive divergence . Southerly populations of walleye and yellow perch thus represent unique genetic sources that may prove valuable for conservation.
Walleye populations exhibit moderate divergence levels across the range, as described by previous studies [30, 41, 69]. In contrast, yellow perch (control region mean F
ST = 0.469, μsat = 0.236)  and smallmouth bass (cytochrome b mean F
ST = 0.412, μsat = 0.232) [64, 70] possess much higher divergence among spawning groups. This may be due to their more limited migration [71, 72]. Walleye have been documented to disperse 50–300 km , followed by yellow perch to 48 km, with occasional individuals travelling 200 km , and smallmouth bass only to ~10 km .
The European perch P. fluviatilis Linnaeus 1758 discriminates kin from non-kin via olfactory cues and schools in family groups, which may reproduce together [74, 75]. This life history pattern remains to be tested for yellow perch, walleye, or smallmouth bass. Kinship tests by our laboratory reveal high proportions of full siblings in some spawning groups of yellow perch (mean = 0.20 ranging to 75%)  and smallmouth bass (0.15, to 67%) , which are greater than those identified here for walleye. Limited lifetime migration, and apparent close association among kin for yellow perch and smallmouth bass, may lead to their higher divergences among proximate populations. Walleye populations display intermediate divergence and higher diversity, attributable to more outbreeding. Thus, biogeographic patterns of these widely distributed freshwater fishes largely result from a combination of extrinsic (i.e., changes in climate and drainage patterns) and intrinsic factors (dispersal capabilities, degree of natal homing, population size, and inbreeding), which are regulated by behavior of the species.
Contemporary genetic diversity patterns
Our data indicate relatively high genetic diversity for most walleye spawning groups; these values mirror those reported across their range using nine μsat loci (H
O = 0.68) . Difference between the mtDNA and nuclear data is attributable to the former having a slower evolutionary rate [51, 77], ¼ effective the population size, and being more influenced by population bottlenecks .
Walleye reproducing in the lower Great Lakes (Lakes St. Clair and Erie) have the highest diversity values, reflecting admixture from glacial refugia and larger population sizes, similar to results from other studies [30, 40]. Some small upper Great Lakes populations display the lowest mtDNA diversities, but average nuclear DNA diversities. Notably, the population from the Moon/Musquash River (site L) has the least control region diversity but higher μsat value. Gatt et al.  also recovered low mtDNA diversity at this location, attributing this to stocking and overexploitation. Since this pattern is restricted to mtDNA alone in our study, it likely instead reflects past bottlenecking and small population size.
Yellow perch populations possess much lower mtDNA control region diversity levels across their range (mean H
D = 0.31) than nuclear DNA variability (H
O = 0.53, 15 μsat loci) . This also is true for the related European perch  and Eurasian ruffe Gymnocephalus cernua (Linnaeus 1758) . This difference among percid species may reflect their respective evolutionary history and behavior. Notably, strong association of European perch in kin groups [74, 75] and high proportions of yellow perch full siblings in spawning groups  may lead to lower diversity from inbreeding. Smallmouth bass spawning groups also have lower genetic diversity than walleye in mtDNA sequences (mean H
D = 0.50)  and eight μsat loci (H
O = 0.46) . This may reflect higher association of kin groups and limited lifetime migration . Thus, population genetic diversity and divergence of smallmouth bass [64, 70] and yellow perch  differ from walleye due to their respective reproductive behavior and life history characters.
Small isolated southern populations of walleye have lower genetic diversity values than those in once glaciated regions. The former contain the oldest and most unique genetic variants, reflecting long-term isolation, smaller population sizes, and likely genetic drift and population bottlenecks . Gulf Coastal yellow perch populations also possess the lowest diversity values, along with unique haplotypes and alleles , similar to walleye. These southern populations comprise a historic source of diversity and an important genetic resource for both species. Their adaptations to warmer habitats may provide a critical genetic reservoir in the face of climate warming.
Genetic patterns shaped by anthropogenic factors in Lake Erie
Genetic diversity values for mtDNA and μsat loci are 94% and 37% lower for historic Lake Erie walleye and “blue pike” from 1923–49 compared with contemporary Lakes Erie and Ontario populations. Genetic diversity appears to have increased over the past 70+ years, which may reflect population recovery.
Europeans settled Lake Erie shores during the 1700–1800s, cutting down the forests and draining the marshlands, which disappeared by 1900 [44, 81]. As the region developed and industrialized, untreated wastes were released into the Lake from saw mills, slaughterhouses, and steel factories. By 1830, Lake Erie walleye comprised an abundant and important commercial fishery. In 1874, construction of a major shipping channel drastically modified the Lakes Huron–Erie Corridor connection (Figure 1). Lake Erie steadily lost much of its fish habitat from 1900–1970s due to draining of wetlands, armoring of shorelines, channelization, dredging, and increased industrialization [82, 83]. During the 1960s, high levels of phosphorus caused massive algal blooms, accompanied by oxygen depletion and marked fish die offs .
Many native Lake Erie fish populations experienced steady declines from 1900–1970s, including the lake trout, lake sturgeon Acipenser fulvescens Rafinesque 1817, and walleye [81, 84]. Industrial outputs resulted in heavy metal contamination and declining fish health, manifested with neoplasms, tumors, and lesions on walleye and other species . In 1970, walleye fisheries from Lakes Huron through Erie were closed due to high mercury levels. By 1978, fisheries managers declared Lake Erie walleye as being in crisis from overfishing and pollution .
Lower allelic numbers and diversity for historic walleye and “blue pike” from 1923–49 likely reflect these population declines. The rare haplotypes and alleles we identify in historic walleye may have disappeared. Most haplotypes and almost all alleles in historic Lake Erie samples are common and widespread today. The historic samples appeared to possess more haplotypes that trace to the Atlantic Refugium, whereas those from the Mississippian Refugium dominate today’s walleye populations [21, 25, 41]. This change merits further testing, but may be a response to climate warming.
Genetic diversity of other Great Lakes fishes similarly was lower in 1927–59 and higher in 1995–2005, attributed to population declines from environmental conditions and overexploitation [86, 87]. Notably, lake whitefish Coregonus clupeaformis (Mitchill 1818) from Lakes Huron and Erie in 1927 had lower diversity at seven μsat loci (0.60) than in 1997–2005 (0.65) . Lake trout from Lakes Superior, Michigan, and Huron were less variable in 1940–59 (0.47) than in 1995–99 (0.51) using five μsat loci . Similarly, Atlantic cod Gadus morhua Linnaeus 1758 declined in diversity and number of alleles from 1954–80s, then increased from 1980–98 according to archived otolith samples and three μsat loci . That study moreover documented that genetic composition changed, similar to the pattern here for Lake Erie walleye, hypothesizing recovered diversity via immigration from a nearby spawning group.
Contemporary Lake Erie walleye populations also may have been influenced by migration and recruitment. Walleye is described to natally home, with chemical cues presumably facilitating recognition of reproductive grounds [89, 90]. Olson and Scidmore  discerned lower homing in areas with high habitat degradation. Some contemporary western Lake Erie spawning group samples appear genetically similar to those from Lake St. Clair, suggesting possible genetic exchange. Walleye movement between these lakes during non-spawning times is well known from tagging [91, 92] and Lake Erie walleye may have migrated to spawn at a recently augmented reef in the Detroit River .
Lake Erie walleye appear to have recovered from declines in diversity and numbers (~10 million in 1978), after the fishery's closure from 1970–76 , implementation of the 1970 Canada Water Act, the 1972 U.S. Clean Water Act, and the 1972 Canada-U.S. Great Lakes Water Quality Agreement . Increasing water temperature –especially in the shallow western basin– changed the Lake Erie fish community from cold water (e.g., lake trout) to warmer water species, favoring walleye and yellow perch. Declining numbers of colder water competitors presumably enhanced walleye abundance . By 1988, Lake Erie had rebounded to ~80 million (8x 1978), decreasing to ~18 million walleye in 2013 . Genetic analyses  revealed temporal consistency in genetic diversity from 1995–2008 for three primary Lake Erie walleye spawning groups (Maumee and Sandusky Rivers in the west, and Van Buren Bay in the east). Whether this continues remains to be discerned, as genetic variability may decline with population sizes.
Other walleye populations have shown a mixture of temporal stability and decline. Walleye spawning in Escanaba Lake, Wisconsin displayed consistent diversity levels from 1952–2002 (mean H
O = 0.76) using eight μsat loci  (six of those here). Garner et al.  likewise described consistent diversity levels for walleye in Lake Superior’s Black Bay from 1966–2010 (mean H
O = 0.62) employing nine μsat loci (six of those here). However, the Escanaba Lake and Black Bay populations were stocked, likely circumventing fluctuations. MtDNA restriction fragment length polymorphism diversity of walleye spawning in Lake Huron’s Georgian Bay declined over three decades (0.50 in the 1960s to 0.15 in the 1990s), attributed to exploitation and stocking . Our study recovers similarly low mtDNA control region variability for walleye spawning in the Moon/Musquash Rivers of Georgian Bay today. However, this bottleneck effect is restricted to mtDNA, since we denote average levels of nuclear DNA variability.
Other alternatives may explain lower diversity levels in the historic samples. Our contemporary samples were adults collected from spring spawning runs at specific spawning sites. In contrast, historic samples were collected from July–November, when walleye intermix. Thus genetic diversity may have been lowered due to population admixture via a Wahlund effect . For historically archived samples, such as ours, Nielsen and Hansen  recommended including positive and negative controls, having a separate laboratory space and separate chemicals, testing for null alleles with Micro-Checker, using samples with complete documentation of biological information, testing for HWE, and applying more than one statistical test to validate patterns. We followed all of these precautions to ensure reliability of data from our formalin fixed historic samples. Some studies have documented issues with historic samples having biased amplification of shorter length alleles [99, 100]. We found slight suggestion of null alleles in historic samples, with shorter allele lengths being more prevalent. Lower template quality may have resulted from DNA shearing with formalin fixation , leading to partial repeat amplification if primer sites were unavailable for binding . However, our Micro-Checker tests and other analyses demonstrate lack of statistical support for such problems. Our mtDNA sequences reveal the same pattern as the μsat analyses. Additional analysis of historic walleye and “blue pike” samples from intermediate decades may help to further interpret temporal population genetic patterns.
Taxonomic status of historic “blue pike” and turquoise-mucus variants
We discern that the historic “blue pike” appears genetically indistinguishable from walleye populations. It has no unique genetic variation in our database, rendering S. vitreus “glaucus” invalid. It fails to meet the criteria of the Evolutionary Species Concept (ESC)  or the Phylogenetic Species Concept (PSC) . The “blue pike” is not “an entity that kept its identity from others over time and space and that had its own independent evolutionary fate and historical tendencies” and possesses no bootstrap or posterior probability support, lacks reciprocal monophyly, diagnosable synapomorphies, and demarcation from the walleye as required by the ESC and PSC [103, 104]. It also does not show interspecific variation 10x greater than the mean intraspecific variation of walleye . In fact, the “blue pike” has no mtDNA sequence differentiation and its μsat variation is identical to that among typical walleye spawning groups and populations. In contrast, many walleye populations across North America possess much more pronounced genetic variation, particularly from the New (X) and North Rivers (Y). Walleye spawning in those southerly locations meet more of the criteria of being distinct taxa. However, we regard those in the New and North Rivers as divergent populations of walleye, and not as separate taxa, and believe that most ichthyologists and systematists would concur. Our findings thus indicate that the “blue pike” does not constitute a separate genetic taxon from walleye, and does not merit species or subspecies recognition.
Slight morphological variations between historic walleye and “blue pike” samples suggest some possible population-level differences. However, single individuals of the historic “blue pike” and walleye cannot be identified morphologically or genetically as either “blue pike” or walleye. Their coloration also is not a reliable identification character as it was/is very variable among historic as well as contemporary walleye [44, 48]. Ichthyologists from the era of the “blue pike” reported a large numbers of intergrades in color, as well as among all morphological traits [44, 48, 49].
Some fishes, including lake trout, whitefish Coregonus spp., and Arctic char Salvelinus alpinus (Linnaeus 1758), have been regarded as multiple morphological races that developed through adaptation to northern proglacial lakes , but possess low genetic divergence . The “blue pike” was reported to inhabit deeper waters, have slower growth , and a larger eye [42,this study] compared to walleye. Slower growth likewise characterizes walleye in eastern Lake Erie today . We find that although “blue pike” and walleye display some slight morphological variation, this is rather negligible, and unaccompanied by population genetic distinction, rendering its subspecies status invalid. “Blue pike” were walleye, and fell within the normal range of walleye population variation.
The turquoise-colored mucus walleye from McKim Lake (site D) do not genetically differ from co-occurring yellow walleye. Stepien and Faber  likewise analyzed several assorted turquoise mucus walleye from a variety of Canadian Shield lakes using entire mtDNA control region sequences and found no genetic distinction from the normal variation of walleye. Paradis and Magnan  morphologically compared sympatric yellow and turquoise mucus walleye in five Canadian Shield lakes near Quebec, reporting longer head lengths and smaller interorbital distances in some of the latter. However, early fishery biologists found that these turquoise mucus walleye did not possess the morphological characteristics of “blue pike” . Laporte et al.  alleged slight genetic difference between turquoise mucus and yellow walleye populations sampled within a lake using amplified fragment length polymorphism markers and assignment tests, but lacked diagnostic alleles and their genetic distance analyses showed no significant bootstrap support. It may be that there are some population-level variants within some lakes across the range of the Canadian Shield; many such distinctions among walleye spawning groups are found in the present study and others [30, 39], but these do not warrant taxonomic recognition.
Occasional steel-blue colored walleye regularly are reported from Lake Erie and other waters, including the Ohio River drainage [44, 48]. Yellow perch that are dark blue in color also co-occur . We analyzed mtDNA control region from a steel-grey/blue walleye individual sampled in western Lake Erie near Sandusky OH and found it had mtDNA haplotype 1, the most common walleye haplotype. A skin scraping revealed no turquoise mucus. Wayne Schaeffer (pers. comm., University of Wisconsin, September 2013) also found no turquoise mucus or sandercyanin in Lake Erie walleye using the methods of Yu et al. . Overall, no diagnosable genetic or morphological characters have been found that distinguish historic “blue pike” from walleye, rendering its subspecies status invalid.
Effects of climate change on walleye populations
Global temperatures are predicted to increase over the next 50 years, with the Great Lakes region rising by 5–5.5°C, becoming more like today’s Gulf Coast . Today, Lake Erie houses the largest walleye abundance  and greatest genetic diversity. Increased temperatures are predicted to shift walleye distribution northward . Fringe populations may experience declines and increased isolation, with bottlenecks and drift reducing genetic variation, accompanied by loss of unique haplotypes and alleles. Hence, valuable genetic resources may disappear as unconnected populations become sequestered.
High connectivity in Great Lakes’ watersheds allows ample dispersal opportunities, which may homogenize gene pools of distinctive spawning groups as they move northward and mix, producing a Wahlund effect. Thus, climate change may lead to decline of divergence patterns from today’s walleye spawning groups. Walleye likely will remain abundant and adapt, but unique variants may be lost. Common alleles may increase in frequency, raising concerns for retaining adaptive potential, which should become a management priority. It might be possible to utilize unique warm-adapted variants to the south (e.g., the North River, Y) and southeast (Ohio/New Rivers, W–X) to aid future walleye populations.