Isotopic niche size of Coregonus artedi (sensu lato) Increases in the presence of Mysis diluviana, expanded habitat use and phenotypic diversity

Abstract Post‐glacial colonization of lakes in Algonquin Park, Ontario, Canada resulted in food webs with cisco (Coregonus artedi sensu lato) and either Mysis diluviana or Chaoborus spp. as the dominant diel migrator. Mysis as prey, its diel movements and benthic occupancy, are hypothesized to be key elements of ecological opportunity for cisco diversity in the Laurentian Great Lakes. If correct, the hypothesis strongly implies that lakes with Mysis would have greater trophic niche size and drive greater adaptive radiation of cisco forms relative to lakes without Mysis. The dichotomy in diel migrator in Algonquin Park lakes was an opportunity to assess the isotopic niche size of cisco (δ15N and δ13C) and determine if niche size expands with Mysis presence. We found the presence of Mysis is necessary to expand isotopic niche size in our study lakes. The use of habitats not typically associated with the ancestral form of cisco (e.g., benthic habitats) and phenotypic diversity (blackfin and cisco) also continue to expand niche size in Mysis‐based food webs. Partial ecological speciation based on a large niche space appears to be present in one lake (Cauchon Lake) where use of alternative habitats is the only real difference in cisco. The presence of blackfin expands niche space in Cedar and Radiant Lakes. This was not matched in Hogan Lake where niche space was relatively smaller with similar forms. Possible reasons for this discrepancy may be related to the asymmetric basin of Hogan Lake and whether the two forms overlap during cool and cold‐water periods of the annual temperature cycle. By comparing trophic niche size among lakes with and without Mysis, we conclude that Mysis provides a key ecological opportunity for cisco diversity in our study lakes and likely more widely.

occupancy, are hypothesized to be key elements of ecological opportunity for cisco diversity in the Laurentian Great Lakes. If correct, the hypothesis strongly implies that lakes with Mysis would have greater trophic niche size and drive greater adaptive radiation of cisco forms relative to lakes without Mysis. The dichotomy in diel migrator in Algonquin Park lakes was an opportunity to assess the isotopic niche size of cisco (δ 15 N and δ 13 C) and determine if niche size expands with Mysis presence. We found the presence of Mysis is necessary to expand isotopic niche size in our study lakes. The use of habitats not typically associated with the ancestral form of cisco (e.g., benthic habitats) and phenotypic diversity (blackfin and cisco) also continue to expand niche size in Mysis-based food webs. Partial ecological speciation based on a large niche space appears to be present in one lake (Cauchon Lake) where use of alternative habitats is the only real difference in cisco. The presence of blackfin expands niche space in Cedar and Radiant Lakes. This was not matched in Hogan Lake where niche space was relatively smaller with similar forms. Possible reasons for this discrepancy may be related to the asymmetric basin of Hogan Lake and whether the two forms overlap during cool and cold-water periods of the annual temperature cycle. By comparing trophic niche size among lakes with and without Mysis, we conclude that Mysis provides a key ecological opportunity for cisco diversity in our study lakes and likely more widely.

K E Y W O R D S
cisco diversity, ecological speciation, isotopes

| INTRODUC TI ON
Ecological opportunity is central to the diversifying process of adaptive radiations. Opportunity can be conceived as competitor-free habitat and abundant resources for colonizers of islands or lakes (Losos, 2010;Martin & Wainwright, 2013;Stroud & Losos, 2016).
Under these conditions, stages of radiation begin with divergence in habitat occupancy followed by trait divergence (Gavrilets & Losos, 2009;Siwertsson et al., 2013). One outcome stemming from ecological opportunity can be expanding trophic niche space allowing for diversification of an ancestral phenotype (Schluter, 1996).
For Anolis, perching and habitat location as a function of vegetation structure represents the behavioral/habitat axis for diversification in repeated settings (Losos, 2010). For ciscoes, diel migration and foraging depth likely represent their axis for diversification repeated in several Laurentian Great Lakes (Blanke et al., 2018;Eshenroder et al., 1998;Hudson et al., 2017;Schmidt et al., 2011).
Historical contingency can also extend to the presence or absence of organisms that affect or disrupt organismal size distributions and trophic transfer. This may be particularly important in aquatic ecosystems where the size spectrum of body size and abundance are near universal features (Kerr & Dickie, 2001;.
The presence or absence of keystone organisms affecting food web structure may in turn determine the presence and/or extent of adaptive radiations. Natural "experiments" in assemblage structure produced via species introductions or post-glacial dispersal can provide insights into aquatic food web function/disruption and potentially adaptive radiations of fish.

Mysis diluviana (hereafter Mysis) is a keystone species function-
ing as a food web engineer capable of exerting strong influence on aquatic food web structure and function. It is a diel migrating, predatory crustacean that structures lake food webs because it governs plankton size structure as a predator, and fish productivity and trophic status as a prey (Almond et al., 1996;Nero & Sprules, 2009;Vander Zanden et al., 1999). Mysis introductions in lakes have disrupted food web function including plankton and planktivorous fish size structure (Devlin et al., 2017;Lasenby et al., 1986), severely reduced herbivorous cladocerans (Spencer et al., 1999), contributed to the extirpation of planktivorous fish (Lasenby et al., 1986;Spencer et al., 1991) and severed aquatic/terrestrial trophic linkages at whole lake scales (Devlin et al., 2017;Spencer et al., 1991).
Mysis has also been identified as a key factor in cisco diversification in the Laurentian Great Lakes (Eshenroder et al., 1998;Eshenroder & Burnham-Curtis, 1999). Indeed, documentation of cisco form radiation in North American lakes has to date been in lakes with Mysis as the dominant diel migrator (e.g., Eshenroder et al., 2016;Etnier & Skelton, 2003;Piette-Lauzière et al., 2019).
Buoyancy maintenance by cisco is hypothesized as the selective mechanism for exploiting Mysis at depth (Eshenroder et al., 1998).
The strong implication of this hypothesis is the greater trophic niche size afforded cisco in Mysis-based food webs that would sustain the adaptive radiation of different cisco forms.
The distribution of Mysis in North American lakes is historically contingent on landscape coverage and drainage of post-glacial lakes that formed after the last glacial maximum. Contemporary lakes with Mysis are relatively deep (depth > 20-30 m) and below elevations of post-glacial lake inundation in drainage networks (Dadswell, 1974;Martin & Chapman, 1965). Above the inundation elevation, Mysis is absent and Chaoborus punctipennis (hereafter Chaoborus) is present as the dominant diel migrating predatory zooplankton in many lakes . In a recent lake survey of one such landscape, Algonquin Park in Ontario, Canada, cisco occur in lakes above and below the inundation elevation (approx. 381 m asl) of the historical drainage outflow for post-glacial Lake Algonquin Ridgway et al., 2017). Above the elevation of inundation, Chaoborus was present, co-occurring cisco were monomorphic, small and occupied the pelagic habitat as would normally be associated with this species (Scott & Crossman, 1998). Below this elevation, Mysis was present, and cisco were either monomorphic in some lakes or showed phenotypic diversity with one form matching pelagic cisco and the other converging on a larger, Mysis predator with distinct black coloration on dorsal and paired fins and greater gill raker numbers Piette-Lauzière et al., 2019). In some Mysis-based food webs, monomorphic ciscoes were captured in the pelagic zone as would be typical of this species while others were captured in bottom set nets that is atypical for this region. Following Eshenroder et al. (2016), the term "form" has been adopted here to describe cisco with distinct phenotypes arising from in situ ecological speciation within C. artedi (sensu lato) as the recognized ancestral form. Here we refer to the ancestral, small pelagic form as cisco and the larger Mysis predator that has evolved in several lakes as blackfin Piette-Lauzière et al., 2019).
Diel movements and size/phenotypes of cisco (sensu lato) show diverse patterns whether Chaoborus or Mysis are the dominant diel migrators. In Chaoborus-based food webs, cisco size is related to zooplankton density with complex diel patterns of movement including no diel movement (Ahrenstorff et al., 2013). In non-Mysis food webs in a Minnesota lake region (Chaoborus may or may not be present), size and shape of cisco are related to overall lake productivity with use of shallow habitat occurring in lakes with inadequate hypolimnetic oxythermal habitat (Jacobson et al., in review).
In response, cisco fin shape and size shifted to accommodate a more substrate-based food source. The typical small pelagic form of cisco was found in oligotrophic lakes with enough hypolimnetic dissolved oxygen.
In Mysis-based food webs, such as the Laurentian Great Lakes, cisco have radiated into different forms along a depth gradient (Eshenroder & Burnham-Curtis, 1999;Eshenroder et al., 2016).
Exploiting Mysis during diel movement or during benthic stages is an important element of Laurentian Great Lake cisco diversity (Ahrenstorff et al., 2011;Hrabik et al., 2006;Jensen et al., 2006;Sierszen et al., 2014). Determining the mechanism(s) governing trade-offs in diel movement and preferred resting depths for cisco in any lake food web are complex and establishing the relative importance of factors such as minimizing predation, maximizing growth potential, or optimizing oxythermal habitat selection all have different levels of support in different lake ecosystems (Ahrenstorff et al., 2013;Hrabik et al., 2006;Jensen et al., 2006). While advances are being made in understanding cisco diversity (e.g., Ahrenstorff et al., 2013;Piette-Lauzière et al., 2019;Turgeon et al., 1999Turgeon et al., , 2016, insight into how food web structure fundamentally drives cisco diversity remains an elusive topic. In this study, we return to the early observation of Mysis in diets of different cisco forms as originally described by Koelz (1929) and the hypothesis that Mysis is important for cisco diversity (Eshenroder et al., 1998). The historical contingency of Mysis presence/absence in lake food webs in Algonquin Park and cisco distribution in the region allows for a comparison of cisco trophic ecology between Mysis and Chaoborus lake food webs. We compare cisco in Mysis versus Chaoborus food webs using isotopic trophic niche  (Jackson et al., 2011;Swanson et al., 2015) and pose several hypotheses. Because cisco also occupy different habitats and/or have different phenotypic forms in Algonquin Park lakes with Mysis Piette-Lauzière et al., 2019), we hypothesize that cisco (sensu lato) in these lake food webs will show greater isotopic niche size corresponding to greater habitat and/or phenotypic diversity (Eshenroderet al., 1998;Eshenroder & Burnham-Curtis, 1999), relative to Chaoborus-based food webs. Finally, we expect to find differences in the trophic niches of sympatric forms or fish found in pelagic versus benthic habitats within Mysis lakes. Given the evidence uncovered in this study, we expand on the Mysis/depth hypothesis for cisco diversity. Figure 1 shows the native distribution of cisco on the Algonquin Park landscape based on recent and historical lake surveys . Cisco occupy lakes within the valley system that drained pro-glacial Lake Algonquin from the Fossmill Outlet in the west to the east (approx. 13,000-12,000 cal yrs BP) as represented by the lower elevation (dark landscape on the gray scale in Figure 1) (Dyke, 2004;Harrison, 1972;Karrow, 2004;Lewis et al., 1994). Cisco above this elevation are native populations (no stocking history) and presumably arrived via upstream movements during Lake Algonquin drainage. Mysis occupancy in lakes below 381 m asl (red dotted line in Figure 1) and absence above this elevation is the "fingerprint" of Lake Algonquin inundation in the valley (Martin & Chapman, 1965;Dadswell, 1974). Because cisco distribution is centered on the Lake Algonquin drainage system, we presume this defines the entry point of cisco to the landscape via the Fossmill Outlet (Mandrak & Crossman, 1992). Lakes used in this study are numbered in Figure 1 and their characteristics are listed in Table 1.

| Study Lakes
Carl Wilson and Catfish lakes, both with Chaoborus, are situated in different watersheds and are directly connected by drainage to the lower elevation lakes (#5 and #7; Figure 1). All other lakes are in the lower elevation region. Manitou and Three Mile lakes have been separated from the other lakes in this drainage system for at least 10,000 cal yrs BP (Harrison, 1972;Ridgway et al., 2017). Similarly, the Mink Lake Sill between Mink Lake (#3; Figure 1) and Cauchon Lake (#4; Figure 1) has separated their respective watersheds for approximately 12,000 cal yrs BP. Hogan Lake (#8; Figure 1) has been separate from all other lakes for a similar length of time (Ridgway et al., 2017). All other lakes drain eastward in the Petawawa R drainage with a series of waterfalls preventing any upstream movement beginning with Grand Lake (#10; Figure 1).

| Netting surveys
Netting surveys to acquire isotope samples followed the same protocol as described in Bell et al. (2019). Briefly, bottom set gillnets were used as part of the park-wide survey of fish in lakes. Bottom net surveys followed a depth stratified random design allocating netting effort in proportion to the aerial surface area of depth strata (1-3 m, 3-6 m, 6-12 m, 12-20m, 20-30 m, 30+ m). Nets (25 m long × 2 m high; 64 m long × 2 m high) were set for 1 or 2 hr and consisted of a mesh series ranging from 38 to 127 mm mesh panels . Some lakes were surveyed for ciscoes using nets with 38, 51, and 64 mm mesh only to increase sample size. Pelagic nets of similar mesh series were set in mid-lake over a deep basin in early evening hours in all study lakes. Fish were removed from the nets and placed on ice for processing on shore. Tissue samples were stored on ice in the field and later frozen.

| Tissue sampling and processing
All samples were obtained between 2010 and 2016 by members of the Harkness Laboratory of Fisheries Research. A sample of back muscle without skin or scales was taken from fish (see Table 2 for sample sizes). Mussels (Elliptio complanata) (N = 8-19) were sampled from each lake in shallow water (≈1 m) during the week of netting.
Mussels were sampled to acquire baseline estimates of carbon (δ 13 C)

| Stable isotopes measurements
Measurements of stable isotopes were reported in parts per thousand (‰) relative to the international standards (Vienna Pee Dee Belemnite for carbon and atmospheric air for nitrogen). Isotopes values were also normalized using secondary standards (see SINLAB procedures) to allow comparisons over multiple years.
A carbon and nitrogen ratio (C:N) between 3.5 and 4.5 (this study: 3.9) can indicate interference of lipids in samples. Several arithmetic corrections have been suggested to correct the δ 13 C values based on the correlation of lipid content and the C:N ratio (Kiljunen et al., 2006;Logan et al., 2008;McConnaughey & McRoy, 1979). A previous study on ciscoes (Schmidt et al., 2011)  We did the same with our samples.

| Baseline correction
Lakes differ in their δ 13 C and δ 15 N at the base of their food chain (e.g., mussel tissue profiles). Thus, before attempting among lake comparisons, corrections of these baselines must be performed. Cabana and Rasmussen (1996) recommended unionid mussels (such as E. complanata) for the baseline correction. E. complanata has a slow nitrogen turnover due to their long life and relatively large size. We used biplots of mussel δ 13 C and δ 15 N by lake to determine centroid points as baselines (Table 3). The lake-centroid δ 13 C and δ 15 N from mussels was subtracted from individual cisco δ 13 C and δ 15 N signatures to obtain relative isotopic values for the consumers (i.e., ciscoes).

| Analysis
Isotopic We first combine all cisco in a lake to represent the overall C. artedi (sensu lato) isotopic niche, regardless of capture location or phenotypic form. This species is widely recognized as diverse and deserving of the taxonomic qualifier "sensu lato." Therefore, combining all cisco allowed for among lake comparisons. Our intention was to describe the C. artedi niche space broadly defined for each lake to assess whether niche space changed across the lake set. We interpreted non-overlap of the 95% credible intervals as a significant or strong difference when comparing isotopic niche size among lakes or between lakes with Chaoborus versus lakes with Mysis. Second, we estimated isotopic niche overlap between ciscoes captured in benthic or pelagic nets (sample sizes in Table 4). This overlap was expressed as a probability using nicheROVER (R package; Swanson et al., 2015). We choose nicheROVER because of the advantages of probability comparison (and 95% credible intervals) over geometric or space comparisons (Swanson et al., 2015). Overlap comparisons (mean, 95% credible intervals) are summarized as, for example, cisco captured in benthic habitat overlapping with pelagic captured cisco from the same lake. Similarly, cisco captured in pelagic habitat overlapping with cisco captured in benthic habitat.
nicheROVER allows for asymmetric comparisons of this kind which is useful in assessing potential directionality in how niche space may have shifted in two-way comparisons.
Third, for lakes with both pelagic and benthic captures of ciscoes (sample sizes in Table 5), we used principle component analysis (PCA) to examine the association between phenotype (length, shape, total including whether individual cisco were captured in pelagic or benthic nets. Landmarks were used to adjust for size and incorporate shape. Briefly, size effects were removed using ordinary least square residuals from all linear measurement's versus standard length for all lakes combined (Reist, 1985). Standard length was based on landmarks 1 and 2 in figure 2 of Piette-Lauzière et al. (2019). Accounting for size in this manner was consistent with earlier publications on cisco morphology (Muir et al., 2013;Reist, 1985). Briefly, shape analysis was based on Procrustes superimposition initially by alignment on landmarks 1 and 2 on each fish sample and then superimposed on their centroids. Procrustes coordinates and shape variation generated a covariance matrix that was then used in the PCA analysis as We determined the number of components to retain from the PCA by the broken stick model (Peres-Neto et al., 2003). This statistical procedure simulates randomly broken K pieces (K = number of variables) from a stick. Pieces sorted in descending order of proportion represent the threshold level of the random distribution of explained variance. Significant PC components were analyzed by ANOVA with habitat (benthic or pelagic capture) and lake (Cedar, Hogan, Radiant, and Cauchon Lks) as factors to assess the relative contribution of each to the multivariate patterns exposed by the PCA. Based on the ANOVA, within lake post hoc comparisons were performed using Tukey HSD between pelagic and benthic-captured cisco. Post hoc comparisons among lake were also conducted to assess any differences that may exist at the lake level.
Finally, given the results of the PCA and the potential association between phenotype and environment among lakes, we further ex-  Note: Gill raker form defined as the low raker form (gill rakers counts ranging from 40 to 51) and high raker form (gill rakers counts = 52+) based on Bell et al. (2019). Not all benthic or pelagic net samples identified in Table 3 were counted for gill rakers. The lower 95% CI of Cedar Lake SEA B barely overlapped with the upper 95% CI of Manitou Lake SEA B .

| RE SULTS
The exception to increases in isotopic niche space among lakes with phenotypic diversity is Hogan Lake (Figure 2). It contains both blackfin and cisco forms with fish captured in both pelagic and benthic deployed nets. The isotopic niche space (med. SEA B = 1.13, 95% CI 0.85, 1.44) was equivalent to other lakes with monomorphic cisco but smaller than the other phenotypically diverse populations.
Four of the Mysis-based cisco populations had enough catches in both pelagic and benthic nets to estimate niche size for fish captured

F I G U R E 2
The Standard Elipse Area (SEA B ; o/oo 2 ) of cisco (C. artedi sensu lato) for the lake set. The dominant diel migrator is either Chaoborus (=C) or Mysis (=M) among the lake set. Single phenotype refers to cisco that are the typical form for the species including pelagic only habitat occupancy, gill raker distribution, and size. Pelagic and benthic habitat refers to cisco captured in bottom set gillnets and pelagic gillnets. Phenotypic diversity refers to either two forms (blackfin and cisco forms; Cedar and Hogan Lakes) or blackfin form primarily with relatively few cisco forms (Radiant Lake). Box plots of posterior estimates for SEA B are presented with the median (point), 50% credible intervals (smallest box), 75% credible intervals, and 95% credible intervals (largest box) in each of the two habitats ( Figure 3). Two lakes with blackfin (Mink and Radiant) did not have enough pelagic catches to make a comparison of niche size between fish in the two habitats. Among the four Mysis lakes, isotopic niche size in general appeared to vary more on the δ 13 C axis than on the δ 15 N axis overall ( Figure 3). All four lakes showed overlap in isotopic niche size with the greatest overlap occurring in Three Mile Lake (Table 6; Figure 3). Both pelagic-and benthic-captured cisco in Three Mile also had similar sized isotopic niches (1.22 and 1.08 SEA B , respectively; Table. 3) accompanying high levels of overlap (Table 6; Figure 3). Although Cauchon Lake is regarded as having a single phenotype of cisco, the probability of overlap for fish captured in benthic and pelagic nets was lowest among the lake set (prob. of overlap with benthic = 33.85; with pelagic = 36.45; Table 6; Figure 3). For both Three Mile and Cauchon Lakes, there was no clear indication of directionality in overlap between cisco captured in different habitats although they strongly differed in the magnitude of overlap (Table 6).
There is directionality in niche space between benthic and pelagic cisco in Cedar and Hogan Lakes with pelagic cisco niche space being smaller and largely contained within the larger niche space of benthic-captured cisco ( The probability of overlap by benthic cisco with pelagic cisco in both lakes was less than 50% (Cedar Lake prob of overlap = 48.9; Hogan Lake prob. = 41.4;  Table 6).
We examined PCA ordination patterns among the lake set in Figure 3 except for Three Mile Lake which showed extensive niche overlap (Table 6; Figure 3). We added Radiant Lake benthic captures of blackfin to the analysis to assess their position relative to lakes with blackfin and cisco forms (Cedar and Hogan Lakes).
The first two components of the PCA accounted for 66.  TA B L E 6 Summary of isotopic niche size as standard ellipse area (SEA B ) and probability of overlap for ciscoes captured in pelagic and benthic nets (see Figure 3) F I G U R E 4 PCA analysis of ciscoes in lakes where pelagic and benthic nets captured ciscoes. Phenotypic (size, shape, and gill raker number) and isotopic niche data are included in the PCA where all variables were recorded for each specimen. Morphological data based on digitized landmarks from Piette-Lauzière et al. (2019). Insert is the pattern of variable loading of PC1 and 2 in the PCA and associated gray scale is the percent of variation accounted for in the loadings than cisco ( Figure 5; blackfin SEA B = 1.04; 95% CI, 0.56-1.69 vs. cisco SEA B = 0.70; 95% CI, 0.5-0.95). The probability of Hogan Lake cisco being in the niche space of Hogan Lake blackfin was 32.0 (95% CI, 12.4-50.2) while the probability of a blackfin being in the niche space of cisco was 46.4 (95% CI, 26.5-75.1), a nearly identical overlap probability for Cedar Lake blackfin in cisco isotope space.

| D ISCUSS I ON
Isotopic niche size of cisco at least doubled in size in lakes with Mysis relative to lakes with Chaoborus, the two dominant diel migrators in the lake set. Use of different habitats (benthic and pelagic) and the presence of different forms (cisco and blackfin) further increased isotopic niche size for cisco in Mysis-based food webs. These results support the hypothesis that Mysis presents ecological opportunity for the ancestral form of cisco leading to increased use of different habitats in some lakes, and arguably phenotypic diversity as well (Eshenroder et al., 1998;Eshenroder & Burnham-Curtis, 1999).
Blackfin in Algonquin Park prey on Mysis , and we found blackfin to have a higher trophic position relative to coexisting cisco indicating that Mysis have inserted a trophic position step relative to pelagic cisco in the same lake (Vander Zanden et al., 1999). The results of this study have three important implications for understanding cisco niche diversity that for decades proved challenging because of the range of forms, complex glacial history, and taxonomic confusion.
First, niches of coregonine fishes following ecological speciation, including isotopic niche, are relatively conservative and repeatable (Harrod et al., 2010;Siwertsson et al., 2013). We found repeatable patterns of relative overlap of isotopic niches between cisco and blackfin in Cedar and Hogan Lakes, two ecosystems functionally separate for the past 12,000 cal yrs BP (Table 6; Figure 5). Furthermore, the benthic blackfin had a larger SEA B than co-occurring cisco, with pelagic cisco greatly overlapping with blackfin (81% and 97% overlap for Cedar and Hogan Lks, respectively; Table 6). Blackfin overlapped much less with cisco in their isotopic niche (49% and 41% for Cedar and Hogan Lakes, respectively) pointing to blackfin as occupying different and wider isotopic niche space relative to cisco. For cisco in relatively small inland lakes, ecological speciation has repeatedly resulted in pelagic cisco, typical of the spe- Niche conservatism has been detected in lakes where loss or recovery of coregonine forms has occurred. In the Laurentian Great Lakes (Blanke et al., 2018) and Lake Constance (Vonlanthan et al., 2012), loss of phenotypic diversity with the disappearance of recognizable forms resulted in contraction of overall isotopic niche space among remaining forms rather than expansion among surviving forms to occupy the vacant isotopic niche. In Lake Constance, coregonine phenotypic diversity and isotopic niche have recovered under improving water quality conditions with the re-occurrence of a form once thought lost in the lake (Jacobs et al., 2019;Vonlanthen et al., 2012). These outcomes point to relative niche conservatism among coregonines as a repeatable element of the phenotype-environment correlation (Harrod et al., 2010), and as found with Anolis lizards (Gavrilets & Losos, 2009;Losos, 2010), cisco diversity is repeatable among lakes under similar environmental conditions.
Second, cisco diversity is associated with Mysis-based lake food webs. To date, descriptions of cisco diversity, including based on gill raker counts (e.g., this study; Eshenroder et al., 2016;Etnier & Skelton, 2003;Turgeon & Bernatchez, 2003;Turgeon et al., 2016), have been from lakes with Mysis as a member of the zooplankton assemblage. In some F I G U R E 5 Plausible isotopic niche space for blackfin and cisco forms in Cedar and Hogan Lakes based on bimodality in gill raker number . Axis values are δ 15 N and δ 13 C relative to baseline values for each lake (δ 15 N fish − δ 15 N base ; δ 13 C fish − δ 13 C base ). Note the scale of relative δ 13 C varies with each figure panel cases, descriptions of Mysis in the diet of some cisco forms are part of the diversity assessment. Cisco isotopic niche size in the two lakes with Chaoborus as the dominant diel migrator was less than 50% of the isotopic niche size of monomorphic cisco in two lakes with Mysis. We found cisco diversity, both behavioral and phenotypic, to be associated with Mysis lakes in our study region of Algonquin Park. This includes two lakes (Three Mile and Cauchon) where cisco without any clear morphological differences were foraging in different habitats (Figures 2   and 3). This agrees with the proposition that behavioral changes, detected here as foraging habitat, precedes morphological divergence . Mysis expands isotopic niche, but not always in the same morphological direction. In White Partridge Lake of Algonquin Park, Turgeon et al. (2016) found another pair of cisco forms that co-occur with Mysis. In this lake, the benthic form appears to converge in many morphological characteristics with shortjaw. However, the presence of Mysis does not necessarily lead to phenotypic or habitat use diversity. Two Mysis lakes in this study (Grand and Manitou Lakes; Park lakes relative to cisco co-occurring in Cedar and Hogan Lakes. Hogan Lake SEA B was lower than Cedar and Radiant Lakes and appeared to be a departure from the pattern of increasing isotopic niche size corresponding to phenotypic diversity ( Figure 2). One possible explanation may be the strongly asymmetric basin of Hogan Lake where one smaller region of the lake has sufficient bathymetric depth to retain a full cold-water community (including ciscoes, lake whitefish and lake trout) with the dominant predator, lake trout, confined to this basin during periods of lake stratification (Dolson et al., 2009;McCann, 2012). For parts of the annual temperature cycle, cold-water fish are excluded from large areas of asymmetric lakes. The annual sequence of seasonal confinement to deeper but limited cold-water habitat in summer, followed by release in winter and access to previously restricted foraging habitat, may generate spatial structure for cisco and blackfin on a seasonal basis that differs from single basin lakes. For an isotopic niche, this may result from seasonal contrasts in prey type, such as summer that includes Mysis, versus winter that allows for foraging in large shallow areas of asymmetric lakes without Mysis and alternative prey including benthic prey.
Cisco in Cauchon Lake forage in the pelagic zone and in benthic habitat. The PCA analysis indicated some differentiation between the two foraging behaviors ( Figure 4) and relatively little overlap in isotopic niche size (Table 6; Figure 3). There was insufficient phenotypic or genetic diversity to conclude two forms are present as occurred in Cedar and Hogan Lakes (Piette-Lauzière et al., 2019). As well, gill raker counts are unimodal and consistent with C. artedi . Cauchon Lake is relatively deep for its surface area (mean depth = 13.5 m; max. with smaller stages migrating vertically at night and larger stages remaining closer to bottom (Bowers, 1988). The complex movement and habitat occupancy by Mysis as a function of depth suggest multiple fitness peaks based on a trade-off minimizing vulnerability to predation and maximizing growth potential at different depth zones.
If the pattern of Mysis distribution detected in Lake Champlain is a general occurrence, then depths of 100 m or greater may provide fitness peaks in the cisco/ Mysis predator prey system that can support cisco diversification as a function of depth-an expansion of the earlier hypothesis of cisco buoyancy evolution to exploit Mysis at depth (Eshenroder et al., 1998;Eshenroder & Burnham-Curtis, 1999).
Multiple fitness peaks for Mysis as a function of depth also suggest multiple fitness peaks for cisco both as predators in pursuit of Mysis as prey and as prey evading predation from species such as lake trout. In this study, lakes were all less than 100 m in depth yet several possessed cisco diversification represented as expanded habitat use or phenotypic diversification stemming from ecological speciation Piette-Lauzière et al., 2019). Other lakes in this study with Mysis did not show any diversification in habitat use or phenotype. The presence or absence of cisco diversity may reflect differences in Mysis density among lakes and/or whether multiple fitness peaks persist for Mysis, perhaps related to basin morphology and depth. The density of predatory fish and their ability to detect prey including cisco at depth are also likely contributing to cisco habitat utilization as shown in other studies (e.g., Ahrenstorff et al., 2011;Hrabik et al., 2006;Jensen et al., 2006).
This study did not incorporate enough lakes to quantitatively assess the strength of these factors in driving cisco diversity. In inland lakes with Mysis, cisco forms may evolve as purely benthic foragers (Turgeon et al., 2016) or as blackfin as we found in some Algonquin Park lakes.
Differences among lakes in cisco forms may reflect differences in Mysis habitat residency or migratory behavior and corresponding ecological opportunities for cisco diversification.
Cisco diel movements can be complex in lakes with or without Mysis. Cisco partial daily vertical migration, no migration, and other types of behavior have been found in lakes with Chaoborus or other planktonic assemblages but without Mysis (Ahrenstorff et al., 2013).
We did not find any indication of expanded habitat use by cisco in the two Chaoborus lakes in this study. Chaoborus and other planktonic assemblages may not provide consistent fitness peaks as a function of depth or prey profitability that is potentially present in Mysis-based food webs. Given the capacity of Mysis to act as a food web engineer (Vander Zanden et al., 1999), the resulting planktonic composition and size structure may provide an adaptive trophic landscape that persists through time (or minimizes variability) that in turn provides ecological opportunity for cisco. This may also define the replicated diversity noted for cisco across inland and large lake ecosystems with Mysis (Piette-Lauzière et al., 2019; Turgeon & Bernatchez, 2003;Turgeon et al., 1999Turgeon et al., , 2016, and not colonization via glacial lake flooding that has historically been invoked, including the assumption of shared phylogenetic history, to account for consistency in cisco diversity across landscapes (Clarke, 1973;Dymond & Pritchard, 1930;Etnier & Skelton, 2003;Smith & Todd, 2004).
In conclusion, ecological opportunity is central in adaptive radiations including expanded habitat use and eventual clade diversification (Stroud & Losos, 2016). This pattern has been shown repeatedly in many cases, including Anolis lizards (Mahler et al., 2010;Martin & Wainwright, 2013), several fish species (Schluter, 1996;Schluter & McPhail, 1992), and other organismal groups. We found Mysis expands cisco isotopic niche size relative to cisco in Chaoborus-based food webs and significantly so in 6 of 7 Mysis lakes. Isotopic niche size continued to increase for cisco with expanded habitat use and phenotypic diversity. This pattern of expanded trophic niche is a necessary condition for the Mysis hypothesis to account for not only diversity in the Laurentian Great Lakes but in other lake ecosystems as well.

ACK N OWLED G M ENTS
We thank Emily Cowie, Nick Lacombe, Krystal Mitchell, Lee Scholl, Darren Smith, Courtney Taylor, and many summer students who assisted in the field and at the laboratory bench. Funding was provided by the Great Lakes Fishery Commission and Ontario Ministry of Natural Resources and Forestry.

CO N FLI C T O F I NTE R E S T
None declared.

DATA AVA I L A B I L I T Y S TAT E M E N T
Sample isotopic data, PCA input data and R code are available in